Differential Geometry of Warped Product Manifolds and Submanifolds

b2530 International Strategic Relations and China’s National Security: World at the Crossroads
This page intentionally left blank
b2530_FM.indd 6 01-Sep-16 11:03:06 AM

Published by
World Scientific Publishing Co. Pte. Ltd.
5 Toh Tuck Link, Singapore 596224
USA office: 27 Warren Street, Suite 401-402, Hackensack, NJ 07601
UK office: 57 Shelton Street, Covent Garden, London WC2H 9HE
Library of Congress Cataloging-in-Publication Data

Names: Chen, Bang-yen.
Title: Differential geometry of warped product manifolds and submanifolds /
by Bang-Yen Chen (Michigan State University, USA).
Description: New Jersey : World Scientific, 2017. |
Includes bibliographical references and indexes.
Identifiers: LCCN 2017017199 | ISBN 9789813208926 (hardcover : alk. paper)
Subjects: LCSH: Geometry, Differential. | Riemannian manifolds. | Submanifolds. | Tensor products.
Classification: LCC QA641 .C468 2017 | DDC 516.3/62--dc23
LC record available at https://lccn.loc.gov/2017017199
British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library.
Copyright © 2017 by World Scientific Publishing Co. Pte. Ltd.

All rights reserved. This book, or parts thereof, may not be reproduced in any form or by any means,
electronic or mechanical, including photocopying, recording or any information storage and retrieval
system now known or to be invented, without written permission from the publisher.
For photocopying of material in this volume, please pay a copying fee through the Copyright Clearance
Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to photocopy
is not required from the publisher.
Printed in Singapore
LaiFun - 10419 - Differential Geometry.indd 1 02-05-17 12:27:33 PM

March 21, 2017 15:26 Diff. Geometry of Warped Product Manifolds-10419 ws-book9x6 page v
In memory of Professors
S. S. Chern, T. Nagano, T. Otsuki and K. Yano
who had the most important influence on my research
v
b2530_FM.indd 6 01-Sep-16 11:03:06 AM

March 21, 2017 15:26 Diff. Geometry of Warped Product Manifolds-10419 ws-book9x6 page vii
Foreword
For almost half a century now I am being a privileged mind witness of the
ongoing exceptional geometrical creativity of Distinguished MSU Professor
Bang-Yen Chen. Similarly as has been attempted in the Forewords to Pro-
fessor Chen’s previous books “Pseudo-Riemannian Geometry, δ-Invariants
and Applications” (2011) and “Total Mean Curvature and Submanifolds of
Finite Type” (2015), also in the Foreword to his present book on the geom-
etry of warped product manifolds and submanifolds I will try to describe a
personal perception of its contents from a somewhat general cultural point
of view.
Quoting from Chern’s Introduction to “Handbook of Differential Geome-
try. Volume 1” (eds. Franki Dillen et al.) that “While algebra and analysis
provide the foundations of mathematics, geometry is at the core.”, it may
be well in this context to look at the following citations (i) from Newton’s
“Philosophiae Naturalis Principia Mathematica” and (ii) from Neumann’s
“The Mathematician” and (iii) from Freudenthal’s “Initiation to Geome-
try”: (i) “Geometry (...) is, in fact, nothing other than that very part of the
totality of mechanics which forms the basis of and precisely determines the
art of measurement.” and (ii) “The most vitally characteristic fact about
mathematics is, in my opinion, its quite peculiar relationship to the natural
sciences, or, more generally, to any science which interprets experience on
a higher than purely descriptive level. Most people, mathematicians and
others, will agree that mathematics is not an empirical science, or at least
that it is practised in a manner which differs in several decisive respects
from the techniques of the empirical sciences. And, yet, its development is
very closely linked with the natural sciences. Some of the best inspirations
of modern mathematics (I believe, the best ones) clearly originated in the
natural sciences. (...). There is a peculiar duplicity in the nature of math-
vii
March 21, 2017 15:26 Diff. Geometry of Warped Product Manifolds-10419 ws-book9x6 page viii
viii Differential Geometry of Warped Product Manifolds and Submanifolds
ematics. One has to realize this duplicity, to accept it, and to assimilate it
into one’s thinking on the subject. This double face is the face of mathemat-
ics, and I do not believe that any simplified, unitarian view of the thing is
possible without sacrificing the essence.” and (iii) “Symmetry in its broad
sense probably was the first mathematical idea that caught the attention
of human beings, in particular before quantifications and countings of any
kind, human beings appear to have been interested in similarities of forms,
of shapes, in its wide sense.” ; (and it seems to me to be rather worthwhile
to be aware of Chern’s above quoted assertion whenever coming in contact
with mathematics).
In its main notions and in its main statements about these notions,
classical geometry – roughly speaking: Euclidean and conformal and pro-
jective geometry – essentially deals with abstractions and generalisations of
some primitive objects and their properties in accordance with the way that
these objects and their properties are perceived by human vision in “our”
surrounding world, statically as well as dynamically and directly as well as
through natural processes like various sorts of projections and of sections.
Caused by the uncomfort brought about by the finding out of the existence
of pairs of mutually incommensurable line segments, likely in the school of
Pythagoras, such as the sides and diagonals in the regular 5-gons with as
ratio the irrational golden section, (which is clearly present for instance in
the vertical : horizontal scale of “the screens” of human’s instantaneous vi-
sual fields and in “the basic construction” of the 10-step units of the human
DNA-molecules), serious attention has been given to the establishment of
the known mathematics in a logically sound deductive way in axiomatical
systems. Next, a more subtle than before appreciation of the axiomatical
method in mathematics and indeed of the very nature of mathematics as
a whole resulted from the solution of the parallel postulate problem of pla-
nar Euclidean geometry by Lobachevsky and Bolyai and Gauss with their
development of the classical non-Euclidean 2D geometries. And also fur-
ther along in this direction of looking for security about “mathematical
truths”, in particular the contributions of Hilbert and Gödel brought us to
the present day understanding of the logical basis of mathematics, which
may be summarised in the concluding sentence “It is better to be aware of
our limitations than to live in a fool’s paradise.” of Stewart’s chapter on
“The Shape of Logic” in his “Taming the Infinite”. At this stage thinking
back at Neumann’s double face of mathematics, simultaneously one may
recall Lemaı̂tre’s closing words of his Antwerp lecture on “Cosmic Radia-
tion and Cosmology” when referring to “the primaeval atom”: “Birth of
March 21, 2017 15:26 Diff. Geometry of Warped Product Manifolds-10419 ws-book9x6 page ix
Foreword ix
space and origin of multiplicity: all the familiar notions on which we reflect
and base our knowledge lose their meaning and disappear; space vanishes
to a point and multiplicity reduces to unity. (...). Let us modestly admit
our essential limitations!”.
At this stage it seems well to repeat the above quotation of Chern and
to add to it its following line in the “Handbook of Differential Geometry”:
“While algebra and analysis provide the foundations of mathematics, ge-
ometry is at the core. This was already recognized by Euclid whose book
contains a geometrical treatment of the number system.”, and in further
trying to prepare the setting in which later to put the spotlights on warped
products – at least in their most special manifestations – next follows Ein-
stein’s “erkenntnistheoretisches Credo” : “Ich sehe auf der einen Seite die
Gesamtheit der Sinnen-Erlebnisse, auf der andern Seite die Gesamtheit
der Begriffe und Sätze, die in Büchern niedergelegt sind. Die Beziehungen
zwischen den Begriffen und Sätzen unter einander sind logischer Art, und
das Geschäft des logischen Denkens ist strikte beschränkt auf die Herstel-
lung der Verbindung zwischen Begriffen und Sätzen unter einander nach
festgesetzten Regeln, mit denen sich die Logik beschäftigt. Die Begriffe
und Sätze erhalten “Sinn” bezw. “Inhalt” nur durch ihre Beziehung zu
Sinnen-Erlebnissen. Die Verbindung der letzteren mit den erstenen ist
rein intuitiv, nicht selbst von logischer Natur. Der Grad der Sicherheit,
mit der diese Beziehung bezw. intuitive Verknüpfung vorgenommen wer-
den kann, und nichts anderes, underscheidet die leere Phantasterei von
der wissenschaftlichen “Wahrheit”. Das Begriffssystem ist eine Schöpfung
des Menschen samt den syntaktischen Regeln, welche die Struktur der Be-
griffssysteme ausmachen. Die Begriffssysteme sind zwar an sich logisch
gänzlich willkürlich, aber gebunden durch das Ziel, eine möglichst sichere
(intuitive) und vollständige Zuordnung zu der Gesamtheit der Sinnen-
Erlebnisse zuzulassen; zweitens entstreben sie möglichste Sparsamkeit in-
bezug auf ihre logisch unabhängingen Elemente (Grundbegriffe und Axiome)
d. h. nicht definierte Begriffe und nicht erschlossene Sätze. Ein Satz ist
richtig, wem er innerhalb eines logischen Systems nach den acceptierten lo-
gischen Regeln abgeleitet ist. Ein System hat Wahrheitsgehalt, entsprechend
der Sicherheit und Vollständigkeit seiner Zuordnungs-Möglichkeit zu der
Erlebnis-Gesamtheit. Ein richtiger Satz erborgt seine “Wahrheit” von dem
Wahrheit-Gehalt des Systems, dem er angehört”, taken from his “Autobi-
ographisches”.
And, last in this preparation, here comes one more citation, “One ex-
pects a mathematical theorem or a mathematical theory not only to describe
March 21, 2017 15:26 Diff. Geometry of Warped Product Manifolds-10419 ws-book9x6 page x
x Differential Geometry of Warped Product Manifolds and Submanifolds
and to classify in a simple and elegant way numerous and a priori disparate
special cases. One also expects “elegance” in its “architectural”, structural
makeup. Ease in stating the problem, great difficulty in getting hold of it
and in all attempts at approaching it, then again some very surprising twist
by which the approach, or some part of the approach, becomes easy, etc.
(...). I think that it is a relatively good approximation to truth – which is
much too complicated to allow anything but approximations – that math-
ematical ideas originate in empirics, although the genealogy is sometimes
long and obscure. But, once they are so conceived, the subject begins to live
a peculiar life of its own and is better compared to a creative one, governed
by almost entirely aesthetic motivations, than to anything else and in par-
ticular, to an empirical science. There is, however, a further point which,
I believe, needs stressing. As a mathematical discipline travels far from its
empirical source, or still more, if it is a second and third generation only
indirectly inspired by ideas coming from “reality”, it is beset with very grave
dangers. It becomes more and more purely aestheticizing, more and more
purely “l’art pour l’art”. This need not be bad, if the field is surrounded by
correlated subjects, which still have closer empirical connections, or if the
discipline is under the influence of men with an exceptionally well-developed
taste. But there is a grave danger that the subject will develop along the
lines of least resistance, that the stream, so far from its source, will separate
into a multitude of insignificant branches, and that the discipline becomes
a disorganized mass of details and complexities. In other words, at a great
distance from its empirical source, or after much “abstract” inbreeding, a
mathematical subject is in danger of degeneration.”, from Neumann’s “The
Mathematician”.
The purpose of this Foreword, then, consists in an attempt to make clear
the importance of warped products in the above described general setting
and to show in particular, against the background of the above mentioned
warning for the danger of mathematical decadence when traveling far from
empirical sources and worse, that actually exactly to the contrary the ge-
ometry of warped products clearly is situated at some of the main empirical
sources of geometry as such, namely at the sources of the very inspirations
of “our space” and of “our spacetime”. The following presentation to this
end will be done in a qualitative way, sometimes including too rough formu-
lations and oversimplifications; for underlying technical considerations, see
the textbook “Differentialgeometrie. Kurven-Flächen-Mannigfaltigkeiten”
of Kühnel (of which there exists an English translation too), Freudenthal
and Steiner’s Chapter 13 on “Group Theory and Geometry” in “Funda-
March 21, 2017 15:26 Diff. Geometry of Warped Product Manifolds-10419 ws-book9x6 page xi
Foreword xi
mentals of Mathematics. Volume II: Geometry” (eds. Behnke-Bachmann-

Fladt-Kunle), O’Neill’s “Semi-Riemannian Geometry – With Applications
to Relativity”, B.-Y. Chen’s Chapter 3 on “Riemannian Submanifolds” in
“Handbook of Differential Geometry. Volume I” and the article on “Natu-
ral Intrinsic Geometrical Symmetries” with Stefan Haesen in the SIGMA
volume “Elie Cartan and Differential Geometry”.
In the Forewords to Professor Chen’s books “Pseudo-Riemannian Ge-
ometry, δ-Invariants and Applications” and “Total Mean Curvature and
Submanifolds of Finite Type” one may find some comments on the geom-
etry of submanifolds of pseudo Euclidean spaces constituting the essence
of geometry and a sketch of the development of geometry in relation with
psychology as the first of the natural sciences – namely with the study of
the human sensations and perceptions of “the world in which we have the
impression to hang around” – and so in particular in relation with human
vision (cfr. a. o. Bronowski’s “The Origins of Knowledge and Imagina-
tion”; and here is a quote in this respect from “Discoveries and Opinions
of Galileo”: “... I believe that vision, the sense eminent above all others
in the proportion of the finite to the infinite, the temporal to the instan-
taneous, the quantitative to the indivisible, the illuminated to the obscure
– that vision, I say, is related to light itself.”). And having the above ci-
tations of Chern, Newton, Neumann, Freudenthal and Einstein in mind,
one may further reflect on the central position of the general theory of sub-
manifolds of pseudo Euclidean spaces in mathematics and in the natural
sciences (and in technology for that matter, recalling here just in passing
and by way of examples the use of warped products in computer vision
and in biomedical sciences). The extrinsic geometry of such submanifolds
actually is the mathematisation that explicitates our awareness of different
concrete shapes in given ambient spaces and the intrinsic geometry of such
submanifolds is proper and semi Riemannian geometry (and the geometry
of 1- or morefold degenerate spaces too).
As Osserman wrote in his “Curvature in the 80ties”: “The notion of
curvature is one of the central topics of differential geometry, one could
argue that it is the central one, distinguishing the geometrical core of the
subject from those aspects that are analytical, algebraic, or topological. In
the words of Marcel Berger, curvature is ‘the number 1 Riemannian invari-
ant and the most natural’.”, (in the 1990 “Geometry” issue of the American
Mathematical Monthly in which a.o. also appeared Chern’s “What is Ge-
ometry?”). In a terminology of Elie Cartan, the (0, 4) Riemann curvature
tensor R of an nD pseudo Riemannian space (M n , g) with metric (0, 2) ten-
March 21, 2017 15:26 Diff. Geometry of Warped Product Manifolds-10419 ws-book9x6 page xii
xii Differential Geometry of Warped Product Manifolds and Submanifolds
sor g = ghk dxh dxk constitutes the holonomy of (tangent 1D) directions on
M after parallel transport fully around infinitesimal co-ordinate parallelo-
grams and similarly the (0, 6) curvature tensor R·R (whereby here the first
R stands for the curvature operator which acts as a derivation on the (0, 4)
curvature tensor R) constitutes the holonomy of the Riemann sectional cur-
vatures after such same transformations. And while the knowledge of the
Riemann curvature tensor R is equivalent to the knowledge of the Riemann
sectional curvatures K(p, π) – for all (non-degenerate) tangent 2-planes π
at the points p of M – similarly the knowledge of the tensor R · R is equiv-
alent to the knowledge of the Deszcz double sectional curvatures L(p, π, π̄)
– for all pairs of (curvature dependent) tangent 2-planes π and π̄ at the
points p of M – . Thus, the locally Euclidean or locally flat spaces (R = 0,
or, equivalently, K(p, π) = 0 for all points p in M and for all proper tan-
gent 2-planes π to M at p) are the spaces (M n , g) which, in the sense of
Weyl, satisfy the symmetry property that all their (tangent) directions at
all their points remain invariant under the parallel transport fully around
all infinitesimal co-ordinate parallelograms cornered at these points. And,
as is well known, by applying projective transformations to locally Euclidean
spaces one obtains real space forms and the class of real space forms is closed
under projective transformations. Here, by real space forms are meant the
spaces (M n , g) of constant sectional curvatures K(p, π) = c, (c = 0, or,
c > 0, or, c < 0), or, still, in abbreviation, the CC-spaces, denoted by
M n (c), and these spaces are characterised by the fact that R = 2c g ∧ g,
i.e. that their (0, 4) Riemann curvature tensor R is constantly proportional
to the Kulkarni-Nomizu square of their metric tensor g. According to the
Lemma of Schur, for dimensions n ≥ 3, the real space forms M n (c) are
characterised as the spaces (M n , g) having the property that at all their
points p their sectional curvatures K(p, π) for all 2-plane sections π at p are
the same, or, for short, by the isotropy of their sectional curvature function
K(p, π) – i.e. the numerical value of this function being independent of
the tangent 2D directions π implies its constancy, or, still, that the com-
mon value K(p, π) for all planes π at any point p for n ≥ 3 automatically
moreover is independent of the points p as well – . And thus further, the
semi symmetric or Szabó symmetric spaces (R · R = 0, or, equivalently,
L(p, π, π̄) = 0 for all points p in M and for all proper pairs of tangent
2-planes π and π̄ to M at p) are the spaces (M n , g) which, in the sense of
Weyl, satisfy the symmetry property that all their Riemann sectional curva-
tures at all their points remain invariant under the parallel transport fully
March 21, 2017 15:26 Diff. Geometry of Warped Product Manifolds-10419 ws-book9x6 page xiii
Foreword xiii
around all infinitesimal co-ordinate parallelograms cornered at these points.

By applying projective transformations to semi or Szabó symmetric spaces
one obtains pseudo symmetric or Deszcz symmetric spaces and the class of
pseudo or Deszcz symmetric spaces is closed under projective transforma-
tions. Here, by pseudo or Deszcz symmetric spaces are meant the spaces
(M n , g) with isotropic Deszcz double sectional curvature function, that is,
the spaces (M n , g) for which at all points p the numerical values of the
double sectional curvature function L(p, π, π̄) are independent of all proper
pairs of tangent 2-planes π and π̄ to M at p. These spaces are characterised
by the fact that R · R = L ∧g ·R for some function L : M → R, i.e. by
the fact that their (0, 6) curvature tensor R · R is functionally proportional
to the (0, 6) tensor ∧g · R – which results from the action of the natural
metrical endomorphism ∧g as a derivation on the (0, 4) tensor R – with in
general different numerical values L(p) at different points p. For the Deszcz
sectional curvature function L(p, π, π̄) there is no analogous property like
given by the Lemma of Schur for the Riemann sectional curvature function
K(p, π); hence, within the class of all Deszcz symmetric spaces of particu-
lar interest are the so-called pseudo symmetric spaces of constant type, i.e.
the pseudo symmetric spaces of constant double sectional curvature L = c,
(c = 0 – for the semi symmetric spaces –, or, c > 0, or, c < 0). When
showing due respect for both the differential structure and the metrical
structure of pseudo Riemannian spaces one may become well aware of the
significance of the parallel transports fully around infinitesimal co-ordinate
parallelograms and thus of the geometrical importance of their correspond-
ing symmetries.
Next we will focus for a while on definite Riemannian spaces. Then
thinking of symmetry in its plainest meaning of our common sense, the real
space forms are the most perfectly symmetric spaces; it are the spaces that
look the same at all of their points and at every point they look the same in
all directions, they are homogeneous and 1D isotropic. In 1854 Riemann de-
fined Riemannian spaces (M n , g) as nD differential manifolds M with local
co-ordinates (x1 , . . . , xn ) endowed with the geometrical structure given by
the determination of the lengths ds of all infinitesimal line elements on M ;
and as simplest example by which to illustrate his general exposition, he
considered the quadratic infinitesimal line elements ds2 = g = ghk dxh dxk ,
herewith generalising to arbitrary dimensions n and treating in absolute
abstraction the intrinsic geometry of surfaces M 2 in Euclidean spaces E 3
as developed by Gauss in 1827, – in which geometry the infinitesimal met-
ric g on a surface M 2 in E 3 is the restriction of the infinitesimal ver-
March 21, 2017 15:26 Diff. Geometry of Warped Product Manifolds-10419 ws-book9x6 page xiv
xiv Differential Geometry of Warped Product Manifolds and Submanifolds
sion of the Theorem of Pythagoras of the ambient space E 3 to the surface

M 2 –. Nowadays, by Riemannian geometry usually is meant the geometry
of spaces M n endowed with such a quadratic line element. The real space
forms are the Riemannian spaces in which, in accordance with our natural
expectations, – à la Cartan – the measurements of all “beings” living in
these spaces do not depend on their actual locations in these spaces nor
on their actual positioning at these locations, or, still, are the Rieamn-
nian spaces (M n , g) which satisfy the axiom of free mobility (cfr. Riemann,
Helmholtz, Lie and Tits). And, the real space forms M n (c), (regardless
c = 0, c > 0, or, c < 0), all equally well do geometrically model “the am-
bient space of our direct visual sense experiences”, since, as for instance
observed by Klein in his “Elementarmathematik vom höheren Standpunkte
aus”, in view of the threshold of our sense perception and the fact that
our space perception is adapted to a limited part of space only, our space
perception can be described as closely as desired by the Euclidean and the
classical non-Euclidean elliptic and hyperbolic space forms alike.
In a way, the historical Euclidean parallel postulate problem can be for-
mulated as follows: do there exist other valid planar geometries besides the
Euclidean geometry of our visual perception? In Gauss’ investigation of
this problem – not in an axiomatical but rather in a visual manner – a de-
cisive step was his insight in the difference between the meanings of locally
isometric and locally conformal surfaces M 2 in E 3 by realising the existence
of isothermal co-ordinates on every surface M 2 in E 3 , and in terms of such
co-ordinates Gauss by then proved his theorema egregium. A bit later, in
the 1822 publication in which he generalised the stereographic projection
and the Mercator projection of spheres onto planes to maps which realise
similarity in their smallest parts between any two surfaces, he announced
that “herewith the way was paved to greater things” and in 1827 appeared
his general theory of curved surfaces; – about this and much more, see
Peter Dombrowski’s “150 years after Gauss’ ‘disquisitiones generales circa
superficies curvas’ ” –. Highlights of this work are a proof of the Gauss
curvature K of surfaces M 2 in E 3 equaling the product of Euler’s princi-
pal normal curvatures k1 and k2 , K = k1 k2 , and based thereupon a proof
in general curvilinear co-ordinates of the invariance of K under surface
isometries. And “a greater thing” that then came within haptic reach was
the realisation of the classical non-Euclidean hyperbolic (K = c < 0) geom-
etry on the pseudo-spheres in E 3 ; – in retrospect, referring to the above
recalled observation of Klein, one may be pleasantly surprised in how un-
foreseeable manners human knowledge and understanding can develop –.
March 21, 2017 15:26 Diff. Geometry of Warped Product Manifolds-10419 ws-book9x6 page xv
Foreword xv
The surfaces of revolution in E 3 with constant Gauss curvatures respec-

tively K = 0, K > 0 and K < 0 are nicely drawn in the second English
edition of Kühnel’s “Differential Geometry. Curves-Surfaces-Manifolds” in
Figures 3.11, 3.12 and 3.13.
In “The geometry and topology of 3-manifolds”, Thurston wrote the fol-
lowing: “What is a geometry? Up till now we have discussed three kinds
of three-dimensional geometry: hyperbolic, Euclidean and spherical. They
have in common the property of being as uniform as possible: their isome-
tries can move any point to any other point (homogeneity), and can take
any orthonormal frame in the tangent space at a point to any other or-
thonormal frame at that point (isotropy). There are more possibilities if
we remove the isotropy condition, allowing the space to have a grain, so to
speak, so that certain directions are geometrically distinguished from oth-
ers. An enumeration of additional three-dimensional geometries depends on
what spaces we wish to consider and what structures we use to define and
to distinguish the spaces. For instance, do we think of geometry as a space
equipped with such notions as lines and planes, or as a space equipped with a
notion of congruence, or as a space equipped with a metric or a Riemannian
metric? There are deficiencies in all of these approaches. (...). For logical
purposes, we must pick one definition. We choose to represent a geometry
as a space equipped with a group of congruences, that is, a (G, X)-space.”,
and then stated his Lie-Klein inspired definition (3.8.1) of a model geome-
try (G, X) as a manifold X together with a Lie group G of diffeomorphisms
of X, satisfying conditions (a)(b)(c)(d), and proved that the Euclidean and
the classical non-Euclidean geometries E 2 , S 2 and H 2 are the three such
model geometries in dimension 2 and that E 3 , S 3 , H 3 , S 2 × E 1 , H 2 × E 1 ,
˜
SL(2, R), H3 and Sol are the eight such model geometries in dimension 3.
Being locally symmetric or Cartan symmetric spaces, the real space forms
E 3 , S 3 and H 3 – of constant sectional curvatures K = 0, K = c > 0 and
K = c < 0 – and S 2 × E 1 and H 2 × E 1 – which do not have constant
sectional curvatures – in particular also are semi symmetric or Szabó sym-
metric spaces and hence all five these spaces have double sectional curvature
L = 0, while the other three are not semi symmetric pseudo symmetric or
˜
Deszcz symmetric spaces of constant type, L = +1 for SL(2, R) and H3 and
L = −1 for Sol. And in Wall’s “Geometries and Geometric Structures in
real dimension 4 and complex dimension 2” is given a list of the Thurston
geometries for dimension 4 organised in terms of the isotropy subgroups.
And here is for all dimensions n > 2 a metric geometrical approach to
“nice spaces” that are not 1D isotropic, i. e. nice spaces in which certain
March 21, 2017 15:26 Diff. Geometry of Warped Product Manifolds-10419 ws-book9x6 page xvi
xvi Differential Geometry of Warped Product Manifolds and Submanifolds
directions are geometrically distinguished from others. To begin with, fol-

lowing the formerly given attention to the naturalness of mathematisations
of some of our kind’s basic sense experiences, among the alternative ways
to think of geometry mentioned by Thurston we confine our interest to
Riemannian manifolds (M n , g). And, for n > 2, from the same former
point of view, certainly a most natural subclass of such Riemannian spaces
in which to look for 1D anisotropy may well be the class of the locally
conformally Euclidean spaces or the locally conformally flat spaces, i. e. the
spaces (M n , g) which are locally conformal to Euclidean spaces, or, still, the
spaces (M n , g) which in their smallest parts are similar to Euclidean spaces,
(for n > 3, also having Cartan’s characterisation of the locally conformally
flat hypersurfaces of Euclidean spaces E n+1 as umbilical or quasi umbilical
hypersurfaces in mind). Namely, as recalled before, the real space forms
M n (c) – of zero, positive and negative constant sectional curvatures alike –
are fine geometrical models for “our ambient physical space” in case of di-
mension n = 3, and, therefore for all dimensions n ≥ 3 are the geometrical
spaces in which mathematicians and scientists can mentally well visualise
basic notions and properties, and, helped by our kind’s primitive appreci-
ation of being alike in the sense of being conformal, the class of the real
space forms smoothly “grows” to the class of the locally conformally flat
spaces. And to end the present paragraph: it is too difficult not to mention
here that the golden rectangles are the only rectangles that gnomonically
do grow to conformal rectangles.
In Riemannian spaces (M n , g) of dimension n > 2 to geometrically com-
pare 1D directions essentially means to compare their Ricci curvatures. The
Ricci curvature Ric(p, d) of (M n , g) in a tangent 1D direction d at a point
p is the average of the sectional curvatures K(p, π) of all tangent 2-planes
π to M at p through d. The spaces (M n , g) – n > 2 – for which at every
point p these curvatures Ric(p, d) are the same for all directions d at p
by definition are the Riemannian Einstein spaces. And all Einstein spaces
have constant scalar curvature function τ ; (the values τ (p) being the aver-
age of Ric(p, d) over all directions d at p, this “Schur-like” also means that
in Einstein spaces all Ricci curvatures moreover are the same at all points
too), and – even stronger for dimension n = 3, by a Theorem of Schouten
and Struik – every Einstein space (M 3 , g) is a space of constant sectional
curvatures, i. e. is a real space form M 3 (c). Denoting the (0, 2) Ricci cur-
vature tensor of (M n , g) by S, the Einstein manifolds are characterised by
the condition S = nτ g; – and for a brief note concerning the pre-origin of the
Ricci curvatures and tensor, see the paper “On the parallel transport of the
March 21, 2017 15:26 Diff. Geometry of Warped Product Manifolds-10419 ws-book9x6 page xvii
Foreword xvii
Ricci curvatures” with Jahanara, Haesen and Sentürk –. In analogy with

the foregoing discussion of the parallel transport of the sectional curvatures
K(p, π) fully around infinitesimal co-ordinate parallelograms cornered at p
which lead to the extension of the Riemann sectional curvatures K(p, π)
to the Deszcz double sectional curvatures L(p, π, π̄), the (0,4) tensor R · S
constitutes the holonomy of the Ricci curvatures Ric(p, d) after such par-
allel transports – and the geometrical information contained in the tensor
R · S is equivalent with the geometrical information contained in the Ricci
curvatures LS (p, d, π̄) of Deszcz, of which the numerical values in general
are dependent of points p and 1D and 2D tangent directions d and π̄ to M
at p – . And, in analogy with the extension of the real space forms being
determined by the isotropy of the sectional curvatures (i. e. the sectional
curvatures K(p, π) at every point p being independent of the planes π) to
the Deszcz symmetric or pseudo symmetric spaces being determined by the
isotropy of the double sectional curvatures (i. e. the double sectional curva-
tures L(p, π, π̄) at every point p being independent of the planes π and π̄),
the Einstein spaces as determined by the isotropy of the Ricci curvatures
(i. e. the Ricci curvatures Ric(p, d) at every point p being independent of
the directions d) may be extended to the Ricci pseudo symmetric or Ricci
Deszcz symmetric spaces which are determined by the isotropy of their Ricci
curvatures of Deszcz (i. e. the Ricci curvatures of Deszcz LS (p, d, π̄) being
independent of the 1D and 2D directions d and π̄). Equivalently, the Ricci
pseudo or Deszcz symmetric spaces are characterised by the functional pro-
portionality of the (0,4) tensors R · S and ∧g · S : R · S = LS ∧g ·S, for some
function LS : M → R.
Thus, the Riemannian spaces (M n , g) of dimension n > 2 which are 1D
isotropic are the Einstein spaces, and, in particular, for dimension 3 it are
the real space forms M 3 (c). And going for nice spaces (M n , g) of dimen-
sion n > 2 which are not 1D isotropic, it seems most appropriate to look
for Ricci pseudo symmetric spaces, such spaces being the most immediate
extension of the Einstein spaces by the consideration of likely the most
natural Riemannian geometrical symmetries. And, in view of the way in
which the vanishing of the conformal curvature tensor C of Weyl links the
curvature tensors R of Riemann and S of Ricci and, since C = 0 automati-
cally for all 3D Riemannian spaces and since – by a Theorem of Schouten
– Riemannian spaces of dimensions > 3 are locally conformally Euclidean
if and only if C = 0, one has the following: for all 3D Riemannian spaces
and for all locally conformally flat spaces (M n , g) of dimensions n > 3,
being Ricci pseudo symmetric is equivalent to being Deszcz symmetric (and
March 21, 2017 15:26 Diff. Geometry of Warped Product Manifolds-10419 ws-book9x6 page xviii
xviii Differential Geometry of Warped Product Manifolds and Submanifolds
the corresponding curvature functions LS and L are the same). In résumé :

in a metrical geometrical search for nice spaces that are not 1D isotropic,
rather naturally one arrives at the Riemannian spaces (M 3 , g) and at the
locally conformally Euclidean spaces (M n , g) of dimension n > 3 which
are Ricci pseudo symmetric or, equivalently, which are Deszcz symmetric,
and, of course, one is especially interested in the non-trivial such spaces,
that is in such spaces beyond the real space forms. And, within intrinsic
Riemannian geometry, the following theorem makes absolutely clear which
spaces then are concerned: Riemannian spaces of dimensions ≥ 3 with van-
ishing Weyl conformal curvature tensor are Deszcz symmetric if and only if
they are Einstein or “partially Einstein”, – partially Einstein spaces being
defined by the condition that their Ricci tensor has precisely two distinct
eigenvalues –. Because the Riemannian manifolds of dimensions ≥ 3 with
vanishing Weyl conformal curvature tensor which are Einstein are the real
space forms of dimensions ≥ 3 hereafter we further only care about such
spaces which are partially Einstein. Their tangent spaces at all points split
up in two orthogonally complementary subspaces, say S1 and S2 of dimen-
sions n1 and n2 (n1 + n2 = n) namely the eigenspaces belonging to the two
distinct Ricci curvatures, say ρ1 and ρ2 with multiplicities n1 and n2 . All
the 1D directions in S1 are geometrically different from all 1D directions
in S2 while all 1D directions in S1 are geometrically equivalent among each
other as also all 1D directions in S2 are geometrically equivalent among each
other and the 1D directions “intermediate” between S1 and S2 in a way
gradually make the Ricci curvature transition between ρ1 and ρ2 . Therefore
it seems not too amiss then to state that the partially Einstein spaces do
realise a mildest possible geometrical 1D anisotropy. And the simplest man-
ifestations of such spaces (M n , g) are the proper quasi Einstein spaces, i. e.
the partially Einstein spaces with n1 = dimS1 = multiplicity ρ1 = 1 and
n2 = dimS2 = multiplicity ρ2 = n − 1, while in case of even-dimensional
such spaces (n = 2m, m > 1) also the “half Einstein spaces”, i. e. the
partially Einstein spaces with n1 = n2 = m deserve special attention.
In the above metrical geometrical approach to geometry, the geometri-
cal comparison between (1D tangent) directions is done by comparing the
Ricci curvatures of these directions. For a Riemannian space of dimension
> 2 the isotropy of such directions then means to be an Einstein space,
asserting that all their tangent directions have equal Ricci curvatures. For
non-1D isotropic Riemannian spaces, the degree of anisotopy is determined
in particular by the number of distinct eigenvalues of their Ricci tensor, the
n eigenvalues of the Ricci tensor of an n dimensional manifold being two by
March 21, 2017 15:26 Diff. Geometry of Warped Product Manifolds-10419 ws-book9x6 page xix
Foreword xix
two distinct corresponding to the wildest possible anisotropy. For spaces of

dimension 3, one thus has three possibilities: Einstein spaces or equivalently
real space forms (S has three equal eigenvalues), mildly anisotropic spaces
(S has two distinct eigenvalues – one with multiplicity 1, the other with
multiplicity 2 –) and completely anisotropic spaces (S has three distinct
eigenvalues), and for spaces of dimension 4, one thus has the following pos-
sibilities: Einstein spaces (S has four equal eigenvalues), mildly anisotropic
spaces (S has two distinct eigenvalues – either with multiplicities 1 and 3 or
both with multiplicity 2 –), not so mildly but wildly anisotropic spaces (S
has three distinct eigenvalues – one with multiplicity 2 and the two other
with multiplicity 1 –) and the wildest possible anisotropic 4D spaces (S has
four distinct eigenvalues), and so on for higher and higher dimensions; (the
five non-isotropic 3D Thurston geometries clearly are mildly anisotropic –
they are proper quasi Einstein spaces – and for the 4D Thurston geometries
one may in this respect go through the list one by one).
By the former developments, for all dimensions n ≥ 3 one had come
to the proper quasi Einstein spaces with vanishing Weyl tensor C as first
class extension of the real space forms among all 1D anisotropic Rieman-
nian manifolds (M n , g). Now it is a real pleasure to be able to state that,
from the intrinsic point of view, these spaces are (what in the mean time
were well named) the QCC-spaces, that is the spaces of quasi constant sec-
tional curvatures, while, from the extrinsic point of view, they are proper
quasi umbilical hypersurfaces M n in the real space forms M n+1 (c) – i. e.
hypersurfaces with exactly two distinct principal curvatures, say, µ and λ,
with respective multiplicities 1 and n − 1 – as was basically shown already
in 1972 at MSU by Chen-Houh-Yano and by Chen-Yano, respectively, (cfr.
e.g. Chapter 5 of B.-Y. Chen’s 1973 book “Geometry of Submanifolds”).
Within Riemannian geometry the beginning of these studies was Kentaro
Yano and Bang-Yen Chen’s generalisation of the 1930 subprojective spaces
of Kagan, Rachevsky and Shapiro to special conformally flat spaces, while
within the geometry of submanifolds, the study of the quasi umbilical sub-
manifolds in real space forms in a way goes back to the very origin of the
interest in warped products. Indeed, when considering the simplest such
hypersurfaces M n in E n+1 , namely the canal hypersurfaces or envelopes
of 1-parameter families of hyperspheres, and then within this class further
considering the simplest ones, namely the hypersurfaces of revolution, we
arrived at the hypersurfaces M n in E n+1 on which the induced Riemannian
metric was the first explicit occurrence of warped product metrics; – for a
sort of implicit such occurrence in the work of Johann Bernoulli, see Peter
March 21, 2017 15:26 Diff. Geometry of Warped Product Manifolds-10419 ws-book9x6 page xx
xx Differential Geometry of Warped Product Manifolds and Submanifolds
Dombrowski’s “The Brachistochrone Problem: A Problem of Elementary

Differential Geometry” in “Geometry and Topology of Submanifolds. VIII”
(eds. Franki Dillen e. a) –.
Now, going back from extrinsic to intrinsic geometry, and to facilitate
the exposition restricting to quasi umbilical hypersurfaces M n in Euclidean
spaces E n+1 , n ≥ 3, such hypersurfaces are proper quasi Einstein spaces
with the principal tangent direction d˜ corresponding to the hypersurfaces’
normal principal curvature µ with multiplicity 1 giving the eigenspace of
dimension 1 of their Ricci tensor and such hypersurfaces are locally con-
formally Euclidean Deszcz symmetric spaces (and Ricci pseudo symmetric
spaces) with double sectional curvature function L = µλ, (so, that at any
point p of such hypersurfaces the Gauss curvature K(p, π̃) of any 2D normal
section of M n corresponding to any tangent 2-plane π̃ to M n at p which
contains d˜ – a surface in E 3 – gives a visualisation of the double sectional
curvature of these hypersurfaces at p). And, in particular, the rotational
hypersurfaces of M n in E n+1 , for n ≥ 3, which are Deszcz symmetric of
constant type, i. e. with constant double sectional curvature function L, do
have the profile curves of the surfaces of revolution M 2 (c) in E 3 of constant
Gauss curvatures c. It is well known that the real space forms E n , S n , and
H n of respectively zero, positive and negative constant sectional curvatures
for all dimensions n do have warped product metrics, (for instance given in
terms of polar co-ordinates). So, in particular, the 3D geometrical spaces of
our immediate visual sense experiences actually are warped product spaces
too. And, finally, as particular indefinite version of the above mentioned re-
alisations of the Riemannian QCC-spaces as loci of spheres, the causal type
preserving Lorentzian hypersurfaces of revolution in 5D Minkowski ambi-
ent spaces E15 are pretty visualisable Friedmann-Lemaı̂tre warped product
geometrical models of relativistic spacetime cosmology; – and with respect
to the above given geometrical interpretations about locally conformally
flat spaces, quasi Einstein spaces and Deszcz symmetrical spaces in a posi-
tive definite Riemannian setting, basically all can be well transferred to the
pseudo Riemannian situation in a standard way –.
This new book of Professor Bang-Yen Chen presents a wealth of basic
information in the broad field of warped product manifolds and submani-
folds, fundamental old and new notions and their properties, resulting from
the research of geometers working at many places spread all over the world,
and, like in his previous books, many of these notions and properties are
due to the author himself, who is one of the most creative and productive
mathematicians of our time. The readers of this book may expect to be
March 21, 2017 15:26 Diff. Geometry of Warped Product Manifolds-10419 ws-book9x6 page xxi
Foreword xxi
rewarded by enjoying their participation in the creation of the results that

they herein will learn about, according to the following quotations (1) from
Huntley’s “The divine proportion - a study in mathematical beauty” and
(2) from Bronowski’s “Science and Human Values”: (1) “...one of the most
intense joys that the soul of man can experience is that of creative activity.
Ask the artist. Ask the scientist. They all know the deep spiritual satis-
faction associated with the orgasm of creation.” and (2) “The discoveries
of science, the works of art are explorations, more, are explosions, of a
hidden likeness. The discoverer or the artist presents in them two aspects
of nature and fuses them into one. This is the act of creation in which
an original thought is born, and it is the same act in original science and
original art. This view alone gives a meaning to the act of appreciation;
for the appreciator must see the movement, wake to the echo which was
started in the creation of the work. In the moment of appreciation we live
again the moment when the creator saw and held the hidden likeness. We
re-enact the creative act, and we ourselves make the discovery again. The
great poem and the deep theorem are new to every reader, and yet are his
own experiences, because he himself re-creates them. They are the marks
of unity in variety, and in the instant when the mind seizes this for itself,
the heart misses a beat.”.
Leopold Verstraelen
De Haan (Belgium); 29 - 10 - 2016.
b2530_FM.indd 6 01-Sep-16 11:03:06 AM

March 21, 2017 15:26 Diff. Geometry of Warped Product Manifolds-10419 ws-book9x6 page xxiii
Preface
Warped products are the most natural and the most fruitful generalization
of Cartesian products. More precisely, a warped product is a manifold
equipped with a warped product metric of the form:
X X
g= gij (y)dy i ⊗ dy j + f (y) gst (x)dxs ⊗ dxt ,
i,j s,t
where the warped geometry decomposes into a product of the “y” geometry
and the “x” geometry, except that the second part is warped, i.e., it is
rescaled by a scalar function of the other coordinates “y”. If one substitutes
the variable y for the time variable t and x for a 3-dimensional spatial space,
then the first part becomes the effect of time in Einstein’s curved space.
How it curves space will define one or the other solution to a spacetime
model. For that reason different models of spacetime in general relativity
are often expressed in terms of warped geometry. Consequently, the notion
of warped products plays very important roles not only in geometry but
also in mathematical physics, especially in general relativity.
The term of “warped product” was introduced by R. L. Bishop and B.
O’Neill in [Bishop and O’Neill (1964)], who used it to construct a large
class of complete manifolds of negative curvature. However, the concept
of warped products appeared in the mathematical and physical literature
before [Bishop and O’Neill (1964)]; for instance, warped products were
called semi-reducible spaces in [Kruchkovich (1957)]. Nevertheless, inspired
by Bishop and O’Neill’s article, many important works on warped products
from intrinsic point of view were done during the last fifty years.
According to the famous Nash embedding theorem published in 1956,
every Riemannian manifold can be isometrically embedded in some Eu-
clidean spaces. Nash’s theorem shows that every warped product N1 ×f N2
can be embedded as a Riemannian submanifold in some Euclidean spaces
xxiii
March 21, 2017 15:26 Diff. Geometry of Warped Product Manifolds-10419 ws-book9x6 page xxiv
xxiv Differential Geometry of Warped Product Manifolds and Submanifolds
with sufficiently high codimension. Due to this fact, the author asked the
following basic question (see, e.g., [Chen (2002a)]).
Question: What can we conclude from an isometric immersion of an ar-
bitrary warped product into a Euclidean space or into a space form with
arbitrary codimension with arbitrary codimension?
The study of warped products from this extrinsic point of view was
initiated around the beginning of this century by the author in a series of
his articles. Since then the study of warped product submanifolds from
extrinsic point of view has become a very active research subject in differ-
ential geometry and many nice results on this subject have been obtained
by many geometers.
The main purpose of this book is thus to provide an extensive and
comprehensive survey on the study of warped product manifolds and sub-
manifolds from intrinsic and extrinsic points of view done during the last
few decades. It is the author’s hope that the reader will find this book
both a good introduction to the theories of warped product manifolds and
of warped product submanifolds as well as a useful reference for recent and
further research of both areas.
In concluding the preface, the author would like to thank World Scien-
tific Publishing for the invitation to undertake this project. He also would
like to express his appreciation to Professors D. E. Blair, I. Dimitric, O.
J. Garay, I. Mihai, M. Petrović-Torgašev, B. Sahin, B. Suceava, J. Van
der Veken, and S. W. Wei for reading parts of the manuscript and offering
many valuable suggestions. In particular, the author thanks Professor L.
Verstraelen for writing an excellent foreword for this book.
November 1, 2016
Bang-Yen Chen
March 21, 2017 15:26 Diff. Geometry of Warped Product Manifolds-10419 ws-book9x6 page xxv
Contents
Foreword vii
Preface xxiii
1. Riemannian and Pseudo-Riemannian Manifolds 1

1.1 Symmetric bilinear forms and scalar products . . . . . . . 1
1.2 Riemannian and pseudo-Riemannian manifolds . . . . . . 3
1.3 Levi-Civita connection . . . . . . . . . . . . . . . . . . . . 4
1.4 Parallel transport . . . . . . . . . . . . . . . . . . . . . . . 7
1.5 Riemann curvature tensor . . . . . . . . . . . . . . . . . . 10
1.6 Sectional, Ricci and scalar curvatures . . . . . . . . . . . . 12
1.7 Indefinite real space forms . . . . . . . . . . . . . . . . . . 15
1.8 Gradient, Hessian and Laplacian . . . . . . . . . . . . . . 16
1.9 Lie derivative and Killing vector fields . . . . . . . . . . . 17
1.10 Concircular and concurrent vector fields . . . . . . . . . . 19
2. Submanifolds 23
2.1 Embedding theorems . . . . . . . . . . . . . . . . . . . . . 24
2.2 Formulas of Gauss and Weingarten . . . . . . . . . . . . . 26
2.3 Equations of Gauss, Codazzi and Ricci . . . . . . . . . . . 30
2.4 Existence and uniqueness theorems of submanifolds . . . . 34
2.5 Reduction theorems . . . . . . . . . . . . . . . . . . . . . 35
2.6 Totally geodesic submanifolds . . . . . . . . . . . . . . . . 37
2.7 Totally umbilical submanifolds . . . . . . . . . . . . . . . 38
2.8 Pseudo-umbilical submanifolds . . . . . . . . . . . . . . . 41
2.9 Cartan’s structure equations . . . . . . . . . . . . . . . . . 46
xxv
March 21, 2017 15:26 Diff. Geometry of Warped Product Manifolds-10419 ws-book9x6 page xxvi
xxvi Differential Geometry of Warped Product Manifolds and Submanifolds
3. Warped Product Manifolds 47

3.1 Warped products . . . . . . . . . . . . . . . . . . . . . . . 47
3.2 Connection of warped products . . . . . . . . . . . . . . . 49
3.3 Curvature of warped products . . . . . . . . . . . . . . . . 50
3.4 Einstein warped product manifolds . . . . . . . . . . . . . 52
3.5 Conformally flat warped product manifolds . . . . . . . . 58
3.6 Multiply warped product manifolds . . . . . . . . . . . . . 59
3.7 Warped product immersions . . . . . . . . . . . . . . . . . 62
3.8 More results for warped product immersions . . . . . . . . 65
3.9 Twisted products . . . . . . . . . . . . . . . . . . . . . . . 71
3.10 Characterizations of twisted products . . . . . . . . . . . 76
3.11 Convolution manifolds . . . . . . . . . . . . . . . . . . . . 78
4. Robertson-Walker Spacetimes and Schwarzschild Solution 81

4.1 Basic properties of Robertson-Walker spacetimes . . . . . 82
4.2 Totally geodesic submanifolds of Robertson-Walker
spacetimes . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.3 Parallel submanifolds of Robertson-Walker spacetimes . . 87
4.4 Totally umbilical submanifolds of Robertson-Walker
spacetimes . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.5 Realizations of Robertson-Walker spacetimes . . . . . . . 93
4.6 Generalized Robertson-Walker spacetimes . . . . . . . . . 94
4.7 Schwarzschild’s solution and black holes . . . . . . . . . . 96
5. Contact Metric Manifolds and Submersions 99

5.1 Contact metric manifolds . . . . . . . . . . . . . . . . . . 100
5.2 Sasakian manifolds . . . . . . . . . . . . . . . . . . . . . . 100
5.3 Submersions . . . . . . . . . . . . . . . . . . . . . . . . . . 102
5.4 O’Neill integrability tensor and fundamental equations . . 103
5.5 Submersions with totally geodesic fibers . . . . . . . . . . 105
5.6 Sasakian space forms . . . . . . . . . . . . . . . . . . . . . 107
5.7 Geometry of horizontal immersions . . . . . . . . . . . . . 111
5.8 Legendre submanifolds via canonical fibration . . . . . . . 112
6. Kähler and Pseudo-Kähler Manifolds 115

6.1 Pseudo-Kähler manifolds . . . . . . . . . . . . . . . . . . . 115
6.2 Concircular vector fields on pseudo-Kähler manifolds . . . 119
6.3 Pseudo-Kähler submanifolds . . . . . . . . . . . . . . . . . 121
March 21, 2017 15:26 Diff. Geometry of Warped Product Manifolds-10419 ws-book9x6 page xxvii
Contents xxvii
6.4 Segre and Veronese embeddings . . . . . . . . . . . . . . . 124

6.5 Purely real submanifolds of pseudo-Kähler manifolds . . . 125
6.6 Totally real and Lagrangian submanifolds . . . . . . . . . 127
6.7 Totally umbilical and H-umbilical submanifolds . . . . . . 129
6.8 Warped products, H-umbilical submanifolds and
complex extensors . . . . . . . . . . . . . . . . . . . . . . 131
6.9 Classification of H-umbilical submanifolds . . . . . . . . . 134
7. Slant Submanifolds 141

7.1 Examples of slant submanifolds . . . . . . . . . . . . . . . 141
7.2 Basic properties and their applications . . . . . . . . . . . 144
7.3 Existence and uniqueness theorems . . . . . . . . . . . . . 151
7.4 A non-existence theorem for compact slant
submanifolds . . . . . . . . . . . . . . . . . . . . . . . . . 158
7.5 A non-minimality theorem for slant submanifolds . . . . . 162
7.6 Topology and cohomology of slant submanifolds . . . . . . 165
7.7 Pointwise slant submanifolds . . . . . . . . . . . . . . . . 171
7.8 Contact slant submanifolds via canonical fibration . . . . 177
8. Generic Submanifolds of Kähler Manifolds 179

8.1 Generic submanifolds . . . . . . . . . . . . . . . . . . . . . 179
8.2 Integrability . . . . . . . . . . . . . . . . . . . . . . . . . . 181
8.3 Parallelism of P and F . . . . . . . . . . . . . . . . . . . . 182
8.4 Totally umbilical submanifolds . . . . . . . . . . . . . . . 187
8.5 Generic products and Segre embedding . . . . . . . . . . . 190
8.6 Generic products in complex projective spaces . . . . . . . 191
8.7 An application to complex geometry . . . . . . . . . . . . 193
9. CR-submanifolds of Kähler Manifolds 195

9.1 CR-submanifolds as CR-manifolds . . . . . . . . . . . . . 195
9.2 Integrability and minimality . . . . . . . . . . . . . . . . . 197
9.3 Cohomology of CR-submanifolds . . . . . . . . . . . . . . 200
9.4 Totally geodesic and totally umbilical CR-submanifolds . 202
9.5 Mixed foliate CR-submanifolds . . . . . . . . . . . . . . . 205
10. Warped Products in Riemannian and Kähler Manifolds 207

10.1 An algebraic lemma . . . . . . . . . . . . . . . . . . . . . 207
10.2 Warped products in real space forms . . . . . . . . . . . . 209
March 21, 2017 15:26 Diff. Geometry of Warped Product Manifolds-10419 ws-book9x6 page xxviii
xxviii Differential Geometry of Warped Product Manifolds and Submanifolds
10.3 Some applications of Theorems 10.1 and 10.2 . . . . . . . 213

10.4 Rotation hypersurfaces in real space forms . . . . . . . . . 215
10.5 Another optimal inequality for warped products . . . . . 217
10.6 Warped products in Kähler manifolds . . . . . . . . . . . 222
10.7 Warped product submanifolds in generalized complex
space forms . . . . . . . . . . . . . . . . . . . . . . . . . . 227
11. Warped Product Submanifolds of Kähler Manifolds 229

11.1 Warped product CR-submanifolds . . . . . . . . . . . . . 229
11.2 CR-warped products and their characterization . . . . . . 231
11.3 Examples of CR-warped products . . . . . . . . . . . . . 233
11.4 A general inequality for CR-warped products . . . . . . . 235
11.5 Twisted product CR-submanifolds . . . . . . . . . . . . . 238
11.6 Warped product submanifolds with a holomorphic
factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
11.7 Warped product hemi-slant submanifolds . . . . . . . . . 244
11.8 Warped product semi-slant submanifolds . . . . . . . . . . 248
11.9 Warped product pointwise semi-slant submanifolds . . . . 251
11.10 Warped product pointwise bi-slant submanifolds . . . . . 252
11.11 Warped products in locally conformal Kähler
manifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
12. CR-warped Products in Complex Space Forms 257

12.1 CR-warped products . . . . . . . . . . . . . . . . . . . . . 257
12.2 A PDE system associated with the basic equality . . . . . 259
12.3 CR-warped products in Cm satisfying basic equality . . . 262
12.4 CR-warped products in CP m and CH m . . . . . . . . . . 270
12.5 CR-warped products with compact holomorphic
factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
13. More on CR-warped Products in Complex Space Forms 283

13.1 Another optimal inequality for CR-warped products . . . 283
13.2 CR-warped products in Cm satisfying the equality . . . . 286
13.3 CR-warped products in CP m satisfying the equality . . . 296
13.4 CR-warped products in CH m satisfying the equality . . . 299
13.5 Irreducibility of real hypersurfaces in non-flat complex
space forms . . . . . . . . . . . . . . . . . . . . . . . . . . 300
13.6 Warped product real hypersurfaces . . . . . . . . . . . . . 314
March 21, 2017 15:26 Diff. Geometry of Warped Product Manifolds-10419 ws-book9x6 page xxix
Contents xxix
14. δ-invariants, Submersions and Warped Products 325

14.1 δ-invariants . . . . . . . . . . . . . . . . . . . . . . . . . . 326
14.2 An inequality for submanifolds in real space forms . . . . 327
14.3 Inequalities for submanifolds in complex space forms . . . 332
14.4 Improved inequalities for Lagrangian submanifolds . . . . 336
14.5 CR-warped products and δ-invariants . . . . . . . . . . . 338
14.6 Anti-holomorphic submanifolds with p ≥ 2 . . . . . . . . . 341
14.7 Anti-holomorphic submanifolds satisfying the equality . . 344
14.8 An optimal inequality for real hypersurfaces . . . . . . . . 346
14.9 Another optimal inequality involving a δ-invariant . . . . 349
14.10 Examples of δ(2)-ideal warped product submanifolds . . . 355
15. Warped Products in Nearly Kähler Manifolds 359

15.1 Nearly Kähler manifolds . . . . . . . . . . . . . . . . . . . 359
15.2 Nearly Kähler structure on S 6 . . . . . . . . . . . . . . . 361
15.3 Complex submanifolds of nearly Kähler manifolds . . . . 363
15.4 Lagrangian submanifolds of nearly Kähler manifolds . . . 366
15.5 CR-submanifolds in nearly Kähler manifolds . . . . . . . 370
15.6 Warped products in nearly Kähler manifolds . . . . . . . 372
15.7 Examples of warped product CR-submanifolds in
nearly Kähler S 6 . . . . . . . . . . . . . . . . . . . . . . . 375
15.8 Non-existence of CR-products in nearly Kähler S 6 . . . . 377
15.9 A special class of warped product submanifolds in nearly
Kähler S 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . 380
16. Warped Products in Para-Kähler Manifolds 383

16.1 Para-Kähler manifolds . . . . . . . . . . . . . . . . . . . . 383
16.2 Non-flat para-Kähler space forms . . . . . . . . . . . . . . 385
16.3 Invariant submanifolds of para-Kähler manifolds . . . . . 387
16.4 Lagrangian submanifolds of para-Kähler manifolds . . . . 389
16.5 P R-submanifolds in para-Kähler manifolds . . . . . . . . 392
16.6 P R-warped products and P -products in para-Kähler
manifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . 396
16.7 P R-products in non-flat para-Kähler space forms . . . . . 398
16.8 Warped product P R-submanifolds . . . . . . . . . . . . . 400
16.9 P R-warped products satisfying the basic equality . . . . . 406
March 21, 2017 15:26 Diff. Geometry of Warped Product Manifolds-10419 ws-book9x6 page xxx
xxx Differential Geometry of Warped Product Manifolds and Submanifolds
17. Warped Products in Sasakian Manifolds 409

17.1 Sasakian manifolds and submanifolds . . . . . . . . . . . . 409
17.2 Warped products in Sasakian manifolds . . . . . . . . . . 412
17.3 Contact CR-submanifolds . . . . . . . . . . . . . . . . . . 414
17.4 CR-warped products with smallest codimension . . . . . . 418
17.5 Another inequality for contact CR-warped products
in Sasakian manifolds . . . . . . . . . . . . . . . . . . . . 420
17.6 Pointwise bi-slant and hemi-slant warped products in
Sasakian manifolds . . . . . . . . . . . . . . . . . . . . . . 424
18. Warped Products in Affine Spaces 427

18.1 Affine spaces and hypersurfaces . . . . . . . . . . . . . . . 427
18.2 Centroaffine hypersurfaces . . . . . . . . . . . . . . . . . . 429
18.3 Graph hypersurfaces . . . . . . . . . . . . . . . . . . . . . 430
18.4 A realization problem for affine hypersurfaces . . . . . . . 432
18.5 Warped products as centroaffine hypersurfaces . . . . . . 437
18.6 Warped products as graph hypersurfaces . . . . . . . . . . 442
18.7 Realization of Robertson-Walker spaces as affine
hypersurfaces . . . . . . . . . . . . . . . . . . . . . . . . . 443
Bibliography 451
General Index 473
Author Index 481
March 21, 2017 15:26 Diff. Geometry of Warped Product Manifolds-10419 ws-book9x6 page 1
Chapter 1
Riemannian and Pseudo-Riemannian

Manifolds
Riemannian geometry was first put forward in generality by B. Riemann

in the middle of nineteenth century. Riemannian geometry, including the
Euclidean geometry and the classical non-Euclidean geometries as the most
special particular cases, deals with a broad range of more general geometries
whose metric properties vary from point to point.
Under the impetus of Einstein’s Theory of General Relativity, the posi-
tiveness of the inner product induced from Riemannian metric was weak-
ened to non-degeneracy. Consequently, one also has the notion of pseudo-
Riemannian manifolds.
Development of Riemannian geometry resulted in a synthesis of diverse
results concerning the geometry of surfaces and the behavior of geodesics
on them, with techniques that can be applied to the study of differen-
tiable manifolds of higher dimensions. It enabled Einstein’s general rela-
tivity theory, made profound impact on group theory and representation
theory, as well as analysis, and spurred the development of algebraic and
differential topology. Since every manifold admits a Riemannian metric,
Riemannian geometry often helps to solve problems of differential topol-
ogy. Most remarkably, by applying Riemannian geometry, G. Y. Perelman
proved Thurston’s geometrization conjecture in 2003; consequently solved
in the affirmative famous Poincaré’s conjecture posed in 1904.
1.1 Symmetric bilinear forms and scalar products
A symmetric bilinear form on a finite-dimensional real vector space V is

a R-bilinear function B : V × V → R such that B(u, v) = B(v, u) for
all u, v ∈ V . A symmetric bilinear form B is said to be positive definite
(resp. positive semi-definite) if B(v, v) > 0 (resp. B(v, v) ≥ 0) for all
1
2 Differential Geometry of Warped Product Manifolds and Submanifolds
v 6= 0. Similarly, a symmetric bilinear form B is called negative definite

(resp. negative semi-definite) if B(v, v) < 0 (resp. B(v, v) ≤ 0) for all
v 6= 0. B is said to be non-degenerate whenever B(u, v) = 0 for all u ∈ V
implies v = 0.
Definition 1.1. The index of a symmetric bilinear form B on V is the
dimension of the largest subspace W ⊂ V on which B|W is negative definite.
Let B be a symmetric bilinear form on V . If we choose a basis v1 , . . . , vn
of V , then the n × n matrix (bij ), bij = B(vi , vj ), is called the matrix of
B with respect to v1 , . . . , vn . Since B is symmetric, the matrix (bij ) is
symmetric. A symmetric bilinear form is non-degenerate if and only if the
matrix of B with respect to one basis is invertible.
Definition 1.2. A scalar product g on a finite-dimensional real vector space
is a non-degenerate symmetric bilinear form. An inner product is a positive
definite scalar product.
By a scalar product space (V, g) we mean a vector space V equipped
with a scalar product g. A subspace U of a scalar product space is called
non-degenerate if g|U is non-degenerate.
Two vectors u, v of a scalar product space V are called orthogonal, which
are denoted by u ⊥ v, if g(u, v) = 0. Two subsets P, Q ⊂ V are said to be
orthogonal, denoted by P ⊥ Q, if g(u, w) = 0 for all u ∈ P and w ∈ Q.
For a subspace U ⊂ V , put U ⊥ = {v ∈ V : v ⊥ U }. Then (U ⊥ )⊥ = U .
Lemma 1.1. A subspace U of a scalar product space V is non-degenerate
if and only if V is the direct sum of U and U ⊥ .
Proof. Since
dim(U + U ⊥ ) + dim(U ∩ U ⊥ ) = dim U + dim U ⊥ = dim V,
U + U ⊥ = V holds if and only if U ∩ U ⊥ = {0} holds. The latter condition
means that U is non-degenerate.
p On a scalar product space V , the norm ||v|| of a vector v is defined to be

|g(v, v)|. When the scalar product space V is positive definite, we simply
denote ||v|| by |v|. In this case |v| is called the length of v.
A vector of norm one is called a unit vector. A set of mutually orthogonal
unit vectors is called an orthonormal set. A set of n orthonormal vectors
e1 , . . . , en of V is called an orthonormal basis whenever n = dim V .
Lemma 1.2. A scalar product space V of positive dimension admits an
orthonormal basis.
Riemannian and Pseudo-Riemannian Manifolds 3
Proof. Since g is non-degenerate, there is a unit vector e1 ∈ V . Let U1

be the subspace spanned by e1 . Then U1⊥ is a non-degenerate subspace.
Thus there is a unit vector e2 ∈ (U1 )⊥ . The pair {e1 , e2 } is an orthonormal
basis of Span{e1 , e2 }. By continuing this process (n − 1)-times, we obtain
an orthonormal basis e1 , . . . , en of V .
For an orthonormal basis e1 , . . . , en of a scalar product space V , we have
g(ei , ej ) = ǫi δij , ǫi = g(ei , ei ) = ±1,
where δij is the Kronecker delta,which is equal to 1 if i = j; and equal to 0
if i 6= j. Every vector v ∈ V can be expressed in a unique way as
n
X
v= ǫi g(v, ei )ei .
i=1
For an orthonormal basis e1 , . . . , en of a scalar product space V , the number

of negative signs in the signature (ǫ1 , . . . , ǫn ) is the index of V .
A linear transformation T : V → W between two scalar product spaces
is called a linear isometry if it preserves the scalar products. Two scalar
product spaces are linear isometric if and only if they have the same di-
mension and the same index.
1.2 Riemannian and pseudo-Riemannian manifolds
A (pseudo-Riemannian) metric tensor g on a manifold M is a symmetric

non-degenerate (0, 2) tensor field on M of constant index, i.e., g assigns to
each point x ∈ M a scalar product gx on Tx M and the index of gx is the
same for all x ∈ M . Very often, we use h , i as an alternative notation for
g. Thus we have g(v, w) = hu, vi.
A pseudo-Riemannian n-manifold is by definition an n-dimensional
manifold equipped with a (pseudo-Riemannian) metric tensor g. The com-
mon value s, 0 ≤ s ≤ n, of index on M is called the index of M . If s = 0,
M is called a Riemannian manifold. In this case, each gx is a positive defi-
nite inner product on Tx M . A pseudo-Riemannian manifold (resp. metric)
is also known as a semi-Riemannian manifold (resp. metric). A pseudo-
Riemannian metric on an even-dimensional manifold M is called a neutral
metric if its index is equal to 12 dim M .
If the index of M is one, M is called a Lorentz manifold and the corre-
sponding metric is called Lorentzian. A manifold of dimension ≥ 2 admits
a Lorentzian metric if and only if it admits a 1-dimensional distribution.
A tangent vector v of a pseudo-Riemannian manifold M is called space-

like (resp. timelike) if v = 0 or hv, vi > 0 (resp. hv, vi < 0). A vector v is
called lightlike or null if hv, vi = 0 and v 6= 0.
The light cone LC of Ens is defined by LC = {v ∈ Ens : hv, vi = 0}. A
curve in a pseudo-Riemannian manifold is called a null curve if its velocity
vector is a lightlike at each point.
A vector in a Lorentzian vector space that is non-spacelike (i.e., either
lightlike or timelike) is called causal . A causal curve in a spacetime is a
curve whose velocity vectors are all non-spacelike.
Let {u1 , . . . , un } be a coordinate system on an open subset U ⊂ M ,
where n = dim M . Then the components gij of the metric tensor g on U
are given by gij = h∂i , ∂j i , 1 ≤ i, j ≤ n, where ∂i = ∂/∂ui . Since g is a
symmetric (0, 2) tensor field, we have gij = gji for 1 ≤ i, j ≤ n. Hence the
metric tensor on U can be written as
Xn
g= gij dui ⊗ duj . (1.1)
i,j=1
At each point x in the Euclidean n-space En , there exists a canonical

linear isomorphism from En onto Tx En . In terms of natural coordinates on
P
En , it sends a vector v to vx = v j ∂j . The inner product on En gives rise
to a metric tensor on En with
n
X
hvx , wx i = vj wj (1.2)
j=1
Pn Pn
with v = j=1 vj ∂j and w = j=1 wj ∂j .
For an integer s ∈ [0, n], if we change the first s plus signs in (1.2) to
minus sign, then it gives rise to a metric tensor
Xs Xn
hvx , wx i = − vj wj + vk wk (1.3)
j=1 k=s+1
of index s. The resulting pseudo-Euclidean space is denoted by Ens .

If s = 0, Ens reduces to the Euclidean n-space En . The En1 is called a
Minkowski n-space. When n = 4 and s = 1, it is the simplest example of a
relativistic spacetime, known as the Minkowski spacetime.
1.3 Levi-Civita connection
Let M be an n-manifold. Denote by F (M ) the set of all smooth real-valued

functions on M . If f1 , f2 are smooth functions on M , so is their sum f1 + f2
and product f1 f2 . The usual algebraic rules hold for these two operations,
which make F (M ) a commutative ring. We denote by X(M ) the set of all
smooth vector fields on M .
For V, W ∈ X(M ), the bracket [V, W ] is defined by
[V, W ]x (f ) = Vx (W f ) − Wx (V f )
at each x ∈ M and f ∈ F (M ). The bracket operation [ , ] on X(M ) is a
R-bilinear and skew-symmetric, which also satisfies the Jacobi identity:
[X, [Y, Z] ] + [Y, [Z, X] ] + [Z, [X, Y ] ] = 0.
These makes X(M ) an infinite-dimensional Lie algebra.
Definition 1.3. An affine connection ∇ on a manifold M is a function

∇ : X(M ) × X(M ) → X(M ) such that
(1) ∇X Y is F (M )-linear in X;
(2) ∇X Y is R-linear in Y ;
(3) ∇X (f Y ) = (Xf )Y + f ∇X Y for f ∈ F (M ).
∇X Y is called the covariant derivative of Y with respect to X.

The torsion tensor T of an affine connection ∇ is a tensor of type (1, 2)
defined by T (X, Y ) = ∇X Y − ∇Y X − [X, Y ].
The following theorem shows that on a pseudo-Riemannian manifold
there exists a unique connection sharing two further properties.
Theorem 1.1. On a pseudo-Riemannian manifold M , there exists a unique

affine connection ∇ such that
(4) ∇ is torsion free, i.e., [Y, Z] = ∇Y Z − ∇Z Y , and

(5) XhY, Zi = h∇X Y, Zi + hY, ∇X Zi
for X, Y, Z ∈ X(M ). This unique affine connection ∇ is called the Levi-

Civita connection of M and it is characterized by the Koszul formula:
2 h∇Y Z, Xi = Y hZ, Xi + Z hX, Y i − X hY, Zi
(1.4)
− hY, [Z, X]i + hZ, [X, Y ]i + hX, [Y, Z]i .
Proof. Let ∇ be an affine connection which satisfies both properties (4)

and (5). Then after applying (4) and (5) on the right-hand side of (1.4) we
obtain 2 h∇Y Z, Xi. Hence ∇ satisfies the Koszul formula. Therefore there
exists only one affine connection on M which satisfies both properties (4)
and (5).
For the existence, let us define F (Y, Z, X) to be the right-hand side of

(1.4). A direct computation shows that the function X 7→ F (Y, Z, X) is a
F (M )-linear for fixed Y, Z. Thus it is a 1-form. Hence there exists a unique
vector field, denoted by ∇Y Z such that 2 h∇Y Z, Xi = F (Y, Z, X) for all X.
Therefore the Koszul formula holds and we can deduce properties (1)-(5)
from it.
Let {u1 , . . . , un } be a local coordinate system on an open subset U

of a pseudo-Riemannian n-manifold M . The Christoffel symbols for the
coordinate system are the real-valued functions Γkij on U such that
n
X
∇∂i ∂j = Γkij ∂k , 1 ≤ i, j ≤ n.
k=1
Since the connection ∇ is not a tensor, the Christoffel symbols do not

obey the usual tensor transformation rule under change of coordinates.
For the Christoffel symbols we have the following.
Proposition 1.1. Let M be a pseudo-Riemannian n-manifold and let

{u1 , . . . , un } be a coordinate system on an open subset U ⊂ M . Then

Pn ∂Yk Pn k
(1) ∇∂i Y = k=1 + j=1 Γij Yj ∂k and
∂ui

k
Pn g kt ∂gjt ∂git ∂gij
(2) Γij = t=1 + − ,
2 ∂ui ∂uj ∂ut
Pn
where Y = j=1 Yj ∂j and (g ij ) is the inverse matrix of (gij ).
Proof. Statement (1) is an immediate consequence of property (3) given

in Definition 1.3.
To prove (2), let us put X = ∂t , Y = ∂i , Z = ∂j in the Koszul formula.
Since the brackets are zero, it leaves
∂gjt ∂git ∂gij
2 h∇∂i ∂j , ∂t i = + − .
∂ui ∂uj ∂ut
But from the definition of Christoffel symbols we have
n
X
2 h∇∂i ∂j , ∂t i = 2 Γkij gkt .
k=1
P
Attacking both equations with t g tk yields the required formula.
1.4 Parallel transport
Let φ : N → M be a smooth map between two manifolds. The differential

at a point x ∈ N is a linear map φ∗x : Tx N → Tφ(x) M defined as follows:
For each X ∈ Tx N , φ∗x X is the tangent vector in Tφ(x) M such that
(φ∗x X)f = X(f ◦ φ), ∀f ∈ F (N ).
We denote the dual of the differential φ∗ by φ∗ .
For any q-form ω on M , define the q-form φ∗ ω on N by
(φ∗ ω)(X1 , . . . , Xq ) = ω(φ∗ X1 , . . . , φ∗ Xq ), X1 , . . . , Xq ∈ Tx N.
A vector field Z on a smooth map φ : P → M between two manifolds
is a mapping Z : P → T M such that π ◦ Z = φ, where π is the projection
T M → M . The simplest case of a vector field on a mapping is a vector
field Z along a curve γ : I → M defined on an open interval I, where Z
smoothly assigns to each t ∈ I a tangent vector to M at γ(t). For instance,
the velocity vector field γ ′ on γ is a vector field on the curve γ. Let V(γ)
denote the set consisting of smooth vector fields of M along γ.
For a pseudo-Riemannian manifold M , there is a natural way to define
the vector rate of change Z ′ of a vector field Z ∈ V(γ).
Proposition 1.2. Let γ : I → M be a curve in a pseudo-Riemannian
7 Z ′ = DZ
manifold M . Then there exists a unique function Z → dt from
V(γ) → V(γ) such that
(1) (aZ1 + bZ2 )′ = aZ1′ + bZ2′ ,

dλ
(2) (λZ)′ = Z + λZ ′ ,
dt
(3) (Vγ )′ (t) = ∇γ ′ (t) V ,
where a, b ∈ R, λ ∈ F (I), V ∈ X(M ) and t ∈ I. Furthermore, we have
d
(4) hZ1 , Z2 i = hZ1′ , Z2 i + hZ1 , Z2′ i.
dt
Proof. For the uniqueness, let us assume that an induced connection
exists which satisfy only the first three properties. We can assume that γ
lies in the domain of a single coordinate system {u1 , . . . , un }. For a vector
Pn Pn
field Z ∈ V(γ), we have Z(t) = j=1 (Z(t)uj )∂j = j=1 (Zuj )(t)∂j .
Let us denote the component function Zuj : I → R by Zj . Then, by
properties (1), (2) and (3), we find
Xn X Xn Xn
dZj dZj
Z′ = ∂j |γ + Zj (∂j |γ )′ = ∂j + Zj ∇γ ′ (∂j ).
j=1
dt j j=1
dt j=1
Thus Z ′ is completely determined by the Levi-Civita connection ∇. This

shows uniqueness. On any subinterval J of I such that γ(J) lies in a
coordinate neighborhood, let us define Z ′ by the formula above. Then
straightforward computations show that all four properties hold. Now, it
follows from the uniqueness that these local definitions of Z ′ gives rise to a
single vector field Z ′ ∈ V(γ).
The Z ′ = DZ/dt in Proposition 1.2 is called the induced covariant
derivative. For a vector field Z along a curve γ, we simply write Z ′ for
∇γ ′ Z and also γ ′′ for ∇γ ′ γ ′ . In terms of Christoffel symbols we have
n
( n
)
X X
′ dZk k d(ui ◦ γ)
Z = + Γij Zj ∂k .
dt i,j=1
dt
k=1
A vector field Z on γ is called parallel if Z ′ = 0 holds identically along
P
γ. Hence Z = k Zk ∂k is a parallel vector field if and only if Z1 , . . . , Zn
satisfy the system of ordinary differential equations:
X n
dZk d(ui ◦ γ)
+ Γkij Zj = 0, k = 1, . . . , n.
dt i,j=1
dt
Proposition 1.3. For γ : I → M , a ∈ I and z ∈ Tγ(a) M , there exists a

unique parallel vector field Z on γ such that Z(a) = z.
Proof. Follows from the fundamental existence and uniqueness theorem

of systems of first order linear equations.
Consider a curve γ : I → M . Let a, b ∈ I and z ∈ Tγ(a) M . The function
γ(b)
P = Pγ(a) (γ) : Tγ(a) M → Tγ(b) M
sending each z ∈ Tγ(a) M to Z(γ(b)) is called parallel translation along γ
from γ(a) to γ(b), where Z is the unique parallel vector field along γ such
that Z(a) = z.
Proposition 1.4. Parallel translation is a linear isometry.
Proof. Let γ : I → M be a curve and x = γ(a), y = γ(b). Let u, v ∈ Tx M

correspond to parallel vector fields U, V . Since U + V is also parallel, we
have
P (u + v) = (U + V )(b) = U (b) + V (b) = P (u) + P (v).
Similarly, we have P (cu) = cP (u). Hence P is a linear map. For U, V as
above, we get
d
hU, V i = hU ′ , V i + hU, V ′ i = 0.
dt
Thus hU, V i is constant. Hence

hP (u), P (v)i = hU (b), V (b)i = hU (a), V (a)i = hu, vi ,
which implies that P is an isometry.
Definition 1.4. A geodesic in a pseudo-Riemannian manifold M is a curve

γ : I → M whose velocity vector field γ ′ is parallel, or equivalently, it
satisfies
X n
d2 (uk ◦ γ) d(ui ◦ γ) d(uj ◦ γ)
+ Γkij (γ) =0 (1.5)
dt2 i,j=1
dt dt
for k = 1, . . . , n.
It follows from the existence and uniqueness theorem of linear system

of ordinary differential equations that, for any given point x ∈ M and any
given tangent vector v ∈ Tx M , there exists a unit geodesic γv such that
γ(0) = x and γ ′ (0) = v. A geodesic with largest possible domain is called
a maximal geodesic.
A pseudo-Riemannian manifold M for which every maximal geodesic
is defined on the entire real line is said to be geodesic complete or simply
complete.
It follows from (1.5) that the geodesic of a pseudo-Euclidean space Em s
are straight lines. Thus every pseudo-Euclidean n-plane Em s is geodesically
complete.
In general, parallel translation from a point x to another point y depends
on the particular curve jointing two points x and y. However, on a pseudo-
Euclidean space Em s the natural coordinate vector fields are parallel and
hence so their restrictions to any curve. Consequently, parallel translation
from a point x to another point y along any curve is just the canonical
isomorphism vx → vy . This phenomenon is called distant parallelism.
For a given v ∈ Tx M , there is a unique geodesic γv such that γv (0) = x
with initial tangent vector γv′ (0) = v. Let Ux be the set of vectors v ∈ Tx M
such that the geodesic γv is defined at least on [0, 1]. For a vector v ∈ Ux
the exponential map is defined by expx (v) = γv (1).
Definition 1.5. A subset S of a vector space is called star shaped about

o if v ∈ S implies tv ∈ S for all t ∈ [0, 1]. For each point x ∈ M there
exists a neighborhood U of o in Tx M on which the exponential map expo
is a diffeomorphism onto a neighborhood U of x on M . If U is starshaped
about o, then U is called a normal neighborhood of x.
Definition 1.6. Let {e1 , . . . , en } be an orthonormal basis of Tx M such that

hei , ej i = ǫi δij . The normal coordinate system {y1 , . . . , yn } determined by
e1 , . . . , en assigns to each point y ∈ U the vector coordinates relative to
e1 , . . . , en of the corresponding point exp−1
x (y) ∈ U ⊂ Tx M.
In other words,
n
X
exp−1
x (y) = yi (y)ei , y ∈ U.
i=1
The following proposition is well-known.
Proposition 1.5. Let {y1 , . . . , yn } be a normal coordinate system about a

point x ∈ M . Then gij (x) = δij and Γijk (x) = 0.
Definition 1.7. Let x be a point in a Riemannian manifold M . Let Sr be

the hypersphere of Tx M with radius r centered at the origin o. Suppose
that r is a sufficiently small positive number such that expx (Sr ) lies in a
normal coordinate neighborhood of x, then expx (Sr ) is called a geodesic
hypersphere. The set Br (x) = {u ∈ M : d(u, x) ≤ r} is called a geodesic
ball of radius r and with center x.
1.5 Riemann curvature tensor
Gauss’ “theorema egregium” shows that the Gauss curvature, defined as

product of two principal curvatures, of a surface in a Euclidean 3-space E3
is an isometric invariant of the surface itself. This lead G. Riemann to his
invention of Riemannian geometry, whose most important feature is the
generalization of Gauss curvature to arbitrary Riemannian manifolds. No
significant changes are required in extending from Riemannian to pseudo-
For a pseudo-Riemannian manifold M with Levi-Civita connection ∇,
the function
R : X(M ) × X(M ) × X(M ) → X(M )
defined by
R(X, Y )Z = ∇X ∇Y Z − ∇Y ∇X Z − ∇[X,Y ] Z (1.6)
is a (1, 3) tensor field, called the Riemann curvature tensor. Sometimes, we
put
R(X, Y ; Z, W ) = hR(X, Y )Z, W i .
Proposition 1.6. The curvature tensor R satisfies the following properties:

R(u, v)w = −R(v, u)w, (1.7)
hR(u, v)w, zi = − hR(u, v)z, wi , (1.8)
R(u, v)w + R(v, w)u + R(w, u)v = 0, (1.9)
hR(u, v)w, zi = hR(w, z)u, vi (1.10)
for vectors u, v, w, z ∈ Tx M, x ∈ M .
Proof. Since both ∇ and the bracket operation on vector fields are local
operations, it suffices to work on any neighborhood of x. Moreover, be-
cause the identities are tensor equations, u, v, w, z can be extended to local
vector fields U, V, W, Z on some neighborhood of x in any convenient way.
In particular, we may choose the extensions in such way that all of their
brackets are zero.
Since R(U, V )W = [∇U , ∇V ]W − ∇[U,V ] W and the bracket operation is
skew-symmetric, (1.7) follows immediately from the definition of the cur-
vature tensor.
For (1.8) we only need to show that hR(u, v)w, zi = 0 by polarization.
By Theorem 1.1(5), we have
hR(U, V )W, W i = h∇U ∇V W, W i − h∇V ∇U W, W i
= h∇U W, ∇V W i − V h∇U W, W i + h∇V W, ∇U W i − U h∇V W, W i
1 1
= U V hW, W i − V U hW, W i = 0.
2 2
This proves (1.8), since [U, V ] = 0. For (1.9) we consider S, the sum of
cyclic permutations of U, V, W , to find
R(U, V )W + R(V, W )U + R(W, U )V
= SR(U, V )W
= S∇U ∇V W − S∇V ∇U W
= S∇U ∇V W − S∇U ∇W V
= S∇U [Y, W ] = 0.
If we put
S(u, v, w, z) = hR(u, v)w, zi + hR(v, w)u, zi + hR(w, u)v), zi ,
then a direct computation shows that
0 = S(u, v, w, z) − S(v, w, z, u) − S(w, z, u, v) + S(z, u, v, w)
= hR(u, v)w, zi − hR(v, u)w, zi − hR(w, z)u, vi + hR(z, w)u, vi .
Thus, by applying (1.7), we obtain (1.10).
Equation (1.9) is called the first Bianchi identity.
Proposition 1.7. The curvature tensor of a pseudo-Riemannian manifold

M satisfies the second Bianchi identity:
(∇W R)(U, V ) + (∇U R)(V, W ) + (∇V R)(W, U ) = 0, (1.11)
where (∇W R)(U, V ) is defined by
((∇W R)(U, V ))Z = ∇W (R(U, V )Z) − R(∇W U, V )Z
(1.12)
− R(U, ∇W V )Z − R(U, V )(∇W Z).
Proof. Clearly, (1.11) is a tensor identity. Let x be a given point in M .
We consider a normal coordinate system on a neighborhood of x.
We choose the extensions U, V, W of vectors u, v, w ∈ Tx M in such way
that not only all brackets vanishes identically, but also the extensions have
constant components with respect to the normal coordinate system. Hence,
by Proposition 1.5, we have
(∇W R)(U, V )Z = ∇W (R(U, V )Z) − R(U, V )(∇W Z)
at x, which gives (∇W R)(U, V ) = [∇W , R(U, V )] = [∇W , [∇U , ∇V ] ] at x.
Thus summing the above formula over the cyclic permutations of U, V, W
yields the required identity at x.
The curvature tensor R also satisfies the following identities:
(∇X ∇Y ω)(Z) − (∇Y ∇X ω)(Z) − (∇[X,Y ] ω)(Z) = −ω(R(X, Y )Z) (1.13)
and
(∇X ∇Y S)(Z, U ) − (∇Y ∇X S)(Z, U ) − (∇[X,Y ] S)(Z, U )
(1.14)
= R(X, Y )S(Z, U ) − S(R)X, Y )Z, U ) − S(Z, R(X, Y )U )
for a tensor field S of type (1, 2) and 1-form ω, where Z, Y, Z, U are arbitrary
vector fields. The formulas (1.13) and (1.14) are called Ricci identities.
1.6 Sectional, Ricci and scalar curvatures
Since the Riemann curvature tensor is rather complicated, we consider a

simpler real-valued function, the sectional curvature, which completely de-
termines the curvature tensor.
At a point x ∈ M , a 2-dimensional linear subspace π of the tangent
space Tx M is called a plane section. For a given basis {v, w} of the plane
section π, we define a real number by
Q(v, w) = hv, vi hw, wi − hv, wi2 .
The plane section π is called non-degenerate if and only if Q(u, v) 6= 0.

Q(u, v) is positive if g|π is definite, and it is negative if g|π is indefinite.
The absolute value |Q(u, v)| is the square of the area of the parallelogram
with sides u and v.
For a non-degenerate plane section π at x, the number
hR(u, v)v, ui
K(u, v) =
Q(u, v)
is independent of the choice of basis {u, v} for π, which is called the sectional
curvature K(π) of π.
A pseudo-Riemannian manifold is said to be flat if its sectional curvature
vanishes identically. It is well-known that a pseudo-Riemannian manifold
M is flat if and only if its curvature tensor vanishes at every point.
For any index s, the pseudo-Euclidean m-space Em s is flat. In fact,
all Christoffel symbols vanish for a natural coordinate system. Hence the
curvature tensor of Em s vanishes identically.
Definition 1.8. A multilinear function

F : Tx M × Tx M × Tx M × Tx M → R
is called curvature-like if F satisfies properties (1.7)-(1.10) for the function
(u, v, w, z) → hR(u, v)w, zi.
For a curvature-like function F we have the following.
Lemma 1.3. Let M be a pseudo-Riemannian manifold and x ∈ M . If F
is a curvature-like function on Tx M such that
F (u, v, v, u)
K(u, v) =
Q(u, v)
whenever u, v span a non-degenerate plane at x, then
hR(u, v)w, zi = F (u, v, w, z)
for all u, v, w, z ∈ Tx M .
Proof. If we put δ(u, v, w, z) = F (u, v, w, z)−hR(u, v)w, zi, then δ is also
curvature-like. Since δ(u, v, v, u) = 0 whenever u, v span a non-degenerate
plane section at x, we obtain δ = 0.
For sectional curvature K of indefinite Riemannian manifolds, we have
the following result [Kulkarni (1979)].
Theorem 1.2. Let M be a pseudo-Riemannian manifold of dimension ≥ 3
and index s > 0. Then, at each point x ∈ M , the following four conditions
are equivalent:
(1) K is constant;
(2) a ≤ K or K ≤ b;
(3) a ≤ K ≤ b on indefinite planes;
(4) a ≤ K ≤ b on definite planes,
where a and b are real numbers.
It follows from Theorem 1.2 that the sectional curvature of an indefinite

Riemannian manifold at each point is unbounded from above and below
unless M has constant sectional curvature.
Definition 1.9. The Ricci tensor of a pseudo-Riemannian n-manifold M ,

denoted by Ric, is a symmetric (0, 2) tensor defined by
Ric(X, Y ) = Tr{Z 7→ R(Z, X)Y },
or equivalently,
n
X
Ric(X, Y ) = ǫℓ hR(eℓ , X)Y, eℓ i , (1.15)
ℓ=1
where e1 , . . . , em is an orthonormal frame.
It is well-known that Ric(X, Y ) is independent of choice of orthonormal

frame e1 , . . . , en . If the Ricci tensor vanishes, then M is called a Ricci flat
manifold.
Definition 1.10. A pseudo-Riemannian manifold M is called an Einstein

manifold if Ric = cg for some constant c. For a unit vector u ∈ T M , the
Ricci curvature Ric(u) is defined by Ric(u) = Ric(u, u).
If M is a pseudo-Riemannian manifold of dimension ≥ 3 which satisfies

Ric = f g for some function f ∈ F (M ), then M is always Einsteinian.
Definition 1.11. The scalar curvature τ of M is defined by

X
τ= K(ei , ej ), (1.16)
i<j
where e1 , . . . , en is an orthonormal frame of M . The scalar curvature τ is

independent of the choice of the orthonormal frame.
Remark 1.1. The Ricci curvature provides one way of measuring the de-
gree to which the geometry determined by the given Riemannian metric
on M differ from that of the Euclidean n-space. The scalar curvature rep-
resents the amount by which the volume of a geodesic ball in a curved
Riemannian manifold deviates from that of the standard ball in Euclidean

space.
In relativity theory, the Ricci tensor is related to the matter content of
the universe via Einstein’s field equations. It is the part of the curvature
of a spacetime that determines the degree to which matter will tend to
converge or diverge in time. The scalar curvature is the Lagrangian density
for the Einstein-Hilbert action, proposed in [Hilbert (1915)], that yields the
Einstein field equations through the principle of least action.
1.7 Indefinite real space forms
A pseudo-Riemannian manifold M is said to have constant curvature if its

sectional curvature is constant. For a constant c the function F defined by
F (u, v, w, z) = c{hu, zi hv, wi − hu, wi hv, zi}
is curvature-like. Thus Lemma 1.3 implies that F (u, v, v, u) = cQ(u, v).
Hence if u, v span a non-degenerate plane section, we have
F (u, v, v, u)
K(u, v) = c = .
Q(u, v)
Consequently, if a pseudo-Riemannian manifold M is of constant curvature
c, then its curvature tensor R satisfies
R(u, v)w = c{hv, wi u − hu, wi v}. (1.17)
Let Ent
be the pseudo-Euclidean n-space equipped with the canonical
pseudo-Euclidean metric of index t given by
Xt Xn
g0 = − dx2i + dx2j , (1.18)
i=1 j=t+1
where (x1 , . . . , xn ) is a rectangular coordinate system of Ent .

Let c be a nonzero real number. We put
n o
1
Ssk (x0 , c) = x ∈ Ek+1 s : hx − x0 , x − x0 i = > 0 , s > 0, (1.19)
c
n o
k k+1 1
Hs (x0 , c) = x ∈ Es+1 : hx − x0 , x − x0 i = < 0 , s > 0, (1.20)
c
n o
k k+1 1
H (c) = x ∈ E1 : hx, xi = < 0 and x1 > 0 , (1.21)
c
where h , i is the associated scalar product and x=(x1 , . . . , xn ).
Ssk (x0 , c) and Hsk (x0 , c) are pseudo-Riemannian manifolds of curvature
c with index s, known as a pseudo sphere and a pseudo-hyperbolic space,
respectively.
The point x0 is called the center of Ssm (x0 , c) and Hsm (x0 , c). If x0 is
the origin o, we simply denote Ssk (o, c) and Hsk (o, c) by Ssk (c) and Hsk (c),
respectively.
The pseudo-Riemannian manifolds Eks , Ssk (c), Hsk (c) are the standard
models of the indefinite real space forms. In particular, Ek1 , S1k (c), H1k (c)
are the standard models of Lorentzian space forms. Topologically, a de
Sitter spacetime S1k is R × S k−1 . Thus when k ≥ 3 a de Sitter spacetime is
simply-connected.
The S14 and H14 are known as the de Sitter spacetime and anti-de Sitter
spacetime, respectively; named after Willem de Sitter (1872-1934), a Dutch
mathematician, physicist and astronomer.
When s = 0, the manifolds Ek , S k (c) and H k (c) are of constant curva-
ture, called real space forms. The Euclidean k-space Ek , the k-sphere S k (c)
and the hyperbolic k-space H k (c) are simply-connected complete Rieman-
nian manifolds of constant curvature 0, c > 0 and c < 0, respectively.
A complete simply-connected pseudo-Riemannian k-manifold, k ≥ 3,
of constant curvature c and index s is isometric to Eks , or Ssk (c) or Hsk (c)
according to c = 0, or c > 0 or c < 0, respectively.
We denote a k-dimensional indefinite space form of curvature c and
index s simply by Rsk (c). We simply denote the indefinite space form R0k (c)
with index s = 0 by Rk (c).
1.8 Gradient, Hessian and Laplacian
We give the following definitions.
Definition 1.12. Let M be a pseudo-Riemannian n-manifold. For f ∈

F (M ), the gradient of f , denote by ∇f (or by grad f ), is the vector field
dual to the differential df . In other word, ∇f is defined by
h∇f, Xi = df (X) = Xf ∀X ∈ X(M ). (1.22)
In terms of a coordinate system {u1 , . . . , un } of M , we have
Xn X
∂f ∂f
df = duj and ∇f = g ij ∂j . (1.23)
j=1
∂uj i,j
∂ui
Definition 1.13. If X ∈ X(M ) and {e1 , . . . , en } is an orthonormal frame,

the divergence of X, denoted by div X, is defined by

n
X
div X = ǫj h∇ei X, ei i . (1.24)
j=1
Pn ∂ Pn
If we put X = j=1X j ∂u j
, Xi = j=1 gij X j , then
n
( n
)
X ∂Xi X j
div X = + Γjk Xk . (1.25)
j=1
∂ui
k=1
Definition 1.14. The Hessian of f ∈ F (M ), denoted by H f , is the second

covariant differential ∇(∇f ), so that
H f (X, Y ) = XY f − (∇X Y )f = h∇X (∇f ), Y i (1.26)
for X, Y ∈ X(M ).
Definition 1.15. The Laplacian of f ∈ F (M ), denoted by ∆f , is defined

by ∆f = −div(∇f ). In terms of a coordinate system {u1 , . . . , un }, we have
n
( n
)
X ∂2f X
k ∂f
∆f = − − Γij . (1.27)
i,j=1
∂ui ∂uj ∂uk
k=1
The Laplacian (or the Laplace operator) is named after Pierre-Simon de

Laplace (1749-1827), who first applied the operator to the study of celestial
mechanics in the 1770s.
In terms of a natural coordinate system {x1 , . . . , xn } of Ens , we have
n
X Xn Xn
∂f ∂Xj ∂2f
∇f = ǫj ∂j , div X = , ∆f = − ǫj 2 .
j=1
∂xj j=1
∂xj j=1
∂xj
1.9 Lie derivative and Killing vector fields
The algebraic definition of the Lie derivative of a tensor field on a manifold

M follows from the following four axioms:
Axiom 1. The Lie derivative of a function is the directional derivative of

the function: LX f = Xf.
Axiom 2. The Lie derivative obeys the following version of Leibniz’s rule:
LX (S ⊗ T ) = (LX S) ⊗ T + S ⊗ (LX T )
for any tensor fields S and T .
Axiom 3. The Lie derivative obeys the Leibniz rule with respect to con-
traction:
LX (T (Y1 , . . . , Yn )) = (LX T )(Y1 , . . . , Yn ) + T (LX Y1 , . . . , Yn )
+ T (Y1 , . . . , LX Yn ).
Axiom 4. The Lie derivative commutes with exterior derivative d on
functions: [LX , d] = 0.
Lemma 1.4. On a pseudo-Riemannian manifold (M, g), we have

(Lv g)(X, Y ) = g(∇X v, Y ) + g(X, ∇Y v) (1.28)
for vector fields X, Y and v on M .
Proof. Follows easily from Axiom 1, Axiom 3 and Theorem 1.1(4).
Definition 1.16. A Killing vector field on a pseudo-Riemannian manifold

is a vector field X for which the Lie derivative of the metric tensor vanishes,
i.e., LX g = 0.
A conformal vector field is a vector field X for which LX g = λg for some
function λ ∈ F (M ).
Under the flow of a Killing vector field X, the metric tensor does not
change. Thus a Killing vector field is an infinitesimal isometry.
Proposition 1.8. On a pseudo-Riemannian manifold M , the following

three conditions on a vector field X of M are equivalent:
(1) X is a Killing vector field.
(2) X hV, W i = h[X, V ], W i + hV, [X, W ]i.
(3) ∇X is a skew-adjoint relative to the metric tensor g, i.e.,
h∇V X, W i = − h∇W X, V i ,
where V and W are arbitrary vector fields on M .
Proof. For all V, W ∈ X(M ), the following are equivalent:

h∇V X, W i + h∇W X, V i = 0;
h∇X V, W i − h[X, V ], W i = h[X, W ], V i − h∇X W, V i ;
X hV, W i = h[X, V ], W i + hV, [X, W ]i .
In view of the product rule, the last one is equivalent to LX g = 0 according
to Lemma 1.4.
Definition 1.17. Let N and M be pseudo-Riemannian manifolds with

metrics gN and gM . An isometry ψ : N → M is a diffeomorphism that
preserves metric tensors, i.e., ψ ∗ (gM ) = gN .
Definition 1.18. A map φ : N → M between two pseudo-Riemannian

manifolds is called a local isometry at x ∈ N if there is a neighborhood
U ⊂ N of x such that φ : U → φ(U ) is a diffeomorphism satisfying
hu, vix = hφ∗x (u), φ∗x (v)iφ(x) , ∀v ∈ Tx N, ∀x ∈ N. (1.29)
Definition 1.19. A pseudo-Riemannian manifold N is said to be locally

isometric to a pseudo-Riemannian manifold M if for each x ∈ N there
exists a neighborhood U of x and a local isometry φ : U → φ(U ) ⊂ M .
Definition 1.20. A map φ : N → M between two pseudo-Riemannian

manifolds is called conformal if φ∗ (gM ) = f gN for some function f ∈ F (N )
such that f > 0 or f < 0. In particular, if the function f is a nonzero real
number, then φ is called a homothety.
Lemma 1.5. Homothety preserves Levi-Civita connection of pseudo-

Proof. Follows immediately from Koszul’s formula.
1.10 Concircular and concurrent vector fields

Definition 1.21. A vector field on a pseudo-Riemannian manifold N is
called a concircular vector field if it satisfies
∇X v = µX (1.30)
for vectors X tangent to N , where ∇ denotes the Levi-Civita connection of
N and µ is a non-trivial function on N . A concircular vector field satisfying
(1.30) is called non-trivial if the function µ is non-constant.
Remark 1.2. It follows immediately from Lemma 1.4 and (1.30) that every
concircular vector field is a conformal vector field.
Traditionally, a concircular vector field v is called a concurrent vector

field if the function µ in (1.30) is equal to one (cf. e.g., [Petrović et. al.
(1989); Shouten (1954); Yano (1940); Yano and Chen (1971)]). However,
in this book, for simplicity we call a concircular vector field v a concurrent
vector field if the function µ in (1.30) is a nonzero constant.
April 18, 2017 12:18 Diff. Geometry of Warped Product Manifolds-10419 ws-book9x6 page 20
First, we give the following examples of Riemannian manifolds endowed

with concircular vector fields.
Example 1.1. Let I be an open interval of the real line R and let ϕ(s), s ∈
I, be a function on I which is nowhere zero. Consider a warped product
manifold of the form:
I ×ϕ(s) F, (1.31)
where F is a Riemannian manifold. The metric tensor g of I ×ϕ(s) F is
given by g = ds2 + ϕ2 (s)gF , where gF is the metric tensor of the second
factor F . Consider the vector field given by
∂
v = ϕ(s) . (1.32)
∂s
It is easy to verify that the vector field v satisfies (1.30) with µ = ϕ′ (s).
Thus v is a concircular vector field.
We give the following necessary and sufficient condition for a gradient

vector field on a Riemannian manifold to be concircular [Chen (2015)].
Lemma 1.6. Let f be a function on a Riemannian manifold M . Then the

gradient ∇f of f is a concircular vector field if and only if the Hessian H f
of f satisfies
H f (X, Y ) = µg(X, Y ) (1.33)
for X, Y tangent to M , where µ is the function on M . Moreover, in such
case the function µ satisfies (1.30) with v = ∇f .
Proof. Let f be a function of a Riemannian manifold M . Assume that

the Hessian of f satisfies (1.33). Then we have
g(X, µY ) = XY f − ∇X Y f
= XhY, ∇f i − h∇X Y, ∇f i (1.34)
= hY, ∇X (∇f )i
for vector fields X, Y tangent to M . Thus we obtain ∇X (∇f ) = µX, which
implies that v = ∇f is a concircular vector field satisfying (1.30).
The converse can be verified in a similar way.
The following result classifies all concircular vector fields on En .
Proposition 1.9. Let v be a nonzero vector field on the Euclidean n-space

En . Then v is a concircular vector field if and only if only if v = bZ , where
b is a nonzero constant and Z is a concurrent vector field.
Proof. If v is a concircular vector field on a Riemannian manifold, then

it follows from (1.30) and a direct computation that
R(X, v)v = ∇X ∇v v − ∇v ∇X v − ∇[X,v] v
(1.35)
= (Xµ)v − (vµ)X
for any vector field X perpendicular to v, where µ is defined by (1.30). It
follows from (1.35) that the gradient ∇µ is parallel to v.
If M is the Euclidean n-space En , we have R = 0. Hence we obtain
from (1.35) that Xµ = vµ = 0, which implies that µ is a nonzero constant,
say b. Consequently, Z = v/b is a concurrent vector field.
The converse is trivial.
The next result determines concircular vector fields on unit spheres.
Proposition 1.10. Let {u1 , . . . , un } be an isothermal coordinate system on

S n (1) so that the metric tensor of S n (1) is given by
Xn
4
g= P n du2 . (1.36)
(1 + j=1 u2j )2 i=1 i
Then a vector field v on S n (1) is concircular if and only if, up to translations
of u1 , . . . , un , v is a gradient vector field given by v = −∇µ, where
Pn
1 − i=1 u2i
µ= Pn . (1.37)
2 + 2 j=1 u2j
Moreover, µ is exactly the function satisfying (1.30).
Proof. Let {u1 , . . . , un } be the isothermal coordinates so that the metric

tensor of S n (1) is given by (1.36). If v is a concircular vector field of S n (1),
then (1.35) holds.
Since S n (1) is of constant curvature one, we also have
R(X, v)v = g(v, v)X − g(X, v)v (1.38)
for X perpendicular to v. Thus we find from (1.35) and (1.38) that
Xµ = −g(X, v), vµ = −g(v, v). (1.39)
Hence v = ∇f is a gradient field with f = −µ.
On the other hand, it follows from a direct computation that the Levi-
Civita connection of (1.36) satisfies
∂ −2 n ∂ X ∂ o
∇∂/∂ui = Pn u i − u j ,
∂ui 1 + i=1 u2i ∂ui ∂uj
j6=i
(1.40)
∂ −2 ∂ ∂
∇∂/∂ui = Pn ui + uj ,
∂uj 1 + i=1 u2i ∂uj ∂ui
for 1 ≤ i 6= j ≤ n. From f = −µ, (1.26) and (1.40), we find

( )
∂ ∂ ∂ 2
µ 2 ∂µ X ∂µ
Hf , =− 2 − Pn ui − uj ,
∂ui ∂ui ∂ ui 1 + i=1 u2i ∂ui ∂uj
j6=i
( ) (1.41)
2
∂ ∂ ∂ µ 2 ∂µ ∂µ
Hf , =− − Pn uj + ui ,
∂ui ∂uj ∂ui ∂uj 1 + i=1 u2i ∂ui ∂uj
for 1 ≤ i 6= j ≤ n. After combining (1.41) and (1.33) of Lemma 1.6, we get
( )
∂2µ −2 ∂µ ∂µ
= Pn uj + ui , (1.42)
∂ui ∂uj 1 + i=1 u2i ∂ui ∂uj
( )
∂2µ −4µ 2 ∂µ X ∂µ
= P − P ui − uj , (1.43)
∂ 2 ui (1 + ni=1 u2i )2 1 + ni=1 u2i ∂ui ∂uj
j6=i
for 1 ≤ i 6= j ≤ n.
After solving system (1.42)-(1.43) via long computation and by applying
suitable translations on ui , we obtain
Pn
1 − i=1 u2i
µ= Pn . (1.44)
2 + 2 j=1 u2j
Thus Lemma 1.6 implies that the concircular vector field v is given by
n
1X
v= uj ∂uj . (1.45)
2 j=1
Conversely, by applying (1.36), (1.40) and a direct computation, we
obtain ∇V v = µV for V tangent to S n (1).
Similarly, we have the following result which determines concircular
fields on the hyperbolic n-space H n (−1) with constant curvature −1.
Proposition 1.11. Let {v1 , . . . , vn } be an isothermal coordinates on

H n (−1) so that the metric tensor of H n (−1) is
Xn
4
g= Pn dv 2 . (1.46)
(1 − j=1 vj2 )2 i=1 i
Then a vector field v on H n (−1) is concircular if and only if, up to trans-
lations of v1 , . . . , vn , v is a gradient vector field given by v = −∇µ, where
Pn
1 + i=1 vi2
µ= Pn . (1.47)
2 − 2 j=1 vj2
Moreover, µ is exactly the function satisfying (1.30).
Proof. This can be done in a similar way as Proposition 1.10.

Chapter 2
Submanifolds
Differential geometry started during the seventeenth century as the study

of curves in the Euclidean plane and of curves and surfaces in the Euclidean
3-space E3 by means of the techniques of differential calculus.
With his 1827 fundamental work
“Disquisitiones generales circa superficies curvas”,
Gauss initiated modern differential geometry [Gauss (1827)]. Of greatest
importance in this work was that Gauss demonstrated the existence of an
intrinsic geometry of surfaces based on the measurements of the lengths of
arcs in these surfaces, besides extrinsic geometry of surfaces.
In his famous inaugural lecture at Göttingen
“Über die Hypothesen welche der Geometrie zu Grunde liegen”,
Riemann discussed the foundations of geometry, introduced n-dimensional
manifolds, formulated the concept of Riemannian manifolds and defined
their curvature [Riemann (1854)].
Under the impetus of Einstein’s Theory of General Relativity (1915) a
further generalization appeared; the positiveness of the inner product was
weakened to non-degeneracy. Consequently, one has the notion of pseudo-
Inspired by Kaluza-Klein’s work in general relativity and string theory
in particle physics, mathematicians and physicists study not only submani-
folds of Riemannian manifolds but also submanifolds of pseudo-Riemannian
manifolds in recent years.
In recent times, submanifold theory also plays an important part in
computer design, image processing, economic modeling, arts and vision as
well as in mathematical physics and in mathematical biology.
23
2.1 Embedding theorems
Let φ : M → N be a map between two manifolds. Then φ is called an

immersion if φ∗x : Tx M → Tφ(x) N is injective for all x ∈ M . If, in addition,
φ is a homeomorphism onto φ(M ), where φ(M ) has the subspace topology
induced from N , we say that φ is an embedding.
For most local questions of geometry, it is the same to work with either
immersions or embeddings. In fact, if φ : M → N is an immersion of a
manifold M into another manifold N , then for each point p ∈ M , there
exists a neighborhood U ⊂ M of x such that the restriction φ : U → N is
an embedding.
For an immersion φ : M → N , we have dim M ≤ dim N . The difference
dim N − dim M is called the codimension of the immersion. Throughout
this book we consider only immersions of codimension ≥ 1 unless mentioned
otherwise.
Definition 2.1. An immersion φ : M → N of a pseudo-Riemannian man-

ifold into another pseudo-Riemannian manifold is called isometric if
hu, vix = hφ∗x u, φ∗x viφ(x) (2.1)
holds for all u, v ∈ Tx M, x ∈ M .
Let φ : M → N be an isometric immersion. Then, for each point x ∈ M ,

there exists a neighborhood U of x such that φ : U → N is an embedding.
Thus, each vector u ∈ Tx U gives rise to a vector φ∗ u ∈ Tφ(x) N . We may
identify u ∈ Tx M with φ∗ u ∈ Tφ(x) N . In this way, each tangent space
Tx M is a non-degenerate subspace of Tφ(x) N . Hence there is a direct sum
decomposition
Tφ(x) N = Tx M ⊕ Tx⊥ M,
where Tx⊥ M is a non-degenerate subspace of Tφ(x) N , which is called the
normal subspace of M at x. Vectors in Tx⊥ M are said to be normal to M
and those in Tx M are tangent to M . Thus, each vector v ∈ Tφ(x) N has a
unit expression
v = tan v + nor v, (2.2)
where tan v ∈ Tx M and nor v ∈ Tx⊥ M . The orthogonal projections
tan : Tφ(x) N → Tx M ; and nor : Tφ(x) N → Tx⊥ M
are R-linear.
Submanifolds 25
One of the most fundamental problems in submanifold theory is the

problem of isometric immersibility. The embedding problem had been
around since Riemann. The earliest publication by L. Schläfli on isometric
embedding appeared in [Schläfli (1873)].
The problem of isometric immersion (or embedding) admits an obvious
analytic interpretation, namely, if gij (u), u = (u1 , . . . , un ), are the compo-
nents of the metric tensor g in local coordinates u1 , . . . , un on a Riemannian
n-manifold M , and x1 , . . . , xm are the standard Euclidean coordinates of
Em , then the condition for an isometric immersion in Em is
Xn
∂xj ∂xk
= gjk (u),
i=1
∂ui ∂ui
i.e., we have a system of 12 n(n + 1) nonlinear partial differential equations

in m unknown functions. If m = 21 n(n + 1), then this system is defi-
nite and so we would like to have a solution. Schläfli asserted that any
Riemannian n-manifold can be isometrically embedded in Euclidean space
of dimension 21 n(n + 1). Apparently, it is appropriate to assume that he
had in mind of analytic metrics and local analytic embeddings. This was
later known as Schläfli’s conjecture. M. Janet published in [Janet (1926)] a
proof of Schläfli’s conjecture which states that a real analytic Riemannian
n-manifold can be locally isometrically embedded into any real analytic
Riemannian manifold of dimension 12 n(n + 1). É. Cartan revised Janet’s
paper with the same title in [Cartan (1927)]; while Janet wrote the problem
in the form of a system of partial differential equations, Cartan applied his
own theory of Pfaffian systems in involution. Both Janet’s and Cartan’s
proofs contained obscurities. C. Burstin got rid of them in [Burstin (1931)].
This result of Cartan-Janet implies that every Einstein n-manifold (n ≥ 3)
can be locally isometrically embedded in En(n+1)/2 .
The Cartan-Janet theorem is dimensionwise the best possible, i.e., there
exist real analytic Riemannian n-manifolds which do not possess smooth
local isometric embeddings into any Euclidean space of dimension strictly
less than 21 n(n + 1). Not every Riemannian n-manifold can be isometrically
immersed in Em with m ≤ 12 n(n + 1). For instance, not every Riemannian
2-manifold can be isometrically immersed in E3 .
A global isometric embedding theorem was proved by J. F. Nash which
states as follows.
Theorem 2.1. [Nash (1956)] Every closed Riemannian n-manifold can be

isometrically embedded in a Euclidean m-space Em with m = 12 n(3n + 11).
Every non-closed Riemannian n-manifold can be isometrically embedded in

Em with m = 12 n(n + 1)(3n + 11).
R. E. Greene improved Nash’s result in [Greene (1970)] and proved
that every non-compact Riemannian n-manifold can be isometrically em-
bedded in the Euclidean m-space Em with m = 2(2n + 1)(3n + 7). Also,
it was proved independently in [Greene (1970)] and [Gromov and Rokhlin
(1970)] that a local isometric embedding from a Riemannian n-manifold
1
into E 2 n(n+1)+n always exist.
Concerning the isometric embedding of pseudo-Riemannian manifolds,
we have the following existence theorem.
Theorem 2.2. [Clarke (1970); Greene (1970)] Any pseudo-Riemannian
n-manifold Mtn with index t can be isometrically embedded in a pseudo-
Euclidean m-space Ems , for m and s large enough. Moreover, this embedding
may be taken inside any given open set in Em s .
2.2 Formulas of Gauss and Weingarten
Let φ : M → N be an isometric immersion. The Levi-Civita connection

of the ambient manifold N will be denoted by ∇. ˜ Since the discussion is
local, we may assume, if we want, that M is embedded in N .
Let X be a vector field tangent to M . A vector field X̃ on N is called
an extension of X if its restriction to φ(M ) is X. If X̃ and Ỹ are extensions
of vector fields X and Y on M , respectively, then [X̃, Ỹ ] |M is independent
of the extensions. Moreover, we have
[X̃, Ỹ ] |M = [X, Y ]. (2.3)
If X and Y are local vector fields of M and X̃ and Ỹ are local extensions
of X and Y to N . Then the restriction of ∇ ˜ Ỹ on M is independent of the
X̃
extensions X̃, Ỹ of X, Y . This can be seen as follows: Let X̃1 be another
extension of X, then we have ∇ ˜
X̃−X̃1 Ỹ = 0 on M since X̃ − X̃1 = 0 on M .
˜ ˜
Thus ∇X̃ Ỹ = ∇X̃1 Ỹ . Moreover, it follows from (2.3) and Theorem 1.1(4)
˜ Ỹ on M is also independent of the extension Ỹ .
that the restriction of ∇ X̃
Let ∇X Y and σ(X, Y ) denote the tangential and normal components
˜ Ỹ , i.e.,
of ∇ X̃
∇X Y = tan (∇ ˜ Ỹ ) and σ(X, Y ) = nor (∇ ˜ Ỹ ), (2.4)
X̃ X̃
so that we have the following formula of Gauss
∇˜ Ỹ = ∇X Y + σ(X, Y ). (2.5)
X̃
Submanifolds 27
Proposition 2.1. Let φ : M → N be an isometric immersion of a pseudo-

Riemannian manifold M into a pseudo-Riemannian manifold N . Then
(i) ∇ defined in (2.4) is the Levi-Civita connection of M and
(ii) σ(X, Y ) is F (M )-bilinear and symmetric.
Proof. To prove statement (i), we verify properties (1)-(3) of Definition
1.3 and properties (4) and (5) of Theorem 1.1.
Properties (1) and (2) of Definition 1.3 follow from the corresponding
˜ on N and linearity of the projection tan : Tφ(x) N → Tx M .
properties of ∇
To verify property (3), let f be a function in F (M ). Then
∇X (f Y ) = (Xf )Y + f ∇X Y.
After taking the tangential components of both sides, we get property (3)
of Definition 1.3. Next, we prove property (4) of Theorem 1.1. Let us write
∇˜ Ỹ = ∇X Y + σ(X, Y ), (2.6)
X̃
˜ X̃ = ∇Y X + σ(Y, X).
∇ (2.7)
Ỹ
˜
Since ∇ is the Levi-Civita connection of N , it follows form (2.3), (2.6) and
(2.7) that
[X, Y ] = ∇X Y − ∇Y X, (2.8)
σ(X, Y ) = σ(Y, X), (2.9)
which imply property (4) and that σ(X, Y ) is symmetric.
To prove property (5) we start with
Xh Ỹ , Z̃ i = h ∇ ˜ X Z̃ i .
˜ X Ỹ , Z̃ i + h Ỹ , ∇ (2.10)
From (2.5) we have
h∇˜ X Ỹ , Z̃ i = h∇X Y, Zi + hσ(X, Y ), Zi = h∇X Y, Zi . (2.11)
Similarly, we have
˜ X Z̃ i = hY, ∇X Zi .
h Ỹ , ∇ (2.12)
Hence we obtain from (2.10)-(2.12) that
X hY, Zi = h∇X Y, Zi + hY, ∇X Zi ,
which is property (5).
Finally, we show that σ(X, Y ) is F (M )-bilinear. The additivity in X
or Y is obvious. Now, for any f ∈ F (M ), we have
˜ f X Ỹ = f ∇
∇f X Y + σ(f X, Y ) = ∇ ˜ X Ỹ
= f (∇X Y + σ(X, Y )),
which implies that σ(f X, Y ) = f σ(X, Y ). By symmetry, we also obtain
σ(X, f Y ) = f σ(X, Y ).
We define σ : T M × T M → T ⊥ M as the second fundamental form of N

for the given immersion. Let e1 , . . . , en and en+1 , . . . , em be orthonormal
bases of the tangent space Tx M and of the normal space Tx⊥ M at a point
x ∈ M.
If we put
r
σij = hσ(ei , ej ), er i ; i, j = 1, . . . , n; r = n + 1, . . . , m,
then we have
n
X
r
σ(ei , ej ) = ǫr σij er , ǫr = her , er i .
r=n+1
r
We call σij
the coefficients of the second fundamental form.
Let φ : M → N be an isometric immersion. If ξ is a normal vector field
of M in N and X is a tangent vector field of M , then we may decompose
˜ X ξ as
∇
˜ X ξ = −Aξ (X) + DX ξ,
∇ (2.13)
where −Aξ (X) and DX ξ are the tangential and normal components of
˜ X ξ. It is easy to verify that Aξ (X) and DX ξ are smooth vector fields on
∇
N whenever X and ξ are smooth.
Remark 2.1. Formula (2.13) is well-known as the formula of Weingarten.

For surfaces in E3 , it was established by J. Weingarten in [Weingarten
(1861)]. The operator A in (2.13) is known as Weingarten map or as shape
operator.

Riemannian manifold M into a pseudo-Riemannian manifold N . Then
(a) Aξ (X) is F (M )-bilinear in ξ and X; thus, at each point x ∈ M , Aξ (X)
depends only on ξx and Xx ;
(b) For a normal vector field ξ and tangent vectors X, Y of M , we have
hσ(X, Y ), ξi = hAξ (X), Y i ; (2.14)
(c) D is a metric connection on the normal bundle T ⊥ M with respect to the
induced metric on T ⊥ M , i.e., DX hξ, ηi = hDX ξ, ηi + hξ, DX ηi holds
for any tangent vector field X and normal vector fields ξ, η.
Proof. (a) Let f, k be two functions in F (M ). We have

˜ f X (kξ) = f ∇
∇ ˜ X (kξ) = f {(Xk)ξ + k ∇
˜ X ξ}
= f (Xk)ξ − f kAξ (X) + f kDX ξ,
Submanifolds 29
which implies that

Akξ (f X) = f kAξ (X), (2.15)
Df X (kξ) = f (Xk)ξ + f kDX ξ. (2.16)
Thus Ax (X) is F (M )-bilinear in ξ and X, since additivity is trivial.
(b) For arbitrary X, Y ∈ X(M ), we have
˜ X Y, ξ i + h Y, ∇
0 = h∇ ˜ Xξ i
= h∇X Y, ξi + hσ(X, Y ), ξi − hY, Aξ (X)i + hY, DX ξi
= hσ(X, Y ), ξi − hY, Aξ (X)i
which gives (2.14).
(c) It follows from (2.16) that D defines an affine connection on T ⊥ M .
Moreover, for any normal vector fields ξ and η, we have
˜ X ξ = −Aξ (X) + DX ξ and ∇
∇ ˜ X η = −Aη (X) + DX η.
Hence
˜ X ξ, η i + h ξ, ∇
h DX ξ, η i + h ξ, DX η i = h ∇ ˜ X η i = Xhξ, ηi.
Thus D is a metric connection on the normal bundle with respect to the

induced metric on T ⊥ M .
Definition 2.2. An isometric immersion φ : M → N is called totally

geodesic if the second fundamental form vanishes identically, i.e., σ ≡ 0.
For a normal vector field ξ on M , if Aξ = ρI for some ρ ∈ F (M ), then
ξ is called an umbilical section, or M is said to be umbilical with respect
to ξ. If the submanifold M is umbilical with respect to every local normal
vector field, then M is called a totally umbilical submanifold .
The mean curvature vector H of M in N is defined by

1
H= Tr σ, (2.17)
n
where Tr stands for trace and n = dim M . If {e1 , . . . , en } is an orthonormal
frame of M , then the mean curvature vector is given by
X n
1
H= ǫj σ(ej , ej ). (2.18)
n j=1
The length of the mean curvature vector is called the mean curvature.
Notation 2.1. Let M be a submanifold of a Riemannian manifold N . We

denote the squared mean curvature by H 2 , i.e., H 2 = hH, Hi .
Definition 2.3. The metric connection D defined by (2.13) is called the

normal connection. A normal vector field ξ on M is said to be parallel
in the normal bundle, or simply parallel if Dξ = 0 holds identically. In
particular, M is said to have parallel mean curvature vector if DH = 0
holds identically.
The second fundamental form of a totally umbilical submanifold satisfies

σ(X, Y ) = hX, Y i H, X, Y ∈ T M. (2.19)
Definition 2.4. A pseudo-Riemannian submanifold M is called a minimal

submanifold if the mean curvature vector H vanishes identically, i.e., H ≡ 0.
A pseudo-Riemannian submanifold M is called quasi-minimal if we have
H 6= 0 and hH, Hi = 0 at each point of M [Rosca (1972)]. Moreover, a
spacelike submanifold in a spacetime is called marginally trapped if it is
quasi-umbilical.
Remark 2.2. The concept of trapped surfaces was introduced in [Penrose

(1965)], see section 3.8 of [Chen (2011b)] for more details.
2.3 Equations of Gauss, Codazzi and Ricci
Let φ : M → N be an isometric immersion of a pseudo-Riemannian mani-

fold M into a pseudo-Riemannian manifold N .
Denote by R and ∇ the Riemann curvature tensor and the Levi-Civita
˜ the corresponding notions
connection of M , respectively; and by R̃ and ∇
of N . Then we have
˜ X∇
R̃(X, Y )Z = ∇ ˜Y Z − ∇
˜Y ∇
˜ XZ − ∇
˜ [X,Y ] Z
for X, Y, Z ∈ X(M ). By applying Gauss’ formula, we find

˜ X (∇Y Z + σ(Y, Z)) − ∇
R̃(X, Y )Z = ∇ ˜ Y (∇X Z + σ(X, Z))
− ∇[X,Y ] Z − σ([X, Y ], Z)
= R(X, Y )Z + σ(X, ∇Y Z) − σ(Y, ∇X Z)
− σ([X, Y ], Z) + ∇X σ(Y, Z) − ∇Y σ(X, Z).
On the other hand, using Weingarten’s formula we find
R̃(X, Y )Z = R(X, Y )Z − Aσ(Y,Z) X + Aσ(X,Z) Y + σ(X, ∇Y Z)
(2.20)
− σ(Y, ∇X Z) − σ([X, Y ], Z) + DX σ(Y, Z) − DY σ(X, Z).
Hence we obtain
Submanifolds 31
Theorem 2.3. Let φ : M → N be an isometric immersion of a pseudo-

Riemannian manifold M into a pseudo-Riemannian manifold N . Then for
vector fields X, Y, Z, W tangent to M , we have
R(X, Y ; Z, W ) = R̃(X, Y ; Z, W ) + hσ(X, W ), σ(Y, Z)i (2.21)
− hσ(X, Z), σ(Y, W )i ,
⊥ ¯ X σ)(Y, Z) − (∇
(R̃(X, Y )Z) = (∇ ¯ Y σ)(X, Z), (2.22)
where R(X, Y ; Z, W ) = hR(X, Y )Z, W i, (R̃(X, Y )Z)⊥ the normal compo-
¯ the covariant derivative of σ with respect to the
nent of R̃(X, Y )Z, and ∇σ
van der Waerden-Bortolotti connection ∇ ¯ = ∇ ⊕ D, i.e.,
¯ X σ)(Y, Z) = DX σ(Y, Z) − σ(∇X Y, Z) − σ(Y, ∇X Z).
(∇ (2.23)
Similarly, the j-th covariant derivative of σ is defined inductively by
(∇ ¯ j−1 σ)(X2 , . . . , Xj+2 ))
¯ j σ)(X1 , . . . , Xj+2 ) = DX1 ((∇
j+2
X
− ¯ j−1 σ)(X2 , . . . , ∇X1 Xi , . . . , Xj+2 ).
(∇
i=2
If ξ and η are normal vector fields of M , we have
R̃(X, Y ; ξ, η) = h ∇ ˜ Y ξ, η i − h ∇
˜ X∇ ˜ X ξ, η i − h ∇
˜Y ∇ ˜ [X,Y ] ξ, η i
˜ Y (Aξ X), η i − h ∇
= h∇ ˜ X (Aξ Y ), η i + h ∇
˜ X DY ξ, η i
− h∇˜ Y DX ξ, η i − h D[X,Y ] ξ, η i
= h σ(Y, Aξ X), η i − h σ(X, Aξ Y ), η i + h DX DY ξ, η i
− h DY DX ξ, η i − h D[X,Y ] ξ, η i .
D
If R denotes the curvature tensor of the normal bundle T ⊥ M , i.e.,
RD (X, Y )ξ = DX DY ξ − DY DX ξ − D[X,Y ] ξ, (2.24)
then
RD (X, Y ; ξ, η) = R̃(X, Y ; ξ, η) + h[Aξ , Aη ](X), Y i , (2.25)
where [Aξ , Aη ] = Aξ Aη − Aη Aξ .
Equations (2.21), (2.22) and (2.25) are called the equations of Gauss,
Codazzi, and Ricci, respectively; known as the fundamental equations.
Proposition 2.3. Let Msn be a pseudo-Riemannian n-manifold with index
s isometrically immersed in an indefinite real space form Rsm (c) of constant
curvature c. Then the Ricci tensor of Msn satisfies
X
Ric (Y, Z) = (n − 1) hY, Zi c − ǫi hσ(Y, ei ), σ(Z, ei )i + n hH, σ(Y, Z)i ,
i
where {e1 , . . . , en } is an orthonormal frame of Msn .
Proof. The equation of Gauss yields

n
X X
Ric (Y, Z) = ǫi R̃(ei , Y ; Z, ei ) − ǫi hσ(Y, ei ), σ(Z, ei )i
i=1 i
+ n hH, σ(Y, Z)i .

Combining this with (1.17) gives the proposition.
An immediate consequence of Proposition 2.3 is the following.
Corollary 2.1. If Msn is a minimal submanifold of the pseudo-Euclidean
space Ems , then Ric ≤ 0, with the equality holding identically if and only if
Msn is totally geodesic.
Remark 2.3. Corollary 2.1 provides a geometric obstruction for a Rieman-
nian manifold to admit a minimal immersion in a Euclidean space in terms
of the Ricci curvature.
Proposition 2.4. Let Mtn be an n-dimensional pseudo-Riemannian sub-
manifold of an indefinite real space form Rsm (c). Then the scalar curvature,
the mean curvature vector, and the second fundamental form of Mtn satisfy
n2 1 n(n − 1)
τ= hH, Hi − Sσ + c, (2.26)
2 2 2
where Sh is defined as
n
X
Sσ = ǫi ǫj hσ(ei , ej ), σ(ei , ej )i ,
i,j=1
with ǫi = hei , ei i and e1 , . . . , en being an orthonormal frame of Mtn .
Proof. Let e1 , . . . , en be an orthonormal frame of Mtn . Then the equation
of Gauss gives
Xn n
X
ǫi ǫj hR(ei , ej )ej , ei i = ǫi ǫj h R̃(ei , ej )ej , ei i
i,j=1 i,j=1
n n (2.27)
X X
+ hǫi σ(ei , ei ), ǫj σ(ej , ej )i − ǫi ǫj hσ(ei , ej ), σ(ei , ej )i .
i,j=1 i,j=1
Since the sectional curvature K of M satisfies
K(ei ∧ ej ) = ǫi ǫj hR(ei , ej )ej , ei i ,
we find from (2.27) that
Xn
2τ = K(ei ∧ ej ) = n(n − 1)c + n2 hH, Hi − Sσ ,
i,j=1
which gives (2.26).
Submanifolds 33
An immediate consequence of Proposition 2.4 is the following.
Corollary 2.2. If Msn is an n-dimensional minimal submanifold of an

indefinite real space form Rsn+r (c), then we have
2τ ≤ n(n − 1)c. (2.28)
n+r
Similarly, if Msn is a minimal submanifold in Rs+r (c), then we have
2τ ≥ n(n − 1)c. (2.29)
Either equality holds identically if and only if Msn is totally geodesic.
Another application of Proposition 2.4 is the following.
Proposition 2.5. Let M be a submanifold of a real space form Rn+r (c) of

constant curvature c. Then the scalar curvature of M satisfies
n(n − 1) 2
τ≤ (H + c), n = dim M, (2.30)
2
with the equality holding at a point x ∈ M if and only if x is a totally
umbilical point.
Proof. Choose an orthonormal basis e1 , . . . , en , en+1 , . . . , en+r at x such

that en+1 is parallel to the mean curvature vector and e1 , . . . , en diagonalize
the shape operator An+1 = Aen+1 .
It follows from Proposition 2.4 that
X n Xr Xn
n2 H 2 = 2τ + a2i + α 2
(σij ) − n(n − 1)c, (2.31)
i=1 α=n+2 i,j=1
where a1 , . . . , an are eigenvalues of An+1 .
On the other hand, it follows from the Cauchy-Schwarz inequality that
Xn
a2i ≥ nH 2 , (2.32)
i=1
with the equality holding if and only if a1 = a2 = · · · = an . By combining
(2.31) and (2.32) we obtain
Xr X n
2 α 2
n(n − 1)H ≥ 2τ − n(n − 1)c + (σij ) , (2.33)
α=n+2 i,j=1
which implies inequality (2.30).

If the equality sign of (2.30) holds at a point x ∈ M , then it follows
from (2.32) and (2.33) that
An+2 = · · · = An+r = 0 a1 = · · · = an .
Therefore x is a totally umbilical point. The converse is trivial.
Proposition 2.5 has some nice applications, e.g., it implies the following.
Corollary 2.3. If the scalar curvature of an n-dimensional Riemannian

submanifold M of Em satisfies τ ≥ 21 n(n − 1) at a point x ∈ M , then every
isometric immersion of M into any Euclidean space satisfies H 2 ≥ 1 at x
regardless of codimension.
In particular, if τ = 21 n(n − 1) on M , then H 2 ≥ 1 holds identically on
M , with the equality holding identically if and only if M is an open part of
a standard unit hypersphere in a totally geodesic En+1 ⊂ Em .
Proof. Inequality H 2 ≥ 1 follows immediately from Proposition 2.5. If

the scalar curvature of M is 21 n(n − 1) at each point and if |H| = 1 holds
identically on M , then M is totally umbilical. Thus M is an open portion
of an ordinary unit n-sphere.
Similar to Proposition 2.5, we also have the following.
Corollary 2.4. Let M be a spacelike submanifold of an indefinite real space

form Rrn+r (c) of constant curvature c. Then
2τ
|H|2 ≤ − c, n = dim M, (2.34)
n(n − 1)
with equality holding at a point x ∈ M if and only if x is a totally umbilical
point.
2.4 Existence and uniqueness theorems of submanifolds
Now, we can state the fundamental theorems of submanifolds as follows.

For the proofs see, for instance, [Eschenburg and Tribuzy (1993); Wettstein
(1978)].
Theorem 2.4. (Existence) Let (Mtn , g) be a simply-connected pseudo-

Riemannian n-manifold with index t. Suppose that there exists an (m − n)-
dimensional pseudo-Riemannian vector bundle ν(Mtn ) with index s − t over
Mtn and with curvature tensor RD and also exists a ν(Mtn )-valued symmet-
ric (0, 2) tensor σ on Mtn . For a cross section ξ of ν(Mtn ), define Aξ by
g(Aξ X, Y ) = hσ(X, Y ), ξi, where h , i is the fiber metric of ν(Mtn ). If they
satisfy (2.21), (2.22) and (2.25), then Mtn can be isometrically immersed
in an m-dimensional indefinite real space form Rsm (c) of constant curva-
ture c in such way that ν(Mtn ) is the normal bundle and σ is the second
fundamental form.
Submanifolds 35
Theorem 2.5. (Uniqueness) Let φ, φ′ : Mtn → Rsm (c) be two isometric im-
mersions of a pseudo-Riemannian n-manifold Mtn into an indefinite space
for Rsm (c) of constant curvature c with normal bundles ν and ν ′ equipped
with their canonical bundle metrics, connections and second fundamental
forms, respectively. Suppose there is an isometry φ : Mtn → Mtn such that
φ can be covered by a bundle map φ̄ : ν → ν ′ which preserves the bundle
metrics, the connections and the second fundamental forms. Then there is
an isometry Φ of Rsm (c) such that Φ ◦ φ = φ′ .
Two submanifolds M1 and M2 of a pseudo-Riemannian manifold N are

said to be congruent if there exists an isometry of N which carries one to
the other.
2.5 Reduction theorems
Let Rni,j denote the affine n-space with the metric whose canonical form is
 
Oj
 −Ii ,
In−i−j
where Ik is the k × k identity matrix and Oj is the j × j zero matrix.
The metric is non-degenerate if and only if j = 0. The j in Rni,j measures
the degenerate part. The metric of Rni,1 = R0 × Ein−1 vanishes on the first
factor R0 and it is the standard pseudo-Euclidean metric with index i on
the second factor Ein−1 . Denote the natural embedding ι : Rni,1 → En+1 i+1 of
Rni,1 into En+1
i+1 given by
ι((x1 , x2 , . . . , xn )) = (x1 , x2 , . . . , xn , x1 ) ∈ En+1
i+1
for (x1 , . . . , xn ) ∈ Rni,1 . Then the light-like vector ζ0 = (1, 0, . . . , 0, 1) is a

normal vector of Rni,1 in En+1 i+1 .
Let φ : M → N be an isometric immersion of a pseudo-Riemannian
manifold into another pseudo-Riemannian manifold. At each point x ∈ M ,
the first normal space N 1 (x) is defined to be the orthogonal complement
of N 0 (x) = {ξ ∈ Tx⊥ M : Aξ = 0}.
Definition 2.5. Let φ : M → N be an isometric immersion of a pseudo-

Riemannian manifold into another pseudo-Riemannian manifold. The first
normal spaces are called parallel if, for any curve σ joining any two points
x, y ∈ M , the parallel displacement of normal vectors along σ with respect
to the normal connection maps N 1 (x) onto N 1 (y).
The following result is the reduction theorem of Erbacher-Magid.

Theorem 2.6. [Erbacher (1971); Magid (1984)] Let φ : Min → Em s be an
n
isometric immersion of a pseudo-Riemannian n-manifold Mi with index i
into Em
s . If the first normal spaces are parallel, then there exists a complete
(n+k)-dimensional totally geodesic submanifold E ∗ such that ψ(Min ) ⊂ E ∗ ,
where k is the dimension of the first normal spaces.
Proof. Under the hypothesis, the dimension of N 1 is a constant, say k.
If ξ is a normal vector field such that ξ ∈ N 1 (x) for each x ∈ Min , then
DX ξ ∈ N 1 (x) for all X ∈ Tx Min . Thus the first normal spaces N 1 (x) form
a parallel normal subbundle. Since D is a metric connection, the subspaces
N 0 (x) are also parallel with respect to the normal connection.
Let x0 be a point of Min . Consider the (n + k)-dimensional subspace E
of Em 0
s through φ(x0 ) which is perpendicular to N (x0 ), i.e.,
n 1
E = Tx0 (Mi ) ⊕ N (x0 ).
Then the degenerate part of E is N 0 (x0 ) ∩ N 1 (x0 ). Now, we claim that
φ(Min ) ⊂ E. This can be proved as follows: Let β(t) be any curve in Min
starting at x0 . For any ξ0 ∈ N 0 (x0 ), let ξt be the parallel displacement of
ξ0 along β(t), so that ξt ∈ N 0 (β(t)). For the pseudo-Euclidean connection
∇, we have ∇β ′ (t) ξt = −dφ(Aξt (β ′ (t))) + Dβ ′ (t) ξt = 0, which means that ξt
is parallel in Em
s . Thus it is a constant vector. Now, we have
d
hφ(β(t)) − φ(x0 ), ξ0 i = hdφ(β ′ (t)), ξ0 i = hdφ(β ′ (t)), ξt i = 0.
dt
Thus φ(β(t)) lies in E. Since this is true for arbitrary curve β(t) in Min ,
φ(Min ) ⊂ E.
In Theorem 2.6, E ∗ = Rn+k
s,t for some s, t and t need not be zero.
Definition 2.6. A pseudo-Riemannian submanifold M of a pseudo-

¯ = 0 identically.
Riemannian manifold is called parallel if ∇σ
Corollary 2.5. Let φ : M → Em s be an isometric immersion of a pseudo-
Riemannian n-manifold M into pseudo-Euclidean m-space Em s . If φ is a
parallel immersion, then there exists a complete (n + k)-dimensional to-
tally geodesic submanifold E ∗ ⊂ Em ∗
s such that φ(M ) ⊂ E , where k is the
dimension of the first normal spaces.
Proof. ¯ = 0. Thus it follows from (2.23) that
If φ is parallel, then ∇σ
DX σ(Y, Z) = σ(∇′X Y, Z) + σ(Y, ∇′X Z)
for X, Y, Z ∈ X(M ). Hence the first normal spaces are parallel. Therefore,
the corollary follows from Theorem 2.6.
Submanifolds 37
2.6 Totally geodesic submanifolds
The simplest submanifolds are totally geodesic submanifolds.

Riemannian manifold M into a pseudo-Riemannian manifold N . Then the
following three statements are equivalent.
(1) M is a totally geodesic submanifold of N ;

(2) Geodesics of M are geodesics of N ;
(3) For any x ∈ M and any v ∈ Tx M , the geodesic γv with γv (0) = x and
γv′ (0) = v lies locally in M .
Proof. Since M is totally geodesic in N , the second fundamental form

vanishes. Thus if γ is a curve of M , then ∇γ ′ γ ′ = ∇ ˜ γ ′ γ ′ , where ∇ and
˜
∇ are the Levi-Civita connections of M and N , respectively. Hence γ is a
geodesic of N if and only if it is geodesic of M . This proves the equivalence
of (1) and (2).
Now, suppose that γ : I → M is a geodesic of M with initial velocity
vector v, then γ is also a geodesic of N . Hence, by the uniqueness of
geodesics, we obtain (3). If (3) holds, Gauss’ formula implies σ(v, v) = 0.
Thus, for vectors v, w ∈ Tx M , we have
0 = σ(v + w, v + w) = σ(v, v) + 2σ(v, w) + σ(w, w) = 2σ(v, w).
Since this is true for any x ∈ N and any v, w ∈ Tx M , we find σ = 0 by
polarization. Thus M is totally geodesic in N .
Proposition 2.7. Up to rigid motions, an n-dimensional totally geodesic
pseudo-Riemannian submanifold of a pseudo-Euclidean space Em
s is an open
portion of a pseudo-Euclidean linear subspace Ent of Em
s .
Proof. Obviously, every pseudo-Euclidean linear subspace Ent of Em s is a

totally geodesic submanifold of Em s .
Now, assume that M is an n-dimensional totally geodesic pseudo-
Riemannian submanifold of Em m
s such that the origin o of Es lies in M .
m
Then To M is a non-degenerate subspace of To Es .
Because geodesics of Em
s are lines, it follows from Proposition 2.6 that M
is an open portion of the pseudo-Euclidean subspace whose tangent space
at o is To M .
Recall that the pseudo m-sphere Ssm (c) is defined by
n o
1
Ssm (c) = x = (x1 , . . . , xm+1 ) ∈ Em+1
s : hx, xi = > 0 .
c
For n < m and 0 ≤ t ≤ s,

n o
(x1 , . . . , xt , 0, . . . , 0, xs+1 , . . . , xn+1 , 0, . . . , 0) ∈ Ssm (c)
defines a pseudo-Riemannian manifold of constant curvature c with index t,

which is totally geodesic in Ssm (c). We call this totally geodesic submanifold
of Ssm (c) a pseudo n-sphere of Ssm (c). Analogously, we call
n o
(x1 , . . . , xt+1 , 0, . . . , 0, xs+1 , . . . , xn+1 , 0, . . . , 0) ∈ Hsm (c)
a pseudo-hyperbolic n-subspace of Hsm (c).

Similar to Proposition 2.7, we have the following.

pseudo-Riemannian submanifold of a pseudo m-sphere Ssm (c) is an open
portion of a pseudo n-sphere of Ssm (c).

pseudo-Riemannian submanifold of a pseudo-hyperbolic m-space Hsm (c) is
an open portion of a pseudo-hyperbolic n-subspace of Hsm (c).
2.7 Totally umbilical submanifolds
The following result is analogous to Proposition 2.6.
Proposition 2.10. [Ahn et al. (1996)] Let φ : Mtn → Nsm be an isometric

immersion of a pseudo-Riemannian n-manifold Mtn with index t ∈ [1, n −
1] into another pseudo-Riemannian manifold Nsm . Then Mtn is a totally
umbilical submanifold if and only if null geodesics of Mtn are geodesics of
Nsm .
Proof. Under the hypothesis, assume that Mtn is totally umbilical in Nsm .
If γ : I → Mtn is a null geodesic of Mtn , then γ ′ (t) is a null vector for each
t ∈ I. Then it follows from (2.19) that σ(γ ′ (t), γ ′ (t)) = 0. Thus ∇ ˜ γ ′ γ ′ = 0,
m
which shows that γ is also a geodesic of Ns .
Conversely, if null geodesics of Mtn are geodesics of Nsm , then σ(v, v) = 0
for null vectors v of Mtn . At a point x ∈ Mtn , let us choose an orthonormal
basis {e1 , . . . , en } of Tx Mtn such that hei , ei i = −1 for i = 1, . . . , t and
hej , ej i = 1 for j = t+1, . . . , n. Then ei ±ej are null vectors for i ∈ {1, . . . , t}
and j ∈ {t + 1, . . . , n}. Thus σ(ei ± ej , ei ± ej ) = 0, which implies that
σ(ei , ej ) = 0, σ(ei , ei ) + σ(ej , ej ) = 0. (2.35)
Submanifolds 39
√
If t ≥ 2, then ei1 + ei2 + 2en is a null vector for 1 ≤ i1 6= i2 ≤ s. Thus we
find σ(ei1 , ei2 ) = 0 by applying (2.35).
Similarly, if n − t ≥ 2, we have
σ(ej1 , ej2 ) = 0, t + 1 ≤ j1 6= j2 ≤ n.
Consequently, we obtain (2.19). Thus Mtn is totally umbilical in Nsm .
Lemma 2.1. Let φ : M → Rsm (c) be an isometric immersion of a pseudo-

Riemannian n-manifold M into an indefinite real space form Rsm (c). If M
is totally umbilical, then
(1) H is a parallel normal vector field, i.e., DH = 0;

(2) hH, Hi is constant;
(3) ¯ = 0 identically on M ;
φ is a parallel immersion, i.e., ∇σ
(4) N is of constant curvature c + hH, Hi;
(5) AH = hH, Hi I;
(6) M is a parallel submanifold.
Proof. Under the hypothesis, we have (2.19). Thus

¯ Z σ)(X, Y ) = hX, Y i DZ H.
(∇ (2.36)
Hence, by Codazzi’s equation, we obtain hX, Y i DZ H = hZ, Y i DX H for
X, Y, Z ∈ T M . Since dim M > 1, this gives DH = 0. Hence we get (1).
Statement (2) follows from (1) and the fact that D is a metric connection.
(3) follows from (1) and (2.36). And (4) is an easy consequence of (2) and
Gauss’ equation. Statement (5) follows from (2.14) and (2.19). Finally,
statement (6) follows from (2.19), (2.23) and statement (1).
Lemma 2.2. If φ : M → Em s is a totally umbilical immersion of a pseudo-

Riemannian n-manifold M into Em s , then M lies in an (n + 1)-dimensional
totally geodesic submanifolds of Em
s as a hypersurface.
Proof. Follows from Lemma 2.1 and Theorem 2.6.
Proposition 2.11. Let φ : M → Em s be an isometric immersion of a

pseudo-Riemannian n-manifold M with index t into Em s . If n > 1 and M
m
is totally umbilical in Es , then M is congruent to an open portion of one
of the following submanifolds:
(1) A totally geodesic pseudo-Euclidean subspace Ent ⊂ Ems ;

n
(2) A pseudo n-sphere St (c) lying in a totally geodesic pseudo-Euclidean
(n + 1)-subspace En+1
t ⊂ Em
s ;
(3) A pseudo-hyperbolic n-space Htn (c) lying in a totally geodesic pseudo-

Euclidean (n + 1)-subspace En+1 t+1 ⊂ Es ;
m
(4) A flat quasi-minimal submanifold defined by

t n t n
!
X X X X
2 2 2 2
xi − xj , x1 , . . . , xt , 0, . . . , 0, xt+1 , . . . , xn , xi − xj .
i=1 j=t+1 i=1 j=t+1
The last case occurs only when s > t.

Proof. Let M be a totally umbilical pseudo-Riemannian submanifold of
Ems . Assume that the index of M is t. Then Lemma 2.1 implies that hH, Hi
is a constant, DH = 0 and φ is a parallel immersion.
Case (a): H = 0. In this case M is totally geodesic. Thus we obtain (1)
by Proposition 2.6.
Case (b): H 6= 0. Corollary 2.1 implies that φ(M ) is contained in a (n + 1)-
dimensional totally geodesic submanifold E ∗ ⊂ Em s . From Lemma 2.1 we
˜
have ∇X H = − hH, Hi X for X ∈ T M .
Case (b.1): hH, Hi = ǫr2 , r > 0, ǫ = ±1. In this case, φ + (ǫ/r2 )H is a
constant vector, say φ0 . By applying a suitable translation we have φ0 = 0.
Thus hφ, φi = ǫ/r2 , which gives case (2) or case (3) according to H is
spacelike or timelike, respectively.
Case (b.2): H is lightlike. In this case, E ∗ is a totally geodesic Rn+1 1,t .
Since H is a constant lightlike vector and M is totally umbilical, (2.19)
and Gauss’ equation imply that M is flat. Thus locally there is a natural
coordinate system {x1 , . . . , xn } such that the metric g0 of M is given by
Xt Xn
g0 = − dx2i + dx2j . (2.37)
i=1 j=t+1
Hence it follows from Proposition 1.1, (2.5), (2.19) and (2.37) that
φxi xi = H, φxk ,xℓ = 0, φxj xj = −H,
for i = 1, . . . , t; j = t + 1, . . . , n; 1 ≤ k 6= ℓ ≤ n.
After solving this system we find
X n t n
HX 2 H X 2
φ = c0 + ck xk + xi − x
2 i=1 2 j=t+1 j
k=1
for some vectors c0 , c1 , . . . , cn ∈ Em
s . Since H is a constant lightlike vector,
without loss of generality we may put H = (2, 0, . . . , 0, 2) ∈ Em s . Hence,
after choosing suitable initial conditions, we obtain (4).
Remark 2.4. For the classification of totally umbilical submanifolds in Ssm
and in Hsm , see [Chen (2011b)].
Submanifolds 41
2.8 Pseudo-umbilical submanifolds
Let φ : M → Ssm (r−2 ) (resp. φ : M → Hsm (−r−2 )) be an isometric

immersion from a pseudo-Riemannian n-manifold M into Ssm (r−2 ) (resp.
Hsm (−r−2 )). Denote by

x = ι ◦ φ : M → Em+1
s , resp. x = ι ◦ φ : M → Em+1
s+1
the composition of φ with the inclusion map ι : Ssm (r−2 ) ⊂ Em+1

s via (1.19)
(resp. ι : Hsm (−r−2 ) ⊂ Em+1 via (1.20)). Let ∇, ∇˜ and ∇′ be the Levi-
s+1
Civita connections of M, Em+t
i and Ssm (r−2 ) (resp. Hsm (−r−2 )). Denote
by D the normal connection of M in Ssm (r−2 ) (resp. in Hsm (−r−2 )) and
′
by D the corresponding quantities of M in Em+1 s (resp. in Em+1

s+1 ).
Lemma 2.3. Let φ : M → Ssm (r−2 ) (resp. φ : M → Hsm (−r−2 )) be an

isometric immersion of a pseudo-Riemannian n-manifold M into Ssm (r−2 )
(resp. M into Hsm (−r−2 )). Then the second fundamental form σ and the
mean curvature vector H of M in Em+1 s (resp. M in Em+1
s+1 ) via x = ι ◦ φ
are related with the second fundamental form σ ′ and the mean curvature
vector H ′ of M in S m (r−2 ) (resp. M in Hsm (−r−2 )) by
ǫ
σ(X, Y ) = σ ′ (X, Y ) − 2 hX, Y i x, (2.38)
r
ǫ
H = H ′ − 2 x, (2.39)
r
where ǫ = 1 or −1, depending on φ is given by φ : M → Ssm (r−2 ) or by
φ : M → Hsm (−r−2 ). Moreover, we have DH = D′ H ′ .
Proof. Under the hypothesis, the positive vector field is a normal vector
˜ Xx = X
field of M which is normal to Ssm (r−2 ) or to Hsm (−r−2 ). Since ∇
for X ∈ T M , the Weingarten formula yields
Ax = −I, DH = D′ H ′ , (2.40)
where A is the Weingarten map in Em+1
s or in Em+1
s+1 . Hence it follows from
the formula of Gauss and (2.40) that
˜ X Y = ∇′ Y − ǫ x = ∇X Y + σ ′ (X, Y ) − ǫ x,
∇ (2.41)
X
r2 r2
which gives (2.38). By taking the trace of (2.38) we get (2.39).
Corollary 2.6. Let φ : M → Ssm (r−2 ) (resp. φ : M → Hsm (−r−2 )) be

an isometric immersion of a pseudo-Riemannian manifold M in Ssm (r−2 )
(resp. M → Hsm (−r−2 )) and ι : Ssm (r−2 ) ⊂ Em+1
s (resp. ι : Hsm (−r−2 ) ⊂
m+1
Es+1 ) is the inclusion map defined in Section 1.6. Then we have:
(1) φ has parallel mean curvature vector if and only if x = ι ◦ φ has parallel
mean curvature vector.
(2) φ is a parallel immersion if and only if x = ι◦φ is a parallel immersion.
(3) φ is totally umbilical if and only if x = ι ◦ φ is totally umbilical.
Proof. Follows immediately from Lemma 2.3 and (2.41).

Definition 2.7. A non-minimal pseudo-Riemannian submanifold M of a
pseudo-Riemannian manifold M is called a pseudo-umbilical submanifold if
there exists a function λ in F (M ) such that
hσ(X, Y ), Hi = λ hX, Y i , X, Y ∈ X(M ). (2.42)
The following three results classify pseudo-umbilical submanifolds with
parallel mean curvature vector in indefinite real space forms.
Proposition 2.12. Let φ : M → Em s be an isometric immersion of a
pseudo-Riemannian submanifold M into a pseudo-Euclidean m-space Em s .
Then φ is a pseudo-umbilical immersion with parallel mean curvature vector
if and only if one of the following three cases occurs:
(1) M is a minimal submanifold of a pseudo hypersphere Ssm−1 (x0 , r−2 )

for some x0 ∈ Em s and r > 0;
(2) M is a minimal submanifold of a pseudo-hyperbolic hyperspace
m−1
Hs−1 (x0 , −r−2 ) for some x0 ∈ Ems and r > 0;
(3) φ is congruent to (f, z, f ), where f ∈ F (M ), ∆f is a nonzero real
number and z : M → Em−2 s−1 is a minimal isometric immersion.
Case (2) occurs only when s ≥ 1 and case (3) occurs only when s ≥ 1 and
m ≥ dim M + 2.
Proof. Assume that M is a pseudo-umbilical submanifold with parallel
mean curvature vector in Em
s . Then
XhH, Hi = 2 hH, DX Hi = 0
for any X ∈ T M . Thus hH, Hi is constant.
Case (a): hH, Hi 6= 0. We put
ǫ
hH, Hi =
, (2.43)
r2
where ǫ = 1 or −1 depending on H is spacelike or timelike. On the other
hand, it follows from (2.42) that AH = λI for some function λ ∈ F (M ).
Thus we find from (2.43) that ǫ = λr2 . Let us put
φ̂ = φ + ǫr2 H. (2.44)
Submanifolds 43
˜ X φ̂ = X − ǫr2 AH X = 0. Thus φ̂ is a constant vector, say x0 . Hence

Then ∇
hφ − x0 , φ − x0 i = r4 hH, Hi = ǫr2 ,
which implies that M lies either in Ssm−1 (x0 , r−2 ) or in Hsm−1 (x0 , −r−2 ),
according to H is spacelike or timelike, respectively. Since H is either
normal to Srm−1 (x0 , r−2 ) or to Hsm−1 (x0 , −r−2 ), it follows from Lemma
2.3 that N is minimal in Srm−1 (x0 , r−2 ) or in Hsm−1 (x0 , −r−2 ).
Case (b): H is lightlike. It follows from (2.42) that AH = hH, Hi I = 0.
Combining this with DH = 0 implies that H is a lightlike constant
vector, say ζ0 ∈ Em+1 s . Thus Xhφ, ζ0 i = 0 for any X ∈ T M . So,
hφ, ζ0 i = c for some real number c. If we put ζ0 = (1, 0, . . . , 0, 1) ∈ Em
s
and φ = (x1 , . . . , xm ), then we obtain xm = x1 + c. By applying a suitable
translation, we find c = 0. Thus the immersion x : N → Em s takes the form
φ = (f, z, f ), (2.45)
where f is function on N and z : N → Em−2
is an isometric immersion.
s−1
Now, by applying the Laplace operator ∆ to (2.45) we find from Beltrami’s
formula that nH = (−∆f, nHz , −∆f ), where Hz is the mean curvature
vector of z. Since H = (1, 0, . . . , 0, 1), we find Hz = 0 and ∆f = −nr.
Thus z is a minimal immersion and ∆f is a nonzero constant.
The converse can be verified easily.
Proposition 2.13. Let φ : M → Ssm (1) be an isometric immersion of a

pseudo-Riemannian submanifold M into the pseudo m-sphere Ssm (1). Then
φ is a pseudo-umbilical immersion with parallel mean curvature vector if
and only if M lies in a non-totally geodesic, totally umbilical hypersurface
of Ssm (1) as a minimal submanifold.
Proof. Let φ : M → Ssm (1) be an isometric immersion of a pseudo-

Riemannian submanifold M into Ssm (1). Then it follows from Lemma 2.3
that M is a pseudo-umbilical submanifold with parallel mean curvature
vector in Ssm (1) if and only if M is a pseudo-umbilical submanifold of Em+1
s
via the composition x = ι ◦ φ, where ι is the inclusion Ssm (1) ⊂ Em+1
s .
Let M be a pseudo-umbilical submanifold with parallel mean curvature
vector in Ssm (1). Then the mean curvature vector H of M in Em+1s satisfies
AH = λI, DH = 0, λ ∈ R. (2.46)
Case (1): H is spacelike. If we put hH, Hi = r−2 , then as in the proof of
Proposition 2.12, we have
x − x0 = −r2 H (2.47)
for some vector x0 . Hence M is a minimal submanifold of the pseudo

hypersphere Ssm (x0 , r−2 ). Since M lies in Ssm (1) ∩ Ssm (x0 , r−2 ), we find
1
hx, xi = 1 and hx − x0 , x − x0 i = , (2.48)
r2
which implies
1 + hx0 , x0 i 1
hx, x0 i = c, c = − 2. (2.49)
2 2r
Since M is non-minimal in Ssm (1), x0 6= 0.
Case (1.1): hx0 , x0 i = k 2 > 0. If we put x0 = (0, . . . , 0, k −1 ) ∈ Em+1
s , then
the first equation in (2.49) implies that the last canonical coordinate xm+1
of M satisfies xm+1 = ck. Hence M is contained in H = Ssm (1) ∩ E, where
E is the hyperplane defined by xm+1 = ck. An easy computation shows
that ξ = c x − x0 is a normal vector field of H in Ssm (1). Moreover, it
follows from Gauss’ formula that Aξ = −cI. Hence H is a totally umbilical
hypersurface of Ssm (1).
Since M is non-minimal in Ssm (1), H must be a non-totally geodesic,
totally umbilical hypersurface. Also, since H lies in Span {x, ξ}, M is min-
imal in H. Therefore M lies in a non-totally geodesic, totally umbilical
hypersurface of Ssm (1) as a minimal submanifold.
Case (1.2): hx0 , x0 i = −k 2 < 0. Putting x0 = (k −1 , 0, . . . , 0) and using the
same arguments as case (1.1), we get the same conclusion as case (1.1).
Case (1.3): x0 is lightlike. We find from (2.49) that hx, x0 i = c ∈ (−∞, 21 ),
which defines a hyperplane E of Em+1s . Hence M lies in H1 = Ssm (1) ∩ E. It
is easy to verify that η = x0 −cx is a normal vector field of H1 in Ssm (1) such
that hη, ηi = −c2 . Since ∇ ˜ X η = −cX for X ∈ T M , we get Aη = cI. This
shows that H1 is totally umbilical in Ssm (1). Because M is non-minimal
in Ssm (1), we have c 6= 0. Moreover, it follows from (2.47) and Lemma 2.3
that the mean curvature vector H ′ of M in Ssm (1) is η. Hence we have the
same conclusion as case (1.1).
Case (2): H is timelike. This can be as case (1).
Case (3): H is lightlike. Just like in case (b) of the proof of Proposition
2.12, H is a constant lightlike vector in Em+1
s , say ζ0 , and that we have
hx, ζ0 i = c for some real number c. Let E denote the hyperplane defined by
hx, ζ0 i = c. Then M is contained in H1 = Ssm (1) ∩ E. It is easy to verify
that η = ζ0 − cx is a normal vector field of H1 in Ssm (1) with hη, ηi = −c2 .
Since ∇ ˜ X η = −cX, we get Aη = cI. Thus H1 is totally umbilical in S m (1).
s
Because M is non-minimal in Ssm (1), we find c 6= 0. Hence H1 is a
non-totally geodesic, totally umbilical hypersurface. Moreover, it follows
Submanifolds 45
from (2.47) and Lemma 2.3 that the mean curvature vector of M in Ssm (1)
is exactly η. Therefore M is minimal in H1 .
The converse is easy to verify.
Proposition 2.14. Let φ : M → Hsm (−1) be an isometric immersion

of a pseudo-Riemannian manifold M into the pseudo-hyperbolic m-space
Hsm (−1). Then φ is a pseudo-umbilical immersion with parallel mean cur-
vature vector if and only if M is contained in a non-totally geodesic, totally
umbilical hypersurface of Hsm (−1) as a minimal submanifold.
Proof. This can be done in the same way as Proposition 2.13.

By applying Proposition 2.13 we have the following.
Corollary 2.7. Let φ : M → Ssm (1) be an isometric immersion of a

pseudo-Riemannian manifold M into Ssm (1). Then φ is a pseudo-umbilical
immersion with unit timelike parallel mean curvature vector if and only
if M lies in a flat totally umbilical hypersurface of Ssm (1) as a minimal
submanifold.
Proof. We follow the same notations as in the proof of Proposition 2.13.

Assume φ : M → Ssm (1) is a pseudo-umbilical immersion with unit timelike
parallel mean curvature vector H ′ . Then the mean curvature vector H of
M in Em+1s is H ′ − x, where x = ι ◦ φ as before.
′
Since H is a unit timelike vector field, as in the proof of Proposition
2.12, we see that H is a lightlike constant vector, say ζ0 ∈ Em+1
s . Clearly,
we have hx, ζ0 i = −1. So M lies in H1 = Ssm (1) ∩ E, where E is defined by
hx, ζ0 i = −1.
Since η = ζ0 + x is a unit timelike normal vector field of H1 in Ssm (1)
and Aη = −I, the second fundamental form σ̂ of H1 in Ssm (1) satisfies
σ̂(X, Y ) = − hX, Y i η, X, Y ∈ T H1 . Hence the equation of Gauss implies
that H1 is a flat totally umbilical hypersurface. Thus by using the same
argument as given in the proof of Proposition 2.13 we conclude that M is
minimal in H1 . The converse is easy to verify.
Corollary 2.8. Let φ : M → Hsm (−1) be an isometric immersion of a

pseudo-Riemannian submanifold M into Hsm (−1). Then φ is a pseudo-
umbilical immersion with unit spacelike parallel mean curvature vector if
and only if M lies in a flat totally umbilical hypersurface of Hsm (−1) as a
minimal submanifold.
Proof. This can be done in the same way as Corollary 2.7.

2.9 Cartan’s structure equations
Let N be a pseudo-Riemannian n-submanifold of a pseudo-Riemannian m-

manifold M . Denote by ∇ the Levi-Civita connection of M . Choose a
local orthonormal frame e1 , . . . , en , en+1 , . . . , em of N such that e1 , . . . , en
are tangent to N and en+1 , . . . , em are normal to N . Let ω 1 , . . . , ω n be the
dual frame of e1 , . . . , en , i.e., ωi (ej ) = δij .
We shall make use of the following convention on the ranges of indices
unless mentioned otherwise:
1 ≤ α, β, γ, . . . ≤ m; 1 ≤ i, j, k, ℓ ≤ n; n + 1 ≤ r, s, t, . . . ≤ m.
Put heα , eβ i = ǫα δαβ ; α, β = 1, . . . , m, and
X X
∇ej = ǫk ωjk ek + ǫr ωjr er ,
j r
X X (2.50)
∇er = ǫk ωrk ek + ǫs ωrs es .
k s
The 1-forms ωαβ
are called the connection forms. We find from (2.50) that
ωjk = −ωkj , ωjr = −ωrj , ωrs = −ωsr . (2.51)
For the second fundamental form σ of N , if we put
X
r
σ(ei , ej ) = ǫr σij er , (2.52)
r
then we derive from (2.13) and (2.14) that
r
ωjr (ei ) = − h∇ei er , ej i = σij .
Combining this with (2.51) gives
X
ωjr = r i
σij ω. (2.53)
i
The Cartan structure equations are then given by
X
dω i = − ǫj ωji ∧ ω j , (2.54)
j
X X
dωji = ǫr (hrik hrjℓ − hriℓ hrjk )ω k ∧ ω ℓ − ǫk ωki ∧ ωjk + Ωij , (2.55)
k,ℓ,r k
X X
dωir = − ǫj hrjk ω k ∧ ωij + ǫs hsij ω j ∧ ωsr + Ωri , (2.56)
j,k j,s
X X
dωsr = ǫi (hrik hsiℓ − hriℓ hsik )ω k ∧ ω ℓ − ǫt ωtr ∧ ωst + Ωrs , (2.57)
i,k,ℓ t
1X α X
Ωα
β = Kβjk ω j ∧ ω k , R(eα , eβ )eγ = α
ǫδ Kβγδ eδ . (2.58)
2
j,k δ
Those Ωα
β are known as the curvature 2-forms of M , restricted to N .
Chapter 3
Warped Product Manifolds
One of the most fruitful generalizations of the notion of Cartesian (or direct)
products is the notion of warped products defined in [Bishop and O’Neill
(1964)]. The concept of warped products appeared in the mathematical
and physical literature before [Bishop and O’Neill (1964)]. For instance,
warped products were called semi-reducible spaces in [Kruchkovich (1957)].
Many exact solutions of the Einstein field equations and modified field
equations are warped products. For instance, the Schwarzschild solution
and Robertson-Walker models are warped products. While the Robertson-
Walker models describes a simply-connected homogeneous isotropic ex-
panding or contracting universe, the Schwarzschild solution is the best rel-
ativistic model that describes the outer space around a massive star or a
black hole. The Schwarzschild model laid the groundwork for the descrip-
tion of the final stages of gravitational collapse and the objects known today
as black holes.
Twisted products and convolution manifolds are two natural extensions
of warped product manifolds. In the last three sections of this chapter we
discuss both twisted products and convolution manifolds.
3.1 Warped products
Let B and F be two pseudo-Riemannian manifolds of positive dimensions

equipped with pseudo-Riemannian metrics gB and gF , respectively, and let
f be a positive smooth function on B.
Consider the product manifold B × F with its natural projection π :
B × F → B and η : B × F → F .
Definition 3.1. The warped product M = B ×f F is the manifold B × F
47
equipped with the pseudo-Riemannian structure such that

hX, Xi = hπ∗ (X), π∗ (X)i + f 2 (π(x)) hη∗ (X), η∗ (X)i
for any tangent vector X ∈ T M . Thus we have
g = gB + f 2 gF . (3.1)
The function f is called the warping function of the warped product.
A warped product B ×f F is called trivial if f is a constant. In this
case, B ×f F is the Riemannian product B × Ff , where Ff is the manifold
F equipped with metric f 2 gF , which is homothetic to gF .
For a warped product B ×f F , B is called the base of the warped product

and F the fiber. The leaves B × {q} = η −1 (q) and the fibers {p} × F =
π −1 (p) are pseudo-Riemannian submanifolds of M . Vectors tangent to
leaves are called horizontal and those tangent to fibers are called vertical.
We denote by H the orthogonal projection of T(p,q) M onto its horizontal
subspace T(p,q) (B × {q}) and by V the projection onto the vertical subspace
T(p,q) ({p} × F ).
If u ∈ Tp B, p ∈ B and q ∈ F , then the lift ū of u to (p, q) is the unique
vector in T(p,q) M such that π∗ (ū) = u. For a vector field X ∈ X(B), the
lift of X to M is the vector field X̄ whose value at each (p, q) is the lift of
Xp to (p, q). The set of all horizontal lifts is denoted by L(B). Similarly,
we denote by L(F ) the set of all vertical lifts.
For X̄, Ȳ ∈ L(B) and V̄ , W̄ ∈ L(F ), we have
[X̄, Ȳ ] = [X, Y ]− ∈ L(B), (3.2)
−
[V̄ , W̄ ] = [V, W ] ∈ L(F ), (3.3)
[X̄, V̄ ] = 0, (3.4)
−
where [X, Y ] denotes the lift of [X, Y ].
Lemma 3.1. If λ ∈ F (B), then the gradient of the lift λ ◦ π of λ to M =

B ×f F is the lift to M of the gradient of λ on B.
Proof. If v is a vertical vector in T M , it follows from π∗ (v) = 0 that

h∇(λ ◦ π), vi = v(λ ◦ π) = π∗ (v)λ = 0.
Thus, ∇(λ ◦ π) is a horizontal vector. Hence, if z is horizontal, we have
hπ∗ (∇(λ ◦ π)), π∗ (z)i = h∇(λ ◦ π), zi = z(λ ◦ π)
= (π∗ z)λ = h∇λ, π∗ zi ,
which implies that at each point, π∗ (∇(λ ◦ π)) = ∇λ.
Warped Product Manifolds 49
3.2 Connection of warped products
Based on Lemma 3.1, we may simply write λ for λ ◦ π and grad λ for
grad(λ ◦ π). The Levi-Civita connection ∇ of M = B ×f F is related with
the Levi-Civita connections of B and F as follows.
Proposition 3.1. For X, Y ∈ L(B) and V, W ∈ L(F ), we have on B ×f F
that
(1) ∇X Y ∈ L(B) is the lift of ∇X Y on B;
(2) ∇X V = ∇V X = (X ln f )V ;
hV, W i
(3) nor(∇V W ) = σ(V, W ) = − ∇f ;
f
(4) tan(∇V W ) ∈ L(F ) is the lift of ∇′V W on F , where ∇′ is the Levi-Civita
connection of F .
Proof. Property (1) can be proved as follows. From Koszul’s formula we
find 2 h∇X Y, V i = hV, [X, Y ]i − V hX, Y i due to [X, V ] = [Y, V ] = 0. Since
X, Y are lifts from B, hX, Y i is constant on fibers. Because V is vertical,
V hX, Y i = 0. But [X, Y ] is tangent to leaves, hV, [X, Y ]i = 0. Hence,
h∇X Y, V i = 0 for all V ∈ L(F ). This shows that ∇X Y is horizonal. Since
each π|B×q is an isometry, we obtain properties (1).
From [X, V ] = 0, we find ∇X V = ∇V X. Since these vector fields are
vertical, we obtain h∇X V, Y i = − hV, ∇X , Y i = 0. Thus, by the Koszul
formula, we get
2 h∇X V, W i = XhV, W i . (3.5)
On the other hand, by the definition of warped product metric, we find
hV, W i(p,q) = f 2 (p) hVq , Wq i .
So, after writing f for f ◦ π, we have hV, W i = f 2 (hV, W i ◦ η). Hence
XhV, W i = X(f 2 (hV, W i ◦ η) = 2f Xf (hV, W i ◦ η) = 2(X ln f ) hV, W i .
Combining this with (3.5) give property (2). By property (2) we find
h∇V W, Xi = − hW, ∇V Xi = −(X ln f ) hV, W i .
Thus, after applying Lemma 3.1, we find Xf = h∇f, Xi on M as on B.
Hence, for any X, we obtain
h∇V W, Xi f = − hV, W i h∇f, Xi
which implies property (3).
Since V and W are tangent to all fibers, tan(∇V W ) is the fiber covariant
derivative applied to the restrictions of V and W to that fiber. Therefore,
we have property (4).
3.3 Curvature of warped products
Consider a warped product M = B ×f F . The lift T̃ of a covariant tensor

T on B to M is its pullback π ∗ (T ) via the projection π : M → B. Let B R
and F R be the lifts to M of the curvature tensors of B and F , respectively.
The next result provides the curvature of a warped product M = B×f F
in terms of its warping function f and the curvature tensors BR and FR of
B and F .
Proposition 3.2. Let M = B ×f F be a warped product of two pseudo-

Riemannian manifolds. If X, Y, Z ∈ L(B) and U, V, W ∈ L(F ), then we
have
(1) R(X, Y )Z ∈ L(B) is the lift of BR(X, Y )Z on B;

H f (X, Y )
(2) R(X, V )Y = V;
f
(3) R(X, Y )V = R(V, W )X = 0;
hV, W i
(4) R(X, V )W = − ∇X (∇f );
f
h∇f, ∇f i
(5) R(V, W )U = FR(V, W )U + {hV, U i W − hW, U i V },
f2
where R is the curvature tensor of M and H f is the Hessian of f .
Proof. Since the projection π : M → B is isometric on each leaf, BR gives

the Riemannian curvature tensor of each leaf. Because leaves are totally
geodesic in M , BR agrees with the curvature tensor R of M on horizontal
vectors. Thus we have (1).
For X, Y ∈ L(B) and V ∈ L(F ), we have [V, X] = 0 and so
R(X, V )Y = ∇X ∇V Y − ∇V ∇X Y.
Hence, by Proposition 3.1, we find
∇X ∇V Y = ∇X ((Y (ln f )V )
= (XY (ln f ))V + (Y (ln f ))∇X V

= XY (ln f ) + (Y f )Xf −1 V + (X(ln f ))(Y (ln f ))V
= (XY (ln f )) V.
Thus we find
R(X, V )Y = (XY (ln f ))V − ∇V ∇X Y. (3.6)
On the other, since ∇X Y ∈ L(B), we have ∇V ∇X Y = (∇X Y (ln f ))V .
Combining this with (3.6) gives (2).
To prove (3) we assume that [V, W ] = 0. Then we have

R(V, W )X = ∇V ∇W X − ∇W ∇V X.
Since
∇V ∇W X = (V X(ln f ))W + (X(ln f ))∇V W = (X(ln f ))∇V W,
X(ln f ) is constant on fibers. Therefore, by Proposition 1.8, we obtain
R(V, W )X = (X(ln f ))∇V W − ∇W ((X(ln f ))V )
= (X(ln f ))(∇V W − ∇W V )
= (X(ln f ))[V, W ] = 0.
To show R(X, Y )V = 0, first we find
R(X, Y ; V, W ) = R(V, W ; X, Y ) = 0
by applying Proposition 3.1. Then, from (1) we find
R(X, Y ; V, Z) = −R(X, Y ; Z, V ) = 0.
Consequently, we obtain R(X, Y )V = 0. This gives (3).
To prove (4), first we notice that R(X, V )W is horizontal, since
R(X, V ; W, U ) = R(W, U ; X, V ) = 0
according to (3). Thus, after applying the first Bianchi identity, we derive
R(X, V )W = R(X, W )V. Therefore, after using (2), we obtain
R(X, V ; W, Y ) = R(V, X; Y, W )
H f (X, Y )
=− hV, W i
f
hV, W i
=− h∇X (∇f ), Y i .
f
Since R(X, V )W is horizontal and equation holds for every Y , we have (4).
For (5), we observe that R(V, W )U is a vertical vector field since
R(V, W ; U, X) = −R(V, W ; X, U ) = 0
by (3). Now, because the projection η : M → F is a homothety on fibers,
F
R(V, W )U ∈ L(F ) is the application to V, W, U of the curvature tensor of
each fiber.
Consequently, FR(V, W )U and R(V, W )U are related by the equation
of Gauss. Finally, by combining this with the fact the second fundamental
form of the fibers satisfies σ(V, W ) = −(hV, W i /f )∇f , we obtain (5).
From Proposition 3.2 we have the following.
Proposition 3.3. On a warped product M = B ×f F with k = dim F > 1,

let X, Y be horizontal vectors and V, W vertical vectors. Then the Ricci
tensor Ric of M satisfies
k f
(1) Ric(X, Y ) = BRic(X, Y ) − H (X, Y );
f
(2) Ric(X, V ) = 0;
∆f h∇f, ∇f i
(3) Ric(V, W ) = FRic(V, W ) − − (k − 1) hV, W i,
f f2
where BRic and FRic are the lifts of the Ricci curvatures of B and of F ,
respectively.
For an isometric immersion φ : N → M and ϕ ∈ F (M ), we denote by

Dϕ the T ⊥ N -component of the gradient ∇ϕ.
Definition 3.2. A smooth function f on a Riemannian manifold is called

strictly convex if, at each point, the Hessian H f is positive definite, and it
is called convex if H f is positive semi-definite at each point.
By applying Proposition 3.2, the following result was proved in [Bishop

and O’Neill (1964)].
Theorem 3.1. Let B and F be Riemannian manifolds and let f > 0 be

a differentiable function B. Denote by KF is the sectional curvature of
F . Then the warped product B ×f F has curvature K < 0 if the following
conditions holds:
(1) dim B = 1 or K < 0 on B;

(2) f is strictly convex;
(3) dim F = 1, or KF < 0 if f has a minimum, or KF ≤ 0 if f does not
have a minimum.
Many complete manifolds of negative curvature were constructed in

[Bishop and O’Neill (1964)] by applying this theorem.
3.4 Einstein warped product manifolds
The following result is a consequence of Proposition 3.3 (cf. e.g. [Besse

(1987)]).
Proposition 3.4. The warped product M = B ×f F is Einstein (with

Ric = λg) if and only if the following three statements hold:
(1) (F, gF ) is Einstein with FRic = µgF ;

k
(2) BRic = λgB + H f ;
f
(3) µ = (k − 1)|∇f |2 + λf 2 − f ∆f with k = dim F .
Condition (1) in Proposition 3.4 gives a condition on (F, gF ) alone,

whereas (2) and (3) are differential equations for f on (B, gB ).
Lemma 3.2. [Kim and Kim (2003)] Let f be a smooth function on a Rie-
mannian manifold B. Then for any vector X, the divergence of the Hessian
tensor H f satisfies
div(H f )(X) = Ric(∇f, X) − ∆(df )(X), (3.7)
where ∆ is the Laplacian on B acting on differential forms.
Proof. The Ricci identity implies
∇2 df (X, Y, Z) − ∇2 df (Y, X, Z) = df (R(X, Y )Z) (3.8)
for vector fields X, Y, Z, where ∇2XY = ∇X ∇Y − ∇∇X Y is a second order

covariant differential operator. Since df is closed, it is easily proved that
∇2 df (X, Y, X) = ∇2 df (X, Z, Y ). (3.9)
For a fixed point x ∈ B we may choose a local orthonormal frame

E1 , . . . , Em of the space B such that ∇Ei Ej (0) = 0 for all i, j. Also, we
may assume ∇Ei Y (x) = 0 for a vector field Y . Taking the trace with
respect to X and Z in (3.8) and using (3.9), we have
m
X
divH f (Y ) = (∇2 df )(Ei , Ei , Y ) = −d∆f (Y ) + Ric(Y, ∇f ) (3.10)
i=1
at x. Thus we have (3.7).
Proposition 3.5. [Kim and Kim (2003)] Let (B m , gB ) be a compact Rie-

mannian manifold of dimension m ≥ 2. If f is a non-constant function on
B satisfying condition (2) of Proposition 3.4 for a constant λ ∈ R and a
natural number k, then f satisfies condition (3) for µ ∈ R. Hence, for a
compact Einstein space (F, gF ) of dimension k with FRic = µgF , we can
make a compact Einstein warped product B ×f F satisfying Ric = λg.
Proof. Taking the trace of both sides of the formula given in Proposition
3.4(2) gives
k
2τ = mλ − ∆f, (3.11)
f
where τ is the scalar curvature of B defined by (1.16).
The second Bianchi identity implies
dτ = div(Ric). (3.12)
From (3.11) and (3.12), we obtain
k
div(Ric(X)) = {(∆f )df − f d(∆f )}(X). (3.13)
2f 2
On the other hand, by definition we have
f
H 1 1
div (X) = − 2 H f (∇f, X) + divH f (X)
f f f
for any vector field X and an orthonormal frame E1 , . . . , Em of B. Since
1
H f (X, ∇f ) = (∇X df )(∇f ) = d(|∇f |2 )(X),
2
the last equation becomes
f
H 1 1
div (X) = − 2 d(|∇f |2 )(X) + divH f (X)
f 2f f
for a vector field X on B. Thus, it follows from (3.7) and condition (2) of
Proposition 3.4 that
f
H 1
div = {(k − 1)d(|∇f |2 ) − 2f d(∆f ) + 2λf df }. (3.14)
f 2f 2

But condition (2) gives div(Ric) = div fk H f . Therefore (3.13) and (3.14)
imply
d(λf 2 − f ∆f + (k − 1)|∇f |2 ) = 0.
Thus condition (3) in Proposition 3.4 holds for some constant µ. Hence the
first part of the proposition is proved. For a compact Einstein k-manifold
(F, gF ) with FRic = µgF , we can construct a compact Einstein warped
product B ×f F by the sufficiencies of Proposition 3.4
The following are examples of Einstein warped product manifolds.
Example 3.1. Consider the punctured Euclidean n-space En∗ = En − {0}

and R+ = {r ∈ R : r > 0}. Then En∗ is the warped product R+ ×s S n−1 (1)
equipped with the warped product metric g = ds2 + s2 g1 , where g1 denotes
the metric of the unit (n − 2)-sphere.
Example 3.2. The warped product M1 = R+ ×sin s S n−1 (1) equipped

with the warped product metric g = ds2 + (sin2 s)g1 is of constant sectional
curvature 1. Similarly, the warped product M−1 = R+ ×sinh s S n−1 (1)
equipped with the warped product metric g = ds2 +(sinh2 s)g1 is of constant
sectional curvature −1.
Notice that En∗ , M1 and M−1 are non-complete Riemannian manifolds, but
they are open dense submanifolds of En , S n (1) and H n (−1), respectively.
For complete Einstein warped products B ×f F with dim B = 1, we
have the following.
Theorem 3.2. [Besse (1987)] Let M be a warped product B ×f F with

dim B = 1 and k = dim F > 1. If M is a complete Einstein manifold, then
either M is a Ricci flat Riemannian product, or B = R, F is Einstein with
non-positive scalar curvature and M has negative scalar curvature.
Proof. If M = B ×f F is a complete Einstein manifold, then the base

manifold B is complete too and f has to be defined on the whole of R.
Moreover, if B is S 1 , f has to be periodic. Then the only possibilities are
the products with λ = µ = 0 and the solutions corresponding to the cases
where fibers has negative or zero scalar curvature.
So far no one has been able to find examples of compact Einstein warped
products with non-constant warping function. Hence A. L. Besse asked in
[Besse (1987)] the following question:
Question 3.1. Does there exist a compact Einstein warped product with
non-constant warping function?
The following result provides a partial solution to this question.
Theorem 3.3. [Kim and Kim (2003)] Let M = B ×f F be an Einstein

warped product manifold with the base B a compact manifold. If M has non-
positive scalar curvature, then the warped product is simply a Riemannian
product.
Proof. Under the hypothesis, condition (3) in Proposition 3.4 becomes

div(f ∇f ) + (k − 2)|∇f |2 + λf 2 = µ. (3.15)
By integrating (3.15) over B we have
Z Z
k−2 λ
µ= |∇f |2 dV + f 2 dV, (3.16)
vol(B) B vol(B) B
where vol(B) denotes the volume of B.
Case (1): k ≥ 3. Let x be a maximum point of f on B. Then we have

f (x) > 0, ∇f (x) = 0 and ∆f (x) ≥ 0. Hence we obtain from (3.16) and
condition (3) in Proposition 3.4 that
0 ≤ f (x)∆f (x) = λf (x)2 − µ
Z Z
2−k λ
= |∇f |2 dV + (f (x)2 − f 2 )dV ≤ 0.
vol(B) B vol(B) B
The last inequality follows from the hypothesis on λ. Thus f is constant.
Case (2): k = 1, 2. We choose q as a minimum point of f on B. Then we

have f (q) > 0, ∇f (q) = 0 and ∆f (q) ≤ 0. Hence we find from (3.16) and
(3) in Proposition 3.4 that
0 ≥ f (q)∆f (q) = λf (q)2 − µ
Z Z
2−k λ
= |∇f |2 dV + (f (q)2 − f 2 )dV (3.17)
vol(B) B vol(B) B
≥ 0.
As in case (1), the last inequality follows from the hypothesis on λ. If
k = 1 or λ < 0, then (3.17) shows that f is constant.
If k = 2 and λ = 0, (3.15) and (3.16) imply that f 2 is harmonic on B,
and hence f is constant. This completes the proof of the theorem.
Another partial solution to Question 3.1 is the following (cf. Theorem
9.119 of [Besse (1987)] or [Kim (2000)]).
Theorem 3.4. Let M be a warped product B ×f F . If B is a 2-dimensional
compact Riemannian manifold, the warped product is simply a Riemannian
product.
Proof. Let (B, gB ) be a 2-dimensional compact Riemannian manifold.
By Theorem 3.3, we may assume that λ > 0. Since B is 2-dimensional, the
Ricci tensor satisfies BRic = GgB , where G denotes the Gauss curvature of
B. Hence condition (2) in Proposition 3.4 becomes
f
H f = (G − λ)gB . (3.18)
k
Let us assume that the warping function f is non-constant and let p, q
denote the minimum and maximum points of f . Then (3.18) implies that
(B − {p, q}, gB ) is isometric to the following warped product metric (cf.
Theorem 21 of [Kühnel (1988)])
ds2 = dt2 + f ′ (t)2 dθ2 (3.19)
on (a, b) × S 1 . Clearly, we have

f ′ (a) = f ′ (b) = 0. (3.20)
Since the metric (3.19) extends to a smooth Riemannian metric on B, we
may assume that (cf. Page 269 of [Besse (1987)])
f ′′ (a) = −f ′′ (b) = 1. (3.21)
′′
Note that ∆f = −1f (t) in the metric (3.19). Therefore condition (3) in
Proposition 3.4 becomes
2f (t)f ′′ (t) + (k − 1)f ′ (t)2 + λf (t)2 = µ. (3.22)
Now, assume that dim F ≥ 2. By integrating (3.22) we find
µ λ
f ′ (t)2 = − f (t)2 + νf (t)1−k . (3.23)
k−1 k+1
Hence
λ k−1
f ′′ (t) = − f (t) − νf (t)−k , (3.24)
k+1 2
where ν is a constant. Now, if we put

µ λx2 λxk+1 µxk−1
g(x) = − + νx1−k = x1−k ν − + , (3.25)
k−1 k+1 k+1 k−1
then we have f ′ (t)2 = g(f (t)) and f ′′ (t) = 12 g ′ (f (t)). If A, B denote the
minimum f (a) = f (p) and maximum f (b) = f (q) of f , respectively, then
(3.20) and (3.21) imply
g(A) = 0, g ′′ (A) = 2, g(B) = 0, g ′′ (B) = −2. (3.26)
From (3.25) and (3.26) we get
−2 p 1 p
ν= 2 1 + λµ + k Ak , A = ( 1 + λµ − 1). (3.27)
k −1 λ
It follows from (3.25) and (3.26) that
1 p
B = ( 1 + λµ + 1). (3.28)
λ
Since g(B) = 0, we conclude from (3.25) and (3.27)-(3.28) that the positive
√
constant y = 1 + λµ is a positive zero of the following polynomial:
hk (y) =(k − 1)(y + 1)k+1 − (k + 1)(y 2 − 1)(y 2 + 1)k−1 + 2(y + k)(y − 1)k .
It is direct to show that hk (y) is a polynomial of degree k − 2 which can be
expressed as
[ k−1
X 2 ]

k + 1 k−2j
hk (y) = 8 j y ,
j=1
2j + 1
where [ · ] denotes the Gaussian integer function. Since all the coefficients
of hk (y) are positive, it cannot have a positive zero. This contradiction
completes the proof of the theorem for k ≥ 2.
If k = 1, then a similar argument provides a contradiction as well.
3.5 Conformally flat warped product manifolds
Let (M, g) be a pseudo-Riemannian n-manifold with Ricci tensor Ric and

scalar curvature τ . Define a (0, 2)-tensor L by
1 τ
L(X, Y ) = − Ric(X, Y ) + g(X, Y ). (3.29)
n−1 (n − 1)(n − 2)
Let N be the (1, 1)-tensor associated with L, i.e., g(N X, Y ) = L(X, Y ).
The Weyl conformal curvature tensor C is the (1, 3)-tensor field defined by
C(X, Y )Z = R(X, Y )Z + L(Y, Z)X − L(X, Z)Y
(3.30)
+ g(Y, Z)N X − g(X, Z)N Y.
It is well-known that the conformal curvature tensor C is invariant under
conformal changes of the metric and it vanishes identically when dim M = 2
or 3 (cf. [Weyl (1918)] or [Chen (1973b)]). Define the tensor field D by
D(X, Y, Z) = (∇X L)(Y, Z) − (∇Y L)(X, Z). (3.31)
Definition 3.3. A metric g on a pseudo-Riemannian manifold M is called

conformally flat if it is locally conformally related with a flat metric. A
manifold with a conformally flat metric is called a conformally flat manifold.
The following result of [Weyl (1918)] is well-known.
Theorem 3.5. A necessary and sufficient condition for a pseudo-

Riemannian n-manifold M to be conformally flat is that C = 0 for n > 3
and D = 0 for n = 3.
The necessary and sufficient condition for the warped product of two
conformally flat manifolds to be conformally flat was obtained in [Ogawa
(1978)]. We also have the following result from [Brozos-Vázquez et al.
(2005)] for conformally flat warped product manifolds.
Theorem 3.6. Let B ×f F be a pseudo-Riemannian warped product. Then

the following hold:
(1) If dim B = 1, then B ×f F is conformally flat if and only if (F, gF ) is
a space of constant curvature.
(2) If dim B > 1 and dim F > 1, then B ×f F is conformally flat if and
only if
(2.1) (F, gF ) is a space of constant curvature cF .
(2.2) The function f : B → R+ defines a global conformal deformation
on B such that (B, f −2 gB ) is a space of constant curvature −cF .
(3) If dim F = 1, then B ×f F is conformally flat if and only if the function

f : B → R+ defines a global conformal deformation on B such that
(B, f −2 gB ) is a space of constant curvature −cF .
Proof. Let g = gB + f 2 gF be the warped product metric on the warped

product manifold B ×f F . Assume that B ×f F is conformally flat. Put
g = f 2 (f −2 gB + gF ).
Since conformally flatness is a conformally invariant property and g is a

conformally flat metric on B×f F , the metric g̃ = f −2 gB +gF is conformally
flat. Clearly, g̃ is a Riemannian product metric on B × F . Thus both factor
manifolds (B, f −2 gB ) and (F, gF ) are conformally flat manifolds. Now, the
theorem follows from the result obtained in [Ogawa (1978)].
The following is an easy consequence of Theorem 3.6.
Corollary 3.1. [Brozos-Vázquez et al. (2005)] Let B ×f F be a conformally

flat semi-Riemannian warped product. Then (B, gB ) is conformally flat and
(F, gF ) is of constant sectional curvature.
3.6 Multiply warped product manifolds
The notion of warped products can be naturally extended to multiply

warped products as follows:
Definition 3.4. Let N1 , . . . , Nℓ be ℓ pseudo-Riemannian manifolds and let

N = N1 × · · · × Nℓ be the Cartesian product of N1 , . . . , Nℓ . Denote by
πi : N → Ni the canonical projection of N onto Ni for i = 1, . . . , ℓ. If
f1 , . . . , fℓ : N1 → R+ are positive-valued functions in F (N1 ), then
ℓ
X
hX, Y i = hπ1∗ X, π1∗ Y i + (fi ◦ π)2 hπi∗ X, πi∗ Y i (3.32)
i=2
defines a pseudo-Riemannian metric g on N , called a multiply warped prod-

uct metric. The product manifold N endowed with this metric g, denoted
by N1 ×f2 N2 × · · · ×fℓ Nℓ , is called a multiply warped product.
Definition 3.5. A foliation on a manifold M is an integrable subbundle F

of the tangent bundle of M , i.e. for any two vector fields X and Y taking
values in F , the Lie bracket [X, Y ] takes values in F as well.
Definition 3.6. A pseudo-Riemannian submanifold of a pseudo-Rieman-

nian manifold is called an extrinsic sphere if it is a totally umbilical sub-
manifold with parallel mean curvature vector.
Definition 3.7. A foliation L on a pseudo-Riemannian manifold M is

called totally umbilical, if every leaf of L is a totally umbilical pseudo-
Riemannian submanifold of M .
If, in addition, the mean curvature vector of every leaf is parallel in the
normal bundle, then L is called a spherical foliation, because in this case
each leaf of L is an extrinsic sphere of M .
If every leaf of L is a totally geodesic submanifold of M , then L is called
a totally geodesic foliation. A totally geodesic foliation is also known as an
autoparallel foliation.
For a multiply warped product N1 ×f2 N2 × · · · ×fℓ Nℓ , let Di denote

the distribution obtained from the vectors tangent to Ni (or more precisely,
vectors tangent to the horizontal lifts of Ni ).
Let Li denote the foliation of N canonically induced on Ni and let
T N → T Li , v 7→ vi denote the vector bundle projection. Obviously, T N
splits orthogonally with respect to h , i. Thus we have T N = ⊕ℓi=1 T Li .
Definition 3.8. Assume that φ : N1 ×f2 N2 ×· · ·×fℓ Nℓ → M is an isometric

immersion of N1 ×f2 N2 × · · · ×fℓ Nℓ into a pseudo-Riemannian manifold
M . Denote by h the second fundamental form of φ. Then φ is called mixed
totally geodesic if σ(Di , Dj ) = {0} holds for all distinct i, j ∈ {1, . . . , ℓ}.
Let ∇0 , R0 , h , i0 , etc., be the Levi-Civita connection, the Riemann

curvature tensor, the scalar product, etc., of the Riemannian product
N1 × N2 × · · · × Nℓ with f2 = · · · = fℓ = 1
and denote by ∇f , Rf , h , if , etc., the corresponding quantities of the

warped product N1 ×f2 N2 × · · · ×fℓ Nℓ .
The next lemma provides relations between the Levi-Civita connections
∇0 and ∇f and the curvature tensors R0 and R̃f [Ponge and Reckziegel
(1993); Nölker (1996); Dobarro and Ünal (2005)].
Lemma 3.3. Let N = N1 ×f2 N2 × · · · ×fℓ Nℓ be a multiply warped product

of ℓ pseudo-Riemannian manifolds N1 , . . . , Nℓ . If we put
Ůi = −∇((ln fi ) ◦ π1 ), (3.33)

then we have
ℓ
X
i i
∇fX Y − ∇0X Y = X , Y Ůi − h Ůi , X i Y i − h Ůi , Y i X i , (3.34)
i=2
ℓ
X
Rf (X, Y ) − R0 (X, Y ) = (∇fX 1 Ůi − h Ůi , X i Ůi ) ∧ Y i (3.35)
i=2
k
X ℓ
X
+ X i ∧ (∇fY 1 Ůi − h Ůi , Y i Ůi ) − h Ůi , Ůj i X i ∧ Y j ,
i=2 i,j=2
for X, Y ∈ X(N ), where h , i = h , if , X i is the Ni -component of X and

X ∧ Y is defined by (X ∧ Y )Z = hZ, Y i X − hZ, Xi Y.
Proof. Since Di is parallel with to h , i0 , we have

(∇X Y )i = ∇X Y i for all X, Y ∈ X(N ). (3.36)
The tensor field S on N given by the right side of (3.34) is symmetric, hence
ˆ = ∇0 + S is a torsion-free affine connection on N . From (3.32), (3.33)
∇
and (3.36) we easily derive that ∇ ˆ is metric with respect to h , i , thus
f
ˆ f
∇ = ∇ by the uniqueness of Levi-Civita connection. This gives (3.34).
Equation (3.35) can be obtained by a lengthy direct calculation.
Corollary 3.2. Under the notations given above, T Li is a spherical fo-

liation with mean curvature vector Ůi and (T Li )⊥ is autoparallel for i =
2, . . . , ℓ. Consequently, T L1 = ∩ℓi=2 (T Li )⊥ is autoparallel.
Proof. Equation (3.34) of Lemma 3.3 immediately implies that T Li is

totally umbilical with mean curvature vector Ůi and that (T Li )⊥ is auto
parallel with respect to h , i for i = 2, . . . , ℓ. Since now for all vector fields
X in T Li and Y in T Lj , we find from (3.33) and (3.34) that h ∇X Ůi , Y i =
h ∇Y Ůi , X i = 0. Thus T Li is a spherical foliation.
The following famous de Rham decomposition theorem is well-known.
Theorem 3.7. [de Rham (1952)] Let M be a Riemannian manifold and

let T M = ⊕ki=1 Ei be an orthogonal decomposition into non-trivial vector
subbundles such that each Ei is an autoparallel foliation. Then we have:
(a) Let y be a point in M and for each i let Mi be a maximal integral

manifold of Ei through y. Then y has an open neighborhood V such
that V = V1 × · · · × Vk , where each Vi is an open neighborhood of y
in Mi and the Riemannian metric in V is the direct product of the
Riemannian metrics in the Vi ’s.
(b) If M is simply-connected and complete, then for every point p̃ ∈ M

there exists an isometry ψ of a direct product M1 ×ρ2 × · · · ×ρk Mk onto
all of M .
The following result from [Hiepko (1979)] is a natural generalization

of de Rham’s decomposition theorem to warped products as a converse of
Corollary 3.2.
Theorem 3.8. Let M be a Riemannian manifold and let T M = ⊕ki=1 Ei

be an orthogonal decomposition into non-trivial vector subbundles such that
each Ei is spherical and (Ei )⊥ is autoparallel for i = 2, . . . , k. Then
(a) For every point p̃ ∈ M there is an isometry ψ of a warped product

M1 ×ρ2 × · · · ×ρk Mk onto a neighborhood of p̃ in M such that
ρ2 (p̃1 ) = · · · = ρk (p̃1 ) = 1, (3.37)
where p̃1 is the component of ψ −1 (p̃) in M1 and such that
ψ({p̃1 } × · · · × {pi−1 } × Mi × {pi+1 } × · · · × {pk }) (3.38)
is an integral submanifold of Ei for i = 1, . . . , k and for all p1 ∈
M 1 , . . . , pk ∈ M k .
(b) If M is simply-connected and complete, then for every point p̃ ∈ M
there exists an isometry ψ of a warped product M1 ×ρ2 × · · · ×ρk Mk
onto all of M with the properties (3.37) and (3.38).
3.7 Warped product immersions
The following definition is due to [Nölker (1996)].
Definition 3.9. Let N1 ×ρ2 N2 × · · · ×ρℓ Nℓ be a multiply warped product

and let φi : Ni → Mi , i = 1, . . . , ℓ, be isometric immersions. Define fi =
ρi ◦ φ1 : N1 → R+ for i = 2, . . . , ℓ. Then the map
φ : N1 ×f2 N2 × · · · ×fℓ Nℓ → M1 ×ρ2 M2 × · · · ×ρℓ Mℓ
defined by φ(p1 , . . . , pℓ ) = (φ1 (p1 ), . . . , φℓ (pℓ )) is an isometric immersion,
called a warped product immersion.
Definition 3.10. A multiply warped product M1 ×ρ2 M2 × · · · ×ρℓ Mℓ is

called a warped product representation of a real space form Rm (k) if the
warped product M1 ×ρ2 M2 × · · · ×ρℓ Mℓ is an open dense subset of Rm (k).
Remark 3.1. The Euclidean 3-space E3 has two famous warped product
representations; namely, the cylindrical and the spherical coordinates. S.
Nölker generalized these these two examples as far as possible: starting
from algebraic initial data, he constructed for every standard space Rn (k) of
constant curvature k an isometry ψ of a warped product N1 ×ρ1 × · · ·×ρk Nk
onto an open dense subset of Rn (k). Such a warped product representation
ψ has many “rotational” symmetries for i = 2, . . . , k.
The product foliation induced by Ni is an orbit foliation with respect
to the action of a suitable subgroup of the isometry group of Rn (k). Every
leaf of this foliation is a complete totally umbilical submanifold of Rn (k).
The foliation induced by N1 is even totally geodesic (but its leaves are in
general not complete).
For the details on possible warped product representations of standard
spaces of constant curvature, see [Nölker (1996)].
Let φ : N1 ×f2 N2 × · · · ×fℓ Nℓ → M be an isometric immersion of a

multiply warped product N1 ×f2 N2 × · · · ×fℓ Nℓ into a pseudo-Riemannian
manifold M . Let σi denote the restriction of the second fundamental form
to Di , i = 1, . . . , ℓ. Denote by Tr σi the trace of σi restricted to Ni , i.e.,
ni
X
Tr σi = σ(eα , eα )
α=1
for an orthonormal frame fields e1 , . . . , eni of Di .

The partial mean curvature vector Hi is defined by
Tr σi
Hi = , i = 1, . . . , ℓ. (3.39)
dim Ni
Definition 3.11. An immersion φ : N1 ×f2 N2 × · · · ×fℓ Nℓ → M is called
Ni -totally geodesic (resp. Ni -minimal ) if σi (resp. Hi ) vanishes identically.
Notation 3.1. Let σ φ denote the second fundamental form of a warped

product immersion
φ = (φ1 , . . . , φℓ ) : N1 ×f2 N2 × · · · ×fℓ Nℓ → M1 ×ρ2 M2 × · · · ×ρℓ Mℓ ,
and let σ 0 denote the second fundamental form of the corresponding direct
product immersion (φ1 , . . . , φℓ ) : N1 × · · · × Nℓ → M1 × · · · × Mℓ .
Denote by ∇0 and ∇f the Levi-Civita connections of N1 × N2 × · · · × Nℓ
and of N1 ×f2 N2 × · · · ×fℓ Nℓ , respectively; and by ∇ ˜ 0 and ∇ ˜ ρ the Levi-
Civita connections of M1 × M2 × · · · × Mℓ and M1 ×ρ2 M2 × · · · ×ρℓ Mℓ ,
respectively.
The second fundamental forms σ φ and σ 0 are related by the following.
Lemma 3.4. Let φ : N1 ×f2 N2 × · · · ×fℓ Nℓ → M1 ×ρ2 M2 × · · · ×ρℓ Mℓ be

a warped product immersion between pseudo-Riemannian multiply warped
products. Then σ φ and σ 0 are related by
σ φ (X, Y ) = σ 0 (X, Y ), X, Y ∈ D1 , (3.40)
φ 0
σ (Z, W ) = σ (Z, W ) − hZ, W i D(ln ρi ), Z, W ∈ Di , (3.41)
φ
σ (Di , Dj ) = {0}, 1 ≤ i 6= j ≤ ℓ. (3.42)
Proof. Follows from Gauss’ formula and (3.34).
Corollary 3.3. Let φ : N1 ×f2 N2 × · · · ×fℓ Nℓ → M1 ×ρ2 M2 × · · · ×ρℓ Mℓ

be a warped product immersion. Then φ is totally geodesic if and only if
the following two conditions are satisfied:
(1) φ1 : N1 → M1 is totally geodesic and
(2) φi : Ni → Mi is a totally umbilical immersion such that the second
fundamental form is given by
σ φi (Z, W ) = hZ, W i D(ln ρi )
for Z, W ∈ T (Ni ), i ∈ {2, . . . , ℓ}.
Proof. Follows from Lemma 3.4 and the definition of totally umbilical
immersions.
The following is the indefinite version of Moore’s lemma [Moore (1971)]
stated in [Magid (1984)].
Lemma 3.5. Let φ : N1 × · · · × Nℓ → Em be an isometric immersion of

the direct product of ℓ pseudo-Riemannian manifolds N1 , . . . , Nℓ into Em j .
Then φ is a mixed totally geodesic immersion if and only if φ is a product
immersion, i.e., there exist an isometry Ψ of Em and isometric immersions
φi : Ni → Emji , 1 ≤ i ≤ ℓ, such that Ψ ◦ φ(p1 , . . . , pℓ ) = (φ1 (p1 ), . . . , φℓ (pℓ )),
i
where pi ∈ Ni , 1 ≤ i ≤ ℓ.
S. Nölker extended Moore’s result to the following [Nölker (1996)].
Theorem 3.9. Let φ : N1 ×f2 N2 ×· · ·×fℓ Nℓ → Rm (k) be an isometric im-

mersion of a multiply warped product into a complete simply-connected real
space form Rm (k) of constant curvature k. If φ is mixed totally geodesic,
then there is an explicitly constructible isometry ψ of M1 ×ρ2 M2 ×· · ·×ρℓ Mℓ
of a multiply warped product onto an open dense subset of Rm (k), where M1
is an open subset of a standard space and M2 , . . . , Mℓ are standard spaces,

and there exist isometric immersions φi : Ni → Mi , i = 1, . . . , ℓ, such that
fi = ρi ◦ φ1 for i = 2, . . . , ℓ and φ = ψ ◦ (φ1 × · · · × φℓ ).
Here, by a standard space we mean a sphere, a Euclidean space or a
hyperbolic space of constant curvature.
In other words, under the assumptions of the theorem there is a warped

product representation ψ (of a big subset) of the ambient space Rm (k) such
that φ is the product of φ1 , . . . , φℓ with respect to ψ.
Definition 3.12. A submanifold N of a Riemannian manifold M̃ is called

semi-parallel it satisfies
RD (X, Y )σ(Z, W ) = σ(R(X, Y )Z, W ) + σ(Z, R(X, Y )W ), (3.43)
D
where R and R are the normal curvature tensor and Riemann curvature
tensor of N (cf. [Deprez (1986)]).
Parallel submanifolds are semi-parallel submanifolds; as well as all flat

submanifolds with flat normal connection (i.e., with R = 0 and RD = 0).
Remark 3.2. By applying warped product decomposition of real space

forms, semi-parallel submanifolds with flat normal connection in real space
forms were completely described in [Dillen and Nölker (1999)].
3.8 More results for warped product immersions
The following results of this section were proved in [Chen (2005c)].
Theorem 3.10. Let φ = (φ1 , φ2 ) : N1 ×f N2 → M1 ×ρ M2 be a warped

product immersion between two warped product manifolds. Then we have:
(a) φ is mixed totally geodesic.
(b) The second fundamental form σ of φ satisfies
||σ||2 ≥ n2 |D ln ρ|2 , n2 = dim N2 , (3.44)
with the equality holding if and only if φ1 : N1 → M1 and φ2 : N2 → M2
are both totally geodesic immersions.
(c) φ is N1 -totally geodesic if and only if φ1 : N1 → M1 is totally geodesic.
(d) φ is N2 -totally geodesic if and only if φ2 : N2 → M2 is totally geodesic
and (∇ ln ρ)|N1 = ∇ ln f holds, i.e., the restriction of the gradient of
ln ρ to N1 is the gradient of ln f , or equivalently, D ln ρ = 0.
(e) φ is a totally geodesic immersion if and only if φ is both N1 -totally

geodesic and N2 -totally geodesic.
Proof. Let N be an n-dimensional Riemannian manifold isometrically

immersed in another Riemannian manifold M̃ . Let ∇ and ∇ ˜ be the Levi-
Civita connections of N and M̃ , respectively.
Let (Nj , gj ) and (Mj , g̃j ), j = 1, 2, be Riemannian manifolds and φj :
Nj → Mj , j = 1, 2, be isometric immersions. Assume that
φ = (φ1 , φ2 ) : N1 ×f N2 → M1 ×ρ M2
is a warped product immersion of the warped product manifold N1 ×f N2
into the warped product manifold M1 ×ρ M2 .
Denote by ∇1 and ∇f the Levi-Civita connections of N1 × N2 equipped
with the direct product metric g0 = g1 + g2 and with the warped product
metric g = g1 + f 2 g2 , respectively. Similarly, denote by ∇ ˜ 1 and ∇ ˜ ρ the
Levi-Civita connections of M1 ×M2 equipped with the direct product metric
g̃0 = g̃1 + g̃2 and with the warped product metric g̃ = g̃1 +ρ2 g̃2 , respectively.
For vector fields U and V on M1 ×ρ M2 , the connections ∇ ˜ 1 and ∇ ˜ ρ are
related by
˜ρ V = ∇
∇ ˜ 1U V − hU2 , V2 i (∇ ln ρ) + h∇ ln ρ, U i V2 + h∇ ln ρ, V i U2 , (3.45)
U
where h , i is the inner product with respect to g̃, ∇ ln ρ is the gradient of

ln ρ on M1 , and U2 and V2 are the natural projections of U and V onto
L(M2 ), respectively. From (3.45) we obtain
˜ρ Y = ∇
∇ ˜ 1 Y, (3.46)
X X
˜ρ W = ∇
∇ ˜ 1 W − hZ, W i (∇ ln ρ) (3.47)
Z Z
for X, Y ∈ L(M1 ) and Z, W ∈ L(M2 ). Thus (3.46) and (3.47) give

σ(X, Y ) = σ 0 (X, Y ), (3.48)
0
σ(Z, W ) = σ (Z, W ) − hZ, W i D ln ρ, (3.49)
σ(X, Z) = 0 (3.50)
for X, Y ∈ L(N1 ) and Z, W ∈ L(N2 ), where σ is the second fundamental
form of the warped product immersion (φ1 , φ2 ) : N1 ×f N2 → M1 ×ρ M2 ,
and σ 0 is the second fundamental form of the direct product immersion
(φ1 , φ2 ) : N1 ×1 N2 → M1 ×1 M2 between two direct Riemannian products.
The restrictions of σ 0 to L(N1 ) and to L(N2 ) are the second fundamental
forms of φ1 : N1 → M1 and φ2 : N2 → M2 , respectively. Hence, σ 0 (X, Y )
and σ 0 (Z, W ) are orthogonal for X, Y ∈ L(N1 ) and Z, W ∈ L(N2 ).
Equation (3.50) is nothing but statement (a). Further, statements (b)

and (c) follows from (3.48) and (3.49).
If φ : N1 ×f N2 → M1 ×ρ M2 is N2 -totally geodesic, then it follows from
(3.49) that σ 0 (Z, W ) = hZ, W i (D ln ρ) for Z, W ∈ L(N2 ). Since D ln ρ and
σ 0 (Z, W ) are orthogonal, we find σ 0 (Z, W ) = 0 and D ln ρ = 0. The first
equation implies that φ2 is totally geodesic and the second equation implies
that (∇ ln ρ)|N1 = ∇ ln f .
Conversely, if φ2 is totally geodesic and (∇ ln ρ)|N1 = ∇ ln f holds, then
it follows from (3.49) that σ(Z, W ) = 0 for Z, W ∈ L(N2 ). Hence φ2 is a
totally geodesic immersion. This proves statement (d).
Statement (e) follows from statements (c) and (d) and equation (3.50).
This completes the proof of the theorem.
Theorem 3.11. A warped product immersion φ = (φ1 , φ2 ) : N1 ×f N2 →

M1 ×ρ M2 between two warped product manifolds is totally umbilical if and
only if we have:
(1) φ1 : N1 → M1 is a totally umbilical immersion with mean curvature

vector given by −D ln ρ, and
(2) φ2 : N2 → M2 is a totally geodesic immersion.
Proof. Assume that φ : N1 ×f N2 → M1 ×ρ M2 is a totally umbilical

immersion. Then we have
σ(X, Y ) = hX, Y i H, σ(Z, W ) = hZ, W i H (3.51)
for X, Y ∈ L(N1 ) and Z, W ∈ L(N2 ).

On the other hand, equations (3.48) and (3.51) imply that H is tan-
gent to the first factor M1 . Hence, it follows from (3.49) and (3.51) that
σ 0 (Z, W ) = 0 for Z, W in L(N2 ), since σ 0 (Z, W ) is always tangent to the
second factor M2 . Therefore φ2 : N2 → M2 is a totally geodesic immersion.
Consequently, we obtain condition (2).
Also from (3.48), (3.49), and (3.51) we find
σ 0 (X, Y ) = hX, Y i H, H = −D ln ρ (3.52)
for X, Y ∈ L(N1 ) which implies condition (1).

Conversely, it is easy to verify that if both conditions (1) and (2) hold,
then φ is a totally umbilical immersion. This gives the theorem.

product immersion between two warped product manifolds. Then we have:
(a) The partial mean curvature vector H1 is equal to the mean curvature
vector of φ1 : N1 → M1 ; thus, φ is N1 -minimal if and only if φ1 :
N1 → M1 is a minimal immersion.
(b) φ is N2 -minimal if and only if φ2 : N2 → M2 is a minimal immersion
and (∇ ln ρ)|N1 = ∇ ln f holds.
(c) φ = (φ1 , φ2 ) is a minimal immersion if and only if φ2 : N2 → M2 is a
minimal immersion and the mean curvature vector of φ1 : N1 → M1 is
given by n−1
1 n2 D ln ρ.
Proof. Since each lift of N1 is totally geodesic in N1 ×f N2 , (3.48) implies

that H1 is nothing but the mean curvature vector of φ1 : N1 → M1 . This
gives statement (a).
If φ : N1 ×f N2 → M1 ×ρ M2 is N2 -minimal, then (3.49) implies
f 2 Tr σ20 = n2 (D ln ρ). (3.53)
0
Because D ln ρ and Tr σ2 are orthogonal, we know that φ2 is a minimal
immersion and (∇ ln ρ)|N1 = ∇ ln f holds.
Conversely, if φ2 is minimal and (∇ ln ρ)|N1 = ∇ ln f holds, then it
follows from (3.49) that Tr σ2 = 0. Hence we obtain statement (b).
Finally, suppose that φ : N1 ×f N2 → M1 ×ρ M2 is a minimal immersion.
Then we have Tr h = 0. Thus, by applying (3.48) and (3.49), we find
0 = Tr σ10 + f 2 Tr σ20 − n2 (D ln ρ). (3.54)
Since Tr σ1 and D ln ρ are both tangent to the first factor M1 and Tr σ20 is
0
tangent to M2 , (3.54) implies that

Tr σ10 = n2 (D ln ρ), Tr σ20 = 0. (3.55)
This shows that φ2 : N2 → M2 is a minimal immersion and the mean
curvature vector of φ1 : N1 → M1 is given by n−1 1 n2 D ln ρ.
Definition 3.13. An immersion φ : N1 ×f N2 → M is called N2 -pseudo
umbilical if its shape operator AH satisfies AH Z = λZ for Z ∈ L(N2 ).
Definition 3.14. A warped product manifold M1 ×ρ M2 is called a warped
product representation of a real space form Rm (c) if the warped product
M1 ×ρ M2 is an open dense subset of Rm (c).
For warped product immersions into a real space form, we have the
following.
product immersion from a warped product N1 ×f N2 into a warped product
representation M1 ×ρ M2 of a real space form Rm (c). Then we have:
(1) The shape operator of φ satisfies

∆f
AH1 Z = −c Z (3.56)
n1 f
for Z ∈ L(N2 ), where ∆ is the Laplacian operator of N1 .
(2) For any X, Y ∈ L(N1 ) and Z ∈ L(N2 ), DZ σ(X, Y ) = 0 holds, where
D is the normal connection of φ. In particular, we have DZ H1 = 0.
(3) The two partial mean curvature vectors H1 and H2 are perpendicular
to each other if and only if the warping function f is an eigenfunction
of the Laplacian operator ∆ with eigenvalue n1 c.
(4) The warping function f is an eigenfunction of ∆ with eigenvalue n1 c
if and only if either φ1 : N1 → M1 is minimal or (∇ ln ρ)|N1 = ∇ ln f
holds.
(5) When c = 0, the two partial mean curvature vectors H1 and H2 are
perpendicular to each other if and only if the warping function f is a
harmonic function.
(6) If φ1 : N1 → M1 is a non-minimal immersion and the two partial
mean curvature vectors H1 and H2 are parallel at each point, then φ is
N2 -pseudo umbilical and φ2 : N2 → M2 is a minimal immersion.
Proof. Suppose that M1 ×ρ M2 is a warped product representation of a

real space form Rm (c) and φ : N1 ×f N2 → M1 ×ρ M2 is a warped product
immersion. Then, from Theorem 3.10(a), we have
σ(X, Z) = 0 (3.57)
for X ∈ L(N1 ) and Z ∈ L(N2 ).
Since N1 ×f N2 is a warped product, we also have
∇X Z = ∇Z X = (X ln f )Z, h∇X Y, Zi = 0 (3.58)
for unit vector fields X, Y ∈ L(N1 ) and Z ∈ L(N2 ). By (3.58) we find
K(X ∧ Z) = h ∇Z ∇X X − ∇X ∇Z X, Z i
1 (3.59)
= (∇X X)f − X 2 f .
f
If we choose a local orthonormal frame e1 , . . . , en1 +n2 in such way that
e1 , . . . , en1 are in L(N1 ) and en1 +1 , . . . , en1 +n2 in L(N2 ), then we derive
from (3.59) that
Xn1
∆f
= K(eα ∧ es ), s = n1 + 1, . . . , n1 + n2 . (3.60)
f α=1
On the other hand, the equation of Gauss implies that the curvature
tensor R of N1 ×f N2 satisfies
hR(X, Y )Z, W i = hσ(X, W ), σ(Y, Z)i − hσ(X, Z), σ(Y, W )i
(3.61)
+ c {hX, W i hY, Zi − hX, Zi hY, W i}
for vectors X, Y, Z, W ∈ T (N1 ×f N2 ). Using (3.57), (3.60), and (3.61), we
obtain
∆f
h H1 , σ(Z, Z) i = −c (3.62)
n1 f
for any unit Z ∈ L(N2 ). Thus, by applying polarization, we find
h H1 , σ(Z, W ) i = 0 (3.63)
for orthonormal vectors Z, W ∈ L(N2 ). Equations (3.62) and (3.63) imply
that the shape operator at H1 satisfies

∆f
AH1 Z = −c Z (3.64)
n1 f
for Z ∈ L(N2 ). Thus we have statement (1).
It follows from (3.57) and (3.58) that the covariant derivative of the
second fundamental form satisfies
¯ X σ)(Y, Z) = DX σ(Y, Z) − σ(∇X Y, Z) − σ(Y, ∇X Z)
(∇ (3.65)
= −σ(∇X Y, Z) = 0,
due to the fact that N1 is totally geodesic in N1 ×f N2 .
On the other hand, by applying (3.57) and (3.58), we also find
¯ Z σ)(X, Y ) = DX σ(X, Y ).
(∇ (3.66)
Therefore, after applying (3.65), (3.66), and the equation of Codazzi, we
obtain statement (2).
By applying equations (3.39) and (3.62) we obtain
∆f
h H1 , H 2 i = −c (3.67)
n1 f
which gives statement (3). It follows from equations (3.48) and (3.49) that
the partial mean curvature vectors H1 and H2 are perpendicular to each
other if and only if we have either (i) H1 = 0 or (ii) (∇ ln ρ)|N1 = ∇ ln f .
According to Theorem 3.12(a), the first case occurs when and only when
φ1 is a minimal immersion. By combining these results with statement
(3), we obtain statement (4). Obviously, statement (5) is a special case of
statement (3). If φ1 is a non-minimal immersion and if the two partial mean
curvature vectors H1 and H2 are parallel, then there exists a function µ

such that H2 = µH1 . In such case the mean curvature vector of φ is related
to the partial mean curvature vector H1 by
n1 + n2 µ
H= H1 .
n
Therefore, after applying (3.64) we may conclude that φ is N2 -pseudo-
umbilical.
Since φ1 is assumed to be a non-minimal immersion, we have H1 6= 0
according to Theorem 3.12(a). Therefore, by applying the parallelism of
H1 and H2 and the orthogonality of σ 0 (X, Y ) and σ 0 (Z, W ) for X, Y ∈
L(N1 ) and Z, W ∈ L(N2 ), we may conclude from (3.48) and (3.49) that
φ2 : N2 → M2 is a minimal immersion.
Remark 3.3. Many results given in this section were extended to doubly
warped product immersions φ : f1 N1 ×f2 N2 → ρ1 M1 ×ρ2 M2 in [Faghfouri
and Majidi (2015)].
3.9 Twisted products
Twisted products are natural extensions of warped products, namely the

warping function of a warped product were replaced by a twisting function
(cf. [Chen (1981a)]).
Definition 3.15. Let B and F be two pseudo-Riemannian manifolds

equipped with pseudo-Riemannian metrics gB and gF , respectively, and let
f be a positive smooth function on M . The twisted product M = B ×f F
is the manifold B × F with the pseudo-Riemannian metric g given by
g(X, Y ) = gB (π∗ (X), π∗ (Y )) + f 2 · gF (η∗ (X), η∗ (Y )) (3.68)

for tangent vectors X, Y ∈ T M .
When f depends only on B, the twisted product B ×f F reduces to

a warped product. In the case that B is a point, the twisted product is
nothing but a conformal change of metric on F .
Just like warped products, B is called the base and F the fiber of the
twisted product B ×f F . Both the leaves B × q = η −1 (q) and the fibers
p × F = π −1 (p) are pseudo-Riemannian submanifolds of B ×f F . Vectors
tangent to leaves are called horizontal and those tangent to fibers are called
vertical . We denote by H the orthogonal projection of T(p,q) M onto its
horizontal subspace T(p,q) (B × q) and by V the projection onto the vertical

subspace T(p,q) (p × F ).
If v is vector tangent to B at p ∈ B and q ∈ F , then the lift v̄ of v to
(p, q) is the unique vector in T(p,q) M such that π∗ v̄ = v. For a vector field
X ∈ X(B), the lift of X to M is the vector field X̄ whose value at each
(p, q) is the lift of Xp to (p, q).
The following result can be found in [Chen (1981a)].
Proposition 3.6. Let M = B ×f F be a twisted product of two pseudo-
Riemannian manifolds. Then
(1) leaves are totally geodesic in M ;
(2) for each p ∈ B, the fiber {p} × F is a totally umbilical submanifold of
M with −(∇(ln f ))H as its mean curvature vector, where (∇(ln f ))H
is the horizontal component of ∇(ln f );
(3) fibers have parallel mean curvature vector if and only if f is the product
of two positive functions λ ∈ F (B) and µ ∈ F (F ).
Proof. Since the projection π : M → B is isometric on each leaf, every
geodesic of a leaf is also a geodesic of M . Hence leaves are totally geodesics
submanifolds of M .
Let X, Y, Z ∈ L(B) and V, W ∈ L(F ). Then [X, V ] = 0, so
∇X V = ∇V X. (3.69)
From (3.68) we find
XhV, W i = X(f 2 hV, W iF ) = 2(X(ln f )) hV, W i , (3.70)
where h , iF is the scalar product associated with the metric gF on F .
On the other hand, it follows from (3.69) that
XhV, W i = h∇V X, W i + hV, ∇W Xi
= − h∇V W, Xi − h∇W V, Xi (3.71)
= −2 hσ(V, W ), Xi ,
where σ is the second fundamental of fibers.
By comparing (3.70) and (3.71) we find
σ(V, W ) = −(∇(ln f ))H hV, W i , (3.72)
which implies (2). Now, let us choose {e1 , . . . , er } to be an orthonormal
frame of horizontal spaces. Then the mean curvature vector of fibers can
be expressed as
Xr
H=− ǫi (ei (ln f ))ei . (3.73)
i=1
Thus, for any vertical vector V , we find

Xr r
X
∇V H = − ǫi (V ei (ln f ))ei − ǫi (ei (ln f ))∇V ei . (3.74)
i=1 i=1
Since the leaves are totally geodesic submanifolds in M , the last term in
(3.74) is vertical. Hence the mean curvature vector of fibers is parallel if
and only if V X(ln g) = 0 for each horizontal vector field X and vertical
vector field V . Consequently, the function f is the product of two positive
functions λ ∈ F (B) and µ ∈ F (F ).
Corollary 3.4. Let M = B ×f F be a warped product of two pseudo-

Riemannian manifolds. Then we have:
(1) Leaves are totally geodesic in M .
(2) Fibers are totally umbilical submanifolds with parallel mean curvature
vector.
Proof. Follows immediately from Proposition 3.6.

Another immediate consequence of Proposition 3.6 is the following.
Corollary 3.5. Every pseudo-Riemannian n-manifold N can be isometri-

cally embedded in some (n + p)-dimensional pseudo-Riemannian manifold
M with arbitrary p ≥ 1 as an extrinsic sphere.
This corollary shows that there do not exist geometric or topological

obstructions for a Riemannian manifold to be isometrically embedded as
an extrinsic sphere in some pseudo-Riemannian manifold.
The following two results characterize warped products among twisted
product manifolds.
Theorem 3.14. [Fernández-López et al. (2001)] Let B ×b F be the twisted

product of pseudo-Riemannian manifolds (B, gB ) and (F, gF ) with a twist-
ing function b and dim F > 1. Then Ric(X, V ) = 0 for all X ∈ L(B) and
V ∈ L(F ) if and only if B ×b F can be expressed as a warped product of
(B, gB ) and (F, gF ), where gF is a metric tensor conformal to gF .
Theorem 3.15. [Kazan and Sahin (2013)] Let B×b F be the twisted product
of pseudo-Riemannian manifolds with dim B > 1 and dim F > 1. Then
B ×b F can be expressed as a warped product of (B, gB ) and (F, gF ) if and
only if B is conformally flat along F , where gF is a metric tensor conformal
to gF .
The following result was proved by applying Theorem 3.14.
Theorem 3.16. [Brozos-Vázquez et al. (2005)] Let B ×f F be a pseudo-

Riemannian twisted product with dim B ≥ 2 and dim F ≥ 2. If B ×f F is
conformally flat, then it can be expressed as a warped product.
Definition 3.16. A nowhere zero vector field T on a (pseudo) Riemannian

manifold satisfying the following two conditions
∇X T = ϕX + α(X)T and α(T ) = 0, (3.75)
is called a torqued vector field (cf. [Chen (2017a)]).
The following result was obtained in [Chen (2017a)].
Theorem 3.17. A Riemannian n-manifold M admits a torqued vector field

if and only if it is locally a twisted product I ×λ F , where I is an open
interval, F is a Riemannian (n−1)-manifold and λ is the twisting function.
Proof. Assume that T is a torqued vector field on a Riemannian n-

manifold M . Let ρ = |T |. Then we have
T = ρe1 , (3.76)
where e1 is a unit vector field on M . It follows (3.75) and (3.76) that
ρϕe1 = ∇T T = (T ρ)e1 + ρ2 ∇e1 e1 . (3.77)
Since ∇e1 e1 is perpendicular to e1 , we find
∇e1 e1 = 0, T (ln ρ) = ϕ. (3.78)
The first equation in (3.78) shows that the integrable curves of e1 are
geodesics in M . Thus, if we put D = Span {e1 }, then D is a totally
geodesic foliation, i.e., D is an integrable distribution whose leaves are
totally geodesic in M . We may extend the unit vector field e1 to a local
orthonormal frame e1 , . . . , en on M . If we put
n
X
∇ej ei = ωik (ej )ek , i = 1, . . . , n, (3.79)
k=1
then we have ωik = −ωki .

Let us put D⊥ = Span {e2 , . . . , en }. Then we derive from (3.75) and
(3.76) that
ϕej + α(ej )ρe1 = ∇ej T = (ej ρ)e1 + ρ∇ej e1 . (3.80)
for j = 2, . . . , n.
We find from (3.79) and (3.80) that

ϕ
ω1k (ej ) = δjk , j, k = 2, . . . , n, (3.81)
ρ
α(ej ) = ej (ln ρ), j = 2, . . . , n. (3.82)
Also, (3.75) and (3.76) give
α(e1 ) = 0. (3.83)
⊥
Equation (3.81) implies that D is an integrable distribution whose
leaves are totally umbilical hypersurfaces of M . Therefore, it follows from
a result of [Ponge and Reckziegel (1993)] that M is locally a twisted product
I ×λ F , where I is an open interval, F is a Riemannian (n − 1)-manifold
and λ is a positive function on I × F , so that the metric tensor g of M
takes the form
g = ds2 + λ2 gF , (3.84)
with e1 = ∂/∂s. It follows from (3.76) and (3.78) that the torqued function
of T satisfies ϕ = ∂ρ/∂s. Also, it follows from (3.82) and (3.83) that the
torqued form α is the dual 1-form of dπF (∇(ln f )), where πF : I ×λ F → F
is the natural projection and ∇(ln f ) is the gradient of ln f .
Conversely, suppose that M is the twisted product I ×λ F of an open
interval and a Riemannian (n − 1)-manifold F so that the metric of M is
given by (3.84). Then we have

∂ ∂ ∂ ln λ
∇∂ = 0, ∇V = V, (3.85)
∂s ∂s ∂s ∂s
for V tangent to F . Let us put
∂
v=λ . (3.86)
∂s
Then it follows from (3.85) and (3.86) that

∂λ ∂
∇∂v= (3.87)
∂s ∂s ∂s

∂ ∂λ ∂
∇V v = (V λ) + V, V ⊥ . (3.88)
∂s ∂s ∂s
Now, let us define a scalar function ϕ on M by
∂λ
ϕ= (3.89)
∂s
and define a 1-form
α on M by
∂ ∂
α = 0 and α(V ) = V (ln λ) if V ⊥ . (3.90)
∂s ∂s
Then we obtain from (3.86)-(3.90) that ∇X v = ϕX + α(X)v with α(v) = 0
for all X ∈ T M . Therefore the twisted product I ×λ F admits a torqued
∂
vector field given by v = λ ∂s .
3.10 Characterizations of twisted products
The notion of twisted products was extended to the notion of doubly twisted
products in [Ponge and Reckziegel (1993)].
Definition 3.17. Let M1 and M2 be two pseudo-Riemannian manifolds

endowed with pseudo-Riemannian metrics g1 and g2 , respectively. If f1 , f2
are two positive functions in F (M1 × M2 ) and πi : M → Mi the canonical
projection for i = 1, 2. Then the doubly twisted product M1 ×(f1 ,f2 ) M2 of
(M1 , g1 ) and (M2 , g2 ) is the manifold M1 × M2 equipped with the pseudo-
Riemannian metric g defined by
g(X, Y ) = f12 · g1 (π1∗ X, π1∗ Y ) + f22 · g2 (π2∗ (X), π2∗ Y ) (3.91)
for tangent vectors X, Y ∈ T (M1 × M2 ).
The following two results were proved in [Ponge and Reckziegel (1993)].
Theorem 3.18. Let g be a pseudo-Riemannian metric on M1 ×M2 . If the

canonical foliations L1 and L2 intersect perpendicularly everywhere, then g
is the metric of
(a) a doubly twisted product M1 ×(f1 ,f2 ) M2 if and only if L1 and L2 are
totally umbilical foliations;
(b) a twisted product M1 ×f M2 if and only if L1 is a totally geodesic and
L2 a totally umbilical foliation;
(c) a warped product M1 ×f M2 if and only if L1 is a totally geodesic and
L2 a spheric foliation;
(d) a direct product of pseudo-Riemannian manifolds if and only if L1 and
L2 are totally geodesic foliations.
Theorem 3.19. Let (M, g) be a simply-connected pseudo-Riemannian

manifold which admits two complementary foliations L and K whose leaves
intersect perpendicularly. If L is totally geodesic and K is totally umbilical,
then (M, g) is isometric to a twisted product B ×f F such that L and K
correspond to the canonical foliations of the product B × F .
The connection and the curvature tensor of a doubly twisted product

M1 ×(f1 ,f2 ) M2 can be expressed in terms of the functions f1 and f2 and
the connections and the curvature tensors of M1 , M2 .
Proposition 3.7. [Ponge and Reckziegel (1993)] Let (M1 , g1 ) and (M2 , g2 )
be two pseudo-Riemannian manifolds and g the metric of doubly twisted
product M1 ×(f1 ,f2 ) M2 . Put Ui = −∇(ln fi2 ), where the gradient of ln fi2
is calculated with respect to g. Then the Levi-Civita connection ∇ and
curvature tensor R of the doubly twisted product M1 ×(f1 ,f2 ) M2 is related
to the Levi-Civita connection ∇ ˜ and curvature tensor R of the direct product
of (M1 , g1 ) and (M2 , g2 ) by
X
∇X Y = ∇ ˜ XY + {g(Pi X, Pi Y )Ui − g(X, Ui )Pi Y − g(Y, Ui )Pi X}
i
X
R(X, Y ) = R̃(X, Y ) + {(∇X Ui − g(X, Ui )Ui ) ∧ Pi Y
X
− (∇Y Ui − g(Y, Ui )Ui ) ∧ Pi X} + g(Ui , Uj )Pi X ∧ Pj Y,
i,j
where u ∧ v is the linear map w 7→ g(v, w)u − g(u, w)v for all u, v ∈ Tx M
and Pi : T M → ζi is the vector bundle projection related to the splitting
T M = ζ1 ⊕ ζ2 with ζ1 = ker(π2∗ ) and ζ2 = ker(π1∗ ).
In particular, we have the following results (see [Fernández-López et al.

(2001)]).
Corollary 3.6. For a doubly twisted product M1 ×(f1 ,f2 ) M2 , we have

∇X Y = ∇M
X Y + X(ln f1 )Y + Y (ln f1 )X − g(X, Y )∇(ln f1 ),
1
(3.92)
∇X V = V (ln f1 )X + X(ln f2 )V, (3.93)
M1
for X, Y ∈ L(M1 ) and V ∈ L(M2 ), where ∇ is the connection of M1 .
Corollary 3.7. The Ricci tensor Ric of M1 ×(f1 ,f2 ) M2 satisfies

f1
Ric(X, Y ) = RicM1 (X, Y ) + Hf1 (X, Y ) + (1 − n1 )HM 1
(X, Y ) (3.94)
+ n1 X(ln f1 )Y (ln f1 ) − g(X, Y ){∆(ln f1 )
f2
+ g(∇(ln f1 ), ∇(ln f1 ))} − n2 HM 1
(X, Y )

+ X(ln f2 )Y (ln f2 ) − X(ln f1 )Y (ln f2 ) − X(ln f2 )Y (ln f1 )
Ric(X, V ) = (1 − n1 )V X(ln f1 ) + (1 − n2 )XV (ln f2 ) (3.95)
+ (n1 + n2 − 2)X(ln f2 )V (ln f1 ),
for X, Y ∈ L(M1 ) and V ∈ L(M2 ), where n1 = dim M1 , n2 = dim M2 , ∆f
is the Laplacian of f and
fi
HM 1
(X, Y ) = XY (ln fi ) − (∇M
X Y )(ln fi ), i = 1, 2,
1
(3.96)
f1
Hf1 (X, Y ) = HM 1
(X, Y ) − X(ln f1 )Y (ln f2 ) − X(ln f2 )Y (ln f1 ) (3.97)
+ g(X, Y )g(∇(ln f1 ), ∇(ln f2 )).
Theorem 3.20. [Fernández-López et al. (2001)] Let B ×λ F be a twisted

product of (B, gB ) and (F, gF ) with dim F > 1. Then Ric(X, V ) = 0 for
all X ∈ L(B) and V ∈ L(F ) if and only if B ×λ F can be expressed as a
warped product B ×φ F̃ of (B, gB ) and (F, g̃F ) with a warping function ϕ,
where g̃F is a conformal metric tensor to (F, gF ).
Proof. Under the hypothesis, if Ric(X, V ) = 0 holds for all X ∈ L(B)

and V ∈ L(F ), then (3.95) gives 0 = XV (ln λ). Thus V (ln λ) only depends
on F , and likely V X(ln λ) = 0 implies that X ln(λ) only depends on B.
Thus ln λ can be expressed as the sum φ(p) + ψ(q) for any (p, q) ∈ B × F .
Hence λ = Φ(p)Ψ(q) with Φ = exp φ and Ψ = exp ψ. Consequently, B ×λ F
can be expressed as a warped product B ×φ F̃ of (B, gB ) and (F, g̃F ) with
a warping function ϕ, where g̃F is a conformal metric tensor to (F, gF ).
The following result of [Fernández-López et al. (2001)] is an immediate
consequence of Theorem 3.20.
Corollary 3.8. Let B ×λ F be a twisted product of (B, gB ) and (F, gF )

with dim F > 1. If B ×λ F is Einstein, then it can be expressed as a warped
product B ×φ F̃ of (B, gB ) and (F, g̃F ) with a warping function as Theorem
3.20.
3.11 Convolution manifolds
The notion of convolution of two Riemannian manifolds (or more generally,

of pseudo-Riemannian manifolds) was introduced in [Chen (2002g)]. Such
notion arises naturally from tensor product immersions studied in [Chen
(1993a); Decruyenaere et al. (1994)]. This notion provides another natural
extension of warped products.
Definition 3.18. [Chen (2003e)] Let (N1 , g1 ) and (N2 , g2 ) be two pseudo-
Riemannian manifolds and let f ∈ F (N1 ) and h ∈ F (N2 ). The symmetric
tensor h g1 ∗ f g2 on N1 × N1 defined by
h g1 ∗ f g2 = h2 g1 + f 2 g2 + 2f hdf ⊗ dh, (3.98)

is called the convolution of g1 and g2 via h and f . The product manifold
N1 ×N2 together with h g1 ∗f g2 , denoted by h N1 ⋆ f N2 , is called a convolution
of (N1 , g1 ) and (N2 , g2 ). If h g1 ∗f g2 is non-degenerate, it defines a pseudo-
Riemannian metric on N1 × N2 , which is called a convolution metric.
When f and h are irrelevant, we simply denote h N1 ⋆ f N2 and h g1 ∗f g2

by N1 ⋆ N2 and g1 ∗ g2 , respectively.
Remark 3.4. If either f = 1 or h = 1, then (3.98) is nothing but a warped

product metric.
We put Em m n n
∗ = E − {0} and Cs∗ = Cs − {0} for 0 ≤ s < n.
The next result from [Chen (2003e)] shows that the notion of convolution
manifolds arises very naturally from tensor product immersions.
Proposition 3.8. Let φ : (N1 , g1 ) → En∗ ⊂ En and ϕ : (N2 , g2 ) → Em

∗ ⊂
Em be isometric immersions. Then the map
ψ = φ ⊗ ϕ : N1 × N2 → En ⊗ Em = Enm (3.99)
defined by ψ(u, v) = φ(u)⊗ϕ(v) gives rise to a convolution manifold N1 ⋆ N2
with the convolution metric
ρ2 g1 ∗ ρ1 g2 = ρ22 g1 + ρ21 g2 + 2ρ1 ρ2 dρ1 ⊗ dρ2 , (3.100)
whenever (3.100) is non-degenerate, where ρ1 = |φ| and ρ2 = |ϕ|.
Proof. For vector fields X, Y ∈ X(N1 ) and Z, W ∈ X(N2 ), we have

dψ(X) = Xψ = X ⊗ ϕ, dψ(Z) = Zψ = φ ⊗ Z. (3.101)
Also, it follows from the definitions of gradient of ρ1 = |φ| that
1
hX, φi = Xhφ, φi = ρ1 (Xρ1 ) = ρ1 dρ1 (X). (3.102)
2
Similarly, we get
hZ, ϕi = ρ2 dρ2 (Z). (3.103)
Now, the proposition follows from (3.101), (3.102) and (3.103).
Example 3.3. Let φ : (N1 , g1 ) → En∗ ⊂ En be an isometric immersion. If

ϕ : (N2 , g2 ) → Em is an isometric immersion such that ϕ(N2 ) lies in the
unit hypersphere S m−1 (1) centered at the origin. Then the convolution
g1 ∗ g2 is the warped product metric g = g1 + |φ|2 g2 .
Proposition 3.9. Let h N1 ⋆ f N2 be the convolution of two Riemannian

manifolds (N1 , g1 ) and (N2 , g2 ) via h and f . Then h g1 ∗f g2 is degenerate if
and only if the following two conditions hold:
(1) The length |∇f | of the gradient of f on (N1 , g1 ) is a nonzero constant,
say c.
(2) The length |∇h| of the gradient of h on (N2 , g2 ) is c−1 .

Proof. By a direct computation we have

det(h g1 ∗ f g2 ) = f 2n1 h2n2 1 − |∇f |2 |∇h|2 , (3.104)
where n1 and n2 denote the dimensions of N1 and N2 , respectively. Thus
the convolution h g1 ∗ f g2 is degenerate if and only if |∇f | · |∇h| = 1.
Since |∇f | and |∇h| depend only on N1 and N2 , respectively, it follows
that |∇f | · |∇f | = 1 holds identically if and only if statements (1) and (2)
of this proposition hold.
Proposition 3.9 implies immediately the following.
Corollary 3.9. Let h N1 ⋆ f N2 be the convolution of Riemannian manifolds
(N1 , g1 ) and (N2 , g2 ) via h and f . Then h g1 ∗ f g2 is a Riemannian metric
on h N1 ⋆ f N2 if and only if |∇f | · |∇f | < 1 holds.
Let x : M → En∗ be an isometric immersion. We decompose the position
vector x of M into x = xT + x⊥ , where xT and x⊥ are the tangential and
normal components.
In views of Propositions 3.8 and 3.9, we give the following.
Proposition 3.10. Let x : M → En be an isometric immersion. Then the
distance function ρ = |x| satisfies |∇ρ| = c for some constant c if and only
if |xT | = c|x|. In particular, if |∇ρ| = c holds, then c ∈ [0, 1].
Proof. Let e1 , . . . , en be an orthonormal frame field on M . Then the
Pn
gradient of ρ is given by ∇ρ = j=1 (ej ρ)ej .
Since ej ρ = hej , xi /|x|, we find
Xn 2
2 hej , xi
|∇ρ| = .
j=1
|x|2
Thus |∇ρ| = c holds identically if and only if |xT | = c|x| holds.
Remark 3.5. Euclidean submanifolds whose position vector field satisfying
|xT | = c|x| for some constant c are called submanifolds of constant ratio.
Such submanifolds were studied initially in [Chen (2001e, 2002i, 2003f)]. For
a nice link between constant ratio submanifolds and biology via D’Arcy
Thompson’s work on “On Growth and Form”, see [Haesen et al. (2012);
Chen (2017b)].
Remark 3.6. For further results on convolutions and their applications,
see [Chen (2001e, 2002b,g,h, 2003e,f)].
Chapter 4
Robertson-Walker Spacetimes and

Schwarzschild Solution
Astronomers were unsure of the size of our galaxy at the beginning of the
20th century. Generally, they believed it was not much greater than a few
tens of thousands of light years across. However, astronomers had noted
fuzzy patches of light in the night sky, which are called nebulae. Some
astronomers thought these could be distant galaxies.
E. Hubble observed in the 1920s that some of these nebulae were indeed
distant galaxies comparable in size to our own Milky Way. Hubble also
made the remarkable discovery that our Universe is expanding. Hubble’s
observations that the light from nebulae showed a red shift increasing with
distance ruled out the possibility that Einstein’s static model1 as well as de
Sitter’s static model without matter represented the real universe. A new
estimate of the mass of our galaxy made in 1927 caused de Sitter to re-
examine his assumption. At a meeting in London of the Royal Astronomical
Society in early 1930, de Sitter admitted that neither his nor Einstein’s
solution to the field equations could represent the observed universe.
In fact, a few astronomers had been looking for other solutions to Ein-
stein’s field equations. Back in 1922 A. Friedmann had published a set of
possible mathematical solutions that gave a non-static universe. In 1927, G.
Lemaı̂tre arrived independently at similar results as Friedmann. During the
1930s, H. Robertson and A. G. Walker explored the problem further. They
rigorously proved in 1935 that the Robertson-Walker metric is the only one
on a spacetime that is spatially homogeneous (all places look the same) and
isotropic (all spatial direction the same). This is the birth of Robertson-
Walker spacetimes. In the last section we discuss Schwarzschild’s solution
to Einstein’s field equations briefly.
1 A static model is a cosmological model in which the universe is both spatially infinite
and temporally infinite, and space is neither expanding nor contracting.
81
4.1 Basic properties of Robertson-Walker spacetimes
Astronomical evidence indicates the universe can be modeled as a space-

time containing a perfect fluid whose “molecules” are the galaxies. The
decisive fact that no large asymmetriy has been observed in the distribu-
tions of the galaxies. At large scale appropriate to cosmology, the universe
looks the same in all direction which is the evidence that the universe is
statistically homogeneous and isotropic. Thus it is possible to build a sim-
ple cosmological model with reasonable chance of being physically realistic.
Robertson-Walker’s spacetimes are such relativistic models.
In general relativity, a Robertson-Walker spacetime is a warped product
L41 (k, f ) := (I × R3 (k), g), g = −dt2 + f 2 (t)gk , (4.1)
3
of an open interval I and a Riemannian 3-manifold (R (k), gk ) of constant
curvature k, while the warping function f describes the expanding or con-
tracting of our Universe.
In the following we consider a Robertson-Walker spacetime as a warped
product
Lm
1 (k, f ) := (I × R
m−1
(k), g), g = −dt2 + f 2 (t)gk , (4.2)
m−1
of an open interval I and a real space form (R (k), gk ), where m is any
integer ≥ 2. Let ∂t denote the first coordinate vector field on Lm 1 (k, f ),
known as the comoving observer field in general relativity.
A Robertson-Walker spacetime possesses two relevant geometrical fea-
tures. On one hand, its fibers have constant curvature. Thus the spacetime
is spatially homogeneous. On the other hand, it has a timelike vector field
K = f (t)∂t which satisfies ∇X K = f ′ (t)X for any X. In particular, we
have LK g = 2f ′ g by (1.28), where LK is the Lie derivative along K. Hence
K is a conformal vector field. These properties of K show a certain sym-
metry of the spacetime metric on Lm 1 (k, f ).
By a rest space or a spacelike slice in a Lorentzian manifold, we mean a
spacelike hypersurface given by t constant. Hence a rest space in Lm−11 (k, f )
is a fiber
S(t0 ) := {t0 } ×f (t0 ) Rm−1 (k), t0 ∈ I.
Thus a rest space S(t0 ) in Lm−11 (k, f ) is a manifold of constant curvature
whose metric tensor is f 2 (t0 )gk .
Definition 4.1. A pseudo-Riemannian submanifold N of a Robertson-

Walker spacetime Lm 1 (k, f ) is called transverse if it is contained in a rest
space S(t0 ) for some t0 ∈ I.
Robertson-Walker Spacetimes and Schwarzschild Solution 83
Definition 4.2. A pseudo-Riemannian submanifold N of Lm 1 (k, f ) is called

a H-submanifold if the comoving observer field ∂t is tangent to N at each
point on N .
For a tangent vector X of a Robertson-Walker spacetime Lm 1 (k, f ), we
decompose
X = ϕX ∂t + X − , (4.3)
where ϕX = − hX, ∂t i and X − is the vertical component of X.
The following two lemmas follow immediately from Proposition 3.1 and
Proposition 3.2.
Lemma 4.1. For V, W ∈ L(Rm−1 (k)), we have
(1) ∇∂t ∂t = 0;
(2) ∇∂t V = ∇V ∂t = (ln f )′ V ;
(3) h ∇V W, ∂t i = − hV, W i (ln f )′ ;
(4) (∇V W )− is the lift of ∇′V W on Rm−1 (k),
where ∇′ is the Levi-Civita connection on Rm−1 (k).

Lemma 4.2. The curvature tensor R of Lm
1 (k, f ) satisfies
f ′′
(1) R(∂t , V )∂t = V;
f
f ′′
(2) R(V, ∂t )W = − hV, W i ∂t ;
f
(3) R(V, W )∂t = 0;
k + f ′2
(4) R(U, V )W = {hV, W i U − hU, W i V }
f2
for U, V, W ∈ L(Rm−1 (k)).

Corollary 4.1. A Robertson-Walker spacetime Lm 1 (k, f ) is of constant cur-
vature if and only if the warping function f satisfies f f ′′ = f ′2 + k.
Proof. Assume Lm 1 (k, f ) is of constant curvature. Then it follows from
Lemma 4.2(1) that the sectional curvature of Lm ′′
1 (k, f ) is f /f .
On the other hand, it follows from Lemma 4.2(4) that the sectional
curvature is also equal to (k + f ′2 )/f 2 . Therefore the warping function
satisfies f f ′′ = f ′2 + k.
The converse can be verified by direct computation.
Remark 4.1. It follows from Corollary 4.1 that

(a) Lm 2
1 (k, f ) is flat if and only if f (t) = at + b, with k = −a ;
(b) Lm 2
1 (k, f ) has constant curvature c > 0 if and only if the warping
function is f (t) = a cosh(ct) + b sinh(ct) with k = c2 (a2 − b2 );
(c) Lm 2
1 (k, f ) has constant curvature −c < 0 if and only if the warping
function is f (t) = a sin(ct) + b cos(ct) with k = −c2 (a2 + b2 ).
Lemma 4.3. Let N be a pseudo-Riemannian submanifold of Lm 1 (k, f ).
Then for u, v, w ∈ Tx N, x ∈ N and ξ, η ∈ Tx⊥ (N ) we have:
n o
k
(1) (R(u, v)w)⊥ = − (ln f )′′
(hu, wi ϕv − hv, wi ϕu )∂t⊥ ;
2 f
(2) (R(u, v)ξ)⊥ = 0;
(3) the Ricci equation in Lm
1 (k, f ) is given by

D
R (u, v)ξ, η = h[Aξ , Aη ]u, vi ;
(4) the sectional curvature K(u ∧ v) of Lm 1 (k, f ) with respect to the plane
section spanned by two orthonormal vectors u, v ∈ Tx (N ) is given by

k + f ′2 k
K(u ∧ v) = + (ǫu ϕ2u + ǫv ϕ2v ) 2 − (ln f )′′ ,
2 f f
⊥
where ( · ) is the normal component of ( · ), ǫu = hu, ui and ǫv = hv, vi.
Proof. Let e1 , . . . , en be a local orthonormal frame field on N . From
(4.3) we find
ej = ϕj ∂t + êj , j = 1, . . . , n, (4.4)
where ϕj = − hej , ∂t i, 1 ≤ j ≤ n. Then we have
hêi , êj i = ǫj δij − ϕi ϕj , ǫj = hej , ej i , i, j = 1, . . . , n. (4.5)
It follows from Lemma 4.2, (4.4) and (4.5) that

f ′′ f ′2 +k
R(ei , ej )ek = ϕi ϕk + (ϕ ϕ
i k − ǫ δ
k ik ) êj
f2 f
n o (4.6)
f ′′ f ′2 +k f ′′
− ϕj ϕk + (ϕj ϕk − ǫ k δ jk ) êi − (δik ϕj −δjk ϕi )ǫk ∂t .
f 2 f f
Thus, by applying (4.4) and (4.6), we find

k
(R(ei , ej )ek )⊥ = (δik ϕj − δjk ϕi )ǫk − (ln f )′′ ∂t⊥ , (4.7)
f2
which implies (1) by linearity. By applying Lemma 4.2(4) we find
f f ′′ + f ′2 + k
R(u, v)ξ = (ϕu v̂ − ϕv û)ϕξ . (4.8)
f2
Combining this with û = u − ϕu ∂t and v̂ = v − ϕv ∂t gives (2). (3) follows
from (2) and Ricci’s equation. Finally, (4) follows from (4.6).
For an H-submanifold N , we put D⊥ = {X ∈ X(N ) : X ⊥ ∂t }.
Lemma 4.4. Let N be an H-submanifold of Rm (k, f ). We have:
(1) the Levi-Civita connection ∇′ of N satisfies

∇′∂t ∂t = 0, ∇′X Y = (∇X Y )− + hX, Y i (ln f )′ ∂t (4.9)
⊥ −
for vector fields X, Y in D , where (∇X Y ) is the vertical component
of ∇X Y ;
(2) the distribution D⊥ is integrable;
(3) locally, N is a warped product I ′ ×fˆ P n−1 , where I ′ is an open subin-
terval of I, P n−1 is a submanifold of Rm−1 (k) and fˆ = f |I ′ ;
(4) the second fundamental form σ of N in Lm
1 (k, f ) satisfies
(4.a) σ(∂t , ∂t ) = σ(∂t , X) = 0, and
(4.b) σ(X̄, Ȳ ) is the lift of σ S (X, Y ) for X, Y ∈ T (P n−1 ), where hS is
the second fundamental form of P n−1 in Rm−1 (k) and X̄, Ȳ are the lift
of X, Y to I ′ ×fˆ N n−1 .
Proof. (1) follows immediately from Lemma 4.1. From (4.3) we get
[X, Y ] = [X, Y ]− ∈ D⊥ . Thus, we obtain (2) by Frobenius’ theorem.
For (3) let us observe that Lemma 4.1(1) implies that the rank-one
distribution spanned by ∂t on N is a totally geodesic distribution of N .
Moreover, it follows from (1) that the integral manifolds of D⊥ are totally
umbilical hypersurfaces of N with constant mean curvature. Thus, Hiepko’s
theorem implies that N is locally the warped product of an open subinterval
I ′ ⊂ I and an integral manifold P n−1 of D⊥ with respect to the warping
function fˆ = f |I ′ .
Since P n−1 is perpendicular to I ′ ⊂ I, P n−1 lies in some rest space
S(t0 ), t0 ∈ I ′ . Without loss of generality, we may assume that P n−1 is a
submanifold of Rm−1 (k). This proves statement (3).
Statement (4) follows from statement (3) and Lemma 4.1.
Lemma 4.5. If N is a transverse submanifold of Lm

1 (k, f ), then
(1) the second fundamental form σ of N in Lm

1 (k, f ) and σ
S(t0 )
of N in
S(t0 ) are related by
σ(X, Y ) = σ S(t0 ) (X, Y ) + hX, Y i (ln f )′ ∂t , X, Y ∈ X(N );
(2) the normal connection D of N in Rm (k, f ) and DS(t0 ) of N in S(t0 )
S(t )
satisfy DX ∂t = 0 and DX ξ = DX 0 ξ for every X ∈ X(N ) and ξ
orthogonal to ∂t ;
S(t0 )
(3) the normal curvature tensor RD of N in Rm (k, f ) and RD of N
D DS(t0 )
in S(t0 ) satisfy R (X, Y )ξ = R (X, Y )ξ for X, Y ∈ X(N ) and ξ
orthogonal to ∂t .
Proof. Statements (1) and (2) follow from Lemma 4.1; and statement
(3) is an easy consequence of statement (2).
The next two corollaries follow immediately from Lemma 4.5(1).
Corollary 4.2. A transverse submanifold N of a Robertson-Walker space-

time Lm 1 (k, f ) is non-totally geodesic unless N lies in a rest space S(t0 )
′
with f (t0 ) = 0 as a totally geodesic submanifold.
Corollary 4.3. If a transverse submanifold of a Robertson-Walker space-

time Lm 1 (k, f ) is totally umbilical, then it lies in a rest space S(t0 ) as a
totally umbilical submanifold.
4.2 Totally geodesic submanifolds of Robertson-Walker

spacetimes
Definition 4.3. A submanifold N of a pseudo-Riemannian manifold M is

called curvature-invariant if the curvature tensor R of M satisfies
R(u, v)(Tx N ) ⊂ Tx N
for all u, v ∈ Tx N at each point x ∈ N .
Lemma 4.6. Assume that Lm 1 (k, f ) = I ×f R

m−1
(k) contains no open sub-
sets of constant curvature. Then a curvature-invariant pseudo-Riemannian
submanifold N with dim N ≥ 2 in Lm 1 (k, f ) is (i) a transverse submanifold
of L1 (k, f ) or (ii) an H-submanifold of Lm
m
1 (k, f ).
Proof. Assume that Lm 1 (k, f ) contains no open subsets of constant cur-

vature. Then Corollary 4.1 implies that the warping function f satisfies
(ln f )′′ 6= k/f 2 on a dense subset of the open interval I.
Let N be a pseudo-Riemannian submanifold of Lm T
1 (k, f ) and let ∂t and
⊥
∂t be the tangential and the normal component of ∂t . If N is a curvature-
invariant submanifold, then Lemma 4.3 yields
(hu, wi hv, ∂t i − hv, wi hu, ∂t i)∂t⊥ = 0
for u, v, w ∈ Tx N, x ∈ N. Hence, at each point x ∈ N , ∂t is either normal
to N (i.e. ∂tT = 0) or ∂t is tangent to N (i.e. ∂t⊥ = 0); otherwise choosing
u = w ⊥ v = ∂tT 6= 0 in the above expression would lead to ∂t⊥ = 0 , a

contradiction. Thus, by continuity, ∂t is either normal to N at each point
on N or tangent to N at each point on N . Hence, we have either statement
(i) or statement (ii).
The following result classifies totally geodesic spacelike submanifolds of
a Robertson-Walker spacetime.
Proposition 4.1. If Lm 1 (k, f ) contains no open subsets of constant cur-

vature, then Lm 1 (k, f ) admits a spacelike totally geodesic submanifold of
dimension ≥ 2 if and only if f has a critical point.
Further, the only spacelike totally geodesic submanifolds of dimension
≥ 2 in Lm ′
1 (k, f ) are either rest spaces S(t0 ) with f (t0 ) = 0 or totally
geodesic submanifolds which lie in some rest spaces.
Proof. Let Lm 1 (k, f ) be a Robertson-Walker spacetime which contains no

open subsets of constant curvature. If N is a totally geodesic spacelike sub-
manifold, then it is curvature-invariant by the equation of Codazzi. Hence,
it follows from Lemma 4.6 that N is a transverse submanifold. Thus N lies
in a rest space S(t0 ) for some t0 . Since N is totally geodesic in Lm 1 (k, f ),
Corollary 4.2 implies f ′ (t0 ) = 0 and N is totally geodesic in S(t0 ).
The converse follows immediately from Lemma 4.5(1).
Proposition 4.2. Suppose that Lm 1 (k, f ) = I ×f R

m−1
(k) is a Robertson-
Walker spacetime which contains no open subsets of constant curvature. If
N is an n-dimensional totally geodesic Lorentzian submanifold of Lm 1 (k, f ),
then N is an open portion of a warped product I ′ ×fˆ P n−1 ⊂ I ×f Rm−1 (k),
where I ′ is an open subinterval of I, P n−1 is a totally geodesic submanifold
of Rm−1 (k) and fˆ = f |I ′ .
Proof. Under the hypothesis, N is curvature-invariant. Thus, by Lemma

4.6, N is an H-submanifold. Hence, ∂t is tangent to N at each point. Thus,
according to Lemma 4.4(3), N is an open portion of I ′ ×fˆ P n−1 for some
submanifold P n−1 of Rm−1 (k). Because N is totally geodesic in Lm1 (k, f ),
P n−1 is totally geodesic in Rm−1 (k).
4.3 Parallel submanifolds of Robertson-Walker spacetimes
Proposition 4.3. If a Robertson-Walker spacetime Lm 1 (k, f ) contains no

open subsets of constant curvature, then a pseudo-Riemannian submanifold
of Lm
1 (k, f ) is a parallel submanifold if and only if it is one of the following:
(a) A transverse submanifold lying in a rest space S(t0 ) of Lm

1 (k, f ) as a
parallel submanifold.
(b) An H-submanifold which is locally a warped product I ×f P n−1 , where
I is an open interval and P n−1 is a submanifold of Rm−1 (k). Further,
(b.1) if f ′ 6= 0 on I, then I ×f P n−1 is totally geodesic in Lm
1 (k, f );
(b.2) if f ′ = 0 on I, then P n−1 is a parallel submanifold of Rm−1 (k).
Proof. Assume that Lm 1 (k, f ) contains no open subsets of constant cur-

vature. Then, by Corollary 4.1, f satisfies (ln f )′′ 6= k/f 2 on I. Suppose
that N is a parallel submanifold, then Codazzi’s equation shows that N
is curvature-invariant. Hence, N is either a transverse submanifold or an
H-submanifold according to Lemma 4.6.
Case (1): N is a transverse submanifold. In this case, N lies in a rest space
S(t0 ) of Lm
1 (k, f ), t0 ∈ I. Let X, Y, Z be tangent to N . Then, Lemma 4.1
implies that the second fundamental form σ of N in Lm 1 (k, f ) is given by
σ(X, Y ) = σ̂(X, Y ) + hX, Y i (ln f )′ ∂t , (4.10)

where ĥ is the second fundamental form of N in S(t0 ). So, after applying
Lemma 4.1, we find
∇X (σ(Y, Z)) = ∇X (σ̂(Y, Z)) + h∇X Y, Zi (ln f )′ ∂t
+ hY, ∇X Zi (ln f )′ ∂t + hY, Zi (ln f )′2 X.
Thus DX (σ(Y, Z)) = DX σ̂(Y, Z) + (h∇′X Y, Zi + hY, ∇′X Zi)(ln f )′ ∂t , which
implies that
(∇¯ X σ)(Y, Z) = DX (σ̂(Y, Z) + hY, Zi (ln f )′ ∂t )
− σ(∇′X Y, Z) − σ(Y, ∇′X Z)
= DX (σ̂(Y, Z)) + h∇′X Y, Zi (ln f )′ ∂t − σ(∇′X Y, Z)
+ hY, ∇′X Zi (ln f )′ ∂t − σ(Y, ∇′X Z)
= DX (σ̂(Y, Z)) − σ̂(∇′X Y, Z) − σ̂(Y, ∇′X Z)
= DX (σ S(t) (Y, Z)) − σ S(t) (∇′X Y, Z) − σ S(t) (Y, ∇′X Z)
¯ X σ S(t) )(Y, Z).
= (∇
Hence, N is a parallel submanifold in Lm
1 (k, f ) if and only if N is a parallel
submanifold in the rest space S(t). This gives (a).
Case (2): N is an H-submanifold. We follow the same notation as Lemma
4.4. Since N is an H-submanifold, the second fundamental form σ of N in
Lm1 (k, f ) lies in the space of lifts of X(R

m−1
(k)). Moreover, according to
Lemma 4.4(3), locally N is a warped product I ′ ×f P n−1 , where I ′ is an
open subinterval of I and P n−1 is a submanifold of Rm−1 (k). Because N
is parallel, (∇ ¯ X σ)(Y, Z) = 0 for X, Y, Z ∈ T N . Thus
0 = (∇ ¯ ∂t σ)(Y, Z) = D∂t σ(Y, Z) − σ(∇′ Y, Z) − σ(Y, ∇′ Z) (4.11)
∂t ∂t
⊥
for Y, Z ∈ D . Now, by applying Lemma 4.1(2), we obtain
D∂t σ(Y, Z) = 2(ln f )′ σ(Y, Z). (4.12)
On the other hand, Weingarten’s formula and Lemma 4.1(2) give
− Aξ (∂t ) + D∂t ξ = ∇∂t ξ = (ln f )′ ξ (4.13)
for ξ normal to N , which yields
Aξ (∂t ) = 0, D∂t ξ = (ln f )′ ξ. (4.14)
Case (2.a): f ′ 6= 0 on I. Combining (4.12) and the second equation of

(4.14) gives σ(Y, Z) = 0, Y, Z ∈ D⊥ . Hence, by using the first equation of
(4.14), we find σ = 0. Thus, N is totally geodesic, which gives (b.1).
Case (2.b): f ′ = 0 on I. It follows from (4.14), Lemmas 4.1 and 4.4 that
∇′∂t ∂t = ∇′∂t X = ∇′X ∂t = D∂t ξ = 0,
(4.15)
∇′X Y = (∇X Y )− , σ(∂t , ∂t ) = σ(∂t , X) = 0
for X, Y ∈ D⊥ . Now, it follows from (4.11) and (4.15) that
(∇ ¯ X σ)(∂t , Y ) = (∇
¯ ∂t σ)( · , · ) = (∇ ¯ X σ)(∂t , ∂t ) = 0, (4.16)
¯ X σ)(Y, Z) = (∇
(∇ ¯ X σ )(Y, Z).
S
(4.17)
Therefore, after applying the assumption ∇σ¯ = 0, we conclude that P n−1
is a parallel submanifold of Rm−1 (k). This gives (b.2).
The converse can be verified easily.
Remark 4.2. Most results given in sections 4.2-4.4 are based on [Chen
and Van der Veken (2007); Chen and Wei (2008b)].
4.4 Totally umbilical submanifolds of Robertson-Walker

spacetimes
Proposition 4.4. Let Lm 1 (k, f ) be a Robertson-Walker spacetime which

contains no open subsets of constant curvature. Then a pseudo-Riemannian
submanifold of Lm 1 (k, f ) is totally umbilical with parallel mean curvature if
and only if it is one of the following two types of submanifolds:
(i) a totally geodesic H-submanifold of Lm1 (k, f ), or

(ii) a submanifold lying in a rest space S(t0 ) as a totally umbilical subman-
ifold with parallel mean curvature vector.
Proof. Under the hypothesis, assume that N is a totally umbilical sub-

manifold of Lm
1 (k, f ) with parallel mean curvature vector. Then the second
fundamental form σ of N satisfies
σ(X, Y ) = hX, Y i H, X, Y ∈ T N. (4.18)
Thus (∇ ¯ X σ)(Y, Z) = hY, Zi DX H = 0 for X, Y, Z ∈ X(N ), which implies
that N is a parallel submanifold. Hence, by Proposition 4.3, N is either a
transverse submanifold or an H-submanifold.
If N is an H-submanifold, then it follows from Lemma 4.5(1) that
σ(∂t , ∂t ) = 0. Combining this with the total umbilicity of N implies that
M is totally geodesic. This gives (i). If N is a transverse submanifold of
Lm1 (k, f ), it lies in a rest space S(t0 ). Thus, by Lemma 4.5, we have
σ(X, Y ) = σ S(t0 ) (X, Y ) + hX, Y i (ln f )′ ∂t .

Thus N is totally umbilical in S(t0 ). We also have H = H S(t0 ) + (ln f )′ ∂t ,
where H S(t0 ) is the mean curvature vector of N in S(t0 ). Therefore we find
σ S(t0 ) (X, Y ) = hX, Y i H S(t0 ) .
By applying Lemma 4.1, we get DX H = DX H S(t0 ) . Thus, N is totally
umbilical with parallel mean curvature vector in S(t0 ). This gives (ii).
The converse can be easily verified.
Proposition 4.5. If a Robertson-Walker spacetime Lm 1 (k, f ) contains no

open subsets of constant curvature, then a pseudo-Riemannian submanifold
N of Lm 1 (k, f ) is totally umbilical with constant mean curvature if and only
if either
(i) N lies in a rest space S(t0 ) as a totally umbilical submanifold with

parallel mean curvature vector, or
(ii) N is a totally umbilical submanifold of Lm 1 (k, f ) with vertical mean
curvature vector field, i.e., H ⊥ ∂t .
Proof. Under the hypothesis on Lm 1 (k, f ), if N is totally umbilical in

m ¯
L1 (k, f ), then we have (∇X σ)(Y, Z) = hY, Zi DX H. Thus, for orthonormal
vectors X, Y , we obtain (∇ ¯ X σ)(Y, Y ) − (∇ ¯ Y σ)(X, Y ) = DX H. Hence it
follows from the equation of Codazzi and Lemma 4.3(1) that

k ′′
DX H = − (ln f ) hY, Y i hX, ∂t i ∂t⊥ (4.19)
2 f
for any X ∈ X(N ). Hence we find

k
X hH, Hi = 2 2 − (ln f )′′ hY, Y i hX, ∂t i h∂t , Hi .
f
Thus if hH, Hi is constant, then hX, ∂t i h∂t , Hi = 0 holds at each point on
N . Thus if we put U = {p ∈ M : h∂t , Hi = 6 0 at p}, then U is an open
subset of N ; moreover, ∂t is perpendicular to U .
If U = ∅, we get (ii). If U = N , then N is transverse. So, it lies in a rest
space, say S(to ). Because N is totally umbilical in a real space form S(to ),
the mean curvature vector is a parallel normal vector field. This gives (i).
If U is neither empty nor N , then each connected component U o of U
is a transverse submanifold. Thus U o lies in a rest space S(to ). Hence, by
o
Lemma 4.5(1), the mean curvature vector H S(t ) of U o in S(to ) satisfies
o
H = H S(t ) + (ln f )′ ∂t . (4.20)
Obviously, the value of (ln f )′ on U o is the constant (ln f )′ (to ).
On the other hand, since the mean curvature vector H on M − U is
perpendicular to ∂t , (4.20) implies that (ln f )′ = 0 on the boundary of
M − U . Thus, by the continuity
of H, (ln f)′ (to ) is zero, which leads to a
o
contradiction, namely, h∂t , Hi = ∂t , H S(t ) = 0 on U o .
Contrast to totally umbilical submanifolds in Rm (c), there exist totally
umbilical submanifolds in Lm
1 (k, f ) with non-constant mean curvature.
Example 4.1. Let f (t) be a positive function with f ′′ > 0 on an open

interval I ∋ 0 and (t, x2 , . . . , xm ) a natural coordinate system of I × Em−1 .
The metric tensor of Lm 1 (0, f ) = I ×f E
m−1
is
Xn
g̃ = −dt2 + f 2 (t) dx2j . (4.21)
j=2
n−1
Consider the immersion φ : I × R → Lm
1 (0, f ) defined by
Z s
dt
φ(s, u2 , . . . , un ) = s, b , u2 , . . . , un , 0, . . . , 0 , R ∋ b > 1. (4.22)
0 f (t)
Then
(j+1)-th
z}|{ (4.23)
−1
φs = (1, bf (s) , 0, . . . , 0), φuj = (0, . . . , 0, 1 , 0, . . . , 0),
for j = 2, . . . , n. The metric tensor induced from (4.21) via φ is
Xn
g = (b2 − 1)ds2 + f 2 (s) du2j , (4.24)
j=2
and
r-th
1 z}|{
ξ1 = √ 2 b, f −1 , 0, . . . , 0 , ξr = f −1 0, . . . , 0, 1 , 0, . . . , 0 ,
b −1
r = n + 2, . . . , m,
are orthonormal normal vector fields to N in Lm 1 (0, f ). A straightforward
computation shows that the second fundamental form of φ satisfies
√
b b2 − 1f ′
σ(∂s , ∂s ) = ξ1 , σ(∂s , ∂ui ) = 0, σ(∂ui , ∂uj ) = 0,
f
(4.25)
bf f ′
σ(∂u2 , ∂u2 ) = · · · = σ(∂un , ∂un ) = √ ξ1 ,
b2 − 1
for 2 ≤ i 6= j ≤ n. Hence φ is a totally umbilical immersion such that
bf ′
σ(X, Y ) = √ hX, Y i ξ1 . (4.26)
b2 − 1f
Since f ′′ is nowhere zero, (4.26) implies that φ is totally umbilical with
non-constant mean curvature. Therefore, we have DH 6= 0.
Proposition 4.6. Assume that Lm 1 (k, f ) contains no open subsets of con-
stant curvature and N is a totally umbilical pseudo-Riemannian submani-
fold of Lm
1 (k, f ) with dim N ≥ 3. Then N is an Einstein manifold if and
only if N is a transverse submanifold.
Proof. If Lm 1 (k, f ) contains no open subsets of constant curvature and
N is a totally umbilical in Lm 1 (k, f ) with dim N ≥ 3, then it follows from
(4.18) and Lemma 4.3(4) that the sectional curvature K N (u ∧ v) of N with
respect to orthonormal vectors u, v is given by

k + f ′2 k
K N (u ∧ v) = hH, Hi + + (ǫ ϕ2
u u + ǫ ϕ
v u
2
) − (ln f ) ′′
.
f2 f2
If we choose an orthonormal basis e1 , . . . , en of Tx N such that e2 , . . . , en
are perpendicular to ∂t , then the Ricci curvature of N satisfies

k + f ′2 2 k ′′
Ric(e1 ) = (n − 1) hH, Hi + + ǫ1 ϕe1 − (ln f ) ,
f2 f2
(4.27)
k + f ′2 2 k ′′
Ric(ej ) = (n − 1) hH, Hi + + ǫ1 ϕe1 − (ln f ) ,
f2 f2
for j = 2, . . . , n. Since n ≥ 3, it follows from (4.27) and the Einstein
condition that ϕe1 = 0. Hence N is a transverse submanifold.
Conversely, if N is a totally umbilical transverse submanifold of
Lm1 (k, f ), then N lies in a rest space S(t0 ) for some t0 ∈ I. Since S(t0 ) is
a real space form, Gauss’ equation implies that N is of constant curvature.
Hence, N is an Einstein manifold.
4.5 Realizations of Robertson-Walker spacetimes
Now, we present the realizations of generalized Robertson-Walker space-

times I ×f F in pseudo-Euclidean spaces.
Case (A): Let ψ1 : F → S m−1 (1) ⊂ Em be an isometric embedding of F
in the unit hypersphere S m−1 (1). Consider the map φ1 : I ×f F → Em+1
1
defined by
Z t
p
φ1 (t, p) = 1 + f ′ (u)2 du, f (t)ψ1 (p) . (4.28)
0
Then φ1 is an embedding. Moreover, we have
p
˜ ∂/∂t φ1 =
∇ 1 + f ′ (t)2 , f ′ (t)ψ1 ,
(4.29)
˜ X φ1 = (0, f (t)ψ1 (X)) , X ∈ T F.
∇ ∗
It follows from (4.28) and (4.29) that the induced metric via φ1 is exactly
the metric of the generalized Robertson-Walker spacetime I ×f F .
Case (B): Let ψ2 : F → H m−1 (−1) ⊂ Em 1 be an isometric embedding of F
into the unit hyperbolic space H m−1 (−1) centered at the origin. Consider
the map φ−1 : I ×f F → Em+1 1 defined by
Z t
p
φ−1 (t, p) = 1−f ′(u)2 du, f (t)ψ2 (p) . (4.30)
0
Then φ−1 is an embedding. It is direct to verify that the induced metric via
φ−1 is the metric of the generalized Robertson-Walker spacetime I ×f F .
As easy consequences of (A) and (B), we conclude that the Robertson-
Walker spacetime Ln1 (1, f ) = I ×f S n−1 (1) can be realized in En+1
1 via
Z t
p
φ1 (t, p) = 1 + f ′ (u)2 du, f (t)ι1 (p) , (4.31)
0
where ι1 : S n−1 (1) → En is the inclusion map.
Similarly, the Robertson-Walker spacetime Ln1 (−1, f ) = I ×f H n−1 (−1)
can be realized in En+1
2 via
Z t
p
φ−1 (t, p) = 1 − f ′ (u)2 du, f (t)ι2 (p) , (4.32)
0
where ι2 : H n−1 (−1) → En1 is the inclusion of H n−1 (−1) in En1 .
For the Robertson-Walker spacetime Ln1 (0, f ) = I ×f En−1 , the map
Z t
p
φ0 (t, u2 , . . . , un ) = 1+(n−1)f ′(u)2 du, f (t) cos u2 ,
0

f (t) sin u2 , . . . , f (t) cos un , f (t) sin un
is a realization of Ln1 (0, f ) = I ×f En−1 in E2n−1

1 .
4.6 Generalized Robertson-Walker spacetimes
Robertson-Walker spacetimes are the more classical and venerable relativis-

tic cosmological models of the expanding universe, but the hypothesis of
constant curvature for their fibers is a quite strong assumption. This prop-
erty even though reasonable for the first approximation of the large scale
structure of the universe, but it could not be appropriate when one consid-
ers a more accurate scale. Thus it is natural to study as well astronomical
models more general than Robertson-Walker models, namely, generalized
Robertson-Walker (GRW) spacetimes. A generalized Robertson-Walker
spacetime is a warped product with base an open interval of the real line en-
dowed with the opposite of its metric and fiber any Riemannian manifold.
Contrary to Robertson-Walker spacetimes, generalized Robertson-Walker
spacetimes are not necessarily spatially-homogeneous. In fact, small defor-
mations of the metric on the fiber of classical Robertson-Walker spacetimes
fit into the class of generalized Robertson-Walker spacetimes.
Generalized Robertson-Walker spacetimes were introduced in [Alı́as et
al. (1995)] and have been studied in [Aledo et al. (2014); Alı́as et al. (1995,
2013); Caballero et al. (2011); Deszcz and Kucharski (1999); Romero et al.
(2013a,b); Sánchez (1999)] among others. In particular, a global character-
ization of GRW spacetimes in term of a timelike and spatially conformal
vector field satisfying certain natural conditions was obtained in [Sánchez
(1999)]. Several characterizations of GRW spacetimes in term of timelike
gradient conformal vector fields were established in [Caballero et al. (2011)].
The following very simple characterization of generalized Robertson-
Walker spacetimes was obtained in [Chen (2014)].
Theorem 4.1. A Lorentzian n-manifold with n ≥ 3 is locally a generalized
Robertson-Walker spacetime if and only if it admits a timelike concircular
vector field.
Proof. Let M be a Lorentzian n-manifold with n ≥ 3. If M admits a
timelike concircular vector field, then v satisfies
∇X v = µX, X ∈ T M. (4.33)
Let us put v = ϕe1 , where e1 is a timelike unit vector field parallel to v. We
extend e1 to an orthonormal frame e1 , e2 , . . . , en on M so that e2 , . . . , en
are orthonormal spacelike vector fields. Define the connection forms ωij by
Xn
∇X ei = ǫj ωij (X)ej , i = 1, . . . , n, (4.34)
j=1
where ǫ1 = −1 and ǫ2 = · · · = ǫn = 1. It follows from (4.33) that

R(ei , v)v = ∇ei ∇v v − ∇v ∇ei v − ∇[ei ,v] v = (ei µ)v − (vµ)ei , (4.35)
for i = 2, . . . , n, where R is the curvature tensor R. From (4.35) we get
e2 µ = · · · = en µ = 0. (4.36)
Thus ∇µ is a timelike vector field parallel to v. From (4.33) with X = e1
and v = ϕe1 we find µe1 = ∇e1 (ϕe1 ) = (e1 ϕ)e1 + ϕ∇e1 e1 , which gives
e1 ϕ = µ, ∇e1 e1 = 0. (4.37)
From (4.37) we know that the integral curves of e1 are geodesics. Thus the
distribution D1 = Span{e1 } is a totally geodesic foliation. Let us define
another distribution by putting D2 = Span{e2 , . . . , en }.
It follows from (4.33) with v = ϕe1 and X = ei with i = 2, . . . , n that
µei = ∇ei (ϕe1 ) = (ei ϕ)e1 + ϕ∇ei e1 , which implies
e2 ϕ = · · · = en ϕ = 0, ϕ∇ei e1 = µei . (4.38)
From (4.34) and (4.38) we obtain ϕωi1 (ej )
= µδij for 2 ≤ i, j ≤ n. Hence
D2 is an integrable distribution whose leaves are totally umbilical in M .
Moreover, the mean curvature of leaves of D2 are given by µ/ϕ. Since
the leaves of D2 are spatial hypersurfaces, it follows from (4.36) and (4.38)
that the mean curvature vector fields of leaves of D2 are parallel in the
normal bundle in N . Thus D2 is a spherical foliation. Consequently, by a
result of [Hiepko (1979)] we conclude that M is locally an open portion of
a warped product I ×f (t) F , where f (t) is a function on I, ∂/∂t = e1 , and
F is a Riemannian (n − 1)-manifold. Therefore the sectional curvature of
M satisfies
K(e1 , ei ) = −f ′′ (t)/f (t), i = 2, . . . , n. (4.39)
On the other hand, we find from v = ϕe1 and (4.35) that
ϕK(e1 , ei ) = −vµ = −µ′ (t), i = 2, . . . , n. (4.40)
Thus, after combining (4.40) with (4.37) and (4.39), we obtain
f ′′ (t) µ′ (t) ϕ′′ (t)
= = .
f (t) ϕ ϕ(t)
Hence, if we choose f (t) = ϕ(t), then M is an open portion of the Lorentzian
warped product manifold I ×f F with f (t) = ϕ(t).
Conversely, consider I ×f F with metric g = −dt2 + f 2 gF , where (F, gF )
is a Riemannian manifold and f is a nowhere zero function. Let us consider
∂
the timelike vector field given by v = f (t) ∂t . It follows from Proposition
3.3 and direct computation that v satisfies (4.33) with µ = f ′ (t). Hence v
is a timelike concircular vector field.
The following result from [Chen (2017a)] can be regarded as a natural

extension of Theorem 4.1.
Theorem 4.2. A Lorentzian n-manifold M admits a time-like torqued vec-

tor field if and only if it is locally a twisted product I ×λ F , where I is an
open interval, F is a Riemannian (n − 1)-manifold so that the metric of M
takes the form g = −ds2 + λ2 gF .
Remark 4.3. A Lorentzian manifold with Ricci tensor of the form

Ric = λg + µω ⊗ ω (4.41)
with functions λ and µ and timelike unit 1-form ω, is often called a perfect
fluid spacetime. It is known that every Robertson-Walker spacetime is a
perfect fluid spacetime, and a 4D generalized Robertson-Walker spacetime
is a perfect fluid if and only if it is a Robertson-Walker spacetime (see
[Mantica et al. (2016a)]).
Definition 4.4. A Lorentzian manifold M is called Ricci simple if its Ricci

tensor satisfies Ric = ρ ω ⊗ ω for some function ρ and timelike unit 1-form
ω [Deszcz et al. (2001)].
Remark 4.4. In a recent paper [Mantica et al. (2016b)], C. A. Mantica,

Y-.J. Suh and U. C. De proved that a Ricci simple manifold with vanishing
divergence of the conformal curvature tensor admits a proper concircu-
lar vector field and hence it is necessarily a generalized Robertson-Walker
spacetime according to Theorem 4.1 (see also [Mantica et al. (2016a)]). It
was also proved in [Mantica and Molinari (2016a)] that timelike concircular
vector fields have many nice properties in GRW spacetimes.
For the most recent survey on GRW spacetimes, see [Mantica and Moli-
nari (2017)].
4.7 Schwarzschild’s solution and black holes
Besides Robertson-Walker spacetime, there is another well-known space-

time called Schwarzschild’s spacetime. The Schwarzschild spacetime was
discovered only a few months after the appearance of Einstein’s general
relativity and was published in [Schwarzschild (1916)].
Schwarzschild spacetime is the simplest relativistic model of a universe
containing a single star. The star is assumed to be static and spherically
symmetric; it is also assumed that the star is the only source of gravitation.
The last condition implies that the spacetime is a vacuum, i.e., it is Ricci
flat. Based on those conditions, Schwarzschild found a coordinate chart
(t, r, θ, φ), known as Schwarzschild’s chart for his spacetime.
First, he derived from the static and spherically symmetry conditions
that the metric is a warped product metric of the form:
g = −F (r)dt2 + G(r)dr2 + r2 (dθ2 + sin2 θdφ2 ),
(4.42)
t ∈ R, r ∈ I, 0 < θ < π, −π < φ < π,
where I is an open interval. Hence, on each rest space given by t constant,
the surface r constant has the line element
dσ 2 = r2 (dθ2 + sin2 θdφ2 ), (4.43)
which is a round sphere with Gauss curvature r−2 and surface area 4πr2 .
Obviously, sufficiently far away from the source of gravitation, the influ-
ence of the star becomes arbitrary small. Therefore one may require that
the metric approaches the physical Minkowski metric (in special relativ-
ity) as r → ∞, i.e., the metric is Minkowski at infinity, which in spherical
coordinates is given by
−c2 dt2 + dr2 + r2 (dθ2 + sin2 θdφ2 ),
where c denotes the speed of light and t is the time coordinate measured
by a stationary clock located infinitely far from the massive body. Hence
F (r) → −c2 and G(r) → +1 as r → ∞. (4.44)
By applying the Ricci flat condition on the spacetime and using (4.44),
one may determine functions F and G as
rs
F = −c2 λ, G = λ−1 , λ(r) = 1 − ,
r
where rs is a constant, known as the Schwarzschild radius. It is known that
rs is related to the mass M of the star by rs = 2GM/c2 , where G is the
gravitational constant. Consequently, the Schwarzschild metric in term of
Schwarzschild’s chart is the following warped product metric
rs 2 rs −1 2
g = −c2 1 − dt + 1 − dr + r2 (dθ2 + sin2 θdφ2 ). (4.45)
r r
Conversely, it was shown in [Jebsen (1921); Birkhoff (1923)] that any
local spherically symmetric solution of vacuum Einstein field equations is a
part of Schwarzschild’s metric. It follows from (4.45) that Schwarzschild’s
spacetime has singularities at r = 0 and r = rs .
Since the Schwarzschild metric is only expected to be valid for radii
larger than the radius R of the gravitating body, there is no problem as
long as R > rs , which is always the case for ordinary stars. For example,
the radius of the Sun is approximately 700,000 km, while the Schwarzschild
radius rs of the Sun is only 3 km.
On the other hand, any physical object whose radius R is less than or
equal to the Schwarzschild radius will undergo gravitational collapse. For
this reason any object whose radius is smaller than its Schwarzschild radius
is often called a black hole and the Schwarzschild radius is often called the
radius of a black hole.
The singularity at r = rs divides the Schwarzschild spacetime into two
disconnected portions. The exterior one with r > rs is the one that is
related to the gravitational field of the star (cf. [Hawking and Ellis (1973)]).
On the other hand, the interior one with 0 ≤ r < rs which contains the
singularity at r = 0, is completely separated from the outer one by the
singularity at r = rs .
The Schwarzschild coordinates give no physical connection between the
exterior and interior portions. Therefore they can be viewed as separate
solutions of Einstein’s field equations. The hypersurface r = rs gives rise
to the event horizon of the black hole. It represents the point past which
light can no longer escape the gravitational field.
The case r = 0 is different. If one asks that the solution be valid for all
r, then it runs into a gravitational singularity at the origin. At this point
the spacetime is no longer well-defined.
The existence of the singularity can be verified by noting that the in-
variant given by the square norm ||R||2 of the Riemann curvature tensor
R is infinite. (||R||2 is also known as the Kretschmann invariant in general
relativity.) A more recent better understanding of general relativity led to
the realization that such singularities were a generic feature of the theory
and not just an exotic special case (see, e.g. [Henry (2000); Cherubini et
al. (2002)]).
Remark 4.5. The Reissner-Nordström spacetime, discovered soon after
Schwarzschild’s article, is a static solution of the Einstein field equations
corresponding to the gravitational field of a charged, non-rotating, spheri-
cally symmetric body with the warped product metric
−1
2 rs e2 2 rs e2
g = −c 1 − + 2 dt + 1 − + 2 dr2 + r2 (dθ2 + sin2 θdφ2 ),
r r r r
where e is the electric charge. If e = 0, it reduces to the Schwarzschild
solution. Therefore Reissner-Nordström’s solution can be regarded as a
generalization of Schwarzschild’s solution of Einstein’s field equations.
Chapter 5
Contact Metric Manifolds and

Submersions
Contact geometry studies a geometric structure on manifolds given by a

hyperplane distribution in the tangent bundle and specified by a 1-form
satisfying a maximum non-degeneracy condition. In view of the Frobenius
theorem, one may recognize the condition as the opposite of the condition
that the distribution be determined by a codimension one foliation on the
manifold.
A contact metric manifold is a contact manifold equipped with an asso-
ciated metric, called a contact metric. An important class of contact met-
ric manifolds is the class of Sasakian manifolds. For general references on
Sasakian manifolds, see, e.g., [Calvaruso and Perrone (2010); Blair (2010);
Chen (2011b)].
Sasakian manifolds have an associated vector field, called the charac-
teristic vector field which generates a one-dimensional foliation. If the
leaves of this foliation are compact, then the space of leaves is a Kähler
orbifold. Sasakian manifolds with Riemannian metric were introduced in
[Sasaki (1960)]. The notion was extended in [Takahashi (1969)] to manifolds
with pseudo-Riemannian metrics.
A submersion π : M → B is a differentiable map between differen-
tiable manifolds whose differential is everywhere surjective. The notion of
submersions is a fundamental concept in differential topology. Rieman-
nian submersions are submersions equipped with compatible Riemannian
metrics. Riemannian submersions are natural generalizations of warped
products. The fundamental equations of Riemannian submersions were de-
rived in [O’Neill (1966)], which are well-known as the O’Neill equations.
One important consequence of O’Neill’s equations is that, for a Rieman-
nian submersion π : M → B, the lower bound for the sectional curvature
of B is at least as big as the lower bound for the sectional curvature of M .
99
5.1 Contact metric manifolds
Definition 5.1. Let M be a (2n + 1)-manifold and φ, ξ and η tensor fields

of type (1, 1), (1, 0) and (0, 1) on M , respectively. The triple (φ, ξ, η) is
called an almost contact structure if the following conditions are satisfied:
η(ξ) = 1, η(φX) = 0,
(5.1)
φ2 (X) = −X + η(X)ξ
for X ∈ T M.
The vector field ξ is called the characteristic vector field (or the Reeb
vector field if the almost contact structure is contact).
Definition 5.2. Let g be a pseudo-Riemannian metric on M . Then
(φ, ξ, η, g, ǫ), (ǫ = 1 or −1) is called an almost contact metric structure
on M if (φ, ξ, η) is an almost contact structure such that
g(ξ, ξ) = ǫ, η(X) = ǫg(ξ, X), ǫ = ±1, (5.2)
η(φ(X), φ(Y )) = g(X, Y ) − ǫη(X)η(Y ), (5.3)
for X, Y ∈ T M. An almost contact metric structure is called a contact
metric structure if it satisfies
dη(X, Y ) = g(X, φY ).
A pseudo-Riemannian manifold with a contact metric structure is called a
contact metric manifold.
Definition 5.3. A contact metric manifold (M, φ, ξ, η, g, ǫ) is said to be
K-contact if its characteristic vector field ξ is a Killing vector field.
A plane section of a contact metric manifold (M, φ, ξ, η, g, ǫ) is called a
φ-section if it is spanned by {v, φv} for some non-null vector v. The section
curvature of a φ-section is called a φ-sectional curvature.
5.2 Sasakian manifolds
Definition 5.4. A contact metric structure (φ, ξ, η, g, ǫ) on M is called a

normal contact metric structure if it satisfies
(∇X φ)Y = g(X, Y )ξ − ǫη(Y )X, X, Y ∈ T M, (5.4)
where ∇ is the Levi-Civita connection of g.
A manifold M endowed with a normal contact metric structure
(φ, ξ, η, g, ǫ) is called a Sasakian manifold.
Contact Metric Manifolds and Submersions 101
The following result is due to [Takahashi (1969)].

Proposition 5.1. For an almost contact metric manifold (M, φ, ξ, η, g),
condition (5.4) implies
(1) ∇X ξ = −φ(X),
(2) ξ is a Killing vector field, and
(3) dη(X, Y ) = g(X, φY ).
Remark 5.1. For an almost contact metric structure (φ, ξ, η, g, ǫ) on M ,
if we put
ḡ = −g, ξ̄ = −ξ, η̄ = −η, φ̄ = φ,
¯ η̄, ḡ, −ǫ) is an almost contact metric structure on M . Hence we
then (φ̄, ξ,
may assume that ǫ = 1. In this case, we simply denote (φ, ξ, η, g, ǫ) by
(φ, ξ, η, g).
If (M, φ, ξ, η, g) is Sasakian and α is a nonzero constant, we put
Dx = {v ∈ Tp M : η(v) = 0}, x ∈ M,
ḡ = αg + (α2 − α)η ⊗ η,
ξ¯ = α−1 ξ, η̄ = αη, φ̄ = φ.
Then (φ̄, ξ̄, η̄, ḡ) is also a Sasakian structure on M . (M, φ, ξ, η, g) is said to
be D-homothetic to (M, φ̄, ξ̄, η̄, ḡ) and D is called the contact distribution.
For a Sasakian manifold, the sectional curvature of the plane section
spanned by {X, φX} with X ⊥ ξ is called a φ-sectional curvature.
The following results can be found in [Takahashi (1969)].
Proposition 5.2. A Sasakian manifold of constant φ-sectional curvature
c 6= −3 is D-homothetic to a Sasakian manifold of constant curvature one.
Proof. If (M, g) is of constant φ-sectional curvature c, we have
c − 3(α − 1)
K̄ = K̄(X, φ̄X) =
α
for any non-lightlike vector X ∈ Dp , p ∈ M . Thus (M, ḡ) is of constant
φ-sectional curvature {c − 3(α − 1)}/α. Hence if c 6= −3 and if we take
α = (c + 3)/4, then (M, ḡ) is of constant φ-sectional curvature one, and
therefore of constant curvature one.
Proposition 5.3. Let Mi = (Mi , φi , ξi , ηi , gi ), i = 1, 2, be complete,
simply-connected Sasakian manifolds. If they have the same index and the
same constant φ-sectional curvature c 6= −3, then they are equivalent; i.e.,
there exists an isometry Φ : M1 → M2 such that Φ∗ (ξ1 ) = ξ2 , Φ∗ η2 = η1
and Φ∗ ◦ φ1 = φ2 ◦ Φ∗ .
5.3 Submersions
The notion of pseudo-Riemannian submersions was defined by B. O’Neill

in his book [O’Neill (1983)] as follows.
Definition 5.5. Let M and B be two pseudo-Riemannian manifolds. A

pseudo-Riemannian submersion is a smooth map π : M → B which is onto
and satisfies the following three axioms:
(S1) π∗ |p is onto for all p ∈ M ;

(S2) the fibers π −1 (b), b ∈ B, are pseudo-Riemannian submanifolds of M ;
(S3) π∗ preserves scalar products of vectors normal to fibers.
Riemannian submersions are natural generalizations of warped prod-

ucts, which occur widely in geometry. For instance, when a Lie group acts
isometrically, freely and properly on a Riemannian manifold M , the pro-
jection π : M → B to the quotient space B = M/G equipped with quotient
metric is a Riemannian submersion.
Let π : M → B be a pseudo-Riemannian submersion with dim M >
dim B. Denote by V and H the vertical and horizontal distributions as
well as for the orthogonal projections of T M on its horizontal and vertical
subspaces, respectively.
A vector field X̄ on M is said to be basic if X̄ is horizontal and π-related
to a vector field X on B. Notice that every vector field X on B has a unique
horizontal lift X̄ to M which is basic.
Since π is a submersion, π∗ gives a linear isomorphism Hp ≈ Tb B.
Moreover, it follows from (S3) of Definition 5.5 that it is a linear isometry.
Lemma 5.1. For X, Y ∈ X(B), we have

(1) X̄, Ȳ = hX, Y i ◦ π;
(2) H[X̄, Ȳ ] = [X, Y ]− ;
(3) H∇X̄ Ȳ = (∇′X Y )− , where ∇′ is the Levi-Civita connection of B.
Proof. (1) follows immediately form (S3).

(2) follows from the fact that [X̄, Ȳ ] is π-related to [X, Y ]; hence H[X̄, Ȳ ]
is also π-related to [X, Y ].
Clearly, (3) holds if both sides have the same scalar product with every
horizontal vector field,

or merely with every horizontal lift Z̄. Thus, by (1),
it suffices to show ∇X̄ Ȳ , Z̄ = h∇′X Y, Zi ◦ π, which follows by expanding
both sides in the Koszul formula. This is due to the fact that, using (1)
and (2), we have

X̄ Ȳ , Z̄ = X̄(hY, Zi ◦ π)
= (π∗ (X̄)) hY, Zi = X hY, Zi ◦ π,

X̄, [Ȳ , Z̄] = X̄, [X, Y ]− = hX, [Y, Z]i ◦ π.
B. O’Neill characterized in [O’Neill (1966)] the geometry of a Rieman-
nian submersion in terms of the two fundamental tensors A, T defined by
AE F = H∇HE VF + V∇HE HF, (5.5)
TE F = H∇VE VF + V∇VE HF (5.6)
for vector fields E, F ∈ X(M ). Here ∇ is the Levi-Civita connection of g.
The letters U, V, W denote vertical vector fields, X, Y, Z horizontal vec-
tor fields. Notice that TU V is the second fundamental form of each fiber.
Associated with a Riemannian submersion π : M → B, we define two
invariants Ăπ and Åπ on M by
b
X m
X X
Ăπ = |AXi Vs |2 , Åπ = |AXi Xj |2 , (5.7)
i=1 s=b+1 1≤i<j≤b
where X1 , . . . , Xb are orthonormal basic horizontal vector fields and
Vb+1 , . . . , Vm are orthonormal vertical vector fields.
The following result is due to O’Neill.
Theorem 5.1. Let π1 , π2 : (M, g) → (B, gB ) be Riemannian submersions.
If π1 , π2 have the same fundamental tensors A and T and π1∗ (p) = π2∗ (p)
at a point p ∈ M , then π1 = π2 .
5.4 O’Neill integrability tensor and fundamental equations
For a pseudo-Riemannian submersion π : M → B (dim B ≥ 2), leaves not

necessary exist, even locally. In view of Frobenius’ theorem, this failure can
be measured by the vector field
V[X, Y ] = AX Y − AY X.
Thus, the tensor A provides a natural obstruction to the integrability of
horizontal distribution H. For this reason the tensor A is called the O’Neill
integrability tensor.
O’Neill’s integrability tensor satisfies the following [O’Neill (1966)].
Lemma 5.2. Let X, Y be horizontal vector fields and E, F be vector fields
on M . Then each of the following holds:
(a) AX Y = −AY X, or equivalently, AX Y = 21 V[X, Y ];

(b) AHE F = AE F ;
(c) AE maps the horizontal subspace into the vertical one and the vertical
subspace into the horizontal one;
(d) g(AX E, F ) = −g(E, AX F );
(e) If X is basic, then AX V = H∇V X for every vertical vector field V ;
(f) g((∇Y A)X E, F ) = g(E, (∇Y A)X F ).
Let g and gB be the metric tensors of M and B, respectively and gF
the induced metric on fiber π −1 (π(p)), p ∈ M . Denote by R, RB and RF
the Riemann curvature tensors of the metrics g, gB and gF respectively.
The following equations, known as O’Neill’s equations, characterize the
geometry of a pseudo-Riemannian submersion [O’Neill (1966)].
Proposition 5.4. For vertical vector fields U, V, W, W ′ and horizontal
vector fields X, Y, Z, Z ′ , we have the following formulas:
(1) R(U, V ; W, W ′ ) = RF (U, V ; W, W ′ ) + g(TU W, TV W ′ ) − g(TV W, TU W ′ );
(2) R(U, V, W, X) = g((∇U T )V W, X) − g((∇V T )U W, X);
(3) R(X, U ; V, Y ) = g((∇X T )U V, Y ) − g(TU X, TV Y ) + g((∇U A)X Y, V )
+ g(AX U, AY V );
(4) R(U, V ; Y, X) = g((∇U A)X Y, V ) − g((∇V A)X Y, U ) + g(AX U, AY V )
− g(AX V, AY U ) − g(TU X, TV Y ) + g(TV X, TU Y );
(5) R(Y, X; Z, U ) = g((∇Z A)X Y, U ) + g(AX Y, TU Z) − g(AY Z, TU X)
− g(AZ X, TU Y );
(6) R(X, Y ; Z, Z ′ ) = RB (π∗ X, π∗ Y ; π∗ Z, π∗ Z ′ ) + 2g(AX Y, AZ Z ′ )
− g(AY Z, AX Z ′ ) + g(AX Z, AY Z ′ ).
Lemma 5.3. Let π : M → B be a pseudo-Riemannian submersion. Then
RB (π∗ X, π∗ Y ; π∗ Y, π∗ X) = R(X, Y ; Y, X) + 3g(AX Y, AX Y ). (5.8)
Moreover, if π has totally geodesic fibers, then we also have
(1) R(U, V ; V, U ) = RF (U, V ; V, U );
(2) R(X, U ; U, X) = g(AX U, AX U ).
Proof. Follows from O’Neill’s equations and Lemma 5.2(a).
Formula (5.8) and Lemma 5.2(a) imply the following [O’Neill (1966)].
Corollary 5.1. For a Riemannian submersion π : M → B, we have
3
KB (X, Y ) = K(X̄, Ȳ ) + |V[X̄, Ȳ ]|2 (5.9)
4
for orthonormal vector fields X, Y on B. Consequently, the lower bound

for the sectional curvature of B is at least as big as the lower bound for the
sectional curvature of M .
5.5 Submersions with totally geodesic fibers
The following lemma is obvious.
Lemma 5.4. Let π : M → B be a Riemannian submersion with totally

geodesic fibers. If B ′ is a submanifold of B, then π|π−1 (B ′ ) : π −1 (B ′ ) → B ′
is a Riemannian submersion with totally geodesic fibers.
Many explicit examples of pseudo-Riemannian submersions with totally

geodesic fibers can be obtained from the following standard construction
(cf. pages 256-257 in [Besse (1987)]):
Let G be a Lie group and let K and H be two compact Lie subgroups
of G with K ⊂ H. Let π : G/K → G/H be the natural fibration with
fiber H/K and structure group H. Let g be the Lie algebra of G and k ⊂ h
the corresponding Lie subalgebras of K and H. We choose an adG (H)-
invariant complement m to h in g, and an adG (K)-invariant complement p
to k in h. Then p ⊕ m is an ADG (K)-invariant complement to k in g.
An adG (H)-invariant scalar product on m defines a G-invariant Rie-
mannian metric g ′ on G/H and an adG (K)-invariant scalar product on p
defines a H-invariant Riemannian metric ĝ on H/K. The orthogonal direct
sum for these scalar products on p ⊕ m defines a G-invariant Riemannian
metric g on G/K.
The following proposition can be found in [Besse (1987)].
Proposition 5.5. The mapping π : (G/K, g) → (G/H, g ′ ) is a Riemannian

submersion with totally geodesic fibers.
By applying this proposition we have the following examples.
Example 5.1. Let

G = SU (m + 1), H = S(U (1) · U (m)), K = SU (m).
Then the corresponding mapping π : (G/K, g) → (G/H, g ′ ) is the Rieman-
nian submersion π : S 2m+1 (1) → CP m (4), where we choose the canonical
metric g on S 2m+1 (1) which is SU (m+ 1)-invariant. π is a submersion with
totally geodesic fibers onto the complex projective m-space CP m (4). This
Riemannian submersion is known as the Hopf fibration.
Example 5.2. Let G = Sp(m + 1), H = Sp(1) · Sp(m), K = Sp(m). Then

we get
π : S 4m+3 (1) → HP m (4)
with totally geodesic fibers S 3 (1), where HP m (4) is the quaternion projec-
tive m-space with constant quaternionic sectional curvature 4.
Example 5.3. Let G = Sp(m + 1), H = Sp(1) · Sp(m), K = U (1)Sp(m).

Then
π : CP 2m+1 → HP m (4)
is Riemannian submersion with totally geodesic fibers S 2
Example 5.4. Let G = Spin(9), H = Spin(8) and K = Spin(7). We have

the Riemannian submersion with totally geodesic fibers:
S 15 (1) = Spin(9)/Spin(7) → S 8 (4) = Spin(9)/Spin(8).
For Riemannian submersions with totally geodesic fibers, we have the

following result of [Escobales (1975); Ranjan (1985)].
Theorem 5.2. Let π : S m (1) → B be a Riemannian submersion with to-

tally geodesic fibers. Then, up to equivalence by a fiber-preserving isometry
of the total space, π is one following standard ones:
(1) π : S 2n+1 (1) → CP n (4) with S 1 as fibers;
(2) π : S 4n+3 (1) → HP n (4) with S 3 as fibers;
(3) π : S 15 (1) → S 8 (4) with S 7 as fibers.
Here CP n (4) and HP n (4) denote the complex projective space and quater-
nionic projective space of real dimensions 2n and 4n, respectively, taking
sectional curvature in the interval [1, 4].
R. Escobales also obtained the following [Escobales (1978)].
Theorem 5.3. A Riemannian submersion π : CP r (4) → B, with complex

totally geodesic fibers and 2 ≤ dim f iber ≤ 2r − 2, is one of the following:
(1) π : CP 2n+1 (4) → QP n (4) with S 2 as fibers;

(2) π : CP 7 (4) → S 8 (4) with CP 3 (4) as fibers.
Escobales did not know whether case (2) in Theorem 5.3 was empty or
not. By applying homotopy theory it was shown in [Ucci (1983)] that case
(2) cannot occur.
Two Riemannian submersions πi : S m → Bi (i = 1, 2) are said to be

metric congruent if there are isometric diffeomorphisms ι1 : S m → S m and
ι2 : B1 → B2 such that
ι
S m −−−1−→ Sm
 

π1 y
π2
y
ι
B −−−2−→ B ′
commutes.
The following result from [Gromoll and Grove (1988); Wilking (2001)]
generalizes completely the result of Escobales and Ranjan, without the
assumption of totally geodesic fibers.
Theorem 5.4. Let π : S m (1) → B be a Riemannian submersion with

connected fibers of positive dimension. Then π : S m (1) → B is metrically
congruent to one of the following fibrations:
(1) π : S 2n+1 (1) → CP n (4) with S 1 as fibers;
(2) π : S 4n+3 (1) → HP n (4) with S 3 as fibers;
(3) π : S 15 (1) → S 8 (4) with S 7 as fibers.
5.6 Sasakian space forms
There exists a simple connection between Sasakian and Kähler manifolds

as follows:
Let π : M → B be a Riemannian submersion whose total space admits
a Sasakian structure (φ, ξ, η, g) such that ξ spans the vertical distribution.
For X, Y ∈ X(B) with horizontal lifts X̄, Ȳ , if we put
gB (X, Y ) ◦ π = g(X̄, Ȳ ) and JX = π∗ (φX),
then (J, gB ) is a Kählerian structure on B [Morimoto (1964)].
Definition 5.6. A Riemannian submersion π : M → B is called a canonical

fibration of a Sasakian manifold if B is Kählerian and M has a Sasakian
structure (φ, ξ, η, g) such that ξ spans the vertical distribution.
Lemma 5.5. Let π : (M, φ, ξ, η, g) → (B, J, g) be a canonical fibration of

Sasakian manifold. Then
(a) the vertical distribution V is spanned by ξ;
(b) g = η ⊗ η + π ∗ g ′ ;
(c) φX = (Jπ∗ (X))− for X ∈ T M , where ¯ denotes the horizontal lift.

Proof. Follows from definitions.
A Sasakian manifold with constant φ-sectional curvature is called a
Sasakian space form. The curvature tensor of Sasakian space forms was
determined in [Ogiue (1964); Takahashi (1969)].
Proposition 5.6. If a Sasakian manifold (M, φ, ξ, η, g, ǫ) has constant φ-
sectional curvature c, then its Riemann curvature tensor satisfies
c + 3ǫ
R̃(X, Y )Z = (hY, Zi X − hX, Zi Y )
4
ǫc − 1
+ η(X)η(Z)Y − η(Y )η(Z)X
4
+ hX, Zi η(Y )ξ − hY, Zi η(X)ξ + hφY, Zi φX

− hφX, Zi φY − 2 hφX, Y i φZ .
The following are the standard models of Sasakian space forms equipped
with Riemannian metrics.
Example 5.5. Let R2n+1 = {(x1 , . . . , xn , y1 , . . . , yn , z) : xi , yi , z ∈ R}.
Consider R2n+1 with its standard contact structure given by
1 Xn ∂
η= dz − yi dxi , ξ = 2 ,
2 i=1 ∂z
1 Xn
g =η⊗η+ (dxi ⊗ dxi +dyi ⊗ dyi ),
4 i=1
!
Xn
∂ ∂ ∂ (5.10)
φ Xi +Yi +Z
i=1
∂xi ∂y i ∂z
Xn X n
∂ ∂ ∂
= Yi − Xi + Yi yi .
∂xi ∂yi ∂z
i=1 i=1
Then R2n+1 is a Sasakian space form of constant φ-sectional curvature −3.
We denote this Sasakian space form by R2n+1 (−3).
Example 5.6. Consider the unit hypersphere S 2n+1 (1) ⊂ Cn+1 . Denote
by x the position vector field of S 2n+1 (1) in the complex Euclidean (n + 1)-
space Cn+1 and by g the induced metric. Let ξ = −Jx, where J is the
complex structure of Cn+1 . Let η be the dual 1-form given by η(X) =
g(ξ, X) and φ the tensor field of type (1, 1) defined by φ = π ◦ J, where π
is the orthogonal projection
from Tp Cn+1 onto Tp S 2n+1 (1), p ∈ S 2n+1 (1).
Then S 2n+1 (1), φ, ξ, η, g is a Riemannian Sasakian space form of constant
φ-sectional curvature one.
Example 5.7. Let B be a simply-connected bounded domain in Cn and

(J, g ′ ) a Kähler structure on B with constant holomorphic sectional cur-
vature c < 0. Since the fundamental 2-form Ω of the Kähler structure is
exact, we have Ω = dω for some real analytic 1-form ω.
Let π : B × R → B denote the canonical projection and t the coordinate
on R. Then B × R with η = π ∗ ω + dt and g = π ∗ g ′ + η ⊗ η is a Sasakian
manifold.
Regarding η as a connection form on B × R and let X̄ be the horizontal
lift of a vector field X on B. Then by applying the Riemannian submersion
technique we have, for orthonormal pair of vectors X, Y ∈ T B such that
K(X̄, Ȳ ) = KB (X, Y ) − 3(η(∇X̄ Ȳ ))2
where KB is the sectional curvature of B. Since
g(∇X̄ Ȳ , ξ) = −g(Ȳ , ∇X̄ ξ) = g(Ȳ , φX̄) = g ′ (X, JX),
B × R is a Sasakian manifold of constant φ-sectional curvature c − 3.
A typical example of canonical fibration of Sasakian manifolds is the
Hopf fibration
π : S 2n+1 (1) → CP n (4).
Another example is the following.
Example 5.8. Consider the map π : R2n+1 (−3) → Cn , where R2n+1 (−3)
is the Sasakian space form of φ-sectional curvature −3, where Cn is the
usual complex Euclidean n-plane Cn , and π is defined by
1
π(x1 , . . . , xn , y1 , . . . , yn , z) = (y1 , . . . , yn , x1 , . . . , xn ). (5.11)
2
Then
π : R2n+1 (−3) → Cn
is a Riemannian submersion with totally geodesic fiber. This gives rise to
a canonical fibration of R2n+1 (−3).
The following are examples of Sasakian space forms equipped with in-
definite metrics.
Example 5.9. For integers n, s satisfying n ≥ 1 and 0 ≤ s ≤ n, let
2n+1
S2s (1) be the unit pseudo-sphere defined by (1.19). Denote by x the
2n+1
position vector field of S2s (1) in Cn+1
s and let ξ = −Jx, where J is
n+1
the complex structure of Cs . Then ξ is a spacelike unit vector field of
2n+1
S2s (1). Let η be the dual 1-form given by η(X) = g(ξ, X) and φ the
tensor field defined by φ = π ◦ J, where π is the orthogonal projection from
2n+1 2n+1
Tp (Cn+1
s ) onto Tp (S2s (1)), p ∈ S2s (1).
The following results are known from [Takahashi (1969)]:

2n+1 2n+1
(a) S2s (1) is diffeomorphic to R2s × S 2n+1−2s . Thus S2s (1) is simply-
connected for s 6= n.
2n+1
(b) S2n (1) is connected with infinite cyclic fundamental group.
2n+1 2n+1 2n+1
(c) Put S̃2s = S2s (1) for s 6= n and let S̃2n (1) be the universal
2n+1
pseudo-Riemannian covering manifold of S2n (1). Then the Sasakian
2n+1 2n+1
structure on S2n (1) induces a Sasakian structure on S̃2n .
2n+1
Consider the pseudo-Riemannian metric g on S2s (1) induced from
the standard metric of Cn+1
s . Then
2n+1

S2s (1), φ, ξ, η, g
is a Sasakian manifold with pseudo-Riemannian metric of constant curva-

ture one.
The following proposition was obtained in [Takahashi (1969)]:
Proposition 5.7. If a Sasakian (2n + 1)-manifold (M, φ, ξ, η, g) is com-

plete, simply connected and of constant φ-sectional curvature c 6= −3, then
2n+1
it is D-homothetic to S̃2s , where 2s = the index of g if c > −3 and
2s = 2n−the index of g if c < −3.
Example 5.10. Let

2n+1
H2s−1 (−1) = {z ∈ C2n+1
s : bs,n+1 (z, z) = −1}, s ≥ 1.
2n+1
Consider the induced metric ḡ on H2s−1 (−1). Let x be the position vector
2n+1 n+1 ¯ ¯ i.e.,
of H2s−1 (−1) in Cs , ξ = Jx and η̄ the dual 1-form of −ξ,
¯ X), X ∈ T H 2n+1 (−1).
η̄(X) = −ḡ(ξ, 2s−1
Let us put φ̄ = π̄ ◦ J, where J is the natural complex structure of

2n+1
Cn+1
s and let π̄ be the orthogonal projection of Tp Cn+1
into Tp H2s−1 (−1)
s
2n+1 2n−1 ¯
with p ∈ H2s−1 (−1). Then H2s−1 (−1), φ̄, ξ, η̄, ḡ is a Sasakian mani-
fold equipped with a pseudo-Riemannian metric of constant curvature −1.
Hence the anti-de Sitter spacetime H12n+1 (−1) is a Sasakian manifold of
constant curvature −1. It is also known that H2s−12n+1 ¯ −η̄, −ḡ
(−1), −φ̄, −ξ,
2n+1
is S2(n−s+1) (1), φ, ξ, η, g .
Remark 5.2. It was proved in [Bejancu and Duggal (1993)] that every
connected Sasakian real hypersurface of E2n+2
2s is either an open part of
n+1 2n+1
S2s or an open part of H2s−1 (−1).
5.7 Geometry of horizontal immersions
In differential geometry there are important examples of manifolds which

occur as base space B of a pseudo-Riemannian submersion π : M → B and
whose geometric structure is obtained by projection the structure of the
bundle space M . For example, the geometric structure of CP n (4) arises
from the Sasakian structure of S 2n+1 (1) via Hopf’s fibration.
In this section we present a close relation between immersed submani-
folds of the bundle space and the base manifold B of a pseudo-Riemannian
submersion. The general reference of this section is [Reckziegel (1985)].
For a pseudo-Riemannian submersion π : M → B, the curvature form
of the horizontal distribution H is the skew-symmetric bilinear form defined
by
Ω(X, Y ) = V[X, Y ] (5.12)
for X, Y ∈ H.
Clearly, distribution H is integrable if and only if Ω = 0.
Lemma 5.6. Let π : M → B be a pseudo-Riemannian submersion. If

X ∈ X(N ) such that φ∗ X is horizontal and if φ : N → M is a smooth map,
then
2V(∇X η) = Ω(φ∗ X, η), (5.13)
π∗ (∇X η) = ∇′X π∗ η (5.14)
for horizontal vector field η along φ, where ∇ and ∇′ are the Levi-Civita
connection of M and B, respectively.
Proof. From (5.5), (5.12) and Lemma 5.2(a) we find

Ω(φ∗ X, η) = V[φ∗ X, η] = 2Aφ∗ X η = 2V∇X η.
This gives (5.13). Equation (5.14) follows from Lemma 5.2(3).
Let π : M → B be a pseudo-Riemannian submersion. An immersion
φ : N → M is called horizontal if φ∗ (Tp N ) ⊂ Hφ(p) for all p ∈ N .
The following result is due to [Reckziegel (1985)].
Theorem 5.5. Let π : M → B be a pseudo-Riemannian submersion

and φ : N → M be an horizontal isometric immersion from a pseudo-
Riemannian manifold N into M . Then
(a) O’Neill’s integrability tensor A vanishes on vector fields tangent along
φ, i.e., Ω(φ∗ (Tp N ), φ∗ (Tp N )) = 0 for each p ∈ N ;
(b) φ̂ = π ◦ φ is an isometric immersion N → B;

(c) the second fundamental form σ φ of φ takes its values in H and π∗ σ φ
equals the second fundamental form σ φ̂ of φ̂ : N → B;
(d) for every normal vector field η of N in M , π∗ η is a normal vector field
of φ̂; in case that η is horizontal, we have
V(DX η) = AX η, (5.15)
′
π∗ (DX η) = DX (π∗ η), (5.16)
where D and D′ are the normal connection of φ and φ̂ and X is a
tangent vector field of N .
Proof. Let X, Y ∈ X(N ). From Gauss’ formula and (5.13), we find

Ω(φ∗ X, φ∗ Y ) = 2Vσ φ (X, Y ).
As the term on the right (resp. left) is symmetric (resp. skew-symmetric),
both terms vanish. Thus we have (a) and the first half of statement (c).
Statement (b) follows trivially from the definition of pseudo-Riemannian
submersions. For the second half of statement (c) we only need to substitute
Gauss’ formula in (5.14).
Formula (5.15) is immediate from (5.13) because ∇X η − DX η is tangent
to N according to Weingarten’s formula; hence it is horizontal.
For the proof of (5.16) we use shape operators Aφ and Aφ̂ of φ and φ̂.
From (c) we find Aφ̂π∗ η = Aφη . Hence
π∗ (∇X η − DX η) = ∇′X (π∗ η) − DX
′
(π∗ η)
by Weingarten’s formula. (5.16) is an easy consequence of (5.14).
The following is an immediate consequence of Theorem 5.5.
Corollary 5.2. Under the hypothesis of Theorem 5.5, the immersion φ̂ =

π ◦ φ : N → B is totally geodesic (resp. minimal, parallel, totally umbilical,
or pseudo-umbilical ) if and only if φ is totally geodesic (resp. minimal,
parallel, totally umbilical, or pseudo-umbilical ).
5.8 Legendre submanifolds via canonical fibration
For a contact manifold M 2n+1 , the condition η ∧ (dη)n 6= 0 implies that

the contact distribution D of M is non-integrable, even locally. However,
it is in general possible to find n-dimensional submanifolds whose tangent
spaces lie inside the contact distribution.
A submanifold N of a contact manifold M 2n+1 with contact form η is

called an integral submanifold if η(X) = 0 for every X ∈ T N .
For any X, Y ∈ T N , we have
1
dη(X, Y ) = (Xη(Y ) − Y η(X) − η([X, Y ]) = 0.
2
Hence, in terms of associated metrics and g(X, φY ) = 0, we know that φ
maps each tangent vector into a normal vector. Because ξ is normal to every
integral submanifold of M 2n+1 , the dimension of an integral submanifold
of a (2n + 1)-dimensional contact metric manifold M 2n+1 is at most n.
On the other hand, by applying Darboux’s theorem, we obtain a system
of local coordinates (x1 , . . . , xn , y1 , . . . , yn , z) with respect to which
n
X
η = dz − yi dxi .
i=1
Therefore, xi = const. and z = const. define an integral submanifold of n
dimension in M 2n+1 .
Definition 5.7. An n-dimensional integral submanifold N of a contact

metric (2n + 1)-manifold M 2n+1 is called a Legendre submanifold.
The following nice link between Legendre submanifolds and Lagrangian

submanifolds via canonical fibration is due to [Reckziegel (1985)].
Proposition 5.8. Let π : M → B be a canonical fibration of a Sasakian

manifold. Then
(1) an isometric immersion f : N → M is horizontal if and only if the
characteristic vector field ξ is normal to N and f is C-totally real;
(2) an isometric immersion i : N → B has local horizontal lifts if and only
if i is totally real.
We may apply this proposition to obtain the following.
Case (i): R2n+1 (-3). Consider the canonical fibration π : R2n+1 (−3) → Cn
defined by
1
π(x1 , . . . , xn , y1 , . . . , yn , z) = (y1 , . . . , yn , x1 , . . . , xn ). (5.17)
2
Let i : N → Cn be a Lagrangian isometric immersion of a Riemannian
n-manifold N into Cn . Then there is a covering map τ : N̂ → N and a
horizontal immersion î : N̂ → R2n+1 (−3) such that i ◦ τ = π ◦ î. Therefore
each Lagrangian immersion can be lifted (locally) (or globally if N is simply-
connected) to a Legendre immersion of the same manifold.
Conversely, let f : N̂ → R2n+1 be a Legendre isometric immersion.

Then i = π ◦ f : N → Cn is a Lagrangian isometric immersion. Under this
correspondence, we have a one-to-one correspondence between Lagrangian
submanifold N̂ of Cn and Legendre submanifolds N of R2n+1 .
Case (ii): CP n (4). Consider Hopf’s fibration π : S 2n+1 → CP n (4). Then
π is a Riemannian submersion. Given z ∈ S 2n+1 , the horizontal space at
z is the orthogonal complement of iz with respect to the metric on S 2n+1
induced from the metric on Cn+1 .
Let i : N → CP n (4) be a Lagrangian isometric immersion. Then there
is a covering map τ : N̂ → N and a horizontal immersion î : N̂ → S 2n+1
such that i◦τ = π ◦ î. Thus each Lagrangian immersion can be lifted locally
(or globally if N is simply-connected) to a Legendre immersion of the same
Riemannian manifold.
Conversely, if f : N̂ → S 2n+1 is a Legendre isometric immersion, then
i = π ◦ f : N → CP n (4) is a Lagrangian isometric immersion. Under this
correspondence the second fundamental forms σ f and σ i of f and i satisfy
π∗ σ f = σ i by Theorem 5.5. Moreover, σ f is horizontal with respect to π.
Case (iii): CH n (−4). Consider the anti-de Sitter spacetime
H12n+1 (−1) = {z ∈ C2n+1 1 : b1,n+1 (z, z) = −1}
with the canonical Sasakian structure given in Example 5.7. Put
Tz′ = {u ∈ Cn+1 : hu, zi = 0}, H11 = {λ ∈ C : λλ̄ = 1},
where h , i is the Hermitian inner product on Cn+1 1 whose real part is
g0 . Then there is an H11 -action on H12n+1 (−1), z 7→ λz and at each point
z ∈ H12n+1 (−1), the vector ξ = −iz is tangent to the flow of the action.
Since the metric g0 is Hermitian, we have hξ, ξi = −1. The quotient space
H12n+1 (−1)/ ∼, under the identification induced from the action, is the
complex hyperbolic space CH n (−4) with constant holomorphic sectional
curvature −4, with the complex structure J induced from the complex
structure J on Cn+1 1 via π : H12n+1 (−1) → CH n (4c).
Just like case (ii), if i : N → CH n (−4) is a Lagrangian immersion, then
there is an isometric covering map τ : N̂ → N and a Legendre immersion
f : N̂ → H12n+1 (−1) such that i ◦ τ = π ◦ f . Thus every Lagrangian
immersion can be lifted locally (or globally if N is simply-connected) to a
Legendre immersion.
Conversely, if f : N̂ → H12n+1 (−1) is a Legendre immersion, then i =
π ◦ f : N → CH n (−4) is a Lagrangian immersion. Similarly, under this
correspondence the second fundamental forms σ f and σ i are related by
π∗ σ f = σ i . Also, σ f is horizontal with respect to π.
Chapter 6
Kähler and Pseudo-Kähler Manifolds
A Kähler manifold M is a 2n-dimensional manifold equipped with a com-

plex structure J, J 2 = −I, and a compatible Riemannian metric g, namely,
g(Ju, Jv) = g(u, v), ∇J = 0, ∀u, v ∈ T M.
Thus a Kähler manifold admits a U (n)-structure satisfying an integrability
condition. If we define a 2-form Ω on M by Ω(u, v) = g(u, Jv), then Ω is
a symplectic structure, i.e., Ω is a non-degenerate closed 2-form. Hence a
Kähler manifold is a Riemannian manifold, a complex manifold, and a sym-
plectic manifold with these three structures all mutually compatible. This
threefold structure corresponds to the presentation of the unitary group as
an intersection:
U (n) = O(2n) ∩ GL(n, C) ∩ Sp(2n).
Without any integrability conditions, the corresponding notion is an almost
Hermitian manifold.
If the Sp-structure is integrable but the complex structure is not nec-
essary integrable, the notion is an almost Kähler manifold; if the complex
structure is integrable but the Sp-structure need not be, the notion is a
Hermitian manifold. When g is a pseudo-Riemannian metric, the corre-
sponding manifold is pseudo-Kählerian.
Both Kähler manifolds and pseudo-Kähler manifolds are important in
algebraic and differential geometry as well as in mathematical physics.
6.1 Pseudo-Kähler manifolds
Let M be a complex manifold of complex dimension n and {z1 , . . . , zn } a

local complex coordinate system on a coordinate neighborhood U . If
√
zj = xj + iyj , i = −1, j = 1, . . . , n,
115
then {x1 , . . . , xn , y1 , . . . , yn } forms a system of local coordinates on U .

Put Xj = ∂/∂xj , Yj = ∂/∂yj . Then X1 , . . . , Xn , Y1 , . . . , Yn form a local
frame of T M .
Let J be the endomorphism of T U defined by
JXj = Yj , JYj = −Xj , j = 1, . . . , n.
Then J 2 = −I. It is easy to verify that J does not depend on the choice of
(z1 , . . . , zn ). J is called the complex structure of the complex manifold M .
Definition 6.1. A pseudo-Riemannian metric g on a complex manifold M

is called pseudo-Hermitian if g and J are compatible, i.e.,
g(JX, JY ) = g(X, Y ), X, Y ∈ Tp M, p ∈ M. (6.1)
A pseudo-Hermitian manifold is a complex manifold equipped with a
pseudo-Hermitian metric. If the pseudo-Hermitian metric is of index zero,
M is called a Hermitian manifold. If the index of the metric is positive, M
is called an indefinite Hermitian manifold.
It follows from (6.1) that the index of g is an even integer 2s with

0 ≤ s ≤ n, n = dimC M . The integer s is called the complex index. The
fundamental 2-form Ω of a pseudo-Hermitian manifold (M, g) is defined by
Ω(X, Y ) = g(X, JY ), X, Y ∈ T M. (6.2)
Definition 6.2. A pseudo-Hermitian manifold (resp. an indefinite Her-

mitian manifold) is called pseudo-Kähler (resp. indefinite Kähler) if its
fundamental 2-form Ω is closed, i.e., dΩ = 0. The corresponding metric is
called pseudo-Kähler (resp. indefinite Kähler). A pseudo-Kähler manifold
with complex index one is called a Lorentzian Kähler manifold.
Proposition 6.1. A pseudo-Hermitian manifold is pseudo-Kählerian if

and only if J is parallel, i.e., ∇J = 0.
Proof. Let (M, g, J) be a pseudo-Hermitian manifold. Then

3dΩ(X, Y, X) = g(X, (∇Z J)Y ) − g(Y, (∇X J)Z) + g(Z, (∇Y J)X)
for X, Y, Z tangent to M . Thus, ∇J = 0 implies dΩ = 0.
Conversely, because
2g((∇X J)Y, Z) = dΩ(X, JY, JZ) − dΩ(X, Y, Z),
dΩ = 0 implies ∇J = 0.
Kähler and Pseudo-Kähler Manifolds 117
Lemma 6.1. The Riemann curvature tensor R of a pseudo-Kähler mani-

fold satisfies
R(X, Y ) ◦ JZ = J ◦ R(X, Y )Z, (6.3)
R(JX, JY )Z = R(X, Y )Z. (6.4)
Proof. Since J is parallel, ∇X JY = J∇X Y . Thus

R(X, Y ) ◦ JZ = [∇X , ∇Y ]JZ − ∇[X,Y ] JZ = J(R(X, Y )Z),
which gives (6.3). By applying (1.7) and (1.10) we have
g(R(JX, JY )V, U ) = g(R(U, V )JY, JX) = g(J(R(U, V )Y ), JX)
= g(R(U, V )Y, X) = g(R(X, Y )V, U ).
Thus we obtain (6.4).
A plane section on a pseudo-Kähler manifold is called holomorphic if
it is spanned by {v, Jv} for some non-null vector v ∈ T M . The sectional
curvature K(v ∧ Jv) of a holomorphic section is called the holomorphic sec-
tional curvature at v, which is denoted by H(v). The holomorphic sectional
curvature H(v) is independent of the choice of v in Span{v, Jv}.
Definition 6.3. A Kähler manifold M (resp. indefinite Kähler manifold )

is called a complex space form (resp. indefinite complex space form) if it
has constant holomorphic sectional curvature.
An indefinite complex space form with complex index one is called a
Lorentzian complex space form.
The following proposition and examples were obtained in [Barros and

Romero (1982)].
Proposition 6.2. The curvature tensor of an (indefinite) complex space

form Msm (4c) of constant holomorphic sectional curvature 4c satisfies
R(X, Y )Z = c {g(Y, Z)X − g(X, Z)Y + g(JY, Z)JX
(6.5)
− g(JX, Z)JY + 2g(X, JY )JZ}.
Conversely, if the curvature tensor of an (indefinite) Kähler manifold
satisfies (6.6), then it is an (indefinite) complex space form of constant
holomorphic sectional curvature 4c.
Proof. Let Ro be a (1, 3) tensor on M defined by

Ro (X, Y )Z = c {g(Y, Z)X − g(X, Z)Y + g(JY, Z)JX
(6.6)
− g(JX, Z)JY + 2g(X, JY )JZ}.
Consider R̂ = R − Ro . Then
g(R̂(v, Jv)Jv, v) = 0 (6.7)
for v ∈ T M with g(v, v) 6= 0. Since {v ∈ Tp M : g(v, v) 6= 0} is dense in
Tp M , g(R̂(v, Jv)Jv, v) = 0 holds in general. Thus, by applying polariza-
tion, we find R̂ = 0.
The converse is trivial.
Example 6.1. Let n and s be integers such that n ≥ 1 and 0 ≤ s ≤ n. The

complex manifold Cn endowed with the real part of the Hermitian form
Xs Xn
bs,n (z, w) = − z̄j wj + z̄j wj (6.8)
j=1 j=s+1
for z, w ∈ Cn . It defines a flat (indefinite) complex space form of index 2s,
denote by Cns .
Example 6.2. For a positive number c and an integer s ≥ 0, there ex-

ists an (indefinite) complex space form CPsn (4c) with complex dimension
n, complex index s and of constant holomorphic sectional curvature 4c.
CPsn (4c) is homotopy equivalent to the standard complex projective space
CP n−s ; and therefore, CPsn is simply-connected.
The underlying complex manifold of CPsn (4c) is the open submanifold
{z ∈ Cn+1 : bs,n+1 (z, z) > 0}/C∗
of CP n = (Cn+1 − {0})/C∗, where C∗ = C − {0}.
Consider the pseudo hypersphere of curvature c defined by
n o
2n+1 1
S2s (c) = z ∈ Cn+1
s : bs,n+1 (z, z) = . (6.9)
c
2n+1
Then π : S2s (c) → CPsn ; z 7→ z · C∗ is a submersion. There is a unique
pseudo-Kähler metric of index 2s on CPsn which make CPsn an (indefinite)
complex space form of constant holomorphic sectional curvature 4c such
that π is a pseudo-Riemannian submersion.
Example 6.3. For a negative number c and a non-negative integer s,

the (indefinite) complex hyperbolic n-space CHsn (4c) is obtained from
n n
CPn−s (−4c) by replacing the metric of CPn−s (−4c) by its negative. In
n n n n
particular, CP = CP0 and CH = CH0 are called complex projective
n-space and complex hyperbolic n-space, respectively.
Remark 6.1. Every complete simply-connected pseudo-Kähler manifold

of complex dimension n, of complex index s and of constant holomor-
phic sectional curvature 4c is holomorphically isometric to CPsn (4c), Cns
or CHsn (4c) according to c > 0, c = 0 or c < 0, respectively.
6.2 Concircular vector fields on pseudo-Kähler manifolds
The following result shows that there exist no non-trivial concircular vector
fields on every pseudo-Kähler manifold of complex dimension > 1.
Theorem 6.1. [Chen (2016a)] We have:

(1) Every pseudo-Kaehler manifold M n with n = dimC M n > 1 does not
admit a non-trivial concircular vector field.
(2) The result is false for pseudo-Kaehler manifolds M n with n = 1.
Proof. It follows from (1.6) and (1.30) that

R(X, v)v = ∇X (µv) − ∇v (µX) − µ∇X v + µ∇v X
(6.10)
= (Xµ)v − (vµ)X,
for any X ∈ X(M n ). By taking the inner product of (6.10) with v we find
(Xµ)g(v, v) = (vµ)g(X, v), ∀X ∈ T M n . (6.11)
In particular, (6.11) implies the following
vµ = 0 whenever g(v, v) = 0 (6.12)
and
Xµ = 0 whenever g(v, v) 6= 0 and g(X, v) = 0. (6.13)
Let X be a vector satisfying g(X, v) = 0. By taking the inner product
of (6.10) with X, we find
g(R(X, v)v, X) = −(vµ)g(X, X) whenever g(X, v) = 0. (6.14)
Similarly, by applying (1.6), (1.30) and (6.3), we also have
R(Y, Jv)Jv = J(R(Y, Jv)v)
= J{∇Y (µJv) − ∇Jv (µY ) − µ∇Y (Jv) + µ∇Jv Y } (6.15)
= −(Y µ)v − ((Jv)µ)JY
for Y ∈ X(M n ). Therefore, by combining (6.15) with (6.11), we obtain
g(R(Y, Jv)Jv, Y ) = 0 (6.16)
for any tangent vector Y satisfying g(Y, v) = 0.
Next, by applying (1.7), (1.10), (6.4) and (6.16) we have
0 = −g(R(Y, Jv)Jv, Y ) = g(R(JY, v)Jv, Y )
= g(R(Jv, Y )JY, v) = −g(R(v, JY )JY, v) (6.17)
= g(R(JY, v)v, JY )
for any Y satisfying g(Y, v) = 0. By combining (6.14) and (6.17) we get

(vµ)g(X, X) = 0 (6.18)
for any tangent vector X satisfying g(X, v) = g(JX, v) = 0. Let p be a
fixed point in M n .
We divide the proof of statement (1) into two cases.
Case (a.i): v(p) is either space-like or time-like. In this case, there exist an
orthonormal basis e1 , . . . , e2n of Tp M n such that
e2 = Je1 , g(ei , ej ) = ǫi δij , v(p) = ce1 , (6.19)
where c 6= 0 and ǫi = ±1. It follows from (6.18) and (6.19) that vµ = 0
whenever n = dimC M n > 1.
On the other hand, we find from (6.13) that e2 µ = · · · = e2n µ = 0.
Hence we have U µ = 0, ∀U ∈ Tp M n .
Case (a.ii): v(p) is light-like. It follows from (1.6), (1.30) and ∇J = 0 that
R(X, Jv)v = (Xµ)Jv − ((Jv)µ)X (6.20)
for X ∈ T M n . Taking the inner product of (6.20) with v gives
0 = ((Jv)µ)g(X, v) (6.21)
Since v(p) is a light-like vector, there exists another light-like vector u at p
such that g(u, v) = −1. Thus we also have (Jv)µ = 0. Now, by combining
this with (6.20) we find
R(X, Jv)v = (Xµ)Jv. (6.22)
Similarly, we find from (1.6), (1.30) and (6.22) that
R(JX, v)v = (JXµ)v. (6.23)
Since R(X, Jv) = −R(JX, v), (6.23) gives
(Xµ)Jv + (JXµ)v = 0.
Thus Xµ = JXµ = 0 since v and Jv are linearly independent. Hence we
also have U µ = 0 for any vector U ∈ Tp M n . Because p can be chosen
to be any point with v(p) 6= 0, this shows that µ is constant. Therefore
M n admits no non-trivial concircular vector fields if n > 1. This proves
statement (a).
To prove statement (b) let us assume that M n is a pseudo-Kähler man-
ifold with n = 1. So, M 1 is either space-like or time-like.
Case (b.i): M 1 is space-like. Consider the complex projective line CP 1 (4)

of constant curvature 4. Let z = x + iy be a local complex coordinate on
CP 1 (4) so that the metric tensor of CP 1 (4) is given by
dzdz̄
g= . (6.24)
(1 + z z̄)2
Consider the function ϕ defined by
1 − z z̄
ϕ= .
1 + z z̄
It is direct to verify that the gradient vector of ϕ satisfies
∇X grad ϕ = −4ϕX.
Therefore grad ϕ is a non-trivial concircular vector field on CP 1 (4).
1
Case (b.ii): M 1 is time-like. Let CP (4) denote the unit disc D = {z ∈ C :
z z̄ < 1} with the time-like metric:
−dzdz̄
g= . (6.25)
(1 − z z̄)2
Consider the function
1 + z z̄
ψ=
1 − z z̄
1
on CP (4). Then, by a direct computation, we derive ∇X grad ψ = µX
with µ = −4ψ. Therefore grad ψ is a non-trivial concircular vector field.
This proves statement (b).
The following corollary is an immediate consequence of Theorem 6.1.
Corollary 6.1. Every concircular vector field on a pseudo-Kaehler mani-
fold M n with n > 1 is a concurrent vector field.
6.3 Pseudo-Kähler submanifolds
Based on the behavior of the tangent bundle of pseudo-Riemannian sub-

manifolds under the action of the complex structure J of a pseudo-Kähler
manifold, there are several typical interesting families of submanifolds,
namely, complex, purely real, totally real, CR and slant submanifolds.
Definition 6.4. Let N be a pseudo-Riemannian submanifold of a pseudo-
Kähler manifold (M̃ , J, g). A point x ∈ N is called a complex point if
J(Tx N ) = Tx N.
A pseudo-Riemannian submanifold N of a pseudo-Kähler manifold is called
a complex submanifold if every point on N is complex.
Proposition 6.3. A complex submanifold of a pseudo-Kähler manifold is

pseudo-Kählerian with respect to its induced structures. Moreover, its sec-
ond fundamental form σ and shape operator A satisfy
σ(JX, Y ) = σ(X, JY ) = Jσ(X, Y ), (6.26)
AJξ = JAξ , JAξ = −Aξ J, (6.27)
for X, Y tangent to N and ξ normal to N .
Proof. Let N be a complex submanifold of a pseudo-Kähler manifold
M with complex structure J and pseudo-Kähler metric g. Obviously, it
follows from (6.1) that N is a pseudo-Hermitian manifold with respect to
the induced metric and the induced complex structure, also denoted by g
and J. respectively. For any vector fields X, Y ∈ X(N ) we have
˜ X (JY ) = ∇X (JY ) + σ(X, JY ),
∇ (6.28)
where ∇ ˜ and ∇ are the Levi-Civita connection of N and M , respectively.
On the other hand, since M is pseudo-Kählerian, ∇J ˜ = 0. Thus
˜ ˜
∇X (JY ) = J(∇X Y ) = J∇X Y + Jσ(X, Y ). (6.29)
Comparing (6.28) and (6.29) gives ∇J = 0 and σ(X, JY ) = Jσ(X, Y ).
Hence, N is also a pseudo-Kählerian manifold. Further, by symmetry of σ
we have (6.26).
It is easy to verify that (6.27) follows from (2.14) and (6.26).
By a pseudo-Kähler submanifold we mean a complex submanifold of a
pseudo-Kähler manifold with its induced pseudo-Kählerian structure.
Definition 6.5. A pseudo-Riemannian submanifold N of a pseudo-Rie-
mannian manifold M is called austere if there exists a local orthonormal
frame {e1 , . . . , en , e1∗ , . . . , en∗ } on N such that the second fundamental
form of N satisfies [Harvey and Lawson (1982)]
ǫi∗ σ(ei∗ , ei∗ ) + ǫi σ(ei , ei ) = 0, i = 1, . . . , n. (6.30)
Obviously, every austere submanifold is minimal.
Proposition 6.4. Every pseudo-Kähler submanifold of a pseudo-Kähler
manifold is austere.
Proof. Assume that N is a pseudo-Kähler submanifold of a pseudo-
Kähler manifold M with n = dimC N . According to (6.1) we may choose
a local orthonormal frame {e1 , . . . , en , e1∗ , . . . , en∗ } such that ei∗ = Jei for
i = 1, . . . , n on N . Then we have
ǫi = hei , ei i = hei∗ , ei∗ i = ǫi∗ . (6.31)
Thus, it follows from (6.27) that (6.30) holds. Hence N is austere.
Definition 6.6. A pseudo-Riemannian submanifold N of a pseudo-Rie-

mannian manifold is called isotropic at a point p ∈ N if hσ(v, v), σ(v, v)i is
independent of the choice of the unit vector v ∈ Tp N at p, where σ is the
second fundamental form of N .
Also, the submanifold N is called isotropic if it is isotropic at each
point. Moreover, N is called null-isotropic if hσ(v, v), σ(v, v)i = 0 for any
unit tangent vector v ∈ T N .
Proposition 6.5. A pseudo-Kähler submanifold of an indefinite complex
space form Msm (4c) has constant holomorphic sectional curvature 4c if and
only if N is null-isotropic.
Proof. Let N be a pseudo-Kähler submanifold of an indefinite complex

space form Msm (c). For any unit vector v ∈ T N , it follows from Gauss’
equation and Proposition 6.3 that the holomorphic sectional curvature of
N satisfies
H(v) = 4c − 2 hσ(v, v), σ(v, v)i . (6.32)
Therefore N has constant holomorphic sectional curvature 4c if and only if
hσ(v, v), σ(v, v)i = 0 for any unit vector v.
It also follows from Gauss’ equation and Propositions 6.3 and 6.4 that
the scalar curvature τ of a pseudo-Kähler submanifold N of Msm (c) satisfies
2τ = 4n(n + 1)c − Sσ , (6.33)
where
X2n
Sσ = ǫi ǫj hσ(ei , ej ), σ(ei , ej )i
i,j=1
and e1 , . . . , e2n is an orthonormal basis of N . Consequently, we also have
the following.
Proposition 6.6. The scalar curvature of a pseudo-Kähler submanifold of

an indefinite complex space form Msm (4c) is equal to 2n(n + 1)c if and only
if Sσ = 0 holds identically.
Remark 6.2. Contrast to Kählerian case, null-isotropic pseudo-Kähler
submanifolds are not necessary totally geodesic. A simple example is the
flat pseudo-Kähler submanifold Cnt embedded in Cn+2 t+1 defined by
n n
!
X X
zj2 , z1 , . . . , zn , zj2 ,
j=1 j=1
where zj = xj + iyj for j = 1, . . . , n. This non-totally geodesic example
satisfies Sσ = 0 identically.
6.4 Segre and Veronese embeddings
There are some well-known examples of Kähler submanifolds in complex

projective spaces via the Veronese and Segre embeddings.
The Veronese embedding vn : CP n (2) → CP n(n+3)/2 (4) is a Kählerian
embedding defined by homogeneous monomials of degree 2:
s !
2
√ 2 αi αj 2
(z0 , . . . , zn ) 7→ z0 , 2z0 z1 , . . . , z z , . . . , zn (6.34)
αi !αj ! i j
with αi + αj = 2. For n = 1, this gives rise to the quadric curve in CP 2 (4)

defined by
( 2
)
X
2 2
Q1 = (z0 , z1 , z2 ) ∈ CP : zj = 0 .
j=0
The Veronese embedding can be extended to α-th Veronese embedding

n+α
vnα : CP n (4/α) → CP ( α )−1 (4) with α ≥ 2 defined by
r !
α √ α−1 α α0 αn α
(z0 , . . . , zn ) 7→ z0 , αz0 z1 , . . . , z · · · zn , . . . , zn
α0 ! · · · αn ! 0
with α0 + · · · + αn = α.
The Segre embedding Shp : CP h (4)×CP p (4) → CP h+p+hp (4) is defined
by (cf. [Segre (1891)])
Shp (z0 , . . . , zh , w0 , . . . , wp )
= (z0 w0 , . . . , z0 wp , . . . , zh w0 , . . . , zh wp ) (6.35)

= zj wt 0≤j≤h,0≤t≤p ,
where (z0 , . . . , zh ) and (w0 , . . . , wp ) are the homogeneous coordinates of

CP h (4) and CP p (4), respectively. It is well-known that the Segre embed-
ding Shp is also a Kählerian embedding.
When h = p = 1, the Segre embedding gives rise to the complex quadric
surface, Q2 = CP 1 (4) × CP 1 (4) in CP 3 (4), defined by
 
 X3 
Q2 = (z0 , z1 , z2 , z3 ) ∈ CP 3 : zj2 = 0 . (6.36)
 
j=0
The Segre embedding can be naturally extended to product embeddings

of arbitrary number of complex projective spaces as follows (see [Chen and
Kuan (1985)]).
Let (z0i , . . . , zni i ), (1 ≤ i ≤ s) denote the homogeneous coordinates of

Qs
CP ni (4) and put N = i=1 (ni + 1) − 1. Consider the map
Sn1 ···ns : CP n1 (4) × · · · × CP ns (4) → CP N (4)
defined by
Sn1 ···ns (z01 , . . . , zn1 1 , . . . , z0s , . . . , zns s )
(6.37)
= (zi11 · · · zisj )1≤i1 ≤n1 ,...,1≤is ≤ns .
This map Sn1 ···ns is a Kählerian embedding, which is the simplest Kähler
embedding from product of s algebraic manifolds into complex projective
spaces.
6.5 Purely real submanifolds of pseudo-Kähler manifolds
Let N be a pseudo-Riemannian submanifold of a pseudo-Kähler manifold

M̃ . For each vector X ∈ T N , we put
JX = P X + F X, (6.38)
where P X and F X denote the tangential and the normal components of
JX, respectively. Then P is an endomorphism of T N and F is a T ⊥ N -
valued 1-form. Similarly, for each normal vector ξ of N , we put
Jξ = tξ + f ξ, (6.39)
where tξ and f ξ are the tangential and the normal components of Jξ,
respectively. Then f is an endomorphism of T ⊥ N and t is a T N -valued
1-form on T ⊥ N . It follows from (6.1) and (6.38) that
hP X, Y i = − hX, P Y i (6.40)
for X, Y ∈ T N . Thus

2

P X, Y = X, P 2 Y = − hP X, P Y i . (6.41)
Define ∇P and ∇F by
(∇X P )Y = ∇X (P Y ) − P (∇X Y ), (6.42)
(∇X F )Y = DX (F Y ) − F (∇X Y ), (6.43)
where ∇ is the Levi-Civita connection of N .
Proposition 6.7. Let N be a pseudo-Riemannian submanifold of a pseudo-

Kähler manifold M̃ . Then ∇P = 0 if and only if the shape operator satisfies
AF Y Z = AF Z Y (6.44)
for vectors Y, Z tangent to N .
Proof. ˜ = 0 that
It follows from (2.5), (2.13), (6.38), (6.39) and ∇J
˜ X (JY ) − J ∇
0 =∇ ˜ XY
= ∇X (P Y ) + σ(X, P Y ) − AF Y X + DX (F Y ) (6.45)
− P (∇X Y ) − F (∇X Y ) − tσ(X, Y ) − f σ(X, Y )
for X, Y ∈ T N .
The tangential components of (6.45) yield
(∇X P )Y = AF Y X + tσ(X, Y ).
Thus ∇P = 0 holds identically if and only if AF Y X + tσ(X, Y ) = 0, which

is equivalent to (6.44).
Proposition 6.8. Let N be a pseudo-Riemannian submanifold of a pseudo-

Kähler manifold M̃ . Then the following three statements are equivalent:
(a) ∇F = 0;
(b) F ∇X Y = DX F Y for X, Y ∈ T N ;
(c) σ(X, P Y ) = f σ(X, Y ) for X, Y ∈ T N .
Proof. The equivalence of (a) and (b) is an easy consequence of (6.43).

By comparing the normal components of (6.45) we find
σ(X, P Y ) + DX (F Y ) − F (∇X Y ) − f σ(X, Y ) = 0. (6.46)
If ∇F = 0, then by (b) and (6.46) we have (c). Conversely, if (c) holds,

then (6.46) reduces to (b). Thus we get ∇F = 0.
Definition 6.7. A pseudo-Riemannian submanifold N of a pseudo-Kähler

manifold M̃ is called purely real if the complex structure J on M̃ carries
the tangent bundle of N into a transversal bundle, i.e., J(T N ) ∩ T N = {0}.
Remark 6.3. A purely real submanifold N of a pseudo-Kähler manifold

contains no complex points. When dim N = 2, the converse is also true.
Next, we present a basic property of purely real surfaces in an indefinite

Kähler manifold. Suppose that N is a Lorentz surface in an indefinite
Kähler manifold M . Let us choose a local frame {e1 , e2 } on N such that
he1 , e1 i = he2 , e2 i = 0, he1 , e2 i = −1. (6.47)
We call a frame a pseudo-orthonormal frame of N .

It follows from (6.1) and (6.38) that hP u, ui = 0 holds for any u ∈ T N .

Therefore, by applying (6.47), we conclude that there is a function α such
that
P e1 = (sinh α)e1 , P e2 = −(sinh α)e2 . (6.48)
If we put
e3 = (sech α)F e1 , e4 = (sech α)F e2 , (6.49)
Je1 = sinh αe1 + cosh αe3 ,
(6.50)
Je2 = − sinh αe2 + cosh αe4 .
By applying J 2 = −I, (6.47) and (6.50), we find
Je3 = − cosh αe1 − sinh αe3 ,
(6.51)
Je4 = − cosh αe2 + sinh αe4 ,
he3 , e3 i = he4 , e4 i = 0, he3 , e4 i = −1. (6.52)
For a Lorentz surface N in a Lorentzian Kähler surface M̃ , a frame
{e1 , e2 , e3 , e4 } given above is called an adapted pseudo-orthonormal frame.
Since cosh α ≥ 1, we conclude the following result from (6.50).
Proposition 6.9. Every Lorentz surface in any indefinite Kähler manifold
is purely real.
For Lorentz surfaces in a Lorentzian Kähler surface, we have
Theorem 6.2. [Chen (2009a)] The equation of Ricci is a consequence of
the equations of Gauss and Codazzi for a Lorentz surface in any Lorentzian
Kähler surface.
Remark 6.4. The same phenomena occurs for purely real surfaces in any
Kähler surface as well [Chen (2010c)].
6.6 Totally real and Lagrangian submanifolds
Definition 6.8. Let N be a pseudo-Riemannian submanifold of a pseudo-

Kähler manifold (M̃ , J, g). A point x ∈ N is called a totally real point if
J(Tx N ) ⊂ Tx⊥ N . The submanifold N is called totally real if every point of
N is a totally real point.
A totally real submanifold N is called Lagrangian if dim N = dimC M̃ .
The same definitions applied to totally real and Lagrangian submanifolds
of an almost Hermitian manifold.
The study of totally real and Lagrangian submanifolds of Kähler man-

ifolds from differential geometric point of view was initiated in the early
1970’s. Such submanifolds have been studied extensively during the last
four decades.
Since every curve in a pseudo-Kähler manifold is totally real, we shall
only consider totally real submanifolds of dimension ≥ 2.
Some basic properties of Lagrangian submanifolds are the following.
Lemma 6.2. Let N be a Lagrangian submanifold of a pseudo-Kähler man-
ifold M̃ . If σ̄ = Jσ, then for X, Y, Z tangent to N we have
(1) g(σ̄(X, Y ), Z) is totally symmetric, i.e.,
g(σ̄(X, Y ), Z) = g(σ̄(Y, Z), X) = g(σ̄(Z, X), Y );
(2) the induced Levi-Civita connection ∇ and the normal connection D of
N satisfy DX (JY ) = J∇X Y ;
(3) J(R(X, Y )Z) = RD (X, Y )JZ;
(4) N is flat if and only N has flat normal connection;
(5) if M̃ is an (indefinite) complex space form, we have
(∇σ̄)(X, Y, Z) := ∇X σ̄(Y, Z) − σ̄(∇X Y, Z) − σ̄(Y, ∇X Z) (6.53)
is totally symmetric.
Proof. Let X, Y, Z be tangent to N . Then
˜ XY = ∇
J∇X Y + Jσ(X, Y ) = J ∇ ˜ X (JY ) = −AJY X + DX JY,
which implies (2) and σ̄(X, Y ) = −AJY X. Thus, after applying (2.14) and
symmetry of σ we obtain (1).
(3) follows from (2); and (4) is an immediate consequence of (3).
If N is a Lagrangian submanifold of an (indefinite) complex space form
M̃sn (4c), then Codazzi’s equation implies that
(∇¯ X σ)(Y, Z) = (∇
¯ Y σ)(X, Z).
Hence, by applying (2) and σ̄ = Jσ, we have (6.53).
The equations of Gauss, Codazzi and Ricci for Lagrangian submanifolds
N in an (indefinite) complex space form M̃sm (4c) are given respectively by

R(X, Y ; Z, W ) = Aσ(Y,Z) X, W − Aσ(X,Z) Y, W (6.54)
+ c (hX, W i hY, Zi − hX, Zi hY, W i),
(∇σ̄)(X, Y, Z) = (∇σ̄)(Y, X, Z), (6.55)
D
R (X, Y ; JZ, JW ) = h[AJZ , AJW ]X, Y i (6.56)
+ c (hX, W i hY, Zi − hX, Zi hY, W i)
for X, Y, Z, W tangent to N .
Due to Lemma 6.2(2), the equation (6.56) of Ricci is nothing but the
equation (6.54) of Gauss. Consequenlty, after applying a result obtained
in [Eschenburg and Tribuzy (1993)], we obtain the following existence and
uniqueness theorems for Lagrangian submanifolds in (indefinite) complex
space forms.
Theorem 6.3. Let (Nsn , g) be a simply-connected pseudo-Riemannian n-

manifold with index s ≥ 0. If σ̄ is a T Nsn -valued symmetric bilinear form
on Nsn such that
(a) g(σ̄(X, Y ), Z) is totally symmetric,
(b) (∇σ̄)(X, Y, Z) is totally symmetric,
(c) R(X, Y )Z = cg(Y, Z)X − cg(X, Z)Y + σ̄(σ̄(Y, Z), X) − σ̄(σ̄(X, Z), Y ),
then there is a Lagrangian isometric immersion L : Nsn → M̃sn (4c) of Nsn
into a complete simply-connected (indefinite) complex space form M̃sn (4c)
whose second fundamental form σ is given by σ(X, Y ) = J σ̄(X, Y ).
Theorem 6.4. Let L1 , L2 : Nsn → M̄sn (4c) be two Lagrangian isometric

immersions of a pseudo-Riemannian manifold Nsn with second fundamental
forms σ1 and σ2 . If
hσ1 (X, Y ), JL1⋆ Zi = hσ2 (X, Y ), JL2⋆ Zi
holds for vector fields X, Y, Z tangent to Nsn , then there exists an isometry
Φ of M̃sn (4c) such that L1 = Φ ◦ L2 .
6.7 Totally umbilical and H-umbilical submanifolds
Another basic properties of Lagrangian submanifold is the following (cf.

[Chen and Ogiue (1974b); Chen (1997a)]).
Proposition 6.10. Every totally umbilical Lagrangian submanifold N of a

Kähler manifold M̃ is totally geodesic.
Proof. Assume that N is totally umbilical submanifold of a Kähler man-

ifold M̃ . Then we have
σ(X, Y ) = hX, Y i H (6.57)
for vectors X, Y ∈ T N . Hence we get
AH = hH, Hi I, (6.58)
where I is the identity map. Since N is Lagrangian, there exists a tangent

vector field U of N satisfying JU = H.
Let V be a tangent vector field perpendicular to U . Then it follows
from AJX Y = AJY X and (6.58) that
0 = hAJV U, V i = hAJU V, V i = hAH V, V i = hH, Hi hV, V i .
Therefore H = 0 holds identically. Consequently, N is a totally geodesic
submanifold of M̃ .
Remark 6.5. Proposition 6.10 is false if N is a totally umbilical, totally
real submanifold of a Kähler manifold M̃ .
For instance, if N is a non-totally geodesic, totally umbilical hypersur-
face of the real projective n-space RP n (1) (resp., Euclidean n-space En or
hyperbolic n-space H n (−1)), then N can be immersed in CP n (4) (resp.,
in Cn or in CH n (−4)) as a non-totally geodesic, totally umbilical, totally
real submanifold of CP n (4) (resp., of Cn or of CH n (−4)) via the standard
totally geodesic Lagrangian embedding RP n (1) ⊂ CP n (4) (resp., En ⊂ Cn
or H n (−1) ⊂ CH n (−4)).
The notion of H-umbilical submanifolds was introduced in [Chen
(1997a)] as follows.
Definition 6.9. A non-totally geodesic Lagrangian submanifold of a
Kaehler manifold M̃ n is called H-umbilical if its second fundamental form
σ takes the following simple form:
σ(e1 , e1 ) = λJe1 , σ(e1 , ej ) = µJej ,
σ(e2 , e2 ) = · · · = σ(en , en ) = µJe1 , (6.59)
σ(ej , ek ) = 0, 2 ≤ j 6= k ≤ n
for some functions λ and µ with respect to some suitable orthonormal local
frame field.
It is easy to see that an H-umbilical submanifold satisfies the following
two conditions:
(a) JH is an eigenvector of the shape operator AH and
(b) the restriction of AH to (JH)⊥ is proportional to the identity map.
Since σ̄ = Jσ of any Lagrangian submanifold of a Kähler manifold M̃
is totally symmetric according to Lemma 6.2(1), H-umbilical submanifolds
are indeed the simplest Lagrangian submanifolds satisfying both conditions
(a) and (b). Thus H-umbilical submanifolds can be regarded as the simplest
Lagrangian submanifolds in M̃ next to the totally geodesic ones.
6.8 Warped products, H-umbilical submanifolds and

complex extensors
There exist ample examples of H-umbilical submanifolds. The simplest

ones can be constructed via complex extensors introduced in [Chen (1997b)]
which are defined as follows.
Definition 6.10. For a given immersion G : M → Em of a manifold M

into the Euclidean m-space Em and for a given unit speed curve F : I → C
in the complex plane, we may extend the immersion G : M → Em to an
immersion φ of I × M into Cm :
φ = F ⊗ G : I × M → C ⊗ Em = Cm (6.60)
defined by
(F ⊗ G)(s, x) = F (s) ⊗ G(x), s ∈ I, x ∈ M. (6.61)
We call this extension the complex extensor of G via F .
Lemma 6.3. Let G : M → Em be an isometric immersion of a Riemannian

manifold into the Euclidean m-space Em and let F : I → C be a unit speed
curve in the complex plane. Then the complex extensor φ = F ⊗ G is totally
real if and only if either G is spherical or F (s) = cf (s) for some c ∈ C and
real-valued function f .
Proof. It is direct to verify that the complex extensor φ = F ⊗ G is

totally real if and only if, for any s ∈ I, x ∈ M and Y ∈ Tx M, we have
Re(iF (s)F̄ ′ (s)) hG(x), Y i = 0, (6.62)
where F̄ ′ is the complex conjugate of F ′ and Re(iF F̄ ′ ) is the real part of
iF F̄ ′ . So, we have either Re(iF (s)F̄ ′ (s)) = 0 for all s ∈ I or hG(x), Y i = 0
for all x ∈ M, Y ∈ Tx M .
If the first case occurs, F = cf (s) for some c ∈ C; if the second case
occurs, G is spherical.
The following result from [Chen (1997b)] shows that there are ample
examples of H-umbilical submanifolds.
Theorem 6.5. Let ι : S n−1 → En be the inclusion of the unit hypersphere

of En (centered at the origin). Then every complex extensor of ι via a unit
speed curve F in C is an H-umbilical submanifold of Cn , unless F (s) =
(s + a)c for some real number a and some unit complex number c; in this
case, φ is totally geodesic.
Proof. Lemma 6.3 implies that every complex extensor of the unit hy-
persphere centered at the origin in En is a Lagrangian submanifold in Cn .
Now we prove that every complex extensor of the unit hypersphere of
En is an H-umbilical submanifold of Cn as follows.
For a unit speed curve F : I → Cn , we may put F ′ (s) = e iζ(s) for some
real-valued function f on I. Thus F takes the following form:
Z s
F (s) = e iζ(t) dt (6.63)
a
for some real number a. Let {x2 , . . . , xn } be a local coordinate system on
S n−1 . Then {s, x2 , . . . , xn } is a local coordinate chart on I × S n−1 .
Because ι is the unit hypersphere, (6.61) and (6.63) imply
φs = e iζ(s) ⊗ ι, Y φ = F ⊗ Y, (6.64)
and
φss = iζ ′ (s)e iζ(s) ⊗ ι, Y φs = e iζ(s) ⊗ Y,
(6.65)
Y Zφ = F ⊗ ∇Y Z − hY, Zi (F ⊗ ι),
where Y, Z are vectors fields tangent to the second component of I × S n−1 .
Since ι is the unit hypersphere in En , (6.64) implies that e1 = ∂/∂s is
a unit vector field tangent to the first component of I × S n−1 ; moreover,
for each Y tangent to the second component of I × S n−1 , φs and Y φ are
orthogonal. Thus, by applying (6.64) and (6.65) we may conclude that
the second fundamental form of the complex extensor satisfies (6.59) with
respect to a suitable orthonormal local frame field. Moreover, we have

iζ
′ e , iF
λ = ζ (s), µ = , (6.66)
hhF, F ii
where hh , ii denotes the canonical scalar product of the complex plane.
Therefore φ = F ⊗ ι is an H-umbilical submanifold unless φ is totally
geodesic, which occurs only when F (s) = (s + a)c for some a ∈ R and some
unit complex number c ∈ C.
Remark 6.6. It follows from (6.64) that the complex extensor F ⊗ ι of
ι : S n−1 → En via the curve F given by (6.63) is the warped product
I ×f S n−1 with warping function f = |F |.
Now, we provide examples of complex extensors which satisfy (6.59)

with λ = 3µ, λ = 2µ, λ = µ and λ = 0, respectively.
Example 6.4. (Whitney’s sphere). Let w : S n → Cn be defined by
1 + iy0
w(y0 , y1 , . . . , yn ) = (y1 , . . . , yn ), (6.67)
1 + y02
for y02 + y12 + · · · + yn2 = 1. Then w is a Lagrangian immersion, called the

Whitney immersion. This immersion is the complex extensor of ι : S n−1 →
En via F , which is an arc-length reparametrization of ψ : I → C given by
sin ϕ + i sin ϕ cos ϕ
ψ(ϕ) = . (6.68)
1 + cos2 ϕ
The Whitney n-sphere is an H-umbilical submanifold satisfying (6.59) with
λ = 3µ. Up to homothetic transformations, the Whitney sphere is the only
H-umbilical submanifold in Cn satisfying λ = 3µ.
A well known result of [Gromov (1970)] states that embedded compact

Lagrangian submanifolds in Cns are not simply-connected (for a complete
proof of this fact, see [Sikorav (1986)]). Example 6.4 shows that this result
is false if the Lagrangian submanifolds were immersed but not embedded.
Example 6.5. (Lagrangian pseudo-spheres). For a real positive number b,

let F : R → C be the unit speed curve given by
e2ibs + 1
F (s) = . (6.69)
2ib
With respect to the induced metric, the complex extensor of ι : S n−1 → En
via F is a Lagrangian immersion of an open portion of S n (b2 ) of constant
curvature b2 in Cn which is simply called a Lagrangian pseudo-sphere. This
gives an H-umbilical submanifold satisfying λ = 2µ.
Remark 6.7. It was proved in [Castro and Urbano (2004)] that Lagrangian
pseudo-spheres are the only branched Lagrangian immersions of an ordinary
sphere in C2 with constant length mean curvature vector. For an extension
of these results in 2-dimensional complex space forms, see [Li et al. (2008)].
Example 6.6. (Lagrangian-umbilical submanifold ) For a 6= 0, put

Z s
F (s) = e−ia ln t dt, (6.70)
Rs
where f (t)dt is an anti-derivative of f (s). The complex extensor of
n−1
ι:S → En via F is an H-umbilical submanifold of Cn satisfying λ = µ.
Such an H-umbilical submanifold is simply called a Lagrangian-umbilical
submanifold in [Chen (1997b)].
Example 6.7. Let a ∈ C and θ be a real number such that ae−iθ ∈ / R.

iθ
Then the complex extensor of the unit hypersphere via F (s) = a + e s is
an H-umbilical submanifold satisfying λ = 0.
6.9 Classification of H-umbilical submanifolds
In this section we provide some classification results for H-umbilical sub-

manifolds in the complex Euclidean n-space.
Theorem 6.6. [Chen (1997b)] Let L : N → Cn be a Lagrangian isometric

immersion. Then, up to rigid motions of Cn , L is a Lagrangian pseudo-
sphere if and only if L is an H-umbilical immersion satisfying
σ(e1 , e1 ) = 2bJe1 , σ(e2 , e2 ) = · · · = σ(en , en ) = bJe1 ,
(6.71)
σ(e1 , ej ) = bJej , σ(ej , ek ) = 0, 2 ≤ j 6= k ≤ n,
for some non-trivial function b with respect to a suitable orthonormal local
frame field. Moreover, in this case, b is a nonzero constant.
Proof. If L : N → Cn is an H-umbilical immersion satisfying (6.71), then

the covariant derivative of the second fundamental form σ of L satisfies

∇ ¯ ej σ (e1 , e1 ) = 2(ej b)Je1 ,
¯ e1 σ (ej , e1 ) = (e1 b)Jej , ∇ (6.72)
for j = 2, . . . , n. Hence Codazzi’s equation and (6.71) imply that b is a
nonzero constant. Thus, by Gauss’s equation N is a real-space-form of
constant curvature b2 . Therefore N is locally an open portion of the warped
π π
product I ×cos(bs)/b S n−1 with I = (− 2b , 2b ) whose metric is given by
cos2 (bs)
g = ds2 +
g0 , (6.73)
b2
where g0 is the standard metric on the unit (n − 1)-sphere S n−1 . With
respect to a spherical coordinate system {u2 , . . . , un } on S n−1 , we have
g0 = du22 + cos2 u2 du23 + · · · + cos2 u2 · · · cos2 un−1 du2n . (6.74)
From (6.73) and (6.74) we obtain
∂ ∂ ∂
∇∂ = 0, ∇ ∂ = −b tan(bs) ,
∂s ∂s ∂s ∂u ∂uk
k
∂ sin(2bs) ∂
∇ ∂ = ,
∂u2 ∂u 2b ∂s
2
∂ ∂
∇ ∂ = − tan ui , 2 ≤ i < j;
∂ui ∂u ∂uj
j (6.75)
j−1
Y
∂ sin(2bs) ∂
∇ ∂ = cos2 uℓ
∂uj ∂uj 2b ∂s
ℓ=2
j−1 j−1
!
X sin(2uk ) Y ∂
+ cos2 ul , j ≥ 3.
2 ∂uk
k=2 l=k+1
By (6.71), (6.75) and Gauss’ formula, we find

∂L ∂ 2L
Lss = 2ib Ls Ls =, Lss = , (6.76)
∂s ∂s2
Y Ls = (ib − b tan(bs))Y, (6.77)
Y ZL = ib hY, Zi Ls + L∗ (∇Y Z), (6.78)
n−1
where Y, Z are vector fields tangent to the second component S of the
n−1
warped product and ∇ is the Levi-Civita connection of S .
Let {u2 , . . . , un } be a spherical coordinate system on S n−1 . By solving
(6.76) we obtain
L(s, u2 , . . . , un ) = A(u2 , . . . , un )e2ibs + B(u2 , . . . , un ), (6.79)
for Cn -valued functions A, B. It follows from (6.77) and (6.79) that
Auj = Buj j = 2, . . . , n, (6.80)
where Auj is the partial derivative of A with respect to uj . Condition
(6.80) gives B = A + b0 where b0 is a constant vector in Cn . By applying
a translation if necessary, we may assume b0 = 0. Hence B = A. Therefore
L(s, u2 , . . . , un ) = A(u2 , . . . , un )(e2ibs + 1), (6.81)
which implies
Ls = 2ibAe2ibs , Lu2 u2 = Au2 u2 (e2ibs + 1). (6.82)
On the other hand, by (6.73)-(6.75), (6.78), (6.81) and (6.82), we find
Lu2 u2 = −A(e2ibs + 1). (6.83)
From (6.82) and (6.83) we get
Au2 u2 = −A. (6.84)
Therefore
A = b1 sin u2 + b2 cos u2 (6.85)
n
for some C -valued functions b1 , b2 of u3 , . . . , un .
If n = 2, then b1 and b2 are constant vectors in C2 . Thus (6.81) and
(6.85) yield
L(s, u2 ) = (e2ibs + 1)(b1 sin u2 + b2 cos u2 ). (6.86)
2
Because M is Lagrangian in C , using (6.73) and (6.74) we may choose the
following initial conditions: Ls (0, 0) = (1, 0) and Lu2 = (0, 1/(2bi)). From
(6.86) and the initial conditions we get
e2ibs + 1
L= (cos u2 , sin u2 ). (6.87)
2bi
Thus L is a 2-dimensional Lagrangian pseudo-sphere.

If n > 2, by putting Y = ∂/∂u2, Z = ∂/∂u3 in (6.78) and Y = Z =
∂/∂u3 in (6.78) and by applying (6.81) and (6.85), we find as before that
b1 = b1 (u4 , . . . , un ),
(6.88)
b2 = b3 (u4 , . . . , un ) sin u3 + b4 (u4 , . . . , un ) cos u3 .
Continuing such procedure (n − 1)-times, we arrive at
n
L = (e2ibs + 1) c1 sin u2 + c2 sin u3 cos u2 + · · ·
n−1
Y n
Y o (6.89)
· · · + cn−1 sin un cos uj + cn cos uj
j=2 j=2
for some vectors c1 , . . . , cn ∈ C . Since N is Lagrangian in Cn , it follows

n
from (6.73) and (6.74) that we may choose the initial conditions:
Ls (0, . . . , 0) = (1, 0, . . . , 0),

1
Lu2 (0, . . . , 0) = 0, ,...,0 ,
2bi
(6.90)
......

1
Lun (0, . . . , 0) = 0, . . . , 0, .
2bi
By using (6.87) and (6.88) we obtain
 
n n−1
e2ibs + 1  Y Y
L= cos uj , sin u2 , sin u3 cos u2 , · · · , sin un cos uj  .
2ib j=2 j=2
Hence, up to rigid motions of Cn , L is a Lagrangian pseudo-sphere.

Conversely, if L is a Lagrangian pseudo-sphere, then it is direct to verify
that L is an H-umbilical submanifold in Cn satisfying (6.59) and (6.66) with
e2ibs + 1
F (s) = , ζ(s) = 2bs,
2ib
which implies (6.71).
Theorem 6.7. [Chen (1997b)] If L : N → Cn is an H-umbilical submani-

fold with n ≥ 3. Then we have
(i) If N is of constant curvature, then either M is a flat space or, up to
rigid motions of Cn , L is a Lagrangian pseudo-sphere.
(ii) If N contains no open subset of constant curvature, then, up to rigid
motions of Cn , L is a complex extensor of the unit hypersphere of En .
Proof. Let n ≥ 3 and L : N → Cn be an H-umbilical immersion whose

second fundamental form satisfies
σ(e1 , e1 ) = λJe1 , σ(e1 , ej ) = µJej ,
σ(e2 , e2 ) = · · · = σ(en , en ) = µJe1 , (6.91)
σ(ej , ek ) = 0, 2 ≤ j 6= k ≤ n,
for functions λ and µ with respect to a suitable orthonormal local frame.
If N is of constant curvature, then (6.91) implies µ(λ − 2µ) = 0. If
µ ≡ 0, then N is a flat space. If µ 6= 0 on N , then we have λ = 2µ 6= 0 on
a nonempty open subset V of N . Therefore, according to Theorem 6.6, λ
and µ are nonzero constants on V . Consequently, by continuity, we obtain
V = N . If we put µ = b, then by applying Theorem 6.6 again we conclude
that, up to rigid motions of Cn , N is a Lagrangian pseudo-sphere. This
proves statement (i).
For statement (ii), let us assume that N contains no open subset of
constant curvature. In this case
U = { p ∈ N : µ(λ − 2µ) 6= 0 at p } (6.92)
is an open dense subset of N .
Let e1 , . . . , en be an orthonormal local frame field on N satisfying (6.91)
and ω 1 , . . . , ω n the dual 1-forms of e1 , . . . , en . Let
A
(ωB ), A, B = 1, . . . , n, 1∗ , . . . , n∗ ,
be the connection forms defined by
n
X n
X ∗
˜ i=
∇e ωij ej + ωij ej ∗ ,
j=1 j=1
n n (6.93)
X X ∗
˜ i∗ =
∇e ωij∗ ej + ωij∗ ej ∗ ,
j=1 j=1
∗
i∗
where ei∗ = Jei , ωij = −ωji , ωij∗ = −ω for i = 1, . . . , n. Then for the
j∗
Lagrangian submanifold N in Cn , we have
n
X
∗ ∗ ∗ ∗
ωji = ωij , ωij = ωij∗ , ωji = hijk ω k . (6.94)
k=1
From (6.91) and (6.94) we find

∗ ∗ ∗
ω11 = λω 1 , ωi1 = µω i , ωii = µω 1 ,
∗ (6.95)
ωji = 0, 2 ≤ i 6= j ≤ n.
By applying (6.91), (6.95) and Codazzi’s equation, we find

e1 µ = (λ − 2µ)ω12 (e2 ) = · · · = (λ − 2µ)ω1n (en ), (6.96)
ej λ = (2µ − λ)ωj1 (e1 ), j > 1, (6.97)
(λ − 2µ)ω1j (ek ) = 0, 1 < j 6= k ≤ n. (6.98)
ej µ = 3µω1j (e1 ), (6.99)
µω1j (e1 ) = 0, j > 1. (6.100)
Notice that (6.98) and (6.100) occur only for the case n ≥ 3. Since n ≥ 3,
(6.96), (6.99) and (6.100) imply

e1 µ
ω1j = ω j , ej λ = ej µ = 0, j = 2, . . . , n. (6.101)
λ − 2µ
ω1j (ek ) = 0, 1 < j 6= k ≤ n, (6.102)
It follows from (6.101) and Cartan’s structure equations that dω 1 = 0 and
∇e1 e1 = 0. Hence the integral curves of e1 are geodesics.
For j, k > 1, (6.102) yields h[ej , ek ], e1 i = ωk1 (ej ) − ωj1 (ek ) = 0. Thus
the distribution D⊥ = Span{e2 , . . . , en } is integrable. Let D denote the
distribution spanned by e1 . Then D is also integrable due to dim D = 1.
Because D and D⊥ are both integrable, there exists a local coordinate
∂
system {x1 , . . . , xn } satisfying (a) D is spanned by { ∂x 1
} and D⊥ is spanned
∂
by { ∂x 2
, . . . , ∂x∂ n } and (b) e1 = ∂x∂ 1 , ω 1 = dx1 ,
From (6.97), (6.99) and (6.100) we know that λ and µ depend only on
s (= x1 ). Furthermore, by (6.111) and the structure equations, we have
µ′
k ′ + k 2 = µ2 − λµ, k = , (6.103)
λ − 2µ
where µ′ denotes the differentiation of µ with respect to s.
From (6.91), Codazzi’s equation and a direct computation, we obtain

mu′
h∇X Y, e1 i = hX, Y i . (6.104)
2µ − λ
Thus D⊥ is a spherical distribution, i.e., D⊥ is an integrable distribution
whose leaves are totally umbilical with parallel mean curvature vector in
M . Moreover, by (6.91), (6.104) and Gauss’ equation, we know that each
leaf of D⊥ is of constant curvature µ2 + k 2 . Hence, by applying a result
of [Hiepko (1979)], U is a warped product I ×f (s) S n−1 , where S n−1 is the
unit (n − 1)-sphere and f (s) is the warping function. Moreover, each vector
tangent to I is in the distribution D and each vector tangent to S n−1 is in
the complementary distribution D⊥ .
With respect to a spherical coordinate system {u2 , . . . , un } on the unit

(n − 1)-sphere S n−1 , the metric on I ×f S n−1 is then given by
n
g = ds2 + f 2 (s) du22 + cos2 u2 du23 + · · ·
o (6.105)
+ cos2 u2 · · · cos2 un−1 du2n .
From (6.105) we obtain
∂
∇∂ = 0,
∂s ∂s
∂ f′ ∂
∇∂ = ,
∂s ∂u f ∂uk
k
∂ ∂
∇∂/∂u2 = −f f ′ ,
∂u2 ∂s
∂ ∂ (6.106)
∇ ∂ = − tan ui , 2 ≤ i < j,
∂ui ∂u ∂uj
j
Y j−1
∂ ∂
∇ ∂ = −f f ′ cos2 uℓ
∂uj ∂uj ∂s
ℓ=2
j−1 j−1
!
X sin 2uk Y ∂
+ cos2 ul , j > 2.
2 ∂uk
k=2 l=k+1
It follows from Codazzi’s equation, (6.91) and (6.106) that

f′ µ′
= k, k = . (6.107)
f λ − 2µ
Thus there is a real number c 6= 0 such that
Z s
f = c exp k(x)dx . (6.108)
By applying (6.105) and (6.106), we know that the sectional curvature of

the plane section spanned by ∂/∂u2, ∂/∂u3 is given by
∂ ∂ Rs
K ∧ = c−2 e−2 k(s)ds − k 2 . (6.109)
∂u2 ∂u3
On the other hand, (6.91) and Gauss’ equation yield
∂ ∂
K ∧ = µ2 . (6.110)
∂u2 ∂u3
Therefore U is an open portion of the warped product I ×f (s) S n−1 with
warping function given by
Z s
1 mu′
f (s) = c exp k(x)dx = p , k= . (6.111)
µ2 + k 2 λ − 2µ
By (6.91), (6.105)-(6.107) and Gauss’ formula, we get

Lss = iλLs , Y Ls = (iµ + k)Y, (6.112)
Y ZL = µ hY, Zi iLs + L∗ (∇Y Z), (6.113)
where Y, Z are vector fields tangent to the second factor S n−1 of the warped
product. Solving the first equation
Z in (6.112), yields
s Rs
L = A(u2 , . . . , un ) ei λ(t)dt
ds + B(u2 , . . . , un ) (6.114)
for some Cn -valued functions A and B. After applying the second equation
∂
in (6.112) with Y = ∂u , we find
Rj Z s
s
Rx
(iµ + k)Buj = e i λ(t)dt − (iµ + k) e−i λ(t)dt dx Auj (6.115)
for j = 2, . . . , n. Since A and B are independent of s, (6.115), implies
B = αA + C for some α ∈ C and C ∈ Cn .
By combining this with (6.114), we conclude that, after applying a suit-
able translation of Cn , we have
Z s Rs
L(s, u2 , . . . , un ) = α+ ei λ(t)dt
ds A(u2 , . . . , un ).
Now by applying the same argument as given in the proof of Theorem
6.6, we conclude that L is of the following form
Z s R
i s λ(t)dt
L = α+ e ds c1 sin u2 + c2 sin u3 cos u2 + · · ·
n−1 n
! (6.116)
Y Y
· · · + cn−1 sin un cos uj + cn cos uj
j=2 j=2
n
for some constant vectors c1 , . . . , cn ∈ C . Since N is Lagrangian, in view
of (6.105), we may choose the initial conditions (6.90). Then, by using
(6.116)we obtain
Z s R
s
L = α+ ei λ(t)dt ds
 
Yn n−1
Y (6.117)
×  cos uj , sin u2 , sin u3 cos u2 , · · · , sin un 
cos uj .
j=2 j=2
Since U is dense in N , (6.117) and continuity imply that, up to rigid motions
of Cn , N is the complex extensor of the unit hypersphere in En .
Remark 6.8. Flat H-umbilical submanifolds in the complex Euclidean
spaces have been completely determined in [Chen (1999c)]. Furthermore,
H-umbilical submanifolds in complex projective spaces and also in complex
hyperbolic spaces have been classified in [Chen (1999a)].
Chapter 7
Slant Submanifolds
Let N be a Riemannian submanifold of an almost Hermitian manifold

(M̃ , g̃, J). For a given vector X ∈ T N , we put
JX = P X + F X, (7.1)
where P X and F X denote the tangential and normal components of JX,
respectively. For a unit vector X ∈ Tx N, x ∈ N , the angle θ(X) between
JX and Tx N is called the Wirtinger angle of X. A submanifold N is called
slant if the Wirtinger angle θ(X) is constant on N , i.e., θ(X) is independent
of choice of 0 6= X ∈ Tx N and of x ∈ N (see [Chen (1990a)]). The constant
angle θ on a slant submanifold N is called the slant angle.
Complex and totally real submanifolds of an almost Hermitian manifold
are slant submanifolds with θ = 0 and θ = π2 , respectively. A slant sub-
manifold is called proper if it is neither complex nor totally real. Earliest
results on slant submanifolds were collected in the book [Chen (1990b)].
7.1 Examples of slant submanifolds
Let E2m be the Euclidean 2m-space with the Euclidean metric. An almost
complex structure J on E2m is called compatible if (E2m , J) is complex
analytically isometric to the complex number space Cm .
We denote by J0 and J1− (when m is even) the compatible almost com-
plex structures on E2m defined respectively by
J0 (a1 , . . . , am , b1 , . . . , bm ) (7.2)
= (−b1 , . . . , −bm , a1 , . . . , am )
−
J1 (a1 , . . . , am , b1 , . . . , bm ) (7.3)
= (−a2 , a1 , . . . , −am , am−1 , b2 , −b1 , . . . , bm , −bm−1 ).
141
Example 7.1. For any θ > 0,

x(u, v) = (u cos θ, v, u sin θ, v, 0)
defines a slant plane with slant angle θ in C2 .
Example 7.2. Let N be a complex surface in C2 = (E4 , J0 ) and consider

the following compatible almost complex structure on E4 :
Jθ (a, b, c, d) = (cos θ)(−c, −d, a, b) + (sin θ)(−b, a, d, −c).
Then for any constant θ, 0 < θ ≤ π2 , N is slant surface in (E4 , Jθ ) with
slant angle θ. This example shows that there exist infinitely many proper
slant minimal surfaces in C2 = (E4 , J0 ).
The following example provides some non-minimal proper slant surfaces

in C2 = (E4 , J0 ).
Example 7.3. For any positive constant k,

x(u, v) = (eku cos u cos v, eku sin u cos v, eku cos u sin v, eku sin u sin v)
defines a complete, non-minimal,
√ pseudo-umbilical proper slant surface with
slant angle θ = cos−1√(k/ 1 + k 2 ) and with non-constant mean curvature
given by |H| = e−ku / 1 + k 2 .
Example 7.4. For any positive number k,

x(u, v) = (u, k cos v, v, k sin v)
defines a complete, flat, non-minimal and√ non-pseudo-umbilical, proper
slant surface with slant angle cos−1 (1/ 1 + k 2 ) and constant mean cur-
vature k/2(1 + k 2 ) and with non-parallel mean curvature vector.
Example 7.5. Let k be any positive number and (g(s), h(s)) a unit speed
plane curve. Then
x(u, s) = (−ks sin u, g(s), ks cos u, h(s))
√
defines a non-minimal, proper slant flat surface with slant angle k/ 1 + k 2 .
Example 7.6. For any nonzero real numbers p and q, we consider the
following immersion form R × (0, ∞) into C2 defined by
x(u, v) = (pv sin u, pv cos u, v sin qu, v cos qu).
Then the immersion x gives rise to a complete flat slant surface in C2 .
Slant Submanifolds 143
Definition 7.1. A proper slant submanifold N of a Kähler manifold is

called Kählerian slant if its canonical endomorphism P defined in (7.1) is
parallel on N , i.e., ∇P = 0.
A Kählerian slant submanifold with slant angle θ is a Kähler manifold

with respect to the induced metric and the almost complex structure given
by J˜ = (sec θ)P .
Example 7.7. For any real number k ∈ (0, 1), the map
x(u, v, w, z) = (u, v, k sin w, k sin z, kw, kz, k cos w, k cos z)

defines a Kählerian slant submanifold in C4 with slant angle cos−1 k.
Example 7.8. Let N be a complex submanifold of the complex number

space C2m = (E 4m , J0 ). For any constant θ we define Jθ by
Jθ = (cos θ)J0 + (sin θ)J1− .
Then Jθ is a compatible complex structure on E4m and N is a Kählerian
slant submanifold with slant angle θ in (E4m , Jθ ).
Example 7.9. Let φ : E3 → E4 be the map from E3 into E4 defined by

φ(x0 , x1 , x2 ) = (x1 , x2 , 2x0 x1 , 2x0 x2 ).
Then φ induces the Whitney immersion w : S 2 → E4 from the unit 2-

sphere S 2 into E4 . The Whitney immersion w : S 2 → E4 is totally real
with respect to two compatible complex structures on E4 (see pages 42-43
of [Chen (1990b)]).
Example 7.10. Let N be the surface in E4 defined by
x(u, v) = (u, v, k cos v, k sin v).

Then N is the Riemannian product of a line and a circular helix in a
hyperplane E3 of E4 . Let J1 , J2 be the compatible complex structures on
E4 defined respectively by
J1 (a, b, c, d) = (−b, a, −d, c), J2 (a, b, c, d) = (b, −a, −d, c).
Then N is slant with respect to the following four complex structures:
J1 , −J1 , J2 , −J2 , with slant angles given respectively by

1 −1 −1 1
cos−1 √ , cos−1 √ , cos−1 √ , cos−1 √ .
1 + k2 1 + k2 1 + k2 1 + k2
By studying the Gauss map of surfaces in the Euclidean 4-space E4 , the

following result was proved in [Chen and Tazawa (1990)].
Theorem 7.1. For an immersion φ : N → E4 , exactly one of the following

four cases occurs:
(a) φ is not slant with respect to every compatible complex structure on E4 .

(b) φ is slant with respect to infinitely many compatible complex structures
on E4 .
(c) φ is slant with respect to exactly two compatible complex structures on
E4 .
(d) φ is slant with respect to exactly four compatible complex structures on
E4 .
The following two results on slant surfaces in C2 were also proved in

[Chen and Tazawa (1990)].
Proposition 7.1. Every complex surface in C2 is slant with respect to

infinitely many compatible complex structures on E4 .
Proposition 7.2. If N is a non-totally geodesic minimal surface in E3 ⊂

E4 , then N is not slant with respect to every compatible complex structures
on E4 .
7.2 Basic properties and their applications
Let φ : N → M̃ be an isometric immersion of a Riemannian n-manifold

into a Kähler manifold. Let P and F be defined by (7.1).
Since M̃ is almost Hermitian, we have
hP X, Y i = − hX, P Y i , X, Y ∈ T N. (7.4)
2
If we put Q = P , then Q is self-adjoint. Hence, each tangent space Tx N
of N admits an orthogonal direct decomposition of eigenspces of Q as
Tx N = Dx1 ⊕ · · · ⊕ Dxk(x) .
Since P is skew-symmetric and J 2 = −I, each eigenvalue λi of Q lies
in [−1, 0]. Moreover, if λi 6= 0, the corresponding eigenspace Dxi is of even
dimension and it is invariant under the endomorphism P , i.e., P (Dxi ) = Dxi .
Also, for each λi 6= −1, dim F (Dxi ) = dim Dxi and the normal subspaces
F (Dxi ), i = 1, . . . , k(x), are mutually perpendicular. From these we have
dim M̃ ≥ 2 dim N − dim Kx
where Kx denotes the eigenspace of Q with eigenvalue −1.
Lemma 7.1. Let N be a submanifold of a Kähler manifold M . Then the

self-adjoint endomorphism Q is parallel, i.e., ∇Q = 0, if and only if
(1) each eigenvalue λi of Q is constant on N ;
(2) each distribution Di (associated with the eigenvalue λi ) is completely
integrable and
(3) N is locally the Riemannian product N1 × · · · × Nk of the leaves of the
distributions.
Proof. Since Q is self-adjoint, there exist n continuous functions

λ1 ≤ λ2 ≤ · · · ≤ λn
such that λi , i = 1, . . . , n, are the eigenvalues of Q at each point x ∈ N .
Let e1 , . . . , en be a local orthonormal frame given by eigenvectors of Q. If
Q is parallel, it follows from
(∇X Q)Y = ∇X (QY ) − Q(∇X Y ) (7.5)
that ∇X (λi ei ) = Q(∇X ei ), i = 1, . . . , n, for X ∈ T N . Thus we get
(Xλi )ei + λi (∇X ei ) = Q(∇X ei ).
Since ∇X ei and Q(∇X ei ) are perpendicular to ei , each eigenvalue of Q is
constant on N . This proves statement (1).
For statements (2) and (3), let λ1 , . . . , λk denote the distinct eigenvalues
of Q. For each i ∈ {1, . . . , k}, let Di denote the distribution defined by the
eigenspaces of Q with eigenvalue λi .
For vector fields X, Y in the distribution Di , (7.5) and statement (1)
imply
Q(∇X Y ) = λi (∇X Y ),
from which we have ∇X Y ∈ Di for X, Y ∈ Di . Hence each distribution
Di is integrable and each integrable submanifold of Di is totally geodesic
in N . Consequently, by de Rham decomposition theorem, N is locally the
Riemannian product N1 × · · · × Nk of the leaves of these distributions.
The converse of this is easy to verify.
Lemma 7.1 implies immediately the following.
Lemma 7.2. Let N be a submanifold of a Kähler manifold M̃ . Then N

is a slant submanifold if and only if P 2 = −(cos2 θ)I for some θ ∈ [0, π2 ],
where I is the identity map.
By using Lemma 7.1 we also have the following simple characterizations

of submanifolds in terms of ∇P = 0 from [Chen (1990a)].
Lemma 7.3. Let N be a submanifold of a Kähler manifold. Then ∇P = 0

holds if and only if N is locally the Riemannian product N1 × · · · × Nk ,
where each Ni is either a complex submanifold, a totally real submanifold
or a Kählerian slant submanifold of the Khler manifold M̃ .
Proof. Let N be a submanifold of a Kähler manifold M̃ . Under the

hypothesis, if P is parallel, then Q = P 2 is parallel. Thus, by applying
Lemma 7.1, we see that N is locally the Riemannian product N1 × · · · × Nk
of leaves of distributions defined by eigenvectors of Q and moreover each
eigenvalue λi is constant on N .
If an eigenvalue λi is zero, then the corresponding leaf Ni is totally
real. If λi is −1, then Ni is a complex submanifold. If λi 6= 0, −1, then
because Di is invariant under P and hP X, Y i = −λi hX, Y i for any X, Y in
√
Di , we have |P X| = −λi |X|. Thus the Wirtinger angle θ(X) satisfying
√
cos θ(X) = −λi , which is a constant 6= 0, −1. Therefore Ni is a proper
slant submanifold. If λi 6= 0, we put Pi = P |T Ni . Then Pi is nothing but
the endomorphism of T Ni induced from the almost complex structure J.
Let ∇i denote the Riemannian connection of Ni . Since Ni is totally
geodesic in N , we have
(∇iX Pi )Y = (∇X P )Y = 0
for X, Y ∈ T Ni . This shows that if Ni is a complex submanifold, then Ni is
a Kähler manifold. And if Ni is proper slant, then Ni is a Kählerian slant
submanifold of M̃ by definition.
The converse can be verified directly.
Lemma 7.2 implies immediately the following
Corollary 7.1. Let N be an irreducible submanifold of a Kähler manifold

M̃ . Suppose that N is neither a complex nor totally real submanifold of M̃ .
Then N is a Kählerian slant submanifold if and only if the endomorphism
P is parallel, i.e., ∇P = 0.
Theorem 7.2. [Chen (1990a)] Let N be a surface in a Kähler manifold

M̃ . Then the following three statements are equivalent:
(1) N is neither complex nor totally real in M̃ and ∇P = 0;
(2) N is a Kählerian slant surface;
(3) N is a proper slant surface.
Proof. Since every proper slant submanifold is even dimensional, Lemma

7.3 implies that if the endomorphism P is parallel, then N is a complex
surface, or a totally real surface or a Kählerian slant surface. Thus if N is
neither totally real nor complex, then statements (1) and (2) are equivalent
by definition. It is obvious that (2) implies (3).
Now, we prove that (3) implies (2). Let N be a proper slant surface in
M̃ with slant angle θ. If we choose an orthonormal frame e1 , e2 tangent to
N such that P e1 = (cos θ)e2 and P e2 = −(cos θ)e1 , then we have
(∇X P )e1 = cos θ(ω21 (X) + ω12 (X))e1 .
Since ω12 = ω21 , we find ∇P = 0.
For any normal vector ξ of N in M̃ , we put
Jξ = tξ + f ξ, (7.6)
where tξ and f ξ denote the tangential and the normal components of Jξ,
respectively.
Proposition 7.3. [Chen (2002f)]. Let N be a submanifold of a Kähler
manifold M̃ . Then we have:
(i) For X, Y ∈ T N ,
(∇X P )Y = tσ(X, Y ) + AF Y X. (7.7)
(ii) ∇P = 0 holds if and only if AF X Y = AF Y X holds for any X, Y ∈ T N .
(iii) For X, Y ∈ T N , we have
(∇X F )Y = f σ(X, Y ) − σ(X, P Y ). (7.8)
Hence ∇F = 0 holds if and only if Af ξ X = −Aξ (P X) for any normal
vector ξ and tangent vector X of N .
Proof. Since M̃ is Kähler manifold, J is parallel. Thus, by using the
formulas of Gauss and Weingarten and formulas (7.1) and (7.6), we get
(∇X P )Y = ∇X (P Y ) − P (∇X Y ) (7.9)
= tσ(X, Y ) + AF Y X,
(∇X F )Y = DX (F Y ) − F (∇X Y ) (7.10)
= f σ(X, Y ) − σ(X, P Y ).
Hence P is parallel if and only if htσ(X, Y ) + AF Y X, Zi = 0 holds which
is equivalent to
hAF Y X, Zi = − htσ(X, Y ), Zi = hσ(X, Y ), F Zi
= hAF Z X, Y i = hAF Z Y, Xi .
This proves statements (i) and (ii). Statement (iii) follows from (7.10).
Remark 7.1. If N is either a totally real or complex submanifold of a

Kähler manifold, then ∇P = ∇F = 0 holds automatically.
Combining Theorem 7.2 and Proposition 7.3 we obtain the following
simple characterization of slant surfaces in terms of Weingarten map.
Corollary 7.2. Let N be a surface in a Kähler manifold M̃ . Then N is
slant if and only if AF Y X = AF X Y holds for any X, Y ∈ T N .
For a proper slant submanifold N in M̃ , the normal bundle of N admits
a complex subbundle ν such that
T ⊥ N = F (T N ) ⊕ ν, νx ⊥ F (Tx N ), ∀x ∈ N.
Another easy consequence of Proposition 7.3 is the following.
Corollary 7.3. If N is a totally umbilical Kählerian slant submanifold of
a Kähler manifold M̃ , then it is totally geodesic.
Proof. If N is totally umbilical, we get Aξ X = µ(ξ)X for any ξ ∈ T ⊥ N
and X ∈ T N , where µ is function. Combining this with AF X Y = AF Y X
shows that the mean curvature vector H lies in ν. Because ν is J-invariant
we find JH ∈ ν. Since M̃ is Kählerian and H, JH ∈ ν, we derive from
Weingarten’s formula and F P X ⊥ ν that
0 = hAJH X, P Xi = hH, Hi hP X, P Xi .
This implies that H = 0. Hence N is totally geodesic.
Theorem 7.3. [Chen (2002f)] Let N be a proper slant submanifold of a
Kähler manifold M̃ . If ∇F = 0, then N is austere.
Proof. Let N be a proper slant submanifold of a Kähler manifold M . If
∇F = 0, then we find from (7.8) that f σ(X, Y ) = σ(X, P Y ). Let X be√any
unit eigenvector of Q = P 2 with eigenvalue λ 6= 0. Then X∗ = P X/ −λ
is a unit vector perpendicular to X. Hence we obtain
σ(X, X) = σ(P X, P X)/λ = −σ(X∗ , X∗ )
which shows that N is austere.
When the ambient space M̃ is a complex space form, then we have the
following reduction theorem.
Theorem 7.4. [Chen (1990a)] Let N be an n-dimensional proper slant
submanifold of a complex m-dimensional complex space form M̃ m (c) with
constant holomorphic sectional curvature c. If ∇F = 0 holds identically,
then N is contained in a complex n-dimensional complex totally geodesic
submanifold M̃ n (c) of M̃ m (c) as an austere submanifold.
Proof. Let N be a proper slant submanifold of M̃ m (c). Assume that

∇F = 0. For any vector field ξ ∈ ν and vector fields X, Y ∈ T N , we have
˜ X Y, Jξ i
hAJξ X, Y i = hσ(X, Y ), Jξi = h ∇
= − hP Y, Aξ Xi + hF Y, DX ξi ,
from which we find
hDX (F Y ), ξi = − hAξ (P Y ) + AJξ Y, Xi . (7.11)
On the other hand, for any ξ ∈ T ⊥ N , if we denote by tξ and f ξ by
using (7.6), then Proposition 7.3 gives
Af ξ Y + Aξ (P Y ) = 0. (7.12)
Since f |ν = J|ν , (7.11) and (7.12) give hDX (F Y ), ξi = 0 for any ξ ∈ ν.
Thus F (T N ) is a parallel normal subbundle in T ⊥ N .
Now, we claim that the first normal subbundle, Im σ, lies in F (T N ).
This can be proved as follows: Since ∇F = 0, Proposition 7.3(iii) implies
hσ(X, Y ), Jξi = − hσ(X, P Y ), ξi
for ξ ∈ ν. Thus for any eigenvector Y of the self-adjoint endomorphism Q
with eigenvalue λ and any ξ ∈ ν we have
hσ(X, Y ), ξi = −λ hσ(X, Y ), ξi .
Since N is proper slant, we have −1 < λ < 0. Thus Im σ ⊂ F (T N ). Hence,
after applying the reduction theorem we obtain the theorem.
When the ambient space M̃ is Cm , we have the following.
Theorem 7.5. Let N be an n-dimensional proper slant submanifold of Cm .

If ∇F = 0, then N lies in a complex linear subspace Cn of Cm as an austere
submanifold.
Since slant surfaces in Kähler manifolds are Kählerian slant by Theorem

7.2, Corollary 7.3 implies immediately the following.
Corollary 7.4. Every totally umbilical proper slant surface N in a Kähler

manifold M̃ is totally geodesic.
Remark 7.2. From Remark 6.5 we see that Corollary 7.4 is false if the
totally umbilical slant surface is non-proper.
Corollaries 7.3 and 7.4 were extended by B. Sahin to the following.
Theorem 7.6. [Sahin (2009b)] Every totally umbilical proper slant sub-
manifold in a Kähler manifold is totally geodesic.
Proof. From totally umbilicity, we obtain σ(P X, P X) = cos2 θ hX, Xi H

for X ∈ T N , where θ is the slant angle. Thus
˜ PXX − ∇
cos2 θ hX, Xi H = J ∇ ˜ P X F X − ∇P X P X,
Hence, after applying the formulas of Gauss and Weingarten, we obtain

cos2 θ hX, Xi H = P ∇P X X + Jσ(P X, X) + AF X P X
(7.13)
− DP X F X − ∇P X P X.
Since N is totally umbilical, we get σ(P X, X) = 0. By comparing the
normal parts of this equation, we find
cos2 θ hX, Xi H = F ∇P X X − DP X F X. (7.14)
By taking the scalar product of (7.14) with F X, we find
cos2 θ hX, Xi hH, F Xi = hF ∇P X X, F Xi − hDP X F X, F Xi . (7.15)
2
Since hF X, F Y i = sin θ hX, Y i, (7.14) gives
cos2 θ hX, Xi hH, F Xi = sin2 θ h∇P X X, Xi − hDP X F X, F Xi . (7.16)
By differentiating hF X, F Xi = sin2 θ hX, Xi, we have
hDP X F X, F Xi = sin2 θ h∇P X X, Xi . (7.17)
Combining (7.16) and (7.17) yields cos2 θ hX, Xi hH, F Xi = 0. Thus H ∈ ν.
Since ν is a complex vector bundle on N , we also have JH ∈ ν.
˜ X JY = J ∇
Similarly, by considering ∇ ˜ X Y , we obtain from the formula
of Gauss and Weingarten that
∇X P Y + hX, P Y i H − AF Y X + DX F Y
(7.18)
= P ∇X Y + F ∇X Y + hX, Y i JH.
Since JH ∈ ν, by taking the scalar product of (7.18) with JH we find
˜ X F Y, JH i = hX, Y i hH, Hi .
h∇ (7.19)
˜ X JH = J ∇
Also, we find from ∇ ˜ X , we have
−AJH X + DX JH = −P AH X − F AH X + tDX H + f DX H.
Hence we get
hDX JH, F Y i = − hF AH X, F Y i + hf DX H, F Y i . (7.20)
Since f DX H ∈ ν and hF Y, JHi = 0, we derive from (7.20) that
˜ X F Y, JH i = sin2 θ hAH X, Y i = sin2 θ hX, Y i , hH, Hi .
h∇ (7.21)
Combining (7.19) and (7.21) gives cos2 θ hX, Y i hH, Hi = 0. Consequently,
H = 0 and N is totally geodesic.
7.3 Existence and uniqueness theorems
In this section we present the following existence and uniqueness theorems

for slant immersions from [Chen and Vrancken (1997)].
Theorem 7.7. (Existence) Let c and θ be two constants with 0 < θ ≤ π2
and let N be a simply-connected Riemannian n-manifold equipped with an
inner product h , i. Suppose that there exist an endomorphism P of the
tangent bundle T N and a symmetric bilinear T N -valued form α on N such
that, for X, Y, Z, W ∈ T N , we have
P 2 = −(cos2 θ)I, (7.22)
hP X, Y i + hX, P Y i = 0, (7.23)
h(∇X P )Y, Zi = hα(X, Y ), Zi − hα(X, Z), Y i , (7.24)
2
R(X, Y ; Z, W ) = csc θ{hα(X, W ), α(Y, Z)i − hα(X, Z), α(Y, W )i}
+ c{hX, W i hY, Zi − hX, Zi hY, W i + hP X, W i hP Y, Zi (7.25)
− hP X, Zi hP Y, W i + 2 hX, P Y i hP Z, W i},
and
(∇X α)(Y, Z) + csc2 θ{P α(X, α(Y, Z)) + α(X, P α(Y, Z))}
(7.26)
+ (sin2 θ)c{hX, P Zi Y + hX, P Y i Z}
is totally symmetric. Then there exists a θ-slant isometric immersion from
N into a complex space form M̃ n (4c) of constant holomorphic sectional
curvature 4c such that the second fundamental form σ is given by
σ(X, Y ) = csc2 θ(P α(X, Y ) − Jα(X, Y )). (7.27)
Proof. Let c, θ be two constants with 0 < θ ≤ π2 and N a simply-
connected Riemannian n-manifold equipped with an endomorphism P and
a symmetric bilinear T N -valued form α satisfying the five conditions stated
in the theorem.
Consider the Whitney sum T N ⊕ T N . For each X ∈ T N , we identify
(X, 0) with X; and also we denote (0, X) by X ∗ . We define the inner
product h , i on T N ⊕ T N by using the product metric. Let Jˆ be the
endomorphism on T N ⊕ T N defined by
ˆ = P X + sin θX ∗ , JX
JX ˆ ∗ = − sin θX − P X ∗ , (7.28)
for X ∈ T M . Then we have
Jˆ2 ((X, 0)) = J(P
ˆ X, sin θX)
= (P 2 X, sin θP X) − sin θ(sin θX, P X)
= −(X, 0).
Similarly, we have Jˆ2 ((0, X)) = −(0, X). Thus, Jˆ2 = −I. By using (7.28),
ˆ JY
it is easy to verify that h JX, ˆ i = hX, Y i . Thus, (J,
ˆ h , i) is a Hermitian
structure on T N ⊕ T N .
Now, we define A, σ and D by
AY ∗ X = csc θ{(∇X P )Y − α(X, Y )}, (7.29)
∗
σ(X, Y ) = −(csc θ)α (X, Y ), (7.30)
∗ ∗ 2 ∗ ∗
DX Y = (∇X Y ) + csc θ{P α (X, Y ) + α (X, P Y )}, (7.31)
for vector fields X, Y tangent to N .
It is easy to verify that each AY ∗ is an endomorphism on T N , σ is a
(T N )∗ -valued symmetric bilinear form on T N , and D is a metric connection
of the vector bundle (T N )∗ over N .
Let ∇ ˆ denote the canonical connection on T N ⊕ T N induced from the
Levi-Civita connection on T N . Then, from (7.28)-(7.31), we have
(∇ ˆ = (∇
ˆ X J)Y ˆ ∗ = 0,
ˆ X J)Y (7.32)
for vector fields X, Y ∈ T N .
Let RD denote the curvature tensor associated with the connection D
on (T X)∗ , i.e.,
RD (X, Y )Z ∗ = DX DY Z ∗ − DY DX Z ∗ − D[X,Y ] Z ∗ , (7.33)
for X, Y, Z ∈ T N . Then, by (7.22), (7.26), (7.31), (7.33) and a simple
computation, we may obtain
RD (X, Y )Z ∗ = (R(X, Y )Z)∗
+ {cP [hY, P Zi X − hX, P Zi Y − 2 hX, P Y i Z]

+ c[ Y, P 2 Z X − X, P 2 Z Y − 2 hX, P Y i P Z] (7.34)
2
+ csc θ[(∇X P )α(Y, Z) − (∇Y P )α(X, Z)
− α(X, (∇Y P )Z) + α(Y, (∇X P )Z) ] }∗ .
Also, (7.29) yields
sin2 θ h[AZ ∗ , AW ∗ ]X, Y i = h(∇Y P )Z, (∇X P )W i
− h(∇X P )Z, (∇Y P )W i + h(∇X P )Z, α(Y, W )i
+ h(∇Y P )W, α(X, Z)i − h(∇Y P )Z, α(X, W )i (7.35)
− h(∇X P )W, α(Y, Z)i + hα(X, W ), α(Y, Z)i
− hα(X, Z), α(Y, W )i .
From (7.23) we have
hα(Y, Z), P W i + hP α(Y, Z), W i = 0. (7.36)
By taking the derivative of (7.36) and using (7.23), we find

hα(Y, Z), (∇X P )W i + h(∇X P )α(Y, Z), W i = 0. (7.37)
Also, by (7.24) we obtain
h(∇X P )Z, (∇Y P )W i = h(∇X P )Z, α(Y, W )i
(7.38)
− hα(Y, (∇X P )Z), W i .
Hence, by (7.34), (7.35), (7.37), (7.38) and a direct computation, we get

D
R (X, Y )Z ∗ , W ∗ − h[AZ ∗ , AW ∗ ]X, Y i

= c sin2 θ(hY, Zi hX, W i − hX, Zi hY, W i) (7.39)

− 2 hX, P Y i hP Z, W i .
From (6.5), (7.22), (7.23) and (7.39) we see that (N, A, D) satisfies Ricci’s
equation for a θ-slant submanifold in M̃ n (4c).
Moreover, (7.25) and (7.26) imply that (N, σ) satisfies equations of
Gauss and Codazzi for a θ-slant submanifold in M̃ n (4c). Therefore the vec-
tor bundle T N ⊕ T N equipped with the product metric, the shape operator
A, the second fundamental form σ, and the connections D and ∇ ˜ satisfy
the structure equations of n-dimensional θ-slant submanifolds in the com-
plex space form M̃ n (4c). Consequently, after applying a result of [Wettstein
(1978)] (or by Theorem 1 of [Eschenburg and Tribuzy (1993)]), we conclude
that there exists a θ-slant isometric immersion of N into M̃ n (4c) with
σ = csc2 θ(P α − Jα)
as its second fundamental form, A as its shape operator, and D as its
normal connection.
Theorem 7.8. (Uniqueness) Let x1 , x2 : N → M̃ n (4c) be θ-slant isometric

immersions, with 0 < θ ≤ π2 and second fundamental form σ 1 and σ 2 , of a
connected Riemannian n-manifold N into the complex space form M̃ n (4c).
If

1

σ (X, Y ), Jx1⋆ Z = σ 2 (X, Y ), Jx2⋆ Z , (H)
holds for all vector fields X, Y, Z tangent to N , and if at least one of the
following conditions is satisfied:
(1) θ = π2 ,
(2) there exists a point p of M such that P1 = P2 ,
6 0,
(3) c =
then P1 = P2 holds and there exists an isometry Φ of M̃ n (4c) such that

x1 = Φ(x2 ).
Proof. If (1) is satisfied, then P1 = P2 holds trivially since P1 = P2 = 0.

If (2) is satisfied, it follows from (7.24) that (∇X (P1 − P2 ))Y = 0. Since
we have P1 = P2 at a point p, we have P1 = P2 everywhere.
In the remaining case, we assume that c 6= 0, P1 6= P2 , and that (1) and
(2) are not satisfied. First, we want to show that P1 = −P2 . In order to
do so, we use c 6= 0 and (7.25) to obtain
hP1 X, W i hP1 Y, Zi − hP1 X, Zi hP1 Y, W i + 2 hX, P1 Y i hP1 Z, W i
= hP2 X, W i hP2 Y, Zi − hP2 X, Zi hP2 Y, W i (7.40)
+ 2 hX, P2 Y i hP2 Z, W i .
Taking X = W and Y = Z, equation (7.40) reduces to
2 2
hP1 Y, Xi = hP2 Y, Xi . (7.41)
Next put e1 = X and e2 = P1 X and suppose that P2 e1 has a component
in the direction of a vector e3 which is orthogonal to both e1 and e2 . Then
a contradiction follows from (7.41) which states that
2 2 2
hP2 e1 , e3 i = hP1 e1 , e3 i = he2 , e3 i = 0.
Thus, by applying (7.22) and (7.23), we get P1 v = ±P2 v for every tangent
vector v.
Now choose a basis of the tangent space {e1 , . . . , en } at a point p. Then
there exists number ǫi ∈ {−1, 1} such that P1 ei = ǫi P2 ei . So we also have
±P1 (ei + ej ) = P2 (ei + ej ) = ǫi P1 ei + ǫj P1 ej .
The above formula shows that all ǫi have to be equal. So, either P1 v = P2 v
for all v ∈ Tp M , or P1 v = −P2 v for all v ∈ Tp M . Since M is connected
this implies that we have either P1 = P2 or P1 = −P2 in case (3).
Let us now assume that we have two immersions with P1 = −P2 . Since
α1 = α2 = α, it follows from (7.24) in this case that
hα(X, Y ), Zi = hα(X, Z), Y i . (7.42)
We denote P1 by P . Writing down the equation (7.26) for both the immer-
sions and using the fact that P2 = −P1 = P , we deduce that
P α(X, α(Y, Z)) − P α(Y, α(X, Z))
+ α(X, P α(Y, Z)) − α(Y, P α(X, Z)) (7.43)
4
+ c sin θ{hX, P Zi Y − hY, P Zi X + 2 hX, P Y i Z} = 0.
Taking the inner product of (7.43) with a vector W and using (7.42),
we deduce that
− hα(Y, Z), α(X, P W )i + hα(X, Z), α(Y, P W )i
+ hα(X, W ), P α(Y, Z)i − hα(Y, W ), P α(X, Z)i
(7.44)
+ c sin4 θ hX, P Zi hY, W i − hY, P Zi hX, W i

+ 2 hX, P Y i hZ, W i = 0.
If α vanishes identically at a point, then a contradiction follows from
(7.44) since c 6= 0. Now, we take a fixed point p ∈ N and look at the
function f defined on the set of all unit tangent vectors U Np at p by
f (v) = hα(v, v), vi .
Since U NP is compact there exists a vector u such that f attains an absolute
maximum at the vector u. Let w be a unit vector orthogonal to u. Then
the function
f (t) = f (g(t)), g(t) = (cos t)u + (sin t)w,
satisfies f (0) = 0 and f ′′ (0) ≤ 0. The first condition implies that
′
hα(u, u), wi = 0 whereas the second one reduces to hα(u, w), wi ≤

1
2 hα(u, u), ui.
Using now the total symmetry of α, it follows that we can choose an
orthonormal basis e1 = u, e2 , . . . , en such that
α(e1 , e1 ) = λ1 e1 , α(e1 , ei ) = λi ei , i>1 (7.45)
1
with λi ≤ 2 λ1 .
Since α is not identically zero, it follows from the total symmetry of
(7.42) that λ1 > 0. Applying (7.42), (7.45), and also (7.44) with X = Z =
W = e1 and Y = ei , we get
3c sin4 θ hP e1 , ei i
= −λi hα(e1 , P e1 ), ei i + λ1 hα(ei , P e1 ), e1 i
(7.46)
+ λ1 he1 , P α(e1 , ei )i − λ1 λi hei , P e1 i
= hP e1 , ei i (−λ2i − λ1 λi ).
Hence
hP e1 , ei i (λ2i + λi λ1 + 3c sin4 θ) = 0. (7.47)
We now want to show that P e1 is an eigenvector of α(e1 , · ). In order
to do so, we apply again (7.45). First, we take X = Z = e1 , Y = ei and
W = ej for , i, j > 1. Then, we get
(λ2i − λ1 λi + λi λj ) hP ej , ei i + λ1 hα(ei , ej ), P e1 i = 0. (7.48)
Interchanging the indices i and j in (7.47), we obtain

(λ2j − λ1 λj + λi λj ) hP ej , ei i − λ1 hα(ei , ej ), P e1 i = 0. (7.49)
Adding (7.48) and (7.49) yields
hP ej , ei i (λi + λj )(λ1 − λi − λj ) = 0. (7.50)
Notice that since λ1 ≥ 2λi , the third term in (7.50) can vanish only if
λi = λj = 12 λ1 .
In case we put X = e1 , Z = ej , Y = ei and W = e1 into (7.44), we get
hα(ei , ej ), α(e1 , P e1 )i − λj hα(ei , ej ), P e1 i
(7.51)
+ λ1 hα(ei , ej ), P e1 i + λi λj hP ej , ei i + c sin4 θ hP ej , ei i = 0.
Interchanging i and j in equation (7.51) we obtain
hα(ei , ej ), α(e1 , P e1 )i − λi hα(ei , ej ), P e1 i
(7.52)
+ λ1 hα(ei , ej ), P e1 i − λi λj hP ej , ei i − c sin4 θ hP ej , ei i = 0.
Subtracting (7.51) from (7.52) yields
(λi − λj ) hα(ei , ej ), P e1 i + 2λi λj hP ej , ei i + 2c sin4 θ hP ej , ei i = 0. (7.53)
Now, we need to combine the previous equations in the correct way.
First, by taking i = j in (7.48) and using hP ei , ei i = 0, we get
hα(ei , ei ), P e1 i = 0. (7.54)
Hence we have hα(v, v), P e1 i = 0 if v is an eigenvector of α(e1 , · ). More-
over, the symmetry of α then implies that hα(ei , ej ), P e1 i = 0 whenever
λi = λj .
We consider four different cases:
(a) λi + λj 6= 0, but not λi = λj = 12 λ1 . In this case (7.50) implies
hP ei , ej i = 0;
(b) λi + λj = 0 and λi 6= 0. In this case (7.48) gives hα(ei , ej ), P e1 i =
λi hP ej , ei i. Substituting this into (7.53), we get 2c sin4 θ hP ej , ei i = 0
which yields hP ej , ei i = 0;
(c) λi + λj = 0 and λi = 0, or equivalently, λi = λj = 0. In this case it
follows from (7.53) that hP ei , ej i = 0;
(d) λi = λj = 12 λ1 .
Therefore, if ei1 , . . . , eik are eigenvectors belonging to an eigenvalue differ-
ent from 12 λ1 , then each P eiℓ , ℓ = 1, . . . , k, can only have a component in
the direction of e1 , say P eiℓ = µℓ e1 . Thus µℓ P e1 = − cos2 θeiℓ .
Consequently, either k = 1 or there does not exist an eigenvector with

eigenvalue different from 21 λ1 . If k = 1, then clearly P e1 is an eigenvector.
In the latter case α(e1 , · ) restricted to the space e⊥
1 only has one eigenvalue,
namely 12 λ1 . Since P e1 is always orthogonal to e1 , P e1 is also an eigenvector
in this case. Hence P e1 is always an eigenvector of α(e1 , · ).
We may assume that e2 is in the direction of P e1 . Then it follows that
α(e1 , P e1 ) = λ2 P e1 ,
where λ2 satisfies the equation
λ22 + λ1 λ2 + 3c sin4 θ = 0 (7.55)
by virtue of (7.47).
If we choose X = Z = e1 , Y = ei and W = P e1 for i > 2, then it follows
from (7.45) that
0 = −λ1 hα(ei , P e1 ), P e1 i = −λ1 hα(P e1 , P e1 ), ei i .
Thus
α(P e1 , P e1 ) = λ2 cos2 θe1 .
Applying (7.45) again with X = Z = W = P e1 and Y = e1 yields
−λ2 λ1 − λ22 + 3c sin4 θ = 0. (7.56)
4
(7.55) and (7.56) imply c sin θ = 0 which is a contradiction. Consequently,
we obtain P1 = P2 .
Let p be any point of N . If necessary by applying an isometry of M̃ n (4c),
we may assume that x1 (p) = x2 (p) and x1⋆ (p) = x2⋆ (p). Let us then take
a geodesic γ through the point p = γ(0). It is sufficient to prove that
γ1 = x1 (γ) and γ2 = x2 (γ) coincide. We already know that γ1 (0) = γ2 (0)
and γ1′ (0) = γ2′ (0).
Let E1 , . . . , En be any orthonormal frame along γ. We can define a
frame along γ1 and γ2 as follows. Take, for i = 1, . . . , n, Ai = x1∗ (Ei ),
Bi = x2∗ (Ei ), An+i = (x1∗ (Ei ))∗ , Bn+i = (x2∗ (Ei ))∗ , where X ∗ is defined
by X ∗ = csc θF X. Now, it is easy to verify that hγ1′ , Ak i = hγ2′ , Bk i and
h∇ ˜ γ ′ Bk , Bℓ i for k, ℓ = 1, . . . , 2n, such that by Proposition
˜ γ ′ Ak , Aℓ i = h ∇
3 of [Reckziegel (1981)], γ1 = γ2 .
Remark 7.3. When c = 0, condition (H) does not imply P1 = P2 in

general. For instance, let N be the Euclidean plane equipped with the
standard coordinates (u, v) and let C2 denote the complex Euclidean plane
given by (E 4 , g, J) where g is the standard metric on E 4 and J is the almost
complex structure on E 4 defined by J(a, b, c, d) = (−b, a, −d, c).
Consider the immersions x1 , x2 : E 2 → C2 defined respectively by

1 v v v
x (u, v) = u, √ , cos √ , sin √ ,
2 2 2

v v v
x2 (u, v) = −u, √ , cos √ , sin √ .
2 2 2
π
Then x1 and x2 are 4 -slant isometric immersions satisfying both P1 = −P2
and condition (H).
Remark 7.4. Let x1 , x2 : N → C2 be two isometric slant immersions of a

connected Riemannian 2-manifold N into the complex Euclidean plane C2
with second fundamental form σ 1 and σ 2 .
If x1 and x2 satisfy condition (H), then there is an isometry φ of C2 such
that x1 = φ(x2 ). In fact, condition (H) and Lemma 4.1 in page 29 of [Chen
(1990b)] imply that the identity map of M is covered by a bundle map
from the normal bundle of x1 onto the normal bundle of x2 which preserves
the bundle metrics, the normal connections and the second fundamental
forms. Thus, according to the fundamental theorems of submanifolds in a
Euclidean space, x1 and x2 must be congruent, although P1 and P2 are not
necessary equal.
7.4 A non-existence theorem for compact slant

submanifolds
Let E2m = (R2m , h , i) and Cm = (E2m , J0 ) be the Euclidean 2m-space

and the complex Euclidean m-space, respectively, with the canonical inner
product h , i and the canonical almost complex structure defined by (7.2).
Denote by Ω0 the Kähler form of Cm , i.e.,
Ω0 (X, Y ) = hX, J0 Y i , X, Y ∈ E2m . (7.57)
For an immersion φ : N → Cm , the Gauss map µ of φ is given by
µ : N → G(n, 2m) ≡ D1 (n, 2m) ⊂ S K−1 ⊂ ∧n (E2m ),
(7.58)
µ(p) = e1 (p) ∧ · · · ∧ en (p), p ∈ N,

where n = dim N, K = 2m n , D1 (n, 2m) is the set of all unit decomposable
n-vectors in ∧n E2m , identified with the real Grassmannian G(n, 2m) in a
natural way, and S K−1 is the unit hypersphere of ∧n (E2m ) centered at
the origin, and {e1 , . . . , e2m } is a local adapted orthonormal tangent frame
along φ(N ).
We need the following lemmas.
Lemma 7.4. For X1 , . . . , X2k ∈ E2m (k < m), we have

(2k)! Ωk0 (X1 ∧ · · · ∧ X2k )
X
= sign(τ )Ω0 (Xτ (1) , Xτ (2) ) · · · Ω0 (Xτ (2k−1) , Xτ (2k) ),
σ∈S2k
where S2k is the permutation group of order 2k and sign(τ ) is the signature
of raw permutation τ .
Proof. Let e1 , . . . , em be an orthonormal frame of E2m with its dual

coframe given by ω 1 , . . . , ω 2m . Then
2m
X
Ω0 = ϕAB ω A ∧ ω B .
A,B=1
Hence by direct computation we have

Ωk0 (X1 , . . . , X2k )
1 X X
= sign (τ ) ϕA1 A2 ω A1 (Xτ (1) )ω A2 (Xτ (2) )
(2k)! τ
X
... ϕA2k−1 A2k ω A2k−1 (Xτ (2k−1) )ω A2k (Xτ (2k) ) .
From these we obtain the lemma.
Lemma 7.5. Let V ∈ G(n, 2m) and πV : E2m → V be the orthogonal

projection. If V is θ-slant in Cm = (E2m , J0 ), i.e., V is slant with slant
angle θ 6= π2 , then the linear endomorphism JV of V defined by
JV = (sec θ)(πV ◦ J0|V ) (7.59)
is a complex structure compatible with the inner product h , i | V . Hence n
is an even integer.
Proof. Put
P = πV ◦ (J |V ) : V → V, (7.60)
⊥ ⊥
P = J |V − P : V → V , (7.61)
2
Q = P : V → V. (7.62)
Then we have
J |V = P + P ⊥ . (7.63)
By simple computation and using (7.61), we have

hQX, Y i = hX, QY i , hP X, Y i = − hX, P Y i (7.64)
for any X, Y ∈ V .
Since V is assumed to be θ-slant, the angle function ∠ satisfies
∠ (JX, V ) = ∠ (JX, P X) = θ
for any nonzero vector X ∈ V . Hence we have
|P X| = (cos θ)|X| (7.65)
for 0 6= X ∈ V .
By (7.64), Q is a self-adjoint endomorphism. Since J02 = −I, (7.60)-
(7.63) imply that each eigenvalue of Q is equal to − cos2 θ which lies in
[−1, 0). Therefore, by using (7.59), we may prove that JV2 = −I and
|JV X|2 = (sec2 θ)|P X|2 = |X|2
for any X ∈ V . This proves the lemma.
Let ζ̂0 be the metrical dual of (−Ω0 )k with respect to the inner product
h , i naturally defined on ∧2k E2m , i.e.,
h ζ̂0 , η i = (−1)k Ωk0 (η) for any η ∈ ∧2k E2m , (7.66)
then we have the following
π
Lemma 7.6. Let V ∈ G(2k, 2m). If V is θ-slant in Cm with θ 6= 2, then
h ζ̂0 , V i = ck cosk θ, (7.67)
where ck is a nonzero constant depending only on k.
Proof. Let JV be the complex structure on V defined by Lemma 7.5.

For a unit vector X ∈ V , we put Y = JV X ∈ V . Then we have
Ω0 (X, JV X) = h−JV Y, J0 Y i = − cos θ. (7.68)
If X, Y ∈ V and Z is perpendicular to JV X, then
Ω0 (X, Z) = cos θ hX, JV Zi = 0. (7.69)
Thus if we choose an orthonormal JV -basis {e1 , . . . , e2k } on V , i.e.,
e2i = JV e2i−1 , i = 1, . . . , k, (7.70)
and
V = e1 ∧ · · · ∧ e2k , (7.71)
via the natural identification of G(2k, 2m) with D1 (2k, 2m), then we derive
Ω0 (ea , eb ) = −δa∗ b cos θ for a < b, (7.72)
∗ ∗
where (2i) = 2i − 1 and (2i − 1) = 2i for i = 1, . . . , k.
By (7.71), Lemma 7.4, and (7.72), we find
(2k)! Ωk0 (V ) = (2k)! Ω0 (e1 ∧ . . . ∧ e2k )
X
= sign(τ ) Ω0 (eτ (1) , eτ (2) ) · · · Ω0 (eτ (2k−1) , eτ (2k) )
τ ∈S2k
2k
X 12···(2k)
= δa1 ···a2k Ω0 (ea1 , ea2 ) · · · Ω0 (ea2k−1 , ea2k )
a1 ,...,a2k =1
2k
X 12······(2k) (7.73)
= δa1 a∗ ···ak a∗ Ω0 (ea1 , ea∗1 ) · · · Ω0 (eak , ea∗k )
1 k
a1 ,... ak =1
X X 12······(2k)
= 2k ··· δa1 a∗ ···ak a∗ Ω0 (ea1 , ea∗1 ) · · · Ω0 (eak , e∗ak )
1 k
a1 <a∗
1 ak <a∗
k
X X 12·······(2k)
= 2k (− cos θ)k ··· δa1 a∗ ···ak a∗
1 k
a1 <a∗
1 ak <a∗
k
= 2k (− cos θ)k k!.

Thus, by using (7.66) and (7.73), we find (7.67) with ck = 2k k!/(2k)! .
Lemma 7.7. Assume that φ : N → Em is an isometric immersion of

an n-dimensional compact oriented manifold N into E m . Then the Gauss
map µ : N → ∧n (Em ) is mass-symmetric in S K−1 with K = m n , i.e., the
center of gravity of µ coincides with the center of the hypersphere S K−1 in
∧n (Em ).
Proof. Let e1 , . . . , en be an oriented orthonormal local frame of T N with

its dual coframe given by ω 1 , . . . , ω n . Then we have
dx = e1 ω 1 + · · · + en ω n . (7.74)
By direct computation we have
dx ∧ · · · ∧ dx = n! (e1 ∧ · · · ∧ en ) ω 1 ∧ · · · ∧ ω n = n! µ(dV ),
where dV denotes the volume element of N and dx on the left-hand side is
repeated n-times. Thus after applying the divergence theorem, we obtain
Z Z Z
n! µdV = dx ∧ · · · ∧ dx = d(x ∧ dx ∧ · · · ∧ dx) = 0.
N N N
This shows that the center of gravity of µ is the origin of ∧n (E m ).
We have the following main theorem of this section.
Theorem 7.9. [Chen and Tazawa (1991)] Let φ : N → Cm be a slant

immersion of an n-dimensional differentiable manifold N into the complex
Euclidean m-space Cm . If N is compact, then φ is totally real.
Proof. Without loss of generality we may assume that N is oriented be-

cause otherwise we may simply replace N by its two-fold covering. Assume
φ is θ-slant with θ 6= π2 . Then, by Lemma 7.5, n is even.
Let us put n = 2k. Since N is compact, Lemma 7.7 implies that the
Gauss map µ is mass-symmetric in S K−1 with K = 2m 2k . Therefore we
obtain
Z
hµ(x), ζi dV = 0 (7.75)
x∈N
for any fixed 2k-vector ζ ∈ ∧2k (E2m ), where dV is the volume element of
N with respect to the metric induced from the immersion f .
Let ζ = ζ̂0 , where ζ̂0 is defined by (7.66). Then Lemma 7.6 and (7.75)
imply that
ck vol(N ) cosk θ = 0.
But this contradicts to the assumption cos θ 6= 0. Therefore we must have
θ = π2 and φ is a totally real immersion.
Theorem 7.9 implies the following non-existence result from [Chen and
Tazawa (1991)].
Corollary 7.5. There do not exist proper slant compact submanifolds in

any complex Euclidean space.
7.5 A non-minimality theorem for slant submanifolds
We know from section 7.1 that there exist ample examples of proper slant
minimal surfaces in C2 . It is also known that there exist many examples
of proper slant surfaces in complex projective plane CP 2 and in complex
hyperbolic plane CH 2 (cf. [Chen (1998b, 1999b)]).
Now, we present the following non-minimality theorem for proper slant
surfaces in CP 2 and in CH 2 .
Theorem 7.10. [Chen and Tazawa (2000)] Every proper slant surfaces in
a complex space form M̃ 2 (4c) with c 6= 0 is non-minimal.
Proof. Assume N is a proper slant minimal surface in a complex space

form M̃ 2 (4c) with c 6= 0. Denote by θ the slant angle. Then θ ∈ (0, π2 ).
Let
e1 , e2 = (sec θ)P e1 , e3 = (csc θ)F e1 , e4 = (csc θ)F e2 (7.76)
be a local slant frame of N in M̃ 2 (4c). Then it follows from (7.76) that
te3 = − sin θe1 , te4 = − sin θe2 , f e3 = − cos θe4 , f e4 = cos θe3 . (7.77)
Since M̃ 2 (4c) is Kählerian, the formulas of Gauss and Weingarten imply
that
∇X (P Y ) + σ(X, P Y ) − AF Y X + DX (F Y )
(7.78)
= P (∇X Y ) + F (∇X Y ) + tσ(X, Y ) + f σ(X, Y ).
Comparing the normal components of both sides of (7.78) yields
DX (F Y ) = F (∇X Y ) + f σ(X, Y ) − σ(X, P Y ). (7.79)
Hence, we find
De1 e3 = (csc θ){ω12 (e1 )F e2 + σ11
3 4
f e3 + σ11 f e4
3 4
(7.80)
− cos θ(σ12 e3 + σ12 e4 )},
r
where σij = hσ(ei , ej ), er i, i, j = 1, 2; r = 3, 4.
On the other hand, for slant surfaces we know from Corollary 7.2 that
AF X Y = AF Y X. (7.81)
3 4 4 3
Thus we get σ12 = σ11 and σ12 = σ22 . Hence (7.76), (7.77), (7.80) and the
4 2
minimality of N yield ω3 (e1 ) = ω1 (e1 ).
Similarly, we have ω34 (e2 ) = ω12 (e2 ). Therefore we obtain
ω34 = ω12 . (7.82)
Let x be a non-totally geodesic point in N . We define a function γx by
γx : U Nx → R : v 7→ γx (v) = hσ(v, v), F vi , (7.83)
where U Nx =: {v ∈ Tx N : hv, vi = 1}. Since U Nx is a compact set, there
exists a vector v in U Nx such that γx attains its absolute minimum at v.
Since x is a non-totally geodesic point, it follows from (7.81) that γx 6= 0.
By linearity, we have γx (v) < 0. Because γx attains an absolute minimum
at v, it follows that hσ(v, v), F wi = 0 for all w orthogonal to v. Thus v
is an eigenvector of the symmetric operator AF v . Hence, by choosing an
orthonormal basis {e1 , e2 } of Tx N with e1 = v, we obtain
σ(e1 , e1 ) = −µF e1 , σ(e1 , e2 ) = µF e2 , σ(e2 , e2 ) = µF e1 (7.84)
for some real number µ.

If x is a totally geodesic point, (7.84) holds trivially for any adapted
orthonormal basis at x. Consequently, there is a local adapted orthonormal
frame e1 , e2 , e3 , e4 such that the second fundamental form σ of the proper
slant minimal surface M in M̃ 2 (4c) satisfies
σ(e1 , e1 ) = −λe3 , σ(e1 , e2 ) = λe4 , σ(e2 , e2 ) = λe3 (7.85)
where λ = µ sin θ. Using (2.23), (7.82) and (7.85), we find
(∇¯ e2 σ)(e1 , e1 ) = −(e2 λ)e3 − 3λω 2 (e2 )e4 , (7.86)
1
¯ e1 σ)(e1 , e2 ) = (e1 λ)e4 − 3λω 2 (e1 )e3 ,
(∇ (7.87)
1
¯ e1 σ)(e2 , e2 ) = (e1 λ)e3 + 3λω12 (e2 )e4 ,
(∇ (7.88)
¯ e2 σ)(e1 , e2 ) = (e2 λ)e4 − 3λω12 (e2 )e3 .
(∇ (7.89)
On the other hand, it is easy to see that the normal component of
R̃(e2 , e1 )e1 and R̃(e1 , e2 )e2 are given by
(R̃(e2 , e1 )e1 )⊥ = 3c sin θ cos θe3 ,
(7.90)
(R̃(e1 , e2 )e2 )⊥ = −3c sin θ cos θe4 .
Thus, by applying the equation of Codazzi, (7.86), (7.87) and (7.90), we
obtain
e2 λ = 3λω12 (e1 ) − 3c sin θ cos θ, (7.91)
e2 λ = 3λω12 (e1 ) + 3c sin θ cos θ. (7.92)
After combining (7.91) and (7.92) we obtain c sin θ cos θ = 0, which is a
contradiction.
As an immediate consequence of Theorem 7.10, we obtain the following
corollary.
Corollary 7.6. Let N be a slant surface of a complex space form M̃ 2 (4c).

If N is minimal, then either we have c = 0, or N is a complex surface, or
N is a Lagrangian surface.
The following result was proved in [Salavessa and Valli (2002)].
Theorem 7.11. Let φ : N → M̃ be a minimal slant immersion of a 2n-

manifold N into an Einstein-Kähler manifold M̃ with dimC M̃ = 2n.
(1) If n = 2 and M̃ has nonzero scalar curvature, then φ is either complex
or Lagrangian immersion.
(2) If n ≥ 3, N compact, orientable and M̃ has nonzero scalar curvature,
then φ is either a complex or Lagrangian immersion.
7.6 Topology and cohomology of slant submanifolds
Let N be an n-dimensional proper slant submanifold with slant angle θ in

a Kähler manifold M of complex dimension m. Then n is even, say n = 2k.
Let e1 be a unit tangent vector of N . We put
e2 = (sec θ)P e1 , e1∗ = en+1 = (csc θ)F e1 , e2∗ = en+2 = (csc θ)F e2 .
If n ≥ 2, then for each ℓ = 1, . . . , k − 1, we may choose a unit vector
e2ℓ+1 ∈ T N such that e2ℓ+1 is perpendicular to e1 , e2 , . . . , e2ℓ−1 , e2ℓ .
Put
e2ℓ+2 = (sec θ)P e2ℓ+1 ,
e(2ℓ+1)∗ = en+2ℓ+1 = (csc θ)F e2ℓ+1 ,
e(2ℓ+2)∗ = en+2ℓ+2 = (csc θ)F e2ℓ+2 .
If N is totally real in M , i.e., θ = π2 , then we can choose e1 , . . . , en to be
any local orthonormal frame of T N and put
en+1 = e1∗ = Je1 , . . . , e2n = en∗ = Jen .
If m > n, then at a given point x ∈ N there exists a subspace νx of
the normal space Tx⊥ N such that νx is invariant under the action of the
complex structure J of M and
Tx⊥ N = F (Tx N ) ⊕ νx , νx ⊥ F (Tx N ). (7.93)
We choose a local orthonormal frame e4k+1 , . . . , e2m of ν such that
e2n+2 = e(4k+1)∗ = Je4k+1 , . . . , e2m = e(2m−1)∗ = Je2m−1 .
We call such an orthonormal frame
e1 , e2 , . . . , e2k−1 , e2k , e1∗ , e2∗ . . . , e(2k−1)∗ , e(2k)∗ ,
(7.94)
e4k+1 , e(4k+1)∗ , . . . , e2m−1 , e(2m−1)∗
an slant frame of N in M .
For an n-dimensional proper slant submanifold N of a Kähler manifold
M , we define a canonical 1-form Θ on N by
Xn
∗
Θ= ωii , (7.95)
i=1
A
where (ωB ) are the connection forms associated with the slant frame (7.94).
Theorem 7.12. [Chen (1990b)] Let N be a proper slant submanifold of

Cn , n = dim N . Then the canonical 1-form Θ defined by (7.95) is closed.
Hence Θ defines a canonical cohomology class
[Θ] ∈ H 1 (N ; R). (7.96)
Proof. In order to prove Theorem 7.12 we need the following lemmas.
Lemma 7.8. Let N be an n-dimensional proper slant submanifold with

slant angle θ in a Kähler manifold M̃ . Then we have
∗
Θ = (tr σ i )ω i (7.97)
Θ(X) = −n(csc θ) htH, Xi , X ∈ T N, (7.98)
where H is the mean curvature vector and σ the second fundamental form.
Proof. Formula (7.97) follows from (2.53) and (7.98). Since

∗
n htH, ej i = −n hH, F ej i = −n(sin θ) hH, ej ∗ i = −(sin θ) tr σ j ,
(7.97) implies (7.98).
Lemma 7.9. Let N be a 2k-dimensional proper slant submanifold of Cn .

Then with respect to an adapted slant frame we have
(2i−1)∗ (2j−1)∗ 2j 2i
ω2j−1 − ω2i−1 = cot θ(ω2i−1 − ω2j−1 ), (7.99)
∗ ∗
(2i−1) (2j) 2j 2j−1
ω2j − ω2i−1 = cot θ(ω2i − ω2i−1 ), (7.100)
(2j)∗ (2i)∗ 2j−1 2i−1
ω2i − ω2j = cot θ(ω2i − ω2j ), (7.101)
∗
(2i−1) (2j−1)∗ (2j)∗ (2i)∗
ω2j−1 − ω2i−1 = cot θ(ω(2i−1)∗ − ω(2j−1)∗ ), (7.102)
∗ ∗ ∗ ∗
(2j) (2i) (2j−1) (2i−1)
ω2i − ω2j = cot θ(ω(2i)∗ − ω(2j)∗ ), (7.103)
∗ ∗ ∗
(2j−1) 2j−1 (2j) (2j−1)
ω(2i−1)∗ − ω2i−1 = cot θ(ω2i−1 − ω2i ) (7.104)
∗ ∗
(2j) 2j 2i−1 (2j)
ω(2i−1)∗ − ω2i−1 = cot θ(ω(2j−1) ∗ − ω2i ) (7.105)
∗ ∗ ∗
(2j) 2j (2j) (2j−1)
ω(2i)∗ − ω2i = cot θ(ω2i−1 − ω2i ), (7.106)
∗ ∗ ∗ ∗
(2i) (2j−1) (2j) (2j−1)
ω2j−1 − ω2i = cot θ(ω(2i)∗ − ω(2i−1)∗ ) (7.107)
for any i, j = 1, . . . , k.
Proof. From the definition of slant frames we have

hJe2i−1 , e2j−1 i = 0, i, j = 1, . . . , k. (7.108)
By taking the derivative of (7.108) with respect to X ∈ T N , we find
0 = h J∇˜ X e2i−1 , e2j−1 i + h Je2i−1 , ∇
˜ X e2j−1 i
= − h∇X e2i−1 , P e2j−1 i − hσ(e2i−1 , X), F e2j−1 i
+ hP e2i−1 , ∇X e2j−1 i + hF e2i−1 , σ(e2j−1 , X)i

= −(cos θ) h∇X e2i−1 , e2j i − h∇X e2j−1 , e2i i

+ (sin θ) e(2i−1)∗ , σ(X, e2j−1 ) − e(2j−1)∗ , σ(X, e2i−1 ) .
Thus we find
2j 2i

(cot θ)(ω2i−1 − ω2j−1 )(X) = Ae(2i−1)∗ e2j−1 − Ae(2j−1)∗ e2i−1 , X .
Hence, by using ωir (X) = hAer ei , Xi, we get (7.99).
Similarly, after taking the derivatives of the following equations:
hJe2i−1 , e2j i = (cos θ)δij , hJe2i , e2j i = 0,

Je(2i−1)∗ , e(2j−1)∗ = 0, Je(2i)∗ , e(2j)∗ = 0,

Je2i−1 , e(2j−1)∗ = (sin θ)δij ,

Je2i−1 , e(2j)∗ = 0,

Je2i , e(2j)∗ = (sin θ)δij , Je(2i)∗ , e(2j)∗ = 0,
we obtain (7.100)-(7.107), respectively.
Lemma 7.10. Let N be a proper slant submanifold of Cn with dim N =
n = 2k. Then with respect to a slant frame we have
(2j)∗ (2j−1)∗ (2i)∗ (2i−1)∗
ω2i + ω2i−1 = ω2j + ω2j−1 , (7.109)
∗ ∗
(2j) (2j−1)
2j 2j−1
ω(2i)∗ − ω(2i−1)∗ = ω2i − ω2i−1 , (7.110)
2i−1 (2i−1)∗ 2j−1 (2j−1)∗
ω2j − ω(2j)∗ = ω2i − ω(2i)∗ (7.111)
for any i, j = 1, . . . , k.
Proof. Formula (7.109) follows from (7.99) and (7.101). (7.110) follows
from (7.104) and (7.106). And (7.111) follows from (7.101) and (7.103).
Now we return to the proof of Theorem 7.12. From the definition we
have
X2k
∗
Θ= ωℓℓ . (7.112)
ℓ=1
Thus from the structure equations we have
Xk Xk
2j (2i)∗ 2j−1 (2i)∗
−dΘ = ω2i ∧ ω2j + ω2i ∧ ω2j−1
i,j=1 i,j=1
k
X k
X
(2j)∗ (2i)∗ (2j−1)∗ (2i)∗
+ ω2i ∧ ω(2j)∗ + ω2i ∧ ω(2j−1)∗
i,j=1 i,j=1
(7.113)
k
X k
X
2j (2i−1)∗ 2j−1 (2i−1)∗
+ ω2i−1 ∧ ω2j + ω2i−1 ∧ ω2j−1
i,j=1 i,j=1
k
X k
X
(2j)∗ (2i−1)∗ (2j−1)∗ (2i−1)∗
+ ω2i−1 ∧ ω(2j)∗ + ω2i−1 ∧ ω(2j−1)∗ .
i,j=1 i,j=1
By using (7.106) we have

k
X k
X
(2j)∗ (2i)∗ 2j−1 (2i−1)∗
ω2i ∧ ω(2j)∗ + ω2i−1 ∧ ω2j−1
i,j=1 i,j=1
k
X k
X
(2j−1)∗ (2i−1)∗ 2j (2i)∗
+ ω2i−1 ∧ ω(2j−1)∗ + ω2i ∧ ω2j
i,j=1 i,j=1
k
X k
X
(2j)∗ (2i)∗ 2j−1 (2i−1)∗
= ω2i ∧ ω(2j)∗ + ω2i−1 ∧ ω2j−1
i,j=1 i,j=1
k
X k
X
(2j−1)∗ 2i−1 2i (2i)∗ 2j (2i)∗
+ ω2i−1 ∧ (ω2j−1 − ω2j + ω(2j)∗ ) + ω2i ∧ ω2j
i,j=1 i,j=1
k
X k
X
(2j)∗ (2i)∗ (2j−1)∗ (2i)∗
= ω2i ∧ ω(2j)∗ + ω2i−1 ∧ ω(2j)∗
i,j=1 i,j=1
k
X k
X
(2j−1)∗ 2i 2j (2i)∗
− ω2i−1 ∧ ω2j + ω2i ∧ ω2j
i,j=1 i,j=1
k
X k
X
(2j−1)∗ (2j)∗ (2i)∗ 2i (2j−1)∗ (2j)∗
= (ω2i−1 +ω2i ) ∧ ω(2j)∗ + ω2j ∧ (ω2i−1 +ω2i ).
i,j=1 i,j=1
(2j−1)∗ (2j)∗ 2j
Since ω2i−1 + ω2i is symmetric in i.j by Lemma 7.10 and ω2i and
∗
(2i)
ω(2j)∗ are skew-symmetric in i, j, we obtain
k
X k
X
(2j)∗ (2i)∗ 2j−1 (2i−1)∗
ω2i ∧ ω(2j)∗ + ω2i−1 ∧ ω2j−1
i,j=1 i,j=1
(7.114)
k
X k
X
(2j−1)∗ (2i−1)∗ 2j (2i)∗
+ ω2i−1 ∧ ω(2j−1)∗ + ω2i ∧ ω2j = 0.
i,j=1 i,j=1
Moreover, by (7.107), we have

k
X k
X
2j−1 (2i)∗ (2j−1)∗ (2i)∗
ω2i ∧ ω2j−1 + ω2i ∧ ω(2j−1)∗
i,j=1 i,j=1
k
X k
X
2j (2i−1)∗ (2j)∗ (2i−1)∗
+ ω2i−1 ∧ ω2j + ω2i−1 ∧ ω(2j)∗
i,j=1 i,j=1
k
X k
X
2j−1 (2i)∗ (2j−1)∗ (2j)∗ 2i 2j
= ω2i ∧ ω2j−1 + ω2i ∧ (ω(2i−1)∗ + ω2j−1 − ω2i−1 )
i,j=1 i,j=1
k
X k
X
2j (2i−1)∗ (2j)∗ (2i−1)∗
+ ω2i−1 ∧ ω2j + ω2i−1 ∧ ω(2j)∗
i,j=1 i,j=1
k
X k
X
2j−1 (2i)∗ 2i−1 (2j−1)∗
= ω2i ∧ ω2j−1 − ω2j ∧ ω2i
i,j=1 i,j=1
k
X k
X
(2j−1)∗ (2i−1)∗ (2j)∗ (2i−1)∗
− ω2i ∧ ω(2j)∗ + ω2i−1 ∧ ω(2j)∗
i,j=1 i,j=1
k
X
2i−1 (2i−1)∗ (2j)∗ (2j−1)∗
= (ω2j − ω(2j)∗ ) ∧ (ω2i−1 − ω2i ).
i,j=1
2i−1 (2i−1)∗
Since ω2j − ω(2j)∗ is symmetric in i, j according to Lemma 7.10 and
(2j)∗ (2j−1)∗
ω2i−1 − ω2i is skew-symmetric in i, j by (7.100) in Lemma 7.9, we find
k
X k
X
2j−1 (2i)∗ (2j−1)∗ (2i)∗
ω2i ∧ ω2j−1 + ω2i ∧ ω(2j−1)∗
i,j=1 i,j=1
(7.115)
k
X k
X
2j (2i−1)∗ (2j)∗ (2i−1)∗
+ ω2i−1 ∧ ω2j + ω2i−1 ∧ ω(2j)∗ = 0.
i,j=1 i,j=1
Now, the theorem follow from (7.113), (7.114) and (7.115).
Remark 7.5. If N is not slant in Cn , the 1-form Θ is not necessary closed.

For example, if S 2 is the unit 2-sphere embedded standardly in E3 ⊂ E4 ,
then S 2 is not slant with respect to any compatible complex structure on
E4 . Moreover, dΘ 6= 0 on S 2 . The later statement can be seen as follow:
Let S 2 is parametrized by
x(θ, ϕ) = (sin ϕ cos θ, sin ϕ sin θ, cos ϕ, 0). (7.116)
Put
e1 = (− sin θ, cos θ, 0, 0),
(7.117)
e2 = (cos ϕ cos θ, cos ϕ sin θ, − sin ϕ, 0).
With respect to the complex structure J0 we have
1 1
F e1 = − sin 2θ sin 2ϕ, − sin2 θ sin 2ϕ, − sin θ cos2 ϕ, cos θ ,
4 2 (7.118)
F e2 = (cos2 θ sin ϕ, sin θ cos θ sin ϕ, cos θ cos ϕ, sin θ cos ϕ).
Let us put
F e1 = |F e1 |e1∗ , F e2 = |F e2 |e2∗ .
Then with respect to the local orthonormal frame e1 , e2 , e1∗ , e2∗ on the
open subset of S 2 on which θ 6≡ π2 and ϕ 6≡ 0 (mod π), we have
∗ ∗ sin θ sin 2ϕdθ − 2 cos θdϕ
Θ = ω11 + ω22 = .
2 − 2 sin2 θ cos2 ϕ
Now it is direct to verify that dΘ 6= 0.
For a proper slant submanifold N of a Kähler manifold M̃ , we put

Λ(X, Y ) = hX, P Y i (7.119)
for X, Y ∈ T N . It follows from Lemmas 7.4, 7.5 and 7.6 that Λ is a non-
degenerate 2-form on N , i.e., Λk 6= 0 with k = 21 dim N .
Definition 7.2. An 2k-dimensional manifold is called a symplectic mani-

fold if it admits a non-degenerate closed 2-form Ψ, i.e., a closed 2-form Ψ
satisfying Ψk 6= 0.
Theorem 7.13. [Chen (1990b)] Let N be an n-dimensional proper slant

submanifold of a Kähler manifold M̃ . Then Λ is closed, i.e., dΛ = 0. Hence
Λ defines a canonical cohomology class of N :
[Λ] ∈ H 2 (N ; R). (7.120)
Consequently, every proper slant submanifold of a Kähler manifold is a
symplectic manifold.
Proof. From the definition of the exterior differentiation we have

1
dΛ(X, Y, Z) = {XΛ(Y, Z) + Y Λ(Z, X) + ZΛ(X, Y )
3
− Λ([X, Y ], Z) − Λ([Y, Z], X) − Λ([Z, X], Y )}.
Thus, by the definition of Λ, we find
1
dΛ(X, Y, Z) = h∇X Y, P Zi + hY, ∇X (P Z)i + h∇Y Z, P Xi
3
+ hZ, ∇Y (P X)i + h∇Z X, P Y i + hX, ∇Z (P Y )i

− h[X, Y ], P Zi − h[Z, X], P Y i − h[Y, Z], P Xi
1
= hY, ∇X (P Z)i + hZ, ∇Y (P X)i + hX, ∇Z (P Y )i
3
+ h∇X Z, P Y i + h∇Y X, P Zi + h∇Z Y, P Xi .
So, after applying the definition of ∇P we obtain
1
dΛ(X, Y, Z) = hX, (∇Z P )Y i + hY, (∇X P )Zi
3 (7.121)

+ hZ, (∇Y P )Xi .
Hence by applying (7.7) and (7.121) we get

1
dΛ(X, Y, Z) = {hX, tσ(Y, Z)i + hX, AF Y Zi + hY, tσ(Z, X)i
3 (7.122)
+ hY, AF Z Xi + hZ, tσ(X, Y )i + hZ, AF X Y i .
Therefore, after applying (2.14), (7.1), (7.6) and (7.122), we obtain (7.120).
As an immediate consequence of Theorem 7.13 we have the following.
Theorem 7.14. [Chen (1990b)] If N is a compact 2k-dimensional proper

slant submanifold of a Kähler manifold, then
H 2i (N ; R) 6= 0 (7.123)
for any i = 1, . . . , k.
Theorem 7.14 implies the following non-existence result from [Chen

(1990b)].
Theorem 7.15. If N is a compact 2k-dimensional differentiable manifold

satisfying H 2i (N ; R) = 0 for some i ∈ {1, . . . , k}, then N cannot be im-
mersed in any Kähler manifold as a proper slant submanifold.
7.7 Pointwise slant submanifolds
The following notion of pointwise slant submanifolds is natural extension

of the notion of slant submanifolds.
Definition 7.3. An immersion φ : N → M̃ from a manifold N into an

almost Hermitian manifold M̃ is called pointwise slant if, at each given
point x ∈ N , the Wirtinger angle θ(X) is independent of the choice of the
nonzero tangent vector X ∈ Tx∗ N . The function θ on N is called the slant
function. A pointwise slant submanifold is called pointwise proper slant if
it contains no totally real points.
Remark 7.6. Pointwise slant submanifolds have been defined and studied
in [Etayo (1998)] under the name of quasi-slant submanifolds.
Example 7.11. Every 2-dimensional submanifold in an almost Hermitian

manifold is pointwise slant.
Example 7.12. Every slant (resp. proper slant) submanifold in an almost

Hermitian manifold is pointwise slant (resp. pointwise proper slant).
Example 7.13. Let E4n = (R4n , g0 ) denote the Euclidean 4n-space en-
dowed with the standard Euclidean metric g0 and let {J0 , J1 } be a pair of
almost complex structures on E4n satisfying J0 J1 = −J1 J0 .
Assume that J0 , J1 are orthogonal almost complex structures, i.e., they
are compatible with the Euclidean metric g0 ; thus
g0 (Ji X, Ji Y ) = g0 (X, Y ), i = 0, 1,
for X, Y ∈ T (E ). Let us denote (R4n , J0 , g0 ) by C2n
4n
0 .
For any real-valued function f : E4n → R, we define an almost complex
structure Jf on E4n by
Jf = (cos f )J0 + (sin f )J1 . (7.124)
Then C2n 4n
= (R , Jf , g0 ) is an almost Hermitian manifold.
f
For any given pointwise slant submanifold (resp. slant submanifold)
M in (E4n , J0 , g0 ), M is also a pointwise slant submanifold (resp. slant
submanifold) of C2n f = (R4n , Jf , g0 ). In particular, if M is a complex
2n
submanifold of C0 , then M is a pointwise slant minimal submanifold in
C2n
f whose slant function θ is the restriction of f on M , i.e., θ = f |M .
Remark 7.7. There do exist pairs of orthogonal almost complex struc-

tures {J0 , J1 } which satisfy the conditions mentioned in Example 7.13. For
instance, let J0 and J1 be the two orthogonal almost complex structures on
E4n defined by
J0 (a1 , . . . , a2n , b1 , . . . , b2n )
= (−b1 , . . . , −b2n , a1 , . . . , a2n ),
(7.125)
J1 (a1 , . . . , a2n , b1 , . . . , b2n )
= (−a2 , a1 , . . . , −a2n , a2n−1 , b2 , −b1 , . . . , b2n , −b2n−1 ).
Then J0 J1 = −J1 J0 . Since there exist many real-valued functions on C2n
and many complex submanifolds in C2n , we may conclude from Example
7.13 that there are infinitely many examples of pointwise slant minimal
submanifolds in almost Hermitian manifolds which are not slant.
Lemma 7.11. An immersion φ : N → M̃ of a manifold N into an almost

Hermitian manifold M̃ is a pointwise slant immersion if and only if we
have P 2 = −(cos2 θ)I for some real-valued function θ defined on the tangent
bundle T N of N .
Proof. If φ : N → M̃ is a pointwise slant immersion with slant function
θ : N → R, then we have
g(P X, P X) = cos2 θ(x)g(X, X)
for X ∈ Tx N . Combining this with (7.4) yields

g(P 2 X, X) = − cos2 θ(x)g(X, X).
Thus, after applying polarization, we obtain P 2 = −(cos2 θ)I on T N .
Conversely, if N is a submanifold of M̃ satisfying P 2 = −(cos2 θ)I for
some function θ on N , then
g(P X, P X) = −g(P 2 X, X) = cos2 θ(x)g(X, X),
which implies that θ is independent of the choice of X ∈ Tx∗ N at each given
point x ∈ N . Therefore, the submanifold is pointwise slant.
The following result is an immediate consequence of Lemma 7.11.
Corollary 7.7. Let φ : N → M̃ be a pointwise slant immersion of an
n-manifold into an almost Hermitian manifold. If φ is not a totally real
immersion, then N is even-dimensional.
The following results are due to [Chen and Garay (2012b)].
Proposition 7.4. Let φ : N → M̃ be an immersion of a manifold N into
an almost Hermitian manifold M̃ . Then φ is a pointwise slant immersion
if and only if P : T N → T N preserves orthogonality, i.e., P carries each
pair of orthogonal vectors into orthogonal vectors.
Proof. Let φ : N → M̃ be an immersion of an n-manifold N into an
almost Hermitian manifold M̃ . Denote by θ : T 1 N → R the Wirtinger
function on the unit tangent bundle T 1 N defined in section 1. Clearly,
with respect to the induced metric, Tx1 N is the unit hypersphere Σx in
Tx N centered at o. At a given point x ∈ N , we have
g(P X, P X) = cos2 θ(X), X ∈ Tx1 N. (7.126)
For each unit vector Y tangent to Σx at X ∈ Σx (hence Y ⊥ X), we have
2g(P X, P Y ) = Y g(P X, P X) = −(Y θ) sin 2θ(X). (7.127)
Consequently, P carries each pair of orthogonal vectors in Tx N into a pair
of orthogonal vectors in Tx N if and only if the Wirtinger function θ is
independent of the choice of X ∈ Tx1 N . This implies the proposition.
The following proposition shows that the notions of pointwise slant sub-
manifolds and slant functions are conformal invariant.
Proposition 7.5. If φ : N → (M̃ , J, g̃) is a pointwise slant immersion of
a manifold N into an almost Hermitian manifold M̃ . Then, for any given
function f on M̃ , φ : N → (M̃ , J, e2f g̃) is pointwise slant with the same
slant function as φ : N → (M̃ , J, g̃).
Proof. Let φ : N → (M̃ , J, g̃) be a pointwise slant immersion with slant

function θ. Then, for any X ∈ Tx N , we have
g̃(P X, P X) = (cos2 θ)g̃(X, X).
Thus if we put ḡ = e2f g̃ for a given function f on M̃ , then

ḡ(P X, P X) = e2f g̃(P X, P X)
= (cos2 θ)e2f g̃(X, X)
= (cos2 θ)ḡ(X, X).
Hence φ : N → (M̃ , J, e2f g̃) is a pointwise slant immersion with the same
slant function as φ : N → (M̃ , J, g̃).
The following conformal property of slant immersions is an immediate
consequence of Proposition 7.5.
Corollary 7.8. If φ : N → (M̃ , J, g̃) is a slant immersion of N into an

almost Hermitian manifold M̃ . Then, for every given function f on M̃ ,
φ : N → (M̃ , J, e2f g̃) is also a slant immersion with the same slant angle.
A Hermitian n-manifold (M̃ , J, g) is called a locally conformal Kähler

manifold if there exist an open cover {Ui }i∈I of M̃ and a family {fi }i∈I of
smooth functions fi : Ui → R such that each local metric g̃i = e−2fi g|Ui is
Kählerian (cf. [Dragomir and Ornea (1998)]).
Another interesting application of Proposition 7.5 is the following.
Corollary 7.9. For each integer n ≥ 1, there exist infinitely many 2n-
dimensional totally umbilical proper slant submanifolds in locally conformal
Kähler 2n-manifolds.
Proof. Let N be an open domain of a proper slant 2n-plane of the com-

plex Euclidean 2n-space (C2n , J, g0 ) and let f be a smooth function defined
on N . Then N is totally geodesic in C2n .
Consider the new metric g ∗ = e2f g0 . Let ξ denote the vector field
associated with the 1-form df with respect to g0 , i.e., g0 (ξ, Z) = df (Z) for
Z ∈ T C2n . Then it follows from [Chen (1974)] that the second fundamental
σ ∗ of N in (C2n , J, g ∗ ) satisfies σ ∗ (X, Y ) = g0 (X, Y )ξ N , where ξ N is the
normal component of the vector field ξ|M . Thus M is a non-totally geodesic,
totally umbilical submanifold of the locally conformal Kähler 2n-manifold
(C2n , J, g ∗ ) whenever ξ N 6= 0. Therefore it follows from Corollary 7.8 that
N is a proper slant submanifold of (C2n , J, g ∗ ).
The following result provides a necessary and sufficient condition for a

pointwise slat submanifold to be slant.
Theorem 7.16. A pointwise slant submanifold N in a Kähler manifold is

slant if and only if the shape operator of N satisfies
AF X P X = AF P X X (7.128)
for all X ∈ T N.
Proof. Let N be a pointwise slant submanifold of a Kähler manifold M̃

with slant function θ. For any unit tangent vector field X of N , we put
P X = (cos θ)X ∗ , where X ∗ is a unit tangent vector field orthogonal to X.
Then, for any tangent vector Y ∈ T N , we have
∇˜ Y (JX) = ∇˜ Y ((cos θ)X ∗ ) + ∇
˜Y FX
= (cos θ)∇Y X ∗ + (cos θ)σ(Y, X ∗ ) (7.129)
∗
− (sin θ)(Y θ)X − AF X Y + DY (F X).
On the other hand, we also have
˜ Y (JX) = P (∇Y X) + F (∇Y X) + tσ(X, Y ) + f σ(X, Y ).
∇ (7.130)
After comparing the tangential components of (7.129) and (7.130), we find
(sin θ)(Y θ)X ∗ = (cos θ)∇Y X ∗ − AF X Y − P (∇Y X) − tσ(X, Y ). (7.131)
Thus, by taking the inner product of (7.131) with X ∗ , we obtain
(sin θ)Y θ = g̃(σ(X, Y ), F X ∗ ) − g̃(σ(Y, X ∗ ), F X).
Consequently, the pointwise slant submanifold is slant if and only if we have
AF X ∗ X = AF X X ∗ for any X ∈ T N , which implies the theorem.
Some easy consequences of Theorem 7.16 are the following.
Corollary 7.10. Every totally geodesic pointwise slant submanifold of any

Kähler manifold is slant.
Proof. Follows from Theorem 7.16 and the fact that the shape operator
A vanishes identically for totally geodesic submanifolds.
Remark 7.8. Corollary 7.10 have been proved in [Etayo (1998)] under an
additional assumption that the totally geodesic pointwise slant submanifold
is complete.
Corollary 7.11. Let N be a 2n-dimensional totally umbilical pointwise

proper slant submanifold of a Kähler 2n-manifold M̃ . If N is non-totally
geodesic, then N is non-slant in M̃ .
Proof. Assume that N is a 2n-dimensional totally umbilical pointwise

proper slant submanifold of a Kähler 2n-manifold M̃ . Then the second
fundamental form σ of N satisfies
σ(X, Y ) = g(X, Y )H
for X, Y ∈ T N . Thus we have
g(AF P X X, Y ) = g̃(σ(X, Y ), F P X) (7.132)
= g̃(F P X, H)g(X, Y ),
g(AF X P X, Y ) = g̃(F X, H)g(P X, Y ). (7.133)
If N is slant, then Theorem 7.16, (7.132) and (7.133) imply that
g̃(F P X, H)X = g̃(F X, H)P X = 0. (7.134)
Since N is non-totally geodesic, totally umbilical in M̃ , N is a non-minimal

submanifold. Therefore N cannot be a complex submanifold of the Kähler
manifold M̃ .
By our assumption, there exists a point x ∈ N with H(x) 6= 0. Hence,
after applying the condition dim N = dimC M̃ , we conclude that there
exists a vector X ∈ T N satisfying F X = H(x) 6= 0. Therefore, by applying
(7.134) we obtain P X = 0. Consequently, x is a totally real point, which
contradicts to the assumption that M is pointwise proper slant. Therefore
N cannot be a slant submanifold.
For totally umbilical surfaces in an arbitrary Kähler manifold, we have
the following.
Proposition 7.6. We have the following:
(1) Every totally geodesic surface in any Kähler surface is slant.

(2) Every non-totally geodesic, totally umbilical surface in any Kähler sur-
face is a non-slant, pointwise slant surface, unless it is Lagrangian.
Proof. Statement (1) is a special case of Corollary 7.10. Statement (2)

follows from Corollary 7.11 and the fact that every complex surface in a
Kähler manifold is minimal.
Remark 7.9. Proposition 7.6(2) shows that each totally umbilical surface
of the Euclidean complex plane is a non-slant, pointwise slant surface.
Remark 7.10. It follows from Corollary 7.9 that Proposition 7.6(2) is false
if the Kähler surface were replaced by a locally conformal Kähler surface.
7.8 Contact slant submanifolds via canonical fibration
Let (M 2m+1 , φ, ξ, η, g) be an almost contact metric manifold with a Rie-

mannian metric g, the almost contact (1, 1)-tensor φ, and the characteristic
vector field ξ.
An immersion i : N → M 2m+1 is called contact θ-slant if
(i) the characteristic vector field ξ of M̃ 2m+1 is tangent to i∗ (T N ) and
(ii) for each X 6= 0 tangent to i∗ (Tp N ) and perpendicular to ξ, the angle
θ(X) between φ(X) and i∗ (Tp N ) is independent of the choice of X.
In this section, we present the method for constructing slant subman-
ifolds in complex space forms via the canonical fibrations given in [Chen
and Tazawa (2000)].
We use the same notations as section 5.2.
Case (1): Slant submanifolds in CP m (4). Consider the Hopf fibration
π : S 2n+1 (1) → CP n (4). Let i : N → CP m (4) be an isometric immersion.
Then N̂ = π −1 (N ) is a principal circle bundle over N with totally geodesic
fibers. The lift î : N̂ → S 2m+1 (1) of i is an isometric immersion such that
π ◦ î = i ◦ π.
Conversely, if ψ : N̂ → S 2m+1 (1) is an isometric immersion, invari-
ant under the action of C∗ = C − {0}, then there is a unique isometric
immersion ψπ : π(N̂ ) → CP m (4) such that π ◦ ψ = ψπ ◦ π.
We call ψπ : π(N̂ ) → CP m (4) the projection of ψ : N̂ → S 2m+1 (1).
Proposition 7.7. An isometric immersion i : N → CP m (4) is θ-slant if

and only if the lift î : N̂ → S 2m+1 (1) is contact θ-slant.
Let σ, σ̂ be the second fundamental forms of i, î, respectively. Then

σ̂(X̄, Ȳ ) = (σ(X, Y ))¯, σ̂(X̄, V ) = (F X)¯, σ̂(V, V ) = 0
for X, Y ∈ X(N ), where F X is the normal component of JX in CP m (4).
Proposition 7.7 shows that in order to obtain the explicit expression of
a desired θ-slant submanifold of CP m (4) with second fundamental form σ,
it is sufficient to construct a contact θ-slant submanifold of S 2m+1 (1) whose
second fundamental form satisfies π∗ σ̂ = σ, and vice versa.
Case (2): Slant submanifolds in CH m (−4). Consider the generalized Hopf
fibration π : H12n+1 (−1) → CH n (−4). Let i : N → CH m (−4) be an
isometric immersion. Then N̂ = π −1 (N ) is a principal circle bundle over N
with totally geodesic fibers and the lift î : N̂ → H12m+1 (−1) is an isometric
immersion with π ◦ î = i ◦ π.
Conversely, if ψ : N̂ → H12m+1 (−1) is an isometric immersion invariant

under the action of C∗ , then there exists a unique isometric immersion
ψπ : π(N̂ ) → CH m (−4) such that π ◦ ψ = ψπ ◦ π.
Similar to Proposition 7.7 we have
Proposition 7.8. An isometric immersion i : N → CH m (−4) is θ-slant

if and only if the lift î : N̂ → H12m+1 (−1) is contact θ-slant.
Denote by ∇ ˆ and ∇˜ the Levi-Civita connections of S 2m+1 (1) and of

m 2m+1
CP (4) (resp., of H1 (−1) and of CH m (−4)), X̄ the horizontal lift of
X, and by σ and σ̂ the second fundamental forms of i and î, respectively.
Denote by ∇ ˘ the Levi-Civita connection of Cm+1 (resp. of Cm+1 ). If N is
1
a θ-slant submanifold of the complex projective m-space CP m (4) (resp. of
CH m (−4)) and if z : S 2m+1 (1) → Cm+1 (resp. z : H12m+1 (−1) → Cm+11 )
denotes the standard inclusion, then
˘ X̄ Ȳ ∗ = (∇X Y )¯ + (σ(X, Y ))¯ + hJX, Y i iz − ε hX, Y i z
∇ (7.135)
˘ X̄ V = ∇
∇ ˘ V X̄ = (JX)¯, (7.136)
˘ V V = εz,
∇ (7.137)
for X, Y ∈ X(N ), where ε = 1 for CP m (4) and ε = −1 for CH m (−4).
One may obtain the desired θ-slant submanifold by solving the system
(7.135)-(7.137). The general construction procedure goes as follows:
First one determines both the intrinsic and extrinsic structures of the θ-
slant submanifold in order to find the precise form of the differential system
(7.135)-(7.137).
Next, one constructs a coordinate system on the associated contact θ-
slant submanifold π −1 (N ). After that one may solve the differential system
via the coordinate system on N̂ to obtain a solution of the system. The
solution of the system gives rise to the explicit expression of the associ-
ated contact θ-slant submanifold of S 2m+1 (1) or H12m+1 (−1) which in turn
provides the representation of the desired θ-slant submanifold via π.
Case (3): Slant submanifolds in Cn . Similarly, one may apply the canonical
fibration π : R2n+1 → Cn in (5.11) to obtain the one-to-one correspondence
between contact slant submanifolds of R2n+1 and slant submanifolds of Cn .
Remark 7.11. For some explicit constructions of slant submanifolds via

the construction method given in this section, see [Chen and Tazawa
(2000)].
Chapter 8
Generic Submanifolds of Kähler

Manifolds
A real submanifold N of an almost Hermitian manifold (M̃ , g̃, J) is called

a generic submanifold if the maximal complex subspace, Tx N ∩ J(Tx N ), of
Tx N is of constant dimension along N . Generic submanifolds are “generic”
in the sense that every real submanifold of an almost Hermitian manifold
is the closure of the union of some generic submanifolds of the almost
Hermitian manifold.
Some basic results on generic submanifolds of Kähler Manifolds can be
found in [Chen (1981a,e)].
8.1 Generic submanifolds
Let N be a submanifold of an almost Hermitian manifold (in particular, of

a Kähler manifold) M̃ . For a given point x ∈ N , we put
Hx = Tx N ∩ J(Tx N ). (8.1)
Then Hx is the maximal complex subspace of the tangent space Tx M̃ which
is contained in Tx N .
Definition 8.1. Let N be a submanifold of a Kähler manifold M̃ . If the
dimension of Hx is constant along the submanifold N , then N is called a
generic submanifold.
A generic submanifold is said to be proper if we have Hx 6= {0} and
Hx 6= Tx N .
The following result shows the differentiability of the distribution H on
generic submanifolds.
Lemma 8.1. For any generic submanifold N of an almost complex mani-
fold (M̃ , J), the distribution H is a differentiable distribution.
179
Proof. Let N be a generic submanifold of an almost complex manifold

(M̃ , J). Then the Whitney sum T N ⊕ T N is a differentiable vector bundle
over N . Let us define the following differentiable map:
φ : T N ⊕ T N → T M̃
by φ(X, Y ) = X − JY . Since N is a generic submanifold, the implicit
function theorem implies that the kernel of φ, φ−1 (0), is a differentiable
submanifold of T N ⊕ T N . Let
ψ : TN ⊕ TN → TN
be the projection given by ψ(X, Y ) = X. Then ψ|φ−1 (0) is a one-to-one
map and we have H = (φ−1 (0)). This prove the lemma.
The distribution H defined above is called the holomorphic distribution
of the generic submanifold N . For the generic submanifold, let H⊥ denote
the orthogonal complimentary subbundle of H in T N . It is obviously that
JH⊥ ∩ H⊥ = {0}. (8.2)
⊥
It follows from Lemma 8.1 that H is also a differentiable distribution on
N , which is called the purely real distribution. A generic submanifold N is
called purely real if dimC Hx = 0 holds for every x ∈ N .
Throughout this book, for a generic submanifold N we put
h = dimC Hx , p = dimR Hx⊥ , x ∈ N. (8.3)
Let N be a generic submanifold of a Kähler manifold M̃ . For each
vector X ∈ T N , we put
JX = P X + F X, (8.4)
respectively. Then P is an endomorphism of T N and F is a normal-bundle-
valued 1-form on T N .
For each vector ξ normal to N , let tξ and f ξ denote the tangential and
normal components of Jξ. Then we have
Jξ = tξ + f ξ. (8.5)
For a generic submanifold N , we have
Hx ⊥ Hx⊥ , P Hx = Hx and P Hx⊥ ⊂ Hx⊥ . (8.6)
Let νx be the vector subspace of the normal space Tx⊥ N given by
νx = Tx⊥ N ∩ J(Tx⊥ N ). (8.7)
Generic Submanifolds of Kähler Manifolds 181
Then ν is a complex vector subbundle of T ⊥ N . Clearly, we have

T ⊥ N = F H⊥ ⊕ ν, t(T ⊥ N ) = H⊥ , and F H⊥ ⊥ ν. (8.8)
Lemma 8.2. Let N be a generic submanifold of a Kähler manifold with
second fundamental form σ. The we have
hJσ(X, U ), ξi = hσ(JX, U ), ξi
for X ∈ H, U ∈ T N and ξ ∈ ν.
Proof. It follows from Gauss’ formula and the third condition in (8.8)
that
˜ U X, ξ i = h ∇
hJσ(X, U ), ξi = h J ∇ ˜ U JX, ξ i = hσ(JX, U ), ξi
holds. Thus we have the lemma.
8.2 Integrability
In this section we discuss the integrability of holomorphic and purely real

distributions on generic submanifolds.
Proposition 8.1. The holomorphic distribution H of a generic submani-

fold N of a Kähler manifold is integrable if and only if
hσ(X, JY ), F Zi = hσ(JX, Y ), F Zi (8.9)
⊥
holds for X, Y ∈ H and Z ∈ H .
Proof. Since M̃ is Kählerian, the formula of Gauss gives
σ(X, JY ) − σ(JX, Y ) = J[X, Y ] + ∇Y JX − ∇X JY (8.10)
for vector fields X, Y in H. If the holomorphic distribution H is inte-
grable, then the right-hand side of (8.10) always lies in T N . Thus we find
σ(X, JY ) = σ(JX, Y ) which implies (8.9).
Conversely, if (8.9) holds, then we find from Lemma 8.2 that σ(X, JY ) =
σ(JX, Y ) for any for X, Y ∈ H. Thus after applying (8.10), we obtain
J[X, Y ] = ∇X JY − ∇Y JX. Since ∇X JY − ∇Y JX is tangent to N , this
shows that [X, Y ] lies in H. Therefore the proposition follows from the
Frobenius theorem.
Proposition 8.2. The purely real distribution H⊥ of a generic submanifold

N of a Kähler manifold is integrable if and only if
∇Z (P W ) − ∇W (P Z) + AF Z W − AF W Z ∈ H⊥ (8.11)
⊥
holds for Z, W ∈ H .
Proof. For any vector fields Z, W ∈ H⊥ we have

J∇Z W + Jσ(Z, W ) = ∇Z (P W ) + σ(Z, P W ) − AF W Z + DZ (F W ),
from which we find
[Z, W ] = P {AF W Z − AF Z W + ∇W (P Z) − ∇Z (P W )}
+ t{σ(W, P Z) − σ(Z, P W ) + DW (F Z) − DZ (F W )}.
Since t(T N ) = H⊥ , this proves the proposition.
⊥

Proposition 8.3. Let N be a generic submanifold of a Kähler manifold.

H is integrable
If the holomorphic distribution and its leaves are totally
geodesic in N , then we have σ(H, H), F H⊥ = {0}.
Proof. Under the hypothesis, we have ∇X Z ∈ H⊥ for vector fields X in

H and Z in H⊥ . Thus we obtain
0 = h∇X Z, JY i = hAF Z X, Y i − h∇X P Z, Y i = hAF Z X, Y i
for any Y ∈ H. This proves the proposition.

If the purely real distribution H⊥ is integrable and its leaves are totally
geodesic in N , then we have

σ(H, H⊥ ), F H⊥ = {0} or equivalently, AF H⊥ H ⊂ H. (8.12)
Proof. Under the hypothesis, we find

˜ Z X, W i = h ∇
0 = h ∇Z X, W i = h ∇ ˜ Z JX, P W i + h ∇
˜ Z JX, F W i
= − h JX, ∇Z P W i + hσ(JX, Z), F W i = hAF W JX, Zi
for vector fields X ∈ H and Z, W ∈ H⊥ . This proves the proposition.
The last two propositions play important roles since they show that if
one imposes suitable intrinsic conditions on generic submanifolds, one may
obtain important information about the submanifolds.
8.3 Parallelism of P and F
For a generic submanifold N of a Kähler manifold M̃ , we have the canonical

structures P and F defined by (8.4). The covariant derivatives ∇P and ∇F
of P and F are defined respectively by
(∇X P )Y = ∇X (P Y ) − P (∇X Y ), (8.13)
(∇X F )Y = DX (F Y ) − F (∇X Y ) (8.14)
for vector fields X, Y ∈ T N . The structures P (respectively, F ) is said to

be parallel if ∇P = 0 (respectively, ∇F = 0).
For a generic submanifold N of a Kähler manifold, we have
∇U (P V ) + σ(U, P V ) − AF V U + DU F V = J∇U V + Jσ(U, V ) (8.15)
for U, V ∈ T N . Thus we have
(∇U P )V = tσ(U, V ) + AF V U. (8.16)
Hence, for any vector fields U, V, W in T N , we get
h(∇U P )V, W i = hAF V W − AF W V, U i ,
which implies the next proposition.

Then P is parallel if and only if AF U V = AF V U holds for U, V ∈ T N .

If P is parallel, then we have
(a) AF U X = 0 for X ∈ H and U ∈ T N ;
(b) the holomorphic distribution H is integrable.
Proof. If ∇P = 0, then Proposition 8.5 and F H = {0} yields AF U X =

AF X U = A0 U = 0 for X ∈ H and U ∈ T N . Thus we get statement (a).
Now, assume ∇P = 0 holds. Then it follows from AF U X = 0 that
hσ(X, JY ), F Zi = 0 for all X, Y ∈ H and Z ∈ H⊥ . Consequently, by
Proposition 8.1 we conclude that H is integrable.

Then F is parallel if and only if Af ξ U = −Aξ (P U ) holds for U ∈ T N and
ξ ∈ T ⊥N .
Proof. From (8.14) and (8.15) we get (∇U F )V = f σ(U, V ) − σ(U, P V ).

So h(∇U F )V, ξi = h−Aξ (P V ) − Af ξ V, U i holds. This gives the result.

If F is parallel, then for vector fields X ∈ H, Z ∈ H⊥ and U, V ∈ T N we
have:
(1) the holomorphic distribution H is integrable and its leaves are totally
geodesics in N ;
(2) F H⊥ and ν are parallel in the normal bundle;
(3) (∇X P )Z = (∇U P )Z = 0.
Proof. Assume ∇F ¯ = 0 and let X ∈ H, Z ∈ H⊥ and U ∈ T N . Then we

˜
find from ∇U (JX) = J ∇ ˜ U X that σ(U, JX) = F (∇U X) + f σ(X, U ). Thus
we have hAξ (P X), U i = − h∇U X, tξi − hAf ξ X, U i for any normal vector
field ξ. By combining this with Proposition 8.7 and also using the second
equation in (8.8) we get ∇U X ∈ H which implies statement (1).
For ξ ∈ ν, we have Jξ ∈ ν. Thus Gauss’ formula and (8.4) imply that
˜ V JU, ξ i = − hσ(V, P U ), ξi − hDV F U, ξi
hσ(U, V ), Jξi = − h ∇
for U, V ∈ T N . Thus we obtain hAf ξ U, V i
= − hAξ PU, V i − hDV F U, ξi .
Combining this with Proposition 8.7 gives DV H⊥ , ν = 0 which implies
statement (2). Statement (3) follows from statement (1) and (8.16).
Lemma 8.3. Let N be a generic submanifold of a Kähler manifold M̃ . If

the holomorphic distribution H is integrable and AF H⊥ H ⊂ H holds, then
we have
H̃B (X, Z) = R(X, JX; P Z, Z) + 2|σ(X, Z)|2
(8.17)
+ |∇X P Z|2 − |P ∇X Z|2 − |AF Z X|2
for unit vectors X ∈ H and Z ∈ H⊥ , where H̃B (X, Z) is the holomorphic
bisectional curvature of M̃ defined by H̃B (X, Z) = R̃(X, JX; JZ, Z).
Proof. Let N be a generic submanifold of a Kähler manifold M̃ . If the

holomorphic distribution H is integrable and AF H⊥ H ⊂ H holds, then for
vector fields X, Y ∈ H and Z, W ∈ H⊥ we have
hσ([X, Y ], Z), F W i = 0 and JH = H.
Thus we find from Codazzi’s equation and Gauss’ formula that
R̃(X, Y ; Z, F W ) − hDX σ(Y, Z) − DY σ(X, Z), F W i
= hσ(X, ∇Y Z), F W i − hσ(Y, ∇X Z), F W i
˜ X JZ i − h AF W X, J ∇
= h AF W Y, J ∇ ˜ Y JZ i
˜ X P Z i − h AF W X, J ∇
= h AF W Y, J ∇ ˜Y PZ i
˜ X F Z i − h AF W X, J ∇
+ h AF W Y, J ∇ ˜Y FZ i (8.18)
˜ Y P Z i − h JAF W Y, ∇
= h JAF W X, ∇ ˜ XP Z i
˜ X F Z i − h AF W X, J ∇
+ h AF W Y, J ∇ ˜Y FZ i
= h JAF W X, ∇Y P Z i − h JAF W Y, ∇X P Z i
+ hAF W X, JAF Z Y i − hAF W Y, JAF Z Xi
for X, Y ∈ H and Z, W ∈ H⊥ .
Since H is integrable and its leaves are Kähler submanifolds, we also

have
AF W JX = −JAF W X
from (8.12). Hence (8.18) gives
R̃(X, JX; Z, F Z) − hDX σ(JX, Z) − DJX σ(X, Z), F Zi
(8.19)
= hJAF Z X, ∇JX P Zi − hAF Z X, ∇X P Zi + 2|AF Z X|2 .
From hJAF Z X, P Zi ∈ h H, P H⊥ i = {0} we get
hJAF Z X, ∇JX P Zi = − h∇JX (JAF Z X), P Zi
= − hJ∇JX AF Z X, P Zi
˜ AF Z X X, P Z i
= h[JX, AF Z X], JP Zi + h ∇ (8.20)
= h∇˜ JX AF Z X, JP Z i
˜ AF Z X X, P Z i ,
= h∇
where the equality holds since we have [JX, AF Z X] ∈ H and JP Z ⊥ H.
Since [AF Z X, X] ∈ H and AF Z X, P Z ⊥ H, we have
˜ AF Z X X, P Z i = h[AF Z X, X], P Zi + h ∇
h∇ ˜ X AF Z X, P Z i
= h∇˜ X (AF Z X), P Z i (8.21)
= − hAF Z X, ∇X P Zi .
By combining (8.19), (8.20) and (8.21) we obtain
(8.22)
= 2|AF Z X|2 − 2 hAF Z X, ∇X P Zi .
From the formulas of Gauss and Weingarten, (8.4) and (8.12), we have
AF Z X = ∇X P Z − P ∇X Z. (8.23)
By substituting this into (8.22) we find
(8.24)
= 2|AF Z X|2 − 2|∇X P Z|2 + 2 hP ∇X Z, ∇X P Zi .
From (8.23) we also have
|AF Z X|2 = |∇X P Z|2 + |P ∇X Z|2 − 2 h∇X P Z, P ∇X Zi .
After combining this with (8.24) we obtain
(8.25)
= |AF Z X|2 − |∇X P Z|2 + |P ∇X Z|2 .
On the other hand, it follows from σ(H, H⊥ ) ⊥ F H⊥ , Proposition 8.1,

Lemma 8.2 and (8.4) that
hDX σ(JX, Z) − DJX σ(X, Z), F Zi
˜ JX F Z i − h σ(JX, Z), ∇
= h σ(X, Z), ∇ ˜ XFZ i
= hσ(X, Z), Jσ(JX, Z)i − hσ(X, Z), σ(JX, P Z)i (8.26)
− hσ(JX, Z), Jσ(X, Z)i + hσ(JX, Z), σ(X, P Z)i
= −2|σ(X, Z)|2 + 2 hJσ(X, Z), σ(X, P Z)i .
Moreover, by Gauss’ equation, Lemma 8.2 and (8.12), we also have
R̃(X, JX; Z, P Z) = R(X, JX; Z, P Z)
− hσ(X, P Z), σ(JX, Z)i + hσ(X, Z), σ(JX, P Z)i (8.27)
= R(X, JX; Z, P Z) − 2 hJσ(X, Z), σ(X, P Z)i .
Now, by applying JZ = P Z + F Z, (8.25), (8.26) and (8.27) we obtain
H̃B (X, Z) = R̃(X, JX; JZ, Z)
= R(X, JX; P Z, Z) + 2|σ(X, Z)|2 + |∇X P Z|2
− |P ∇Z|2 − |AF Z X|2
for unit vectors X ∈ H and Z ∈ H⊥ , which proves the lemma.
Definition 8.2. A generic submanifold N of a Kähler manifold is said to

be mixed totally geodesic if its second fundamental form satisfies
σ(H, H⊥ ) = {0}.
Clearly, every totally umbilical generic submanifold of a Kähler manifold

is mixed totally geodesic.
Lemma 8.4. Let N be a generic submanifold of an m-dimensional complex

space form M̃ m (4c) with integrable holomorphic distribution H. If either
N is mixed totally geodesic or F H⊥ is parallel in the normal bundle and
(8.12) holds, then we have
|∇X P Z|2 = 2c|F Z|2 + |P ∇X Z|2 + |AF Z X|2 (8.28)
⊥
for unit vectors X ∈ H and Z ∈ H .
Proof. Because the ambient space M̃ m (4c) is a complex space form of

constant holomorphic sectional curvature 4c, we have
R̃(X, JX; Z, F Z) = −2c hX, Xi hF Z, F Zi .
Hence, after combining this (8.25) and using the hypothesis of the lemma,
we obtain (8.28).
Theorem 8.1. Let N be a proper generic submanifold of a complex space

form M̃ m (4c). If F is parallel, then c = 0.
Proof. If ∇F = 0 holds, then it follows from Proposition 8.8 that H

is integrable and its leaves are totally geodesic in N . Also, it follows from
Proposition 8.8 that F H⊥ is parallel in the normal bundle and (8.12) holds.
Hence we obtain (8.28) from Lemma 8.4. Moreover, from statements (1)
and (3) of Proposition 8.8 we also have
∇X P Z = P ∇X Z, AF Z X = 0
for any X ∈ H and Z ∈ H⊥ . Therefore we obtain c|F Z| = 0 from (8.28).
Consequently, we must have c = 0 since N is assumed to be proper. This
proves the theorem.
Theorem 8.2. Let N be a proper generic submanifold of a complex space

form M̃ h+p (4c). If P is parallel, then c = 0.
Proof. Under the hypothesis, we know from Proposition 8.6 that H is

integrable and AF (T N ) H = {0} holds. So, it follows from the assumption on
the codimension that the generic submanifold N is mixed totally geodesic.
Thus we get (8.26) from Lemma 8.4.
Moreover, from ∇P = 0 we obtain ∇X P Z = P ∇X Z. Hence we obtain
c |F Z|2 = 0 from Lemma 8.4. Consequence, we must have c = 0 since N is
proper.
Remark 8.1. Theorem 8.2 is false if the ambient space is a complex space
form with arbitrary codimension.
8.4 Totally umbilical submanifolds
If N is totally umbilical submanifolds in Kähler manifolds, then we have

σ(U, V ) = hU, V i H,
(8.29)
¯
(∇σ)(U, V ) = hU, V i DH,
for U, V ∈ T N .
Proposition 8.9. Let N be a totally umbilical generic submanifold of a

Kähler manifold M̃ . Then we have:
(1) The purely real distribution H⊥ is integrable and its leaves are totally
geodesic in N .
(2) If N is not purely real, then H ∈ F H⊥ and both F H⊥ and ν are parallel
in the normal bundle.
Proof. If follows from the formulas of Gauss and Weingarten that
−JAξ U + JDU ξ = ∇U tξ + σ(U, tξ) − Af ξ U + DU f ξ
for vector fields U ∈ T N and ξ ∈ T ⊥ N . Thus we have
∇U tξ = hf ξ, Hi U − hξ, Hi P U + tDU ξ, (8.30)

DU f ξ = −F Aξ U + f DU ξ − σ(U, tξ). (8.31)
From (8.6) and (8.8), we have
P H⊥ = H⊥ , T ⊥ N = F H⊥ ⊕ ν, t(T ⊥ N ) = t(F H⊥ ) = H⊥ .
Thus (8.30) implies that ∇Z W ∈ H⊥ for vector fields Z, W ∈ H, which

shows that H⊥ is integrable and its leaves are totally geodesic in N . This
gives statement (1).
Now, let us assume that N is not a purely real submanifold. So we have
H 6= {0}. Let η be a normal vector field in ν and X is a tangent vector
field in H. Then, by total imbecility of N , we have
˜ X X, Jη i = hσ(X, JX), Jηi = 0.
hX, Xi hH, ηi = h J ∇
Thus we obtain H ⊥ ν, i.e., H ∈ F H⊥ . For any vector field ξ ∈ ν, we have

tξ = 0. Since H ∈ F H⊥ and N is totally umbilical, we get Aξ = 0, Hence
(8.31) gives DU Jξ = f DU ξ.
On the other hand, we obtain from H ∈ F H⊥ that
˜Uξ = ∇
JDU ξ = J ∇ ˜ U Jξ = DU Jξ. (8.32)
Now, by combining DU Jξ = f DU ξ with (8.32) we obtain tDU ξ = 0. This

shows that ν is parallel in the normal bundle. Since F H⊥ is the orthogonal
complementary subbundle of ν in T ⊥ N , we conclude that F H⊥ is also
parallel in the normal bundle.
A Riemannian manifold M is said to be positively curved (resp., nega-

tively curved ) if it has positive (resp., negative) sectional curvatures.
Theorem 8.3. [Chen and Verheyen (1983)] Every totally umbilical sub-
manifold N of a positively (or negatively) curved Kähler manifold M̃ is
either a totally geodesic complex submanifold or a purely real submanifold.
Proof. Let N be a totally umbilical submanifold of a positively (or neg-

atively) curved Kähler manifold M̃ . Suppose that N is a totally umbilical
complex submanifold of M̃ . Then N is minimal in M̃ since M̃ is Kählerian.
Hence N is a totally geodesic complex submanifold.
Now, assume N is neither a complex or a purely real submanifold of
M̃ . Then there is a nonempty open submanifold N ′ of N such that N ′ is a
totally umbilical proper generic submanifold of M̃ . Now let us restrict our
study on N ′ . Then, by applying (2.21) and Codazzi’s equation, we have
R̃(JX, Z; X, F Z) = hZ, Xi hDJX H, F Zi − hJX, Xi hDZ H, F Zi = 0
for any vector field X ∈ H and Z ∈ H⊥ . Thus we have
R̃(F Z, X; X, F Z) + R̃(P Z, X; X, F Z) = 0.
Hence we derive from Codazzi’s equation that
|X|2 |F Z|2 K̃(X, F Z) = −R̃(P Z, X; X, F Z)
(8.33)
= −|X|2 hDP Z H, F Zi .
On the other hand, we also have
|X|2 |Z|2 K̃(X, JZ) = −R̃(P Z, X; JX, Z)
= R̃(P Z, X; X, P Z) + hDP Z H, F Zi .
Combining this with (8.33) we obtain
|F Z|2 {K̃(X, F Z) + K̃(X, JZ)}
(8.34)
= |P Z|2 {K̃(X, P Z) − K̃(X, JZ)}.
So, if M̃ is positively curved, the (8.34) gives K̃(X, P Z) > K̃(JX, Z) for
any X ∈ H and Z ∈ H⊥ . Thus, after replacing Z by P Z we obtain
K̃(X, P 2 Z) > K̃(JX, P Z) > K̃(X, Z). (8.35)
Hence we have
hZ, P W i = hZ, JW i = − hP Z, W i
for any Z, W ∈ H . Thus P |H⊥ is a skew-symmetric endomorphism of H⊥ .
⊥
Hence (P |H⊥ )2 is a symmetric endomorphism of H⊥ .

If P 2 = 0 on H⊥ , then

hP Z, P Zi = − P 2 Z, Z = 0
for Z ∈ H⊥ . Hence, by (8.34) we get F Z = 0, which is impossible due to
the definition of H⊥ . If P 2 6= 0 on H⊥ , then there is a unit vector Z ∈ H⊥
such that P 2 Z = λZ with λ 6= 0. This contradicts (8.35). This completes
the proof for the case that M̃ is positively curved.
Similar argument applies to negatively curved M̃ .
8.5 Generic products and Segre embedding
Definition 8.3. A real submanifold N of a Kähler manifold M̃ is called

a generic product if locally it is the Riemannian product N T × N ⊥ of a
complex submanifold N T and a purely real submanifold N ⊥ of M̃ .
Lemma 8.5. If N is a generic product of a Kähler manifold, then we have
AF Z X = 0, (8.36)
(∇X P )Z = 0 (8.37)
for X ∈ H and Z ∈ H⊥ .
Proof. If N is a generic product of a Kähler manifold, then both the
holomorphic distribution H and the purely distribution H⊥ are integrable
and their leaves are totally geodesic in N . Hence we obtain (8.36) follows
from Propositions 8.3 and 8.4. Moreover, it is direct to verify that (8.37)
follows from (8.12) and (8.36).
Now, we provide some examples of generic products in complex projec-
tive spaces.
Example 8.1. Consider the Segre embedding
Sh,p : CP n (4) × CP p (4) → CP h+p+hp (4)
defined by (6.35). Let N ⊥ be a p-dimensional purely real submanifold of
CP p (4). Then CP σ (4) × N ⊥ is a generic product in CP h+p+hp (4) via the
Segre embedding Sh,p in which CP σ (4) is embedded in CP h+p+hp (4) as a
totally geodesic complex submanifold.
Definition 8.4. A generic product N = N T × N ⊥ in a CP m (4) is called
a standard generic product if the following two conditions are satisfied:
(a) N lies a totally geodesic complex submanifold CP h+p+hp (4) of CP m (4);
(b) N T is embedded in CP m (4) as a totally geodesic complex submanifold,
where h denotes the complex dimension of N T and p denotes the real
dimension of N ⊥ .
For generic submanifolds in a Kähler manifold, we have the following
general result from [Chen (1981e)].
Theorem 8.4. Let N be a generic product in a Kähler manifold M̃ . Then
for any unit vector fields X ∈ H and Z ∈ H⊥ we have
H̃B (X, Z) = 2|σ(X, Z)|2 . (8.38)
Proof. If N is a generic product in a Kähler manifold M̃ , then

R(X, JX; P Z, Z) = 0 (8.39)
⊥
for unit vector fields X ∈ H and Z ∈ H . Also, it follows from (8.37) that
∇X P Z = P ∇X Z. (8.40)
Thus we obtain (8.38) from (8.36), (8.39), (8.40) and Lemma 8.3.
Corollary 8.1. Let M̃ be a Kähler manifold with negative holomorphic

bisectional curvature. Then every generic product in M̃ is either a complex
submanifold or a purely real submanifold.
Theorem 8.5. [Chen (1981e)] A generic product in the complex Euclidean

m-space Cm is locally the direct product of a complex submanifold N T in a
linear complex subspace Cq and a purely real submanifold N ⊥ of a Cm−q ,
i.e., locally, N = N T × N ⊥ ⊂ Cq × Cm−q = Cm .
Proof. Suppose that N = N T × N ⊥ is a generic product in Cm . Then

it follows from Theorem 8.4 that N is mixed totally geodesic. Hence by
Moore’s lemma [Moore (1971)] the generic product N is locally the direct
product of a complex submanifold and a purely real submanifold.
8.6 Generic products in complex projective spaces
For a generic submanifold N in a Kähler manifold, we put h = dimC Hx ,

p = dimR Hx⊥ , x ∈ N, as before.
Theorem 8.6. [Chen (1981e)] Let N be a generic product in the complex

projective space CP m (4). Then the squared norm of the second fundamental
form σ satisfies
||σ||2 ≥ 4hp. (8.41)
The equality sign of (8.41) holds identically if and only if N is locally
the Riemannian product of a totally geodesic complex submanifold CP h (4)
and a totally geodesic totally real submanifold RP p (1) of CP m (4).
Proof. If N is a generic product in CP m (4), then we have Z ⊥ X, JX

for any unit vectors X ∈ H and Z ∈ H⊥ . Hence Span{X, Z} is a totally
real section. So, we have H̃B (X, Z) = 2. Hence, by Theorem 8.4 we find
|σ(X, Z)| = 1. (8.42)
Thus we have
2h
X p
X
||σ||2 = 4hp + |σ(Xi , Xj )|2 + |σ(Zα , Zβ )|2 , (8.43)
i,j=1 α,β=1
where {X1 , . . . , X2h } (respectively, {Z1 , . . . , Zp }) is an orthonormal frame

of H (respectively, of H⊥ ). Clearly, (8.43) implies statement (1).
Next, let us assume that ||σ||2 = 4hp holds identically. Then it follows
from (8.43) that
σ(H, H) = σ(H⊥ , H ⊥ ) = {0}.
Therefore the generic product N is locally the direct product of a totally
geodesic complex submanifold and a totally geodesic totally real submani-
fold of the complex projective space CP m (4).
Theorem 8.7. [Chen (1981e)] Let N = N T × N ⊥ be a generic product in

the complex projective m-space CP m (4). Then we have:
(1) m ≥ h + p + hp;
(2) Every generic product N in CP m (4c) with m = h+ p+ hp is a standard
generic product.
Proof. Let N be a generic product in the complex projective m-space

CP m (4). Then (8.35) holds. So, after applying linearity we find
hσ(Xi , Z), σ(Xj , Z)i = 0, i 6= j, (8.44)
where X1 , . . . , X2h and Z1 , . . . , Zp are orthonormal basis for Hx and
Hx⊥ , x ∈ N , respectively.
If p = 1, statement (1) follows immediately from (8.12) and (8.44).
If p ≥ 2, then we find from (8.44) and linearity that
hσ(Xi , Zα ), σ(Xj , Zβ )i + hσ(Xi , Zβ ), σ(Xj , Zα )i = 0 (8.45)
for i 6= j, α 6= β. On the other hand, since N is the Riemannian product
of N T and N ⊥ , we have R(Xi , Xj ; Zα , Zβ ) = 0. Thus it follows from the
equation of Gauss that
hσ(Xi , Zα ), σ(Xj , Zβ )i = hσ(Xi , Zβ ), σ(Xj , Zα )i = 0. (8.46)
Now, by combining (8.35), (8.45) and (8.46), we know that
{σ(Xi , Zα ) : i = 1, . . . , 2h; α = 1, . . . , β} (8.47)
are orthonormal normal vectors in ν which is perpendicular to H⊥ . Thus
we obtain m ≥ h + p + hp which gives statement (1).
Next, let us assume that m = h + p + hp. Then for any unit vector
X, Y ∈ H and Z ∈ H⊥ the equation of Gauss and the curvature tensor of
CP m (4c) imply that
hσ(X, Y ), σ(JX, Z)i = hσ(X, Z), σ(JX, Y )i . (8.48)
On the other hand, since σ(H, H⊥ ) ⊥ F H⊥ , Lemmas 8.2 and 8.5 give
hσ(X, Z), σ(JX, Y )i = hσ(X, Z), Jσ(X, Y )i
= − hJσ(X, Z), σ(X, Y )i
(8.49)
= − hJσ(X, Z), Jf σ(X, Y )i
= − hσ(JX, Z), σ(X, Y )i .
By combining (8.48) and (8.49) we find
hσ(X, Y ), σ(JX, Z)i = hσ(X, Z), σ(JX, Y )i = 0. (8.50)
Hence, after applying linearity we derive from (8.50) that
hσ(X, Z), σ(Y, W )i = hσ(Y, Z), σ(X, W )i = 0 (8.51)
for X, Y, W ∈ H and Z ∈ H⊥ . Since m = h + p + hp and (8.48) spans ν, we
obtain from (8.51) that σ(H, H) ⊂ F H⊥ . After combining this with (8.36)
we find σ(H, H) = {0}. Thus N T is immersed as a totally geodesic complex
submanifold in CP m (4). Consequently, M T is an open submanifold of a
complex projective h-space CP h (4).
8.7 An application to complex geometry
By applying Theorem 8.7 we may obtain the following result from [Chen
(1981b,e)].
Theorem 8.8. Let M = M1n1 × M2n2 be a Riemannian product of two

Kähler manifolds with dimC M2h = n1 and dimC M2p = n2 . Then we have:
(1) M admits no Kähler immersion into CP m (4) with m < n1 +n1 +n1 n2 ;
(2) If M admits a Kähler immersion ψ into CP n1 +n1 +n1 n2 (4), then
(2.1) M1n1 and M2n2 are open submanifolds of CP n1 (4) and CP n2 (4),
respectively;
(2.2) the Kähler immersion ψ is given by the Segre embedding Sn1 n2 .
Proof. Let M = M1n1 × M2n2 be a Riemannian product of two Kähler

manifolds. Assume that M admits a Kähler immersion ψ into CP m (4). Let
N2⊥ be a n2 -dimensional purely real submanifold of CP n2 , then M1n1 × N2⊥

is a generic product in CP m (4) via ψ. Thus by applying Theorem 8.7(1)
we obtain m ≥ n1 + n1 + n1 n2 . This gives statement (1).
To prove statement (2), let us assume that M does admit a Kähler
immersion ψ into CP n1 +n1 +n1 n2 (4). Then M1n1 × N2⊥ is a generic product
in CP n1 +n1 +n1 n2 (4). Thus it follows from Theorem 8.7(2) that M1n1 is a
totally geodesic complex submanifold of CP n1 +n1 +n1 n2 (4). Hence M1n1 is
an open submanifold of CP n1 (4).
Similarly, we also know that M2n2 is an open submanifold of CP n2 (4).
This gives statement (2.1). Now statement (2.2) follows from Calabi’s rigid-
ity theorem of Kähler immersions obtained in [Calabi (1953)].
Theorem 8.8 was extended to the following.
Theorem 8.9. [Chen and Kuan (1985)] Let M = M1n1 × · · · × Msns be the
product of s Kähler manifolds with dimC Mini = ni (i = 1, . . . , s). Then the
following statements hold:
(1) M does not admit a Kähler immersion into CP m (4) for any m < N
Qs
with N = i=1 (ni + 1) − 1;
(2) If M admits a Kähler immersion ψ into CP N (4), then
(2.1) Mini is an open submanifold of CP ni (4) for i = 1, . . . , s;
(2.2) the Kähler immersion ψ is given by
Sn1 ···ns (z01 , . . . , zn1 1 , . . . , z0s , . . . , zns s ) = (zi11 · · · zisj )1≤i1 ≤n1 ,...,1≤is ≤ns .
The Segre embedding Sn1 ···ns can also be characterized in terms of
¯ ℓ σ|| for any ℓ ∈ {0, 1, . . . , 2 − s} as follows.
||∇
Theorem 8.10. [Chen and Kuan (1985)] Let M = M1n1 × · · · × Msns be the
product of Kähler manifolds and ψ : M → CP m (4) be a Kähler immersion.
Then the ℓ-covariant derivative ∇ ¯ ℓ σ of the second fundamental form σ of
ψ satisfies
X
¯ ℓ σ||2 ≥ 2ℓ+2 (ℓ + 2)!
||∇ ni1 · · · niℓ (8.52)
i1 <···<iℓ
for ℓ = 1, . . . , s − 2.
The equality sign of (8.52) holds identically for some ℓ ∈ [1, . . . , s − 2]
if and only if
(a) M1n1 × · · · × M ns is an open portion of CP n1 (4) × · · · × CP ns (4);
(b) up to rigid motions of CP m (4), the immersion ψ is given by the Segre
embedding Sn1 ···ns .
Chapter 9
CR-submanifolds of Kähler Manifolds
A generic submanifold of an almost Hermitian manifold (M̃ , g̃, J) is called

a CR-submanifold if its purely real distribution H⊥ is a totally real distri-
bution, i.e., JHx⊥ ⊂ Tx⊥ N for each x ∈ N .
The notion of CR-submanifolds was introduced in [Bejancu (1978)].
Complex submanifolds and totally real submanifolds are special cases of
CR-submanifolds. A CR-submanifold is called proper if it is neither a
complex nor a totally real submanifold.
The main purpose of this chapter is to present basic results on proper
CR-submanifolds.
9.1 CR-submanifolds as CR-manifolds
Let N be a submanifold of an almost Hermitian manifold (M̃ , g̃, J). For

each point x ∈ N , let Hx denote the maximal holomorphic subspace of
Tx N , i.e., Hx = Tx N ∩ J(Tx N ). When dim Hx is the same for all x ∈ N ,
then {Hx : x ∈ N } defines the holomorphic distribution H.
Definition 9.1. A submanifold N of an almost Hermitian manifold M̃ is

called a CR-submanifold if there exist a holomorphic distribution H and a
totally real distribution H⊥ on N , i.e., J(Hx⊥ ) ⊂ Tx⊥ M , such that
T N = H ⊕ H⊥ .
A CR-submanifold is called proper if T N 6= H or H⊥ . A CR-submanifold
N is called anti-holomorphic if JH⊥ = T ⊥ N .
Definition 9.2. For any n-manifold N , let T C N denote the complexified

tangent bundle of N , i.e.,
TxC N = Tx N ⊗R C ≡ Tx N ⊕ i(Tx N ), x ∈ N.
195
Let D be a complex subbundle of complex dimension h. A CR-manifold

of real dimension n and CR-dimension h is a pair (N, D) such that D is
involutive and Dx ∩ D̄x = 0 [Greenfield (1968)].
If (N, D) is a CR-manifold, then there exists a unique subbundle K of

T N such that KC = D ⊕ D̄, and a unique bundle map J : K → K such
that J 2 = −I and D = {X − iJ X : X ∈ K}.
If N is a CR-submanifold of a Hermitian manifold, let P denote the
natural projections of T N to H and let Q denote the natural projections of
T N to H⊥ . Since the holomorphic distribution H is J-invariant, we may
put J = J ◦ P and define a complex subbundle D on T C N by
D = {X − iJ X : X ∈ H}. (9.1)
Lemma 9.1. Let N be a CR-submanifold of a Hermitian manifold

(M̃ , g̃, J). Then we have
(a) −J (X − iJ X) ∈ Dx for every X ∈ Tx N ;
(b) Q([JX, Y ] + [X, JY ]) = 0 for X, Y ∈ H.
Proof. Statement (a) follows easily from definitions. Since M is Hermi-

tian, the Nijenhuis tensor [J, J] of J vanishes. Hence we have
0 = [J, J](JX, Y ) = J([X, Y ] − [JX, JY ]) − [JX, Y ] − [X, JY ],
but [X, Y ] and [JX, JY ] are tangent to N , and hence J[X, Y ] − J[JX, JY ]
has no component in H⊥ . Hence [JX, Y ] + [X, JY ] has no H⊥ -component.
This proves statement (b).
The following fundamental result from [Blair and Chen (1979)] shows
that CR-submanifolds appear as embedded CR manifolds.
Theorem 9.1. Every CR-submanifold of a Hermitian manifold is a CR-

manifold.
Proof. Let X, Y ∈ H. Then, by using [J, J] = 0 and Lemma 9.1, we find

[X − iJ X, Y − iJ Y ] = [X, Y ] − [JX, JY ] − i[JX, Y ] − i[X, JY ]
= −J[JX, Y ] − J[X, JY ] − iP [JX, Y ] − iP [X, JY ]
= −J [JX, Y ] − JQ[JX, Y ] − J [X, JY ]
− JQ[X, JY ] + iJ 2 [JX, Y ] − iQ[JX, Y ]
+ iJ 2 [X, JY ] − iQ[X, JY ]
= −J ([JX, Y ] − iJ [JX, Y ]) − J ([X, JY ] − iJ [X, JY ])
− JQ([JX, Y ] + [X, JY ]) − iQ([JX, Y ] + [X, JY ]).
CR-submanifolds of Kähler Manifolds 197
According to Lemma 9.1(a) the first two terms belong to D. Also according
to Lemma 9.1(b) the last two terms vanish. Hence the complex subbundle
D defined by (9.1) is an involutive distribution.
9.2 Integrability and minimality
Let N be a CR-submanifold of a Kähler manifold M̃ . There there exist

a holomorphic distribution H and a totally real distribution H⊥ such that
T N = H ⊕ H⊥ . As before we put h = dimC Hx and p = dimR Hx⊥ .
Lemma 9.2. Let N be a CR-submanifold of a Kähler manifold. Then, for

any vector fields X ∈ H, Z, W ∈ H⊥ , ξ ∈ ν and U ∈ T N , we have
h∇U Z, Xi = hJAJZ U, Xi , (9.2)
AJZ W = AJW Z, (9.3)
AJξ X = −Aξ JX. (9.4)
˜ = 0. Hence from the formulas
Proof. Since M̃ is Kählerian, we have ∇J
of Gauss and Weingarten we get
J∇U Z + Jσ(U, Z) = −AJZ U + DU JZ, (9.5)
˜ Y JX, ξ i = hJσ(X, Y ), ξi .
hσ(JX, Y ), ξi = h ∇ (9.6)
Now, by taking the inner product of (9.5) with X ∈ H, we obtain (9.2).
Also, by taking the inner product of (9.5) with W ∈ H⊥ , we have (9.3).
Finally, (9.4) is an immediate consequence of (9.6).
Theorem 9.2. The totally real distribution of a CR-submanifold N of a

Kähler manifold is an integrable distribution.
Proof. If N is a CR-submanifold of a Kähler manifold, then we have

P H⊥ = {0}. So, condition (8.11) in Proposition 8.2 holds automatically
due to (9.3). Thus Proposition 8.2 implies that the totally real distribution
H⊥ of the CR-submanifold N is always integrable.
Clearly, Theorem 9.2 implies the following.
Corollary 9.1. Every CR-submanifold of a Kähler manifold M is foliated

by totally real submanifolds of M .
Let us suppose that (M, g, J) is a Hermitian manifold and let Ω be the

fundamental 2-form of M , i.e., Ω(U, V ) = g(U, JV ). It is well-known that
M is Kählerian if and only if dΩ = 0. Here however let us consider a slightly

larger class, namely those for which dΩ = Ω ∧ ω for some 1-form ω, called
the Lee form. When ω is closed, these manifold are call locally conformal
symplectic manifolds (cf. [Vaisman (1976)]).
Theorem 9.2 was extended to the following in [Blair and Chen (1979)].
Theorem 9.3. The totally real distribution of a CR-submanifold N of a

locally conformal symplectic manifold is integrable.
Proof. Let N be a CR-submanifold of a locally conformal symplectic

manifold. Then, for vector fields X ∈ H and Z, W ∈ H⊥ , we have
Ω(X, Z) = Ω(Z, W ) = 0. Thus Ω ∧ ω(X, Z, W ) = 0 and hence
0 = 3dΩ(X, Z, W )
= XΩ(Z, W ) − ZΩ(X, W ) + W Ω(X, Z)
− Ω([X, Z], W ) − Ω([W, X], Z) − Ω([Z, W ], X)
= −g([Z, W ], JX),
but X and hence JX is arbitrary in H and [Z, W ] is tangent to N . Therefore
we have [Z, W ] ∈ H⊥ .
Remark 9.1. It was shown in [Blair and Chen (1979)] that Theorem 9.2
is false for general CR-submanifolds in Hermitian manifolds. In fact, Blair
and Chen constructed examples of CR-submanifolds of a Hermitian mani-
fold on which their totally real distributions are not integrable.
An immediate consequence of Proposition 7.1 is the following [Bejancu

(1978); Chen (1981b)].
Proposition 9.1. The holomorphic distribution H of a CR-submanifold

N of a Kähler manifold is integrable if and only if
hσ(X, JY ), JZi = hσ(JX, Y ), JZi (9.7)
⊥
holds for X, Y ∈ H and Z ∈ H .
Let V be a distribution of a Riemannian manifold N and let V ⊥ be the

orthogonal complementary distribution of V. Put
σ̊(X, Y ) = (∇′X Y )⊥ (9.8)
for vector fields X, Y in V, where (∇′X Y )⊥
is the V -component of ∇′X Y .
⊥
Then σ̊ is a well-defined V ⊥ -valued (0, 2)-tensor field. Moreover, it follows

from Frobenius’ theorem that V is integrable if and only if σ̊ is symmetric.
Let e1 , . . . , ek be an orthonormal basis of V. If we put

k
1 1X
H̊ = Tr σ̊ = ǫj σ̊(ej , ej ), (9.9)
k k j=1
then, up to sign, H̊ is a well-defined vector field, which is called the mean

curvature vector of V. If H̊ = 0 holds identically, V is called a minimal
distribution. In particular, if σ̊ = 0 holds identically, V is called a totally
geodesic distribution. Every totally geodesic distribution on N is an inte-
grable distribution whose leaves are totally geodesic submanifolds on N .
Another fundamental properties of CR-submanifolds is the following
result from [Chen (1981c)].
Theorem 9.4. If N is a CR-submanifold of a Kähler manifold, then the

holomorphic distribution H of N is a minimal distribution.
Proof. Let N be a CR-submanifold of a Kähler manifold. Then we find

from Lemma 9.2 that h[X, Y ], JZi = 0 for any vector fields X, Y in H
and Z in H⊥ . Thus we obtain hZ, ∇X Xi = hAJZ X, JXi . Hence we find
hZ, ∇JX JXi = − hAJZ X, JXi . After combining these two equations we
obtain h∇X X + ∇JX JX, Zi = 0, which implies the minimality of H.
Definition 9.3. A CR-submanifold N of an almost Hermitian manifold

M̃ is called a CR-product if both distributions H and H⊥ are integrable
and N is locally a Riemannian product N T × N ⊥ , where N T is a leaf of H
and N ⊥ is a leaf of H⊥ . A CR-product with H = 6 {0} and H⊥ 6= {0} is a
called a proper CR-product.
The following simple characterization of CR-products was proved in

[Chen (1981b)].
Theorem 9.5. A CR-submanifolds of a Kähler manifold is a CR-product

if and only if P is parallel, i.e. ∇P = 0.
Proof. If N is a CR-submanifold of a Kähler manifold with ∇P = 0,

then Proposition 8.6 implies that H is integrable and
AJZ X = 0, X ∈ H, Z ∈ H⊥ . (9.10)
After applying (9.2) of Lemma 9.2 and (9.10), we find h∇Y Z, Xi = 0 for
X, Y ∈ H. Thus leaves of H are totally geodesic in N .
Also, it follows (9.2) and (9.10) that
h∇W Z, Xi = − hAJZ W, JXi = − hW, AJZ JXi = 0.
Thus the leaves of H⊥ is integrable are totally geodesic in N . Consequently,

N is a CR-product in M̃ .
Conversely, if N is a CR-product, we get ∇U Y ∈ H for Y ∈ H and
U ∈ T N . Thus Jσ(U, U ) = σ(U, JY ). From this we get (∇U P )Y = 0.
Similarly, from ∇U Z ∈ H⊥ , Z ∈ H⊥ , we may obtain (∇U P )Z = 0. Hence
we have ∇P = 0.
From the proof of Theorem 9.5 we have the following.
Lemma 9.3. A CR-submanifold N in a Kähler manifold is a CR-product

if and only if AJH⊥ H = {0}.
Remark 9.2. Corollary 8.1, Theorem 8.5 and Theorem 8.7 imply the fol-
lowing.
(a) There do not exist proper CR-products in any complex hyperbolic
space.
(b) Every CR-product N T × N ⊥ in Cm is locally the direct product of a
complex submanifold and a totally real submanifold of some complex
Euclidean spaces.
(c) For every CR-product N T × N ⊥ in CP m , we have m ≥ h + p + hp,
where h = dimC N T and p = dim N ⊥ .
9.3 Cohomology of CR-submanifolds
For a CR-submanifold N of a Kähler manifold M̃ , we choose an orthonor-

mal local frame {e1 , . . . , eh , Je1 , . . . , Jeh } of the holomorphic distribution
H. Let {ω 1 , . . . , ω h , ω h+1 , . . . , ω 2h } the 1-forms on N satisfying
ω i (Z) = 0 and ω i (ej ) = δji , i, j = 1, . . . , 2h, (9.11)
⊥
where Z ∈ H and eh+α = Jeα , α = 1, . . . , h. Consider θ defined by
θ = ω 1 ∧ · · · ∧ ω 2h . (9.12)
This 2h-form is a well-defined global form on N since H is orientable.
Theorem 9.6. For each closed CR-submanifold of a Kähler manifold M̃ ,

there is a canonical de Rham cohomology class given by
c(N ) ≡ [θ] ∈ H 2h (N ; R). (9.13)
This cohomology class is non-trivial if the holomorphic distribution H is
integrable and the totally real distribution H⊥ is minimal.
Proof. From (9.10) we obtain

2h
X
dθ = (−1)i ω 1 ∧ · · · ∧ dω i ∧ · · · ∧ ω 2h . (9.14)
i=1
It follows from (9.11) and (9.14) that dθ = 0 if and only if

dθ(Z1 , Z2 , X1 , . . . , X2h−1 ) = 0, (9.15)
dθ(Z1 , X1 , . . . , X2h ) = 0 (9.16)
for Z1 , Z2 ∈ H⊥ and X1 , . . . , X2h ∈ H.
It follows from direct computation that (9.15) holds when and only when
H⊥ is integrable and (9.16) holds if and only if H is a minimal distribu-
tion. But for a CR-submanifold of a Kähler manifold both conditions hold
automatically according to Theorems 9.2 and 9.4. Therefore the 2h-form
θ is a closed form. Consequently, it defines the de Rham cohomology class
c(N ) given by (9.13).
Let {e2h+1 , . . . , e2h+p } be an orthonormal local frame of H⊥ and let
{ω 2h+1 , . . . , ω 2h+p } be the 1-forms on N satisfying
ω α (X) = 0 and ω α (eβ ) = 0 (9.17)
for any X ∈ H, where α, β = 2h+ 1, . . . , 2p. So by a similar argument for θ,
we may conclude that if H is integrable and H⊥ is a minimal distribution,
then the p-form θ⊥ = ω 2h+1 ∧ · · · ∧ ω 2h+p is a closed form. Hence the
2h-form is co-closed, i.e., δθ = 0. Because N is a closed submanifold, θ is
a harmonic 2h-form. Since θ is non-trivial, the cohomology class c(N ) is
non-trivial in H 2h (N ; R).
Remark 9.3. The cohomology class given in (9.13) is also known in the
literature as the Chen class (see, e.g. [Dragomir and Ornea (1998)]).
An important consequence of Theorem 9.6 is the following.
Theorem 9.7. [Chen (1981d)] Let N be closed CR-submanifold of a Kähler

manifold M̃ . If
H 2k (N ; R) = 0 (9.18)
for some natural number k ≤ dimC H, then either the holomorphic distribu-
tion H is not integrable or the totally real distribution H⊥ is not minimal.
Proof. Let us choose a local field of orthonormal frame

e1 , . . . , eh , eh+1 , . . . , eh+p , eh+p+1 , . . . , em , Je1 , . . . , Jem
in M̃ in such a way that, restricted to N , e1 , . . . , eh , Je1 , . . . , Jeh are in H

∗ ∗
and eh+1 , . . . , eh+p are in H⊥ . We denote by {ω 1 , . . . , ω m , ω 1 , . . . , ω m }
the dual frame of {e1 , . . . , em , Je1 , . . . , Jem }. Let us put
∗ ∗
φA = ω A + iω A , φ̄A = ω A − iω A , A = 1, . . . , m.
Restricting these forms to N , we have
φα = φ̄α = ω α , φr = φ̄r = 0
for α = h + 1, . . . , h + p and r = h + p + 1, . . . , m. The fundamental form Ω̃
P A
of M̃ is given by Ω̃ = 2i φ ∧ φ̄A . Hence the induced fundamental form
on N is given by
i X i
Ω = ι∗ Ω̃ = hφ ∧ φ̄i .
2 i=1
From this we conclude that the canonical class c(N ) and the fundamental
class [Ω] of N are related by
[Ω]h = (−1)h (h!)c(N ). (9.19)
Now, if H is integrable and H⊥ is minimal in N , then Theorem 9.6 and
(9.19) imply H 2k (N ; R) 6= 0 for k = 1, . . . , h.
Remark 9.4. Let RP p (1) be a totally geodesic, totally real submanifold

of CP p (4). Then N = CP h (4) × RP p (1) ⊂ CP h (4) × CP p (4) is a CR-
submanifold of M̃ = CP h (4) × CP p (4). Clearly, the holomorphic distribu-
tion H of N is integrable and the totally real distribution H⊥ is minimal
in N . Therefore the assumption on the cohomology group in Theorem 9.7
is necessary.
9.4 Totally geodesic and totally umbilical CR-submanifolds
The next two results were proved in [Blair and Chen (1979)].
Theorem 9.8. Every totally geodesic CR-submanifold of a Kähler mani-

fold M̃ is a CR-product N T × T ⊥ in which N T is immersed as a totally
geodesic complex submanifold and N ⊥ as a totally geodesic totally real sub-
manifold of M̃ .
Proof. Let N be a totally geodesic CR-submanifold of a Kähler manifold.

Then it follows from Theorem 9.2 and Proposition 9.1 that H and H⊥ are
both integrable. Suppose that N T is a leaf of H and N ⊥ is a leaf of H⊥ .
It follows from (9.2) that h∇Y X, Zi = 0 for X, Y ∈ H and Z ∈ H⊥ .

Hence N T is totally geodesic in N .
Similarly, we find from (9.2) that hX, ∇W Zi = 0 holds for X ∈ H and
Z, W ∈ H⊥ . Thus N ⊥ is also totally geodesic in N . Consequently, N
is a CR-product. The remaining part follows from the fact that N is a
CR-product and N is totally geodesic in M̃ .
Theorem 9.9. Let N be a totally geodesic CR-submanifold of a Kähler

manifold M̃ . Then K̃(X, Z) = 0 for any unit vectors X ∈ H and Z ∈ H⊥ .
Thus the CR-sectional curvatures of M̃ vanishes.
Proof. Since N is a totally umbilical CR-submanifold of a Kähler man-

¯
ifold M̃ , we have (∇h)(U, V ) = hU, V i DH for U, V tangent to N . Thus
Codazzi’s equation yields
R̃(X, Z; JX, JZ) = hZ, JXi hDX H, JZi − hX, JXi hDZ H, JZi = 0
for vectors X ∈ H and Z ∈ H⊥ . Since M is Kählerian, we also have
R̃(X, Z; JZ, JX) = R̃(X, Z, Z, X) = K̃(X, Z).
Consequently, all CR-sectional curvatures of M̃ vanishes.
The following result is an immediate consequence of Theorem 7.3.
Corollary 9.2. There are no totally umbilical proper CR-submanifolds in

any positively (or negatively) curved Kähler manifold.
Now, we present the following classification result from [Bejancu (1980);

Chen (1981f)].
Theorem 9.10. Let N be a totally umbilical CR-submanifold of a Kähler

manifold M̃ . Then we have:
(a) N is totally geodesic in M̃ or;
(b) N is totally real in M̃ ; or
(c) the totally real distribution H⊥ is of rank one.
Proof. Let N be a totally umbilical CR-submanifold of a Kähler manifold

M . Then we have
hσ(U, U ), JZi = hU, U i hH, JZi (9.20)
⊥
for U ∈ T N and Z ∈ H . On the other hand, from the formulas of Gauss
and Weingarten, we have
J∇U Z + Jσ(U, Z) = −AJZ U + DU JZ. (9.21)
If H⊥ is of rank ≥ 2, then by (9.20) and (9.21) we obtain

hH, JZi = hAJZ U, U i = hσ(U, Z), JU i = 0 (9.22)
for vectors Z, U ∈ H⊥ with Z ⊥ U . So, H ⊥ JH⊥ . Thus either H⊥ is of
rank one or the mean curvature vector is perpendicular to JH⊥ .
Now, let us assume that H ⊥ JH ⊥ and H is non-trivial. Then for
X, Y ∈ H and Z ∈ H⊥ , we have
˜ X JY i = h ∇
hZ, ∇X JY i = h Z, ∇ ˜ X JZ, Y i = − hAJZ X, Y i = 0. (9.23)
The last equation holds because H is perpendicular to JZ. From (9.23) we
find ∇X Y ∈ H for X, Y ∈ H. Hence H is integrable and leaves of H are
totally geodesic in N . Because H is holomorphic, leaves of H are complex
submanifolds of M̃ .
On the other hand, since leaves of H are totally umbilical in M̃ and
M̃ is Kählerian, the mean curvature vector of the leaves of H are totally
geodesic in M̃ . Therefore N is totally geodesic in M̃ . Consequently, one of
the three cases (a), (b) and (c) must occur.
For CR-submanifolds with rank H⊥ = 1, we have the following results
from [Chen (1981f)].
Lemma 9.4. Let N be a totally umbilical CR-submanifold of a Kähler

manifold M̃ with rank H⊥ = 1. Then we have:
(a) DJη H ⊥ JH⊥ ,
(b) If dim N ≥ 5, we have DX H = 0, X ∈ H,
where η is a unit vector field in H⊥ .
Proof. Under the hypothesis, we find from Codazzi’s equation that

R̃(Jη, X; JX, Jη) = R̃(Jη, X; JX, η) = 0 (9.24)
for X ∈ H. Thus by linearity we find
R̃(Jη, X; Y, Jη) = 0 (9.25)
for X, Y ∈ H. In particular, we have
0 = R̃(Jη, X; JX, Jη) = R̃(Jη, X; X, η) = hX, Xi hDHη H, ηi .
This proves (a).
If dim N ≥ 5, then for any given X ∈ H, there is a unit vector Y ∈ H
such that hX, Y i = hX, JY i = 0. Hence by the equation of Codazzi, we get
0 = R̃(JX, Y ; JY, ξ) = 0 (9.26)
for any ξ normal to N .

˜ = 0 and Codazzi’s equation that
On the other hand, we find from ∇J
0 = R̃(JX, Y ; JY, ξ) = R̃(JY, X; JY, ξ) = −|Y |2 hDX H, ξi . (9.27)
Thus by (9.26) and (9.27) we obtain (b).
Theorem 9.11. [Chen (1981f)] Let N be a totally umbilical hypersurface
of a Kähler manifold M̃ with dimC M̃ ≥ 3. Then N has constant mean
curvature.
Proof. If N is a totally umbilical hypersurface of a Kähler manifold M̃ ,
then N is a CR-submanifold with rank H⊥ = 1 and T ⊥ N = JH⊥ . Thus it
follows from Lemma 9.4 that H is parallel in the normal bundle. Therefore
N has constant mean curvature |H|.
Remark 9.5. It is known in [Chen (1979)] that every non-totally geodesic,
totally umbilical hypersurface of CP 1 (4) × CH 1 (−4) has a non-constant
mean curvature. Since there exist ample examples of non-totally geodesic,
totally umbilical hypersurfaces in CP 1 (4) × CH 1 (−4), we know that The-
orem 9.11 is false whenever dimC M̃ = 2.
9.5 Mixed foliate CR-submanifolds
Definition 9.4. A CR-submanifold N in a Kähler manifold is called mixed

foliate if the following two conditions hold:
(a) its holomorphic distribution H is integrable and
(b) its second fundamental form σ satisfies σ(H, H⊥ ) = {0}.
Proposition 9.2. Let N be a mixed foliate CR-submanifold in a Kähler
manifold M̃ . Then we have
H̃B (X, Z) = −2|AJZ X|2 (9.28)
⊥
for any unit vectors X ∈ H and Z ∈ H .
Proof. If N is a mixed foliate CR-submanifold in a Kähler manifold M̃ ,
then we have
σ(H, H⊥ ) = {0}, [H, H] ⊂ H, σ(X, JY ) = σ(JX, Y ) (9.29)
for vector fields X, Y ∈ H. Thus the equation of Codazzi gives
H̃B (X, Z) = hσ(JX, ∇X Z), JZi − hσ(X, ∇JX Z).JZi
= hAJZ JX, ∇X Zi − hAJZ X, ∇JX Zi
for any X ∈ H and Z ∈ H⊥ . Hence Lemma 9.2 yields

H̃B (X, Z) = hAJZ JX, JAJZ Xi − hAJZ X, JAJZ JXi = −2|AJZ X|2 .
This proves the proposition.
Proposition 9.2 implies the following.
Theorem 9.12. [Chen (1981b)] Let M̃ be a Kähler manifold with positive

holomorphic bisectional curvature. Then M̃ admits no mixed foliate CR-
submanifold.
Corollary 9.3. [Bejancu et al. (1981)] Every complex space form M̃ m (c)
with c > 0 admits no mixed foliate CR-submanifold.
Remark 9.6. Geodesic spheres in CP m (4) are real hypersurfaces with

σ(H, H⊥ ) = {0}. Thus geodesic spheres in CP m (4) are mixed totally
geodesic CR-submanifolds, but not mixed foliate.
Theorem 9.13. [Chen (1981b)] Let N be a CR-submanifold of the complex

Euclidean m-space Cm . Then N is mixed foliate if and only if N is a CR-
product. Therefore N is locally a CR-product given by
N T × N ⊥ ⊂ CN × Cm−N , (9.30)
T N ⊥
where N is a complex submanifold of C and N is totally real in Cm−N .
Proof. Let N be a CR-submanifold of Cm . If N is mixed foliate. Then

Proposition 9.2 yields
AJH⊥ H = {0}. (9.31)
Thus, by Lemma 9.3, we know that N is a CR-product. Therefore, after
applying Theorem 6.5 we conclude that locally the immersion of N is given
by (9.30).
Conversely, if N is a CR-product in Cm , then (9.31) holds. Hence, by
Lemma 9.3 and Proposition 9.2 we obtain σ(H, H⊥ ) = {0}. Therefore N
is mixed foliate.
Remark 9.7. For anti-holomorphic submanifolds, Proposition 9.2 and

Theorem 9.13 are due to [Bejancu et al. (1981)].
For mixed foliate CR-submanifolds in complex hyperbolic spaces, we
have the following result from [Chen and Wu (1988)].
Theorem 9.14. Every mixed foliate CR-submanifold in a complex hyper-

bolic m-space CH m (−4) with arbitrary m is non-proper.
Chapter 10
Warped Products in Riemannian and

Kähler Manifolds
According to the famous embedding theorem of J. F. Nash done in [Nash

(1956)], every Riemannian manifold can be isometrically embedded in some
Euclidean spaces with sufficiently high codimension. The Nash embedding
theorem implies that every warped product N1 ×f N2 can be isometrically
embedded as Riemannian submanifolds in Euclidean spaces with sufficiently
high codimension.
In this chapter we present many important results concerning the geo-
metry of warped product manifolds immersed in Riemannian manifolds or
in Kähler manifolds.
10.1 An algebraic lemma
Definition 10.1. Let n be an integer ≥ 2 and let n1 , . . . , nk be k natural

numbers. Then (n1 , . . . , nk ) is called a partition of n if n1 + · · · + nk = n.
We need the following lemma from [Chen (2002j)] for later use.
Lemma 10.1. Let a1 , . . . , an be real numbers and k be an integer satisfying
2 ≤ k ≤ n − 1. Then, for any partition (n1 , . . . , nk ) of n, we have
X X
ai1 aj1 + ai2 aj2 + · · ·
1≤i1 <j1 ≤n1 n1 +1≤i2 <j2 ≤n1 +n2
X
+ aik ajk (10.1)
n1 ···+nk−1 +1≤i1 <j1 ≤n
1
≥ (a1 + · · · + an )2 − k(a21 + · · · + a2n ) ,
2k
with the equality holding if and only if
a1 + · · · + an1 = · · · = an1 +···+nk−1 +1 + · · · + an . (10.2)
207
Proof. Under the hypothesis we find

(
X X
2k ai1 aj1 + ai2 aj2 + · · ·
1≤i1 <j1 ≤n1 n1 +1≤i1 <j1 ≤n1 +n2
) X
n 2 n
X X
+ aik ajk − aα + k a2α
n1 +···+nk−1 +1≤i1 <j1 ≤n α=1 α=1
(
X X
= 2k ai1 aj1 + ai2 aj2 + · · ·
1≤i1 <j1 ≤n1 n1 +1≤i1 <j1 ≤n1 +n2
) n
X X X
+ aik ajk + (k−1) a2α − 2 aα aβ
n1 +···+nk−1 +1≤i1 <j1 ≤n α=1 1≤α<β≤n
( 2
X X
= ai1 − ai2
1≤ai1 ≤n1 n1 +1≤ai2 ≤n1 +n2
( )2
X X
+ ai1 − ai3 + ···
1≤ai1 ≤n1 n1 +n2 +1≤ai3 ≤n1 +n2 +n3
( )2
X X
+ aik−1 − aik
n1 +···+nk−2 +1≤ai1 ≤n1 +···+nk−1 n1 +···+nk−1 +1≤aik ≤n
≥ 0,
with equality holding if and only if (10.2) holds.
An immediate consequence of Lemma 10.1 is the following.
Corollary 10.1. Let a1 , . . . , an , η be n + 1 real numbers such that

2
(a1 + · · · + an ) = (n − 1) η + a21 + · · · + a2n . (10.3)
Then 2a1 a2 ≥ η, with equality holding if and only if
a1 + a2 = a3 = · · · = an .
Proof. Let a1 , . . . , an be real numbers. Then (10.3) is equivalent to

(a1 + · · · + an )2
η= − (a21 + · · · + a2n ).
n−1
By choosing k = n − 1, n1 = 2 and n2 = · · · = nk = 1, (10.1) becomes
2a1 a2 ≥ η, with the equality holding if and only if a1 + a2 = a3 = . . . = an
according to Lemma 10.1.
Warped Products in Riemannian and Kähler Manifolds 209
10.2 Warped products in real space forms
The following result provides the first solution to the following basic ques-
tion proposed in [Chen (2002a)].
Basic Question 10.1: What can we conclude from an arbitrary isometric

immersion of a warped product into a Euclidean space (or more generally
in a Riemannian manifold) with arbitrary codimension?
Theorem 10.1. [Chen (2002a)] Let φ : N1 ×f N2 → Rm (c) be an isometric

immersion of a warped product into a Riemannian m-manifold of constant
sectional curvature c. Then we have
∆f (n1 + n2 )2 2
≤ H + n1 c, (10.4)
f 4n2
where ni = dim Ni , i = 1, 2, H 2 = hH, Hi is the squared mean curvature of
φ, and ∆ is the Laplacian operator of N1 .
The equality sign of (10.4) holds if and only if φ : N1 ×f N2 → Rm (c)
is a mixed totally geodesic immersion with Tr σ1 = Tr σ2 , where Tr σ1 and
Tr σ2 denote the trace of σ restricted to N1 and N2 , respectively.
Proof. Assume that φ : N = N1 ×f N2 → Rm (c) is an isometric im-

mersion of a warped product N1 ×f N2 into a Riemannian manifold of
constant sectional curvature c. Denote by n1 , n2 and n the dimensions of
N1 , N2 and N1 × N2 , respectively.
Since N1 ×f N2 is a warped product, we have
∇X Z = ∇Z X = (X ln f )Z (10.5)
for unit vector fields X, Z tangent to N1 , N2 , respectively. Hence we find
K(X ∧ Z) = h∇Z ∇X X − ∇X ∇Z X, Zi
1n o (10.6)
= (∇X X)f − X 2 f .
f
If we choose a local orthonormal frame e1 , . . . , en such that e1 , . . . , en1 are
tangent to N1 and en1 +1 , . . . , en are tangent to N2 , then we have
Xn1
∆f
= K(ej ∧ es ) (10.7)
f j=1
for each s = n1 + 1, . . . , n.
From the equation of Gauss, it follows that the scalar curvature τ and
the squared mean curvature H 2 of N satisfy
2τ = n2 H 2 − ||σ||2 + n(n − 1)c, (10.8)
where ||σ||2 is the squared norm of the second fundamental form σ of N in

Rm (c).
Let us put
n2 2
δ = 2τ − n(n − 1)c − H . (10.9)
2
Then (10.8) becomes
n2 H 2 = 2δ + 2||σ||2 . (10.10)
If we choose an orthonormal frame en+1 , . . . , em of the normal bundle
so that en+1 is in the direction of the mean curvature vector, then (10.10)
becomes
n
!2
X
n+1
σii
i=1
  (10.11)
n
X X m
X n
X
= 2 δ + n+1 2
(σii ) + n+1 2
(σij ) + r 2
(σij ) .
i=1 i6=j r=n+2 i,j=1
Equation (10.11) is equivalent to

h X
2 n+1 2
(ā1 + ā2 + ā3 ) = 2 δ + ā21 + ā22 + ā23 + 2 (σij )
1≤i<j≤n
m
X n
X X
r 2 n+1 n+1
+ (σij ) −2 σjj σkk (10.12)
r=n+2 i,j=1 2≤j<k≤n1
X i
n+1 n+1
−2 σss σtt ,
n1 +1≤s<t≤n
where
n+1 n+1
ā1 = σ11 , ā2 = σ22 + · · · + σnn+1
1 n1
,
ā3 = σnn+1
1 +1n1 +1
n+1
+ · · · + σnn .
Applying Corollary 10.1 to (10.11) yields
X X
n+1 n+1 n+1 n+1
σjj σkk + σss σtt
1≤j<k≤n1 n1 +1≤s<t≤n
X m n (10.13)
δ n+1 2 1 X X r 2
≥ + (σαβ ) + (σαβ ) ,
2 2 r=n+2
1≤α<β≤n α,β=1
with equality holding if and only if we have

n+1
σ11 + · · · + σnn+1
1 n1
= σnn+1
1 +1n1 +1
n+1
+ · · · + σnn . (10.14)
From the equation of Gauss and (10.7), we have

n2 ∆f X X
=τ− K(ej ∧ ek ) − K(es ∧ et )
f
1≤j<k≤n1 n1 +1≤s<t≤n
m
X X
n1 (n1 − 1) r r r 2

=τ− c− σjj σkk − (σjk ) (10.15)
2 r=n+1 1≤j<k≤n1
Xm X
n2 (n2 − 1) r r r 2

− c− σss σtt − (σst ) .
2 r=n+1 n1 +1≤s<t<n
Therefore, by (10.9), (10.13) and (10.15), we obtain

n2 ∆f n(n − 1) X
n+1 2
≤ τ− c + n1 n2 c − (σjt )
f 2
1≤j≤n1 ; n1 +1≤t≤n
m n m
1 X X r 2 X X
r 2 r r

− (σαβ ) + (σjk ) − σjj σkk
2 r=n+2 r=n+2
α,β=1 1≤j<k≤n1
m
X X
r 2 r r
δ
+ (σst ) − σss σtt −
r=n+2 n1 +1≤s<t<n
2
(10.16)
Xm X X
n(n − 1) r 2
= τ− c + n1 n2 c − (σjt )
2 r=n+1 1≤j≤n1 n1 +1≤t≤n
1 X X r 2 1 X 2
m m X
r δ
− σjj − σtt −
2 r=n+2 2 r=n+2 2
1≤j≤n1 n1 +1≤t≤n
2
n(n − 1) δ n 2
≤ τ− c + n1 n2 c − = H + n1 n2 c,
2 2 4
which proves inequality (10.4). From (10.14) and (10.16) we see that the
equality sign of (10.4) holds if and only if we have
r
σjt = 0, n + 1 ≤ r ≤ m, 1 ≤ j ≤ n1 ; n1 + 1 ≤ t ≤ n
r
σ11 + · · · + σnr 1 n1 (10.17)
= σnr 1 +1n1 +1 + ···+ r
σnn = 0, n + 2 ≤ r ≤ m.
Condition (10.17) shows that the second fundamental form σ of N1 ×f N2
in Rm (c) satisfies σ(D1 , D2 ) = {0}. Thus the immersion φ is mixed totally
geodesic. Hence, by using a result of [Nölker (1996)], we know that locally
there exists a warped product representation M1 ×ρ M2 of Rm (c) such that
φ : N1 ×f N2 → M1 ×ρ M2 = Rm (c) is a warped product immersion of
φ1 : N1 → M1 and φ2 : N2 → M2 ; so that we have
φ(x1 , x2 ) = (φ1 (x1 ), φ(x2 )) for x1 ∈ N1 , x2 ∈ N2 .
Moreover, it follows from (10.11) and (10.15) that

n1
X nX
1 +n2
σ(ej , ej ) = σ(es , es ). (10.18)

j=1 s=n1 +1
Hence, we have Tr σ1 = Tr σ2 .
Conversely, if φ : N1 ×f N2 → M1 ×ρ M2 = Rm (c) is a mixed totally
geodesic immersion with Tr σ1 = Tr σ2 , then all the inequalities in (10.13)
and in (10.16) become equalities. Hence we obtain the equality sign of
(10.4) by (10.16).
Theorem 10.1 was extended in [Chen and Dillen (2008a)] as follows.
Theorem 10.2. Let φ : N1 ×f2 N2 × · · · ×fℓ Nℓ → M be an isometric
immersion of a multiply warped product N1 ×f2 N2 × · · · ×fℓ Nℓ into an
arbitrary Riemannian manifold M , where f2 , . . . , fℓ are positive functions
on N1 . Then
Xℓ
∆fj n2 (ℓ − 1) 2
nj ≤ H + n1 (n − n1 ) max K̃, (10.19)
j=2
fj 2ℓ
Pℓ
where n = j=1 nj and max K̃(p) denotes the maximum of the sectional
curvature K̃ of M restricted to plane sections in Tp N at p ∈ N .
The equality sign of (10.19) holds identically if and only if the following
two conditions hold:
(1) φ is mixed totally geodesic such that Tr σ1 = · · · = Tr σℓ ;
(2) At each point p ∈ N , K̃ satisfies K̃(u, v) = max K̃(p), ∀u ∈ Tp11 N1 and
1
∀v ∈ T(p 2 ,··· ,pk )
(N2 × · · · × Nℓ ).
This theorem was obtained by modifying the proof of Theorem 10.1.
Definition 10.2. A doubly warped product is a product manifold which is of

the form f2 N1 ×f2 N2 with the metric g = f22 g1 ⊕ f12 g2 , where f1 : N1 → R+
and f2 : N2 → R+ are positive smooth maps and g1 and g2 are the metrics
of N1 and N2 , respectively.
Doubly warped products are natural generalization of ordinary warped

products. For doubly warped products, we have the following.
Lemma 10.2. [Ünal (2000)] Let f2 N1 ×f1 N2 be a doubly warped product.
Then
∇X Z = ∇Z X = (X ln f1 )Z + (Z ln f2 )X (10.20)
for any vector field X tangent to N1 and Z tangent to N2 .
Proof. Follows from direct computation of Christoffel symbols.

By applying Lemma 10.2 and by modifying the proof of Theorem 10.1,
the following result was obtained in [Olteanu (2010a)].
Theorem 10.3. Let φ : f2 N1 ×f1 N2 → M be an isometric immersion
of a doubly warped product φ : f2 N1 ×f1 N2 into an arbitrary Riemannian
manifold M , where fi is a positive function on Ni . Then
∆1 f1 ∆2 f2 (n1 + n2 )2 2
n2 + n1 ≤ H + n1 n2 max K̃, (10.21)
f1 f2 4
where ni = dim Ni and ∆i is the Laplacian of Ni for i = 1, 2.
The equality sign of (10.21) holds identically if and only if the following
two conditions hold:
(1) φ is mixed totally geodesic such that Tr σ1 = Tr σ2 .
(2) At each point x = (x1 , x2 ) ∈ N , K̃ satisfies K̃(u, v) = max K̃(x) for
each unit vector u ∈ Tx11 N1 and each unit vector v ∈ T(x1
2)
N2 .
10.3 Some applications of Theorems 10.1 and 10.2
As applications of Theorems 10.1 and 10.2 we have the following results

from [Chen (2002a); Chen and Dillen (2008a)].
Corollary 10.2. If N1 ×f N2 is a warped product of Riemannian manifolds
whose warping function f is a harmonic function, then
(1) N1 ×f N2 admits no isometric minimal immersion into any Riemannian

manifold of negative curvature;
(2) every isometric minimal immersion from N1 ×f N2 into a Euclidean
space is a warped product immersion.
Proof. Let φ : N1 ×f N2 → M be an isometric minimal immersion of

N1 ×f N2 into a Riemannian manifold M . If f is a harmonic function
on N1 , then inequality (10.19) in Theorem 10.2 implies max K̃ ≥ 0 on
N = N1 ×f N2 . This shows that N1 ×f N2 does not admit any isometric
minimal immersion into any Riemannian manifold of negative curvature.
If the ambient space M is a Euclidean space, then the minimality of
N1 ×f N2 and the harmonicity of f imply that the equality sign of (10.2)
in Theorem 10.1 holds identically. Hence φ is a mixed totally geodesic
submanifold according to Theorem 10.1. Therefore, by a result of [Nölker
(1996)], φ is locally a warped product immersion.
Corollary 10.3. Let f be an eigenfunction of Laplacian ∆ on N1 with

eigenvalue λ > 0. Then every Riemannian warped product N1 ×f N2 does
not admit any isometric minimal immersion into any Riemannian manifold
of non-positive curvature.
Proof. Let f be an eigenfunction of ∆ on N1 with eigenvalue λ > 0.

Suppose that N1 ×f N2 admits an isometric minimal immersion into M .
Then (10.19) implies that n1 max K̃ ≥ λ > 0. Therefore M cannot be
non-positively curved.
Corollary 10.4. Let N1 be a compact manifold. Then

(1) every Riemannian warped product N1 ×f N2 does not admit an iso-
metric minimal immersion into any Riemannian manifold of negative
curvature;
(2) every Riemannian warped product N1 ×f N2 does not admit an isometric
minimal immersion into a Euclidean space.
Proof. Assume N1 is compact. Let φ : N1 ×f N2 → M be an isometric

minimal immersion of N1 ×f N2 into a non-positively curved Riemannian
manifold M . Then inequality (10.19) implies that
∆f
≤ n1 max K̃ ≤ 0.
f
Since the warping function f is positive, we find ∆f ≤ 0. Hence it follows
from Hopf’s lemma and the compactness of N1 that f is a positive constant.
Consequently, we have max K̃ = 0, which implies statement (1).
If M s a Euclidean space, then the equality case of (10.19) holds. Hence
φ is mixed totally geodesic. Therefore Moore’s lemma implies that φ is a
product immersion, say
φ = (φ1 , φ2 ) : (N1 , g1 ) × (N2 , f 2 g2 ) → Em1 × Em2 = Em .
Since φ is minimal, φ1 : N1 → Em1 is also minimal. But is impossible since
N1 is compact.
Remark 10.1. The same applications applied to doubly warped products

as well (cf. [Faghfouri and Majidi (2015)]).
Example 10.1. There exist many minimal immersions of a warped product

N1 ×f N2 with harmonic warping function f into a Euclidean space. For
instance, if N2 is a minimal submanifold of the unit (m − 1)-hypersphere
S m−1 (1) ⊂ Em centered at the origin, then the minimal cone C(N2 ) over
N2 with vertex at the origin of Em is a warped product R+ ×s N2 whose

warping function f = s is a harmonic function. Here s is the coordinate
function of the positive real line R+ . This provides many examples of
minimal warped products in Em which satisfy the equality case of (10.4).
In contrast, if f is an Lq function on N1 for some q > 1, then for any
Riemannian manifold N2 the warped product N1 ×f N2 does not admit
any isometric minimal immersion into any Riemannian manifold with non-
positive sectional curvature (see Theorem 6.2 and Remark 7.1 of [Chen and
Wei (2009)]).
Example 10.1 illustrates that Theorems 10.1 and 10.2 and Corollary
10.2 are optimal. The next examples show that Corollaries 10.3 and 10.4
are also optimal.
Example 10.2. There are many isometric minimal immersions of warped
products N1 ×f N2 into a hyperbolic space such that the warping function
f is an eigenfunction with negative eigenvalue. For example, R ×ex En−1
admits an isometric minimal immersion into H n+1 (−1).
Example 10.3. There exist many minimal immersions of N1 ×f N2 into
Euclidean space with compact N2 . For examples, a hypercaternoid in En+1
is a minimal hypersurfaces isometric to a warped product R ×f S n−1 . Also,
for a compact minimal submanifold N2 of S m−1 ⊂ Em , the minimal cone
C(N2 ) is a warped product R+ ×s N2 .
Example 10.4. Contrast to Euclidean and hyperbolic spaces, the unit m-
sphere S m admits warped product minimal submanifolds N1 ×f N2 such
that N1 and N2 are both compact. The simplest such examples are minimal
Clifford tori Mk,n−k in S n+1 defined by
r r
k n−k
Mk,n−k = S k × S n−k , k = 2, . . . , n − 1.
n n
10.4 Rotation hypersurfaces in real space forms
As before, let S n+1 (c) (c > 0) denote the hypersphere of radius c−1 in the
Euclidean space Rn+2 centered at the origin, i.e.,

n+1 n+2 2 2 1
S (c) = x ∈ R : x1 + · · · + xn+2 = ,
c
P
where Rn+2 equips with the standard Euclidean metric ds2 = n+2 2
i=1 dxi .
n+1
Then S (c) has constant positive sectional curvature c.
Suppose that Rn+2 is the Lorentzian space equipped with the Lorentzian
metric ds2 = dx21 + · · · + dx2n+1 − dx2n+2 . Let us consider the hypersurface
H n+1 (c) (c < 0) defined by

1
H n+1 (c) = x ∈ Rn+2 : x21 + · · · + x2n+1 − x2n+2 = , xn+2 > 0 .
c
Then H n+1 (c) with the induced Riemannian metric has constant negative
curvature c.
In this section, let Rn+1 (c) denote a complete simply-connected Rie-
mannian manifold with constant sectional curvature c. Thus Rn+1 (c) will
be the Euclidean space En+1 for c = 0; the hypersphere S n+1 (c) for c > 0;
and the hyperbolic space H n+1 (c) for c < 0.
Definition 10.3. A rotation hypersurface in En+1 is an O(n)-invariant

hypersurface, where O(n) is considered as a subgroup of isometries of En+1 .
Following [do Carmo and Dajczer (1983)] we explain what is a rotation

hypersurface of a real space form Rn+1 (c) forc 6= 0. We shall always con-
sider Rn+1 (c) as a hypersurface in Rn+2 , ds2 . Let P 3 be a 3-dimensional
linear subspace of Rn+2 that intersects Rn+1 (c). We denote the intersection
by R2 (c); if c < 0 we take only the upper part.
Let P 2 be any linear subspace in P 3 . Notice that any isometry of
n+1
R (c) isthe restriction to Rn+1 (c) of an orthogonal transformation of
n+2
R , ds2 , and conversely. Let O(P 2 ) be the group of orthogonal trans-
formations (with positive determinant) that leaves P 2 pointwise fixed. We
take any curve α in R2 (c) which does not intersect P 2 . The orbit of α
under O(P 2 ) is called the rotation hypersurface with profile curve α and
axis P 2 . The orbit of α(s) for a fixed s is a sphere, and if c < 0, then
this sphere is elliptic, hyperbolic or parabolic according to P 2 respectively
being Lorentzian, Riemannian or degenerate.
In order to give a parametrization of a rotation hypersurface of the
different types, we introduce the vector u ∈ P 3 such that P 2 coincides with
u⊥ = {v ∈ P 3 | hv, ui = 0}. We can always assume that u has length
1, −1 or 0, according to P 2 respectively being Lorentzian, Riemannian or
degenerate, and that hu, α′ i > 0.
Let ǫ = hu, ui. We define the map Q as the orthogonal projection of P 3
on u⊥ if ǫ 6= 0 and as the identity map of P 3 if ǫ = 0. Further, let P n−1
be the orthogonal complement of P 3 in Rn+2 and let P n be the linear
space, spanned by P n−1 and u. If ǫ = 1 (respectively, ǫ = −1), then P n is
Riemannian, (respectively, Lorentzian) and we can define a mapping φ of
M n−1 (ǫ) into P n by considering M n−1 (ǫ) as a unit hypersphere in P n . If

ǫ = 0, then we can define a mapping φ of M n−1 (0) into P n by identifying
M n−1 (0) and P n−1 and defining
1
φ(p) = p − hp, pi u. (10.22)
2
Then a parametrization of the rotation hypersurface with profile curve α
around the axis P 2 is given by
f (s, p) = Q(α(s)) + hα(s), ui φ(p).
If we assume that s is the arc length of α, then it follows immediately that
the rotation hypersurface is intrinsically the warped product I ×f M n−1 (ǫ),
where I is an open interval of R and f is defined by f (s) = hα(s), ui.
The second fundamental form σ of the rotation hypersurface M n satis-
fies (cf. [do Carmo and Dajczer (1983)])
!
∂ ∂ f ′′ + cf
σ , = p , (10.23)
∂s ∂s ǫ − cf 2 − f ′ 2
p
ǫ − cf 2 − f ′ 2
σ(X, Y ) = − hX, Y i (10.24)
f
for X and Y tangent to M n−1 (ǫ).
10.5 Another optimal inequality for warped products
Besides Theorems 10.1 and 10.2, there is another general optimal inequality
for a warped product N1 ×f N2 in a real space form Rm (c).
Theorem 10.4. [Chen (2004b)] For any given isometric immersion φ :

N1 ×f N2 → Rm (c), the scalar curvature τ of the warped product N1 ×f N2
satisfies
∆f n2 (n − 2) 2 1
τ≤ + H + (n + 1)(n − 2)c, (10.25)
n1 f 2(n − 1) 2
with n = n1 + n2 , where ni = dim Ni , i = 1, 2.
If n = 2, the equality case of (10.25) holds automatically.
If n ≥ 3, the equality sign of (10.25) holds identically if and only if one
of the following two statement occurs.
(1) N1 ×f N2 is of constant curvature c, the warping function f is an eigen-

function with eigenvalue c, i.e., ∆f = cf , and N1 ×f N2 is immersed
as a totally geodesic submanifold in Rm (c);
(2) Locally, N1 ×f N2 is immersed as a rotational hypersurface in a totally

geodesic submanifold Rn+1 (c) of Rm (c) with a geodesic of Rn+1 (c) as
its profile curve.
Proof. Let φ : N1 ×f N2 → Rm (c) be an isometric immersion of a warped

product N1 ×f N2 into a Riemannian manifold Rm (c) of constant sectional
curvature c. From the equation of Gauss we have
2τ = n2 H 2 − ||σ||2 + n(n − 1)c. (10.26)
Let
n2 (n − 2) 2
η = 2τ − H − (n + 1)(n − 2)c. (10.27)
n−1
Then (10.26) and (10.27) yield
n2 H 2 = (n − 1)||σ||2 + (n − 1)(η − 2c). (10.28)
Let X and Z be two unit local vector fields tangent to N1 and N2 ,
respectively. We choose an orthonormal frame e1 , . . . , em such that e1 =
X, en1 +1 = Z and en+1 is parallel to the mean curvature vector. Then
(10.28) gives
n
!2 ( n
X X X
n+1 n+1 2 n+1 2
σii = (n − 1) (σii ) + (σij )
i=1 i=1 i6=j
) (10.29)
m
X n
X
r 2
+ (σij ) + η − 2c .
r=n+2 i,j=1
By applying Corollary 10.1 to (10.29) we obtain

X Xm Xn
n+1 n+1 n+1 2 r 2
2σ11 σn1 +1n1 +1 ≥ (σij ) + (σij ) + η − 2c, (10.30)
i6=j r=n+2 i,j=1
from which we get

m
X X r 2
2
K(e1 ∧ en1 +1 ) ≥ (σ1j ) + σnr 1 +1 j
r=n+1 j∈Ω1n1 +1
i6=j
X m
1 n+1 2 1 X X
r
2
+ (σij ) + σij
2 2 r=n+2 (10.31)
i,j∈Ω1n1 +1 i,j∈Ω1n1 +1
m
X
1 r
2 η
+ σ11 + σnr 1 +1n1 +1 +
2 r=n+2 2
η
≥ ,
2
where Ω1n1 +1 = {1, . . . , n} r {1, n1 + 1}.

Since N1 ×f N2 is a warped product, we have ∇X Z = ∇Z X = (X ln f )Z
for unit vector fields X, Z tangent to N1 , N2 respectively. Hence we find
1n o
K(X ∧ Z) = h∇Z ∇X X − ∇X ∇Z X, Zi (∇X X)f − X 2 f . (10.32)
f
Combining (10.27), (10.31), and (10.32) yields
1n o n2 (n − 2) 1
τ≤ (∇e1 e1 )f − e21 f + H 2 + (n + 1)(n − 2)c. (10.33)
f 2(n − 1) 2
If the equality sign of (10.33) holds, then all inequalities in (10.30) and
(10.31) become equalities. Thus we have
n+1 n+1 n+1
σ1j = 0, σjn1 +1
= 0, σij = 0, i 6= j;
r r r r
σ1j = σjn1 +1
= σij = 0, σ11 + σnr 1 +1n1 +1 = 0, (10.34)
i, j ∈ Ω1n1 +1 , r = n + 2, . . . , m.
In other words, the shape operators take the following forms:
 
a 0 ··· 0
0 
 
. 
 .. µI


n1 −1 0 


0 
An+1 =  ,
 b 0 · · · 0
 
 0 
 
 .. 
 0 . µIn2 −1 
0
 r r

σ11 0 . . . 0 σ1n 1 +1
0 ... 0
 0 0 
 
 . .. 
 ..


0 . 0


 0 0 
Ar =  r r  , r = n + 2, . . . , m
σ1n1 +1 0 . . . 0 −σ11 0 . . . 0
 
 0 0 
 
 .. .. 
 . 0 . 0
0 0
with a + b = µ, where Ik denotes the identity matrix of order k.
Similar to (10.33), we also have
1n o n2 (n − 2) 1
τ≤ (∇eα eα )f − e2α f + H 2 + (n + 1)(n − 2)c (10.35)
f 2(n − 1) 2
for α = 1, 2, . . . , n1 . Hence, by summing up α from 1 to n1 , we obtain

∆f n1 n2 (n − 2) 2 1
n1 τ ≤ + H + n1 (n + 1)(n − 2)c, (10.36)
f 2(n − 1) 2
If the equality sign of (10.25) holds identically, then the equality sign of
(10.35) holds for each α ∈ {1, . . . , n1 }. Thus, for each α ∈ {1, . . . , n1 } and
each t ∈ {n1 , . . . , n}, we have
n+1 n+1 n+1
σαj = 0, σtj = 0, σij = 0, i 6= j;
r r r r r
(10.37)
σαj = σtj = σij = 0, σαα + σtt =0
for i, j ∈ Ωαt = {1, . . . , n} r {α, t}; r = n + 2, . . . , m.
If n = 2, then n1 = n2 = 1. Thus, by (10.32), we have τ = (∆f )/f .
Hence, the equality sign of (10.25) holds automatically. Next, suppose that
n = n1 + n2 ≥ 3. Then (10.37) implies An+2 = · · · = Am = 0. Moreover,
from n ≥ 3 and (10.37), we find (a) An+1 = 0, or (b) a = 0, b = µ 6= 0 or
a = µ 6= 0, b = 0.
Case (a): An+1 = 0. In this case, N1 ×f N2 is totally geodesic in Rm (c).
Hence N1 ×f N2 is a real space form of constant sectional curvature c which
implies that the scalar curvature of N1 ×f N2 is given by τ = 12 n(n − 1)c.
Since a totally geodesic submanifold is minimal, the equality sign of
(10.25) and τ = 12 n(n − 1)c imply ∆f = cf . Thus f is either a harmonic
function or an eigenfunction of the Laplacian with eigenvalue c, according
to c = 0 or c 6= 0, respectively.
Conversely, suppose N1 ×f N2 is a warped product decomposition of a
real space form Rn (c) such that the warping function f satisfies ∆f = cf .
Clearly, for each integer m > n, N1 ×f N2 can be locally isometrically
immersed in a real space form Rm (c) of the same curvature as a totally
geodesic submanifold. Now, it is easy to verify that such a totally geodesic
immersion satisfies the equality case of inequality (10.25).
Case (b): Either a = 0, b = µ 6= 0 or a = µ 6= 0, b = 0. In this case, An+1
has two distinct eigenvalues 0, µ with multiplicities 1, n − 1, respectively, on
U = {p ∈ N1 ×f N2 : H 2 (p) > 0}. The first normal subbundle, Im h, is of
rank one on U . Without loss of generality, we may assume that
σ(e1 , e1 ) = 0, σ(e2 , e2 ) = · · · = σ(en , en ) = µen+1 , (10.38)
σ(ei , ej ) = 0, 1 ≤ i 6= j ≤ n
on U . From (10.38) we find

¯ e σ (ej , ej ) = (ek µ)en+1 + µDe en+1 ,
∇ k k
j (10.39)
¯ k
∇ej σ (ej , ek ) = ω (ej ) + ω (ej ) en+1
j k
for j ∈ {2, . . . , n} and k ∈ {1, . . . , n}. Therefore the equation of Codazzi

implies that

µDek en+1 + ωjk (ej ) + ωkj (ej ) + ek µ en+1 = 0, j 6= k, j > 1. (10.40)
Since DX en+1 is perpendicular to en+1 , (10.40) implies that DX en+1 = 0

for X being one of e1 , . . . , en . Hence the first normal subbundle Im σ =
Span{en+1 } is a parallel normal subbundle on U . Therefore, by applying a
result of [Erbacher (1971)], we know that M has essential codimension one.
Thus, in a neighborhood of each point of U , the warped product N1 ×f N2 is
isometrically immersed in a totally geodesic submanifold Rn+1 (c) of Rm (c).
Because the shape operator of U has one eigenvalue of multiplicity n − 1
and the other eigenvalue is zero, it follows from a result of [do Carmo and
Dajczer (1983)] that M is a rotation hypersurface whose profile curve is a
geodesic of Rn+1 (c). Therefore, after applying the continuity of the squared
mean curvature H 2 , we may conclude that U is a dense open subset of
N1 × f N2 .
Conversely, assume that M is rotation hypersurface in Rn+1 (c) whose
profile curve α is a geodesic of Rn+1 (c). Let us assume that the profile
curve α is parametrized by an arc length function s. Then M is isometric
to a warped product I ×f M n−1 (ǫ), where I is an open interval of R and
the shape operator of M has exactly two distinct eigenvalues 0 and µ of
multiplicities 1 and n − 1, respectively. Thus, the squared mean curvature
is given by
(n − 1)2 2
H2 = µ . (10.41)
n2
Moreover, by applying the equation of Gauss, we know that the scalar
curvature of the rotation hypersurface is given by
n(n − 1) (n − 1)(n − 2) 2
τ= c+ µ . (10.42)
2 2
Because the profile curve α = α(s) is a geodesic in Rn+1 (c) parametrized by
an arc length function, Formula (10.25) implies that the warping function
f satisfies the differential equation: f ′′ (s) + cf (s) = 0. Hence we get
∆f = cf. (10.43)
Now, we can easily verify from (10.41), (10.42) and (10.43) that the rotation
hypersurface I ×f M n−1 (ǫ) in Rn+1 (c) satisfies the equality case of (10.25)
identically.
10.6 Warped products in Kähler manifolds
For arbitrary warped products submanifolds in a complex hyperbolic space,

we have the following result from [Chen (2002d)].
Theorem 10.5. Let φ : N1 ×f N2 → CH m (−4) be an isometric immer-

sion of a warped product N1 ×f N2 into the complex hyperbolic m-space
CH m (−4) of constant holomorphic sectional curvature −4. Then we have
∆f n2 2
≤ H − n1 , (10.44)
f 4n2
where ni = dim Ni and n = n1 + n2 .
The equality sign of (10.44) holds if and only if the following three
conditions hold.
(a) φ is mixed totally geodesic,

(b) Tr σ1 = Tr σ2 , and
(c) JD1 ⊥ D2 , where J is the almost complex structure of CH m (−4).
Proof. Let φ : N1 ×f N2 → CH m (−4) be an isometric immersion of a

warped product N1 ×f N2 into CH m (−4) with ni = dim Ni .
Put n = n1 + n2 . We use the following convention on the range of
indices unless mentioned otherwise:
j, k, ℓ = 1, . . . , n; α, β = 1, . . . , n1 ; s, t = n1 + 1, . . . , n.
If we choose a local field of orthonormal frame e1 , . . . , en1 +n2 such that
e1 , . . . , en1 ∈ D1 and en1 +1 , . . . , en ∈ D2 , then (10.6) yields
Xn1
∆f
= K(eα ∧ es ), s = n1 + 1, . . . , n. (10.45)
f α=1
Let R denote the Riemannian curvature tensor of N = N1 ×f N2 . Then

the equation of Gauss is given by
hR(X, Y )Z, W i = hσ(X, W ), σ(Y, Z)i − hσ(X, Z), σ(X, Z)i
− {hX, W i hY, Zi − hX, Zi hY, W i + hJY, Zi hJX, W i (10.46)
− hJX, Zi hJY, W i + 2 hX, JY i hJZ, W i}.
It follows from (10.46) that the scalar curvature and the squared mean
curvature H 2 of N satisfy
2τ = n2 H 2 − ||σ||2 − n(n − 1) − 3 ||P ||2 , (10.47)
where ||σ||2 is the squared norm of the second fundamental form σ and
Pn 2
||P ||2 = i,j=1 hei , P ej i , where P is defined by defined by (6.38).
If we put
n2 2
η = 2τ − H + n(n − 1) + 3||P ||2 , (10.48)
2
then we obtain from (10.47) and (10.48) that
n2 H 2 = 2η + 2||σ||2 . (10.49)
If we choose a local field of orthonormal frame en+1 , . . . , e2m of the normal
bundle so that en+1 is in the direction of the mean curvature vector, then
(10.49) becomes
!2
Xn n Xn
X X2m Xn
o
n+1 n+1 2 n+1 2 r 2
σii =2 η+ σii + σij + σij ,
i=1 i=1 i6=j r=n+2 i,j=1
which can be restated as

n X 2
2 n+1
(a1 + a2 + a3 ) = 2 η + a21 + a22 + a23 + 2 σij
1≤i<j≤n
2m
X n
X X
r
2 n+1 n+1
+ σij −2 σαα σββ (10.50)
r=n+2 i,j=1 2≤α<β≤n1
X o
n+1 n+1
−2 σss σtt ,
n1 +1≤s<t≤n
where
n+1 n+1
a1 = σ11 , a2 = σ22 + · · · + σnn+1
1 n1
,
(10.51)
a3 = σnn+1
1 +1n1 +1
n+1
+ · · · + σnn .
Thus, by applying Corollary 10.1 to equation (10.50) we find
n+1 n+1

2σ11 σ22 + · · · + σnn+1
1 n1
X 2m
X n
X
n+1
2 r
2
≥η+2 σij + σij
(10.52)
1≤i<j≤n r=n+2 i,j=1
X X
n+1 n+1 n+1 n+1
−2 σαα σββ − 2 σss σtt
2≤α<β≤n1 n1 +1≤s<t≤n
which is nothing but
X X η
n+1 n+1 n+1 n+1
σαα σββ + σss σtt ≥
2
1≤α<β≤n1 n1 +1≤s<t≤n
X 2m n (10.53)
n+1
2 1 X X r 2

+ σjk + σjk .
2 r=n+2
1≤j<k≤n j,k=1
The equality sign of (10.53) holds if and only if we have

n+1
σ11 + · · · + σnn+1
1 n1
= σnn+1
1 +1n1 +1
n+1
+ · · · + σnn . (10.54)
Now, by applying equations (10.45) and (10.46) of Gauss, we get
n2 ∆f X X
=τ− K(eα ∧ eβ ) − K(es ∧ et )
f s<t
α<β
n1 (n1 − 1) n2 (n2 − 1) X 2
=τ+ + + 3 hP eα , eβ i
2 2
α<β
X 2m
X X (10.55)
2 r r r 2

+ 3 hP es , et i − σαα σββ − (σαβ )
s<t r=n+1 α<β
2m
X X
r r r 2

− σss σtt − (σst ) .
r=n+1 s<t
Therefore, after applying (10.48), (10.53) and (10.55), we obtain

n2 ∆f n(n − 1) X X
n+1 2
≤ τ+ − n1 n2 − σαt + 3 hP es , et i2
f 2 α,t s<t
m n 2m X
1 X X r 2
X
r

− σjk + (σαβ )2 − σαα
r r
σββ
2 r=n+2 r=n+2
j,k=1 α<β
2m X
X X
r
2 r r 2 η
+ σst − σss σtt + 3 hP eα , eβ i −
r=n+2 s<t
2
α<β
(10.56)
2m X
X
n(n − 1) r
2
=τ+ − n1 n2 − σαt
2 r=n+1 α,t
1 X X r 2 1 X X r 2
2m 2m
− σαα − σ
2 r=n+2 α 2 r=n+2 t tt
X 2
X 2 η
+ 3 hP eα , eβ i + 3 hP es , et i − ,
s<t
2
α<β
where the equality case of the inequality holds if and only if condition
(10.54) is satisfied. From (10.56) we find
n2 ∆f n(n − 1) η
≤τ + − n1 n2 −
f 2 2
X 2
X 2
(10.57)
+ 3 hP eα , eβ i + 3 hP es , et i
α<β s<t
with the equality holding if and only if we have

X X
r r
h(D1 , D2 ) = {0}, σαα = σtt = 0, (10.58)
α t
for r = n + 2, . . . , 2m.
Finally, by applying (10.48) and (10.57), we derive that
n2 ∆f n2 2 X 2 n2 2
= H − n1 n2 − 3 hP eα , et i ≤ H − n1 n2 (10.59)
f 4 α,t
4

If the equality sign of (10.44) holds, then clearly all of the inequalities in
(10.53), (10.57) and (10.59) become equalities. Therefore we obtain (10.54),
(10.58) and hP eα , et i = 0. Consequently, we have conditions (1), (2) and
(3) of the theorem.
Conversely, if conditions (1), (2) and (3) hold, then the equality case
of inequalities in (10.56), (10.57) and (10.59) become equalities. Hence we
obtain the equality case of (10.44). This proves the theorem.
By using Theorem 10.5, the next three corollaries from [Chen (2002d)]
can be proved in the same way as Corollaries 10.2, 10.3 and 10.4.
Corollary 10.5. Let N1 ×f N2 be a Riemannian warped product whose

warping function f is harmonic. Then N1 ×f N2 does not admit any iso-
metric minimal immersion into any complex hyperbolic space.
Corollary 10.6. If f is an eigenfunction of Laplacian on N1 with eigen-

value λ > 0, then N1 ×f N2 does not admits an isometric minimal immer-
sion into any complex hyperbolic space.
Corollary 10.7. If N1 is compact, then every Riemannian warped product

N1 ×f N2 does not admit an isometric minimal immersion into any complex
hyperbolic space.
Similarly, for warped products submanifolds in the complex projective

m-space CP m (4), we have the following result from [Chen (2003c)].
Theorem 10.6. Let φ : N1 ×f N2 → CP m (4) be an isometric immersion of

a Riemannian warped product into the complex projective m-space CP m (4).
Then we have
∆f n2 2
≤ H + 3 + n1 . (10.60)
f 4n2
The equality sign of (10.60) holds identically if and only if we have
(a) n1 = n2 = 1,
(b) f is an eigenfunction of the Laplacian of N1 with eigenvalue 4, and
(c) φ is totally geodesic and holomorphic immersion.
Proof. This theorem can be proved in a similar way as Theorem 10.5 by

making some minor modification.
An immediate applications of Theorem 10.6 is the following.
Corollary 10.8. If f is a positive function on a Riemannian n1 -manifold

N1 such that ∆f > (3 + n1 )f at a point x ∈ N1 , then, for any Riemannian
manifold N2 , the warped product N1 ×f N2 does not admit any minimal
immersion into CP m (4) for any m.
For totally real submanifolds in CP m , Theorem 10.6 can be sharpen as

the following.
Theorem 10.7. Let φ : N1 ×f N2 → CP m (4) be a totally real immersion

of a warped product into CP m (4). Then we have
∆f n2 2
≤ H + n1 . (10.61)
f 4n2
The equality sign of (10.61) holds identically if and only if the immer-
sion φ : N1 ×f N2 → CP m (4) is mixed totally geodesic with Tr σ1 = Tr σ2 .
Proof. Since Gauss’ equation for totally real immersions in CP m (4) is

the same as the Gauss equation for immersions in RP m (1), the proof of
this theorem can be done in exactly the same way as Theorem 10.1.
Corollary 10.9. If f is a positive function on a Riemannian n1 -manifold

N1 such that ∆f > n1 f at a point x ∈ N1 , then, for any Riemannian
manifold N2 , the warped product N1 ×f N2 does not admit any totally real
minimal immersion into CP m (4) for any m.
The following examples show that Theorem 10.6, Corollary 10.8 and
Corollary 10.9 are sharp.
Example 10.5. Let N1 = I =: (− π4 , π4 ), N2 = S 1 (1) and f = 12 cos 2s.

Then the warped product I ×(cos 2s)/2 S 1 (1) has constant sectional curvature

∂
4. If we define a complex structure J on the warped product by J ∂s =
∂ 1
2(sec 2s) ∂t , then (I ×(cos 2s)/2 S , g, J) is holomorphically isometric to a
dense open subset of CP 1 (4).
Let φ : CP 1 (4) → CP m (4) be a totally geodesic holomorphic embed-

ding of CP 1 (4) into CP m (4). Then the restriction of φ to I ×(cos 2s)/2 S 1 (1)
is an isometric minimal immersion of I ×(cos 2s)/2 S 1 (1) into CP m (4) which
satisfies ∆f = (3 + n1 )f identically. This example shows that the assump-
tion “∆f > (3 + n1 )f at a point in N1 ” given in Theorem 10.6 is best
possible.
Example 10.6. Consider the same warped product I ×(cos 2s)/2 S 1 (1) as in
Example 10.5. Let φ : CP 1 (4) → CP m (4) be a standard totally geodesic
embedding of CP 1 (4) into CP m (4). Then the restriction of φ to I ×(cos 2s)/2
S 1 (1) gives rise to a minimal isometric immersion of I ×(cos 2s)/2 S 1 (1) into
CP m (4) which satisfies the equality case of (10.61) on I ×(cos 2s)/2 S 1 (1)
identically.
Example 10.7. Let g1 be the standard metric on the unit (n − 1)-sphere
S n−1 (1) and let N1 ×f N2 be the warped product with N1 = (− π2 , π2 ), N2 =
S n−1 (1) and f = cos s. Then the warping function of this warped product
satisfies ∆f = n1 f identically.
Moreover, it is direct to verify that this warped product is isometric to
a dense open subset of S n (1). Let
projection totally geodesic
φ : S n (1) −−−−−−→ RP n (1) −−−−−−−−−−→ CP n (4)
2:1 totally real
be a standard totally geodesic Lagrangian embedding of S n (1) into CP n (4).

Then the restriction of the immersion φ to N1 ×f N2 is a totally real minimal
immersion. This example illustrates that the assumption “∆f > n1 f at a
point in N1 ” given in Theorem 10.7 is also sharp.
10.7 Warped product submanifolds in generalized complex

space forms
Definition 10.4. An almost Hermitian manifold (M̃ , J, g) is called an RK-

manifold if its curvature tensor R̃ is invariant by J, i.e.,
R̃(JX, JY ; JZ, JW ) = R̃(X, Y ; Z, W ), (10.62)
for X, Y, Z, W ∈ T M̃ .
Definition 10.5. An almost Hermitian manifold (M̃ , J, g) is said to be of
pointwise constant type if for any x ∈ M̃ and X ∈ Tx M̃ we have
λ(X, Y ) = λ(X, Z), (10.63)
with λ(X, Y ) = R̃(X, Y ; JX, JY ) − R̃(X, Y, X, Y ), whenever the planes

defined by X, Y and X, Z are totally real and g(Y, Y ) = g(Z, Z).
An almost Hermitian manifold M̃ is said to be of constant type if for
any unit vector fields X, Y on M̃ with hX, Y i = hJX, Y i = 0, λ(X, Y ) is a
constant function.
Definition 10.6. A generalized complex space form is an RK-manifold of
constant holomorphic sectional curvature and of constant type.
A generalized complex space form of constant holomorphic sectional
curvature c and constant type α is denoted by M̃ (c, α). Every complex
space form is obviously a general complex space form, but the converse
is not true. The 6-sphere S 6 endowed with the standard nearly Kähler
structure is an example of generalized complex space form which is not a
complex space form.
The curvature tensor R̃ of a general complex space form M̃ (c, α) has
the following expression [Vanhecke (1975)]:
c + 3α
R̃(X, Y )Z = {hY, Zi X − hX, Zi Y }
4 (10.64)
c−α
+ {hX, JZi JY − hY, JZi JX + 2 hX, JY i JZ}.
4
For warped product in a generalized complex space form, we have the
following result from [Mihai (2005c)].
Theorem 10.8. Let φ : N1 ×f N2 → M̃ (c, α) be an isometric immersion of
a warped product into a generalized complex space form. Put ni = dim Ni
for i = 1, 2.
(1) If c ≤ α, then we have
∆f (n1 + n2 )2 2 (c + 3α)n1
≤ H + . (10.65)
f 4n2 4
The equality sign of (10.65) holds identically if and only if φ is a mixed
totally geodesic immersion with n1 H1 = n2 H2 (in addition J(T N1 ) ⊥
T N2 for c < α), where Hi is the partial mean curvature vector of Ni .
(2) If c < α, then we have
∆f (n1 + n2 )2 2 (c + 3α)n1 3(c − α)
≤ H + + ||P ||2 . (10.66)
f 4n2 4 8
The equality sign of (10.66) holds identically if and only if φ is a mixed
totally geodesic immersions, n1 H1 = n2 H2 , and N1 , N2 are totally real
submanifolds.
Chapter 11
Warped Product Submanifolds of

Kähler Manifolds
11.1 Warped product CR-submanifolds
A warped product N1 ×f N2 is called proper if its warped function f is a

non-constant function.
For warped product CR-submanifolds in any Kähler manifold we have
the following result from [Chen (2001a)].
Theorem 11.1. If N ⊥ ×f N T is a warped product submanifold of a Kähler

manifold M̃ such that N ⊥ is a totally real submanifold and N T is a complex
submanifold of M̃ , then N ⊥ ×f N T is always non-proper, i.e., the warping
function f is constant.
Proof. Let M =: N ⊥ ×f N T be a warped product CR-submanifold in

a Kähler manifold M̃ such that N ⊥ is a totally real submanifold and N T
is a complex submanifold of M̃ . Since the metric tensor of M is given by
g = gN ⊥ + f 2 gN T , N ⊥ is a totally geodesic submanifold of M . Thus, for
any vector fields Z, W tangent to N ⊥ and X tangent to N T , we have
h∇Z W, Xi = 0. (11.1)
Since the ambient space M̃ is Kählerian, the formulas of Gauss and
Weingarten give
−AJW Z + DZ (JW ) = J(∇Z W ) + Jσ(Z, W ). (11.2)
Hence, by taking the inner product of (11.2) with JX, we find
hAJW Z, JXi = − h∇Z W, Xi . (11.3)
Also, by combining (11.1) and (11.3) we obtain

σ(H, H⊥ ), JH⊥ = 0, (11.4)
229
where H and H⊥ are the complex and totally real distributions on M ,

respectively.
From Proposition 3.1(2), we have
∇X Z = ∇Z X = (Z ln f )X, (11.5)
⊥ T
for any vector fields X ∈ H and Z ∈ H . Thus, if we denote by σ and
AT the second fundamental form and the shape operator of N T in M , then
we obtain from the formulas of Gauss and Weingarten that

T

σ (X, Y ), Z = ATZ X, Y = − h∇X Z, Y i = −(Z ln f ) hX, Y i (11.6)
for any X, Y ∈ H and Z ∈ H⊥ . Hence we find
σ T (X, Y ) = −∇(ln f ) hX, Y i , (11.7)
where ∇(ln f ) is the gradient of ln f . It follows from (11.7) that N T is
totally umbilical in M .
Let σ̂ denote the second fundamental form of N T in M̃ . Then we have
σ̂(X, Y ) = σ T (X, Y ) + σ(X, Y ). (11.8)
By applying (11.7) and (11.8) we find
hσ̂(X, X), Zi = −Z(ln f ) hX, Xi . (11.9)
T
Since N is a complex submanifold of M̃ , we also have
σ̂(X, JY ) = σ̂(JX, Y ) = J σ̂(X, Y ). (11.10)
Hence, after combining (11.9) and (11.10), we derive that
hσ̂(X, X), Zi = − hσ̂(JX, JX), Zi = Z(ln f ) hX, Xi . (11.11)
It follows from (11.9) and (11.11) that Z(ln f ) = 0. Thus we find from
(11.6) and (11.8) that

hσ̂(X, Y ), Zi = σ T (X, Y ), Z = 0 (11.12)
for any X, Y ∈ H and Z ∈ H⊥ . Therefore, by applying (11.8), (11.10) and
(11.12), we get
hσ(X, Y ), JZi = hσ̂(X, Y ), JZi = − hσ̂(X, JY ), Zi = 0. (11.13)
Consequently, we obtain

σ(H, H), JH⊥ = 0. (11.14)
Conditions (11.4) and (11.14) imply AJH⊥ H = 0. Therefore, after applying
Lemma 9.3, we may conclude that N ⊥ ×f N T is non-proper.
Remark 11.1. Warped product lightlike submanifolds in pseudo-
Riemannian manifolds were defined in [Duggal and Sahin (2010)]. A result
analogous to Theorem 11.1 for warped product lightlike submanifolds was
proved in [Duggal and Sahin (2010)].
Warped Product Submanifolds of Kähler Manifolds 231
11.2 CR-warped products and their characterization
Theorem 11.1 shows that there exist no warped product CR-submanifolds

of the form N ⊥ ×f N T such that N T is a complex submanifold and N ⊥ is
a totally real submanifold of M̃ besides CR-products. For this reason the
author introduced the following definition in [Chen (2001a)].
Definition 11.1. A warped product submanifold in a Kähler manifold M̃

is called a CR-warped products if it is of the form N T ×f N ⊥ , where N T
is a complex submanifold and N ⊥ is a totally real submanifold of M̃ . A
CR-warped product is called trivial if its warping function is constant.
In the following, by a CR-warped product we mean a non-trivial one

unless mentioned otherwise.
Definition 11.2. A warped product N1 ×f N2⊥ in an almost Hermitian

manifold M̃ is called a generic warped product if N1 is a generic submanifold
and N2⊥ is a totally real submanifold of M̃ . A generic warped product is
called trivial if its warping function f is constant.
A trivial generic warped product N1 ×f N ⊥ is a generic product N1 ×Nf⊥ ,

where Nf⊥ is the manifold with the metric f 2 gN ⊥ homothetic to the original
metric gN ⊥ on N ⊥ .
For CR-warped products in Kähler manifolds we have the following.
Lemma 11.1. For a CR-warped product N T ×f N ⊥ in a Kähler manifold

M̃ , we have

(1) σ(H, H), JH⊥ = 0;
(2) ∇X Z = ∇Z X = (X ln f )Z;
(3) hσ(JX, Z), JW i = (X ln f ) hZ, W i;
(4)
DX (JZ) = J∇X Z, whenever σ(H, H⊥ ) ⊂ JH⊥ ;
(5) σ(H, H⊥ ), JH⊥ = 0 if and only if N T ×f N ⊥ is a trivial CR-warped
product,
(6) σ(JX, Z) = (X ln f )JZ + Jσν (X, Z),
(7) σν (JX, Z) = Jσν (X, Z),
where X, Y are vector fields tangent to N T , Z, W are vector fields tangent
to N ⊥ and σν is the ν-component of the second fundamental form σ.
Proof. Since M̃ is Kählerian, we have

J∇X Z + Jσ(X, Z) = −AJZ X + DX JZ, (11.15)
for any vector fields X, Y tangent to N T and Z tangent to N ⊥ . Thus, by

taking the inner product of (11.15) with JY , we find
h∇X Z, Y i = − hAJZ X, JY i = − hσ(X, JY ), JZi . (11.16)
On the other hand, since M = N T ×f N ⊥ is a warped product, N T
is totally geodesic in M . Thus, we also have h∇X Z, Y i = 0. Combining
this with (11.16), we obtain statement (1). Statement (2) is nothing but
Proposition 3.1(2).
By applying Proposition 3.1(2), Lemma 9.2 and statement (2), we get
hσ( JX, Z), JW i = − hJAJW Z, Xi = − h∇Z W, Xi
(11.17)
= h∇Z X, W i = (X ln f ) hZ, W i
for X tangent to N T and Z, W tangent to N ⊥ . This proves statement (3).

Since N T is a totally geodesic submanifold of M , ∇X Z ∈ H⊥ holds.
Thus J∇X Z ∈ JH⊥ . On the other hand, it follows from σ(H, H⊥ ) ⊂ JH⊥
that Jσ(X, Z) ∈ T M . Therefore, by applying (11.15), we obtain statement
(4). Statement (5) follows from statement (3) and the definition of trivial
CR-warped products.
From statement (2) we have
˜ + ZJX = (X ln f )JZ + Jσ(X, Z),
∇Z (JX) + σ)Z, JX) = ∇
which implies statement (6). Finally, statement (7) is an easy consequence
of statement (6).
The following is a simple characterization of CR-warped products.
Theorem 11.2. A proper CR-submanifold M of a Kähler manifold M̃ is

locally a CR-warped product if and only if
AJZ X = ((JX)µ)Z, X ∈ H, Z ∈ H⊥ , (11.18)
for some function µ on M satisfying W µ = 0 for W ∈ H⊥ .
Proof. If M is a CR-warped product N T ×f N ⊥ in M̃ , then statements

(1) and (2) of Lemma 11.1 imply
AJZ X = −((JX) ln f )Z, X ∈ H, Z ∈ H⊥ .
Since f is a function on N T , we also have W (ln f ) = 0 for W ∈ H⊥ .
Conversely, assume that M is a proper CR-submanifold of M̃ satisfying
AJZ X = ((JX)µ)Z, X ∈ H, Z ∈ H⊥ , (11.19)
for some function µ with W µ = 0, W ∈ H⊥ . Then we have

σ(H, H), JH⊥ = 0, (11.20)
hσ(JX, Z), JW i = −(Xµ) hZ, W i , (11.21)
⊥
for X, Y ∈ H and Z, W ∈ H . Thus Proposition 9.1 and (11.20) imply
that the holomorphic distribution H is integrable and its leaves are totally
geodesic in M .
On the other hand, it follows from Lemma 9.2 and (11.21) that
h∇Z X, W i = − h∇Z W, Xi = − hJAJW Z, Xi
(11.22)
= hσ(JX, Z), JW i = −(Xµ) hZ, W i
for X ∈ H and Z, W ∈ H⊥ . Since the totally real distribution H⊥ of a
CR-submanifold of a Kähler manifold is always integrable by Theorem 9.2,
(11.22) and the condition W µ = 0, W ∈ H⊥ , imply that each leaf of H⊥ is
an extrinsic sphere in M . Hence, by applying a result of [Hiepko (1979)],
we conclude that M is locally the warped product N T ×f N ⊥ of a complex
submanifold and a totally real submanifold N ⊥ of M , where N T is a leaf
of H and N ⊥ is a leaf of H⊥ and f is a certain warping function. This
completes the proof of the theorem.
11.3 Examples of CR-warped products
We provide some examples of CR-warped products in Kähler manifolds.

Example 11.1. Put C∗ = C − {0}. Let ι : C∗ → C∗ be the identity map,
ϕ : F → Ck a complex submanifold and w : Q → SrN ⊂ EN +1 an isometric
immersion of a Riemannian p-manifold Q into the hypersphere SrN of EN +1
with radius r and centered at the origin.
Consider the map:
ψ : C∗ × F × Q → CN +1+k
defined by
ψ(z, u, v) 7→ (zw(v), ϕ(u)), z ∈ C∗ , u ∈ F, v ∈ Q.
It is direct to verify that the induced metric on C∗ × F × Q is the warped
metric given by
g = gN T ⊕ f 2 gN ⊥ ,
with f = |z|, where N T = C∗ × F and N ⊥ = Q.
It is direct to verify that N T is immersed in CN +k+1 as a complex
submanifold and N ⊥ as a totally real submanifold. Therefore, N T ×|z| N ⊥
is isometrically immersed in CN +k+1 as a CR-warped product submanifold.
Example 11.2. In Example 11.1, if we choose F to be the complex (h−1)-

plane Ch−1 , Q the unit hypersphere S p (1) ⊂ Ep+1 , ϕ the identity map of
Ch−1 and w the standard embedding of S p+1 (1), then the map
ψ̂ : C∗ × Ch−1 × S p+1 (1) → Ch+p ,
defined by
ψ(z, u, v) 7→ (zw(v), ϕ(u))
h−1
for z ∈ C∗ , u ∈ C , v ∈ S p (1), is a CR-warped product in Ch+p . Also,
by a straightforward computation, we know that this CR-warped product
of Ch+p satisfying the equality:
||σ||2 = 2p|∇(ln f )|2 .
Example 11.3. Put E2∗ = E2 − {(0, 0)}. Let

φ : T p = S 1 (1) × · · · × S 1 (1) → Cp
denote the standard Lagrangian embedding of T p into Cp . Then the map
Ψ : E2∗ × T p → Cp defined by
Ψ(x, y, u) 7→ (xφ(u), yφ(u)), (x, y) ∈ E2∗ , u ∈ T p ,
p
is a CR-warped product in Cp with warping function f = x2 + y 2 .
Example 11.4. Let Cm be the complex Euclidean m-space equipped with

a natural coordinate system {z1 , . . . , zm }. Put Cm m
∗ = C − {0}.
Suppose that z : N T → Cm ∗ ⊂ C
m
is a Kähler immersion of a Kähler
manifold of complex dimension h into Cm ∗ and w : N
⊥
→ S q (1) ⊂ Eq+1
is an isometric immersion of a Riemannian p-manifold into the unit hyper-
sphere S q (1) centered at the origin.
For a natural number α ≤ h, we define a map
(α)
Ŝhp : N T × N ⊥ → Cm × S q (1) → Cm+αq (11.23)
by
(α)
Ŝhp (u, v) = w1 (v)z1 (u), . . . , wq+1 (v)z1 (u), . . . ,
(11.24)
w1 (v)zα (u), . . . , wq+1 (v)zα (u), zα+1 (u), . . . , zm (u)
for u ∈ N T and v ∈ N ⊥ . Then (11.24) induces an isometric immersion
(α)
Ŝhp : N T ×f N ⊥ → Cm+αq (11.25)
qP
α
from the warped product N T ×f N ⊥ with f = 2
j=1 |zj (u)| into C
m+αq
as a CR-warped product.
We put

Ch(α) = (z1 , . . . , zh ) ∈ Ch : |z1 |2 + · · · + |zα |2 6= 0 . (11.26)
If z : N T = Ch∗ ֒→ Ch and w : S p (1) ֒→ Ep+1 are the inclusion maps, then
(11.25) is the map
(α)
Shp : Ch(α) ×f S p (1) → Ch+αp (11.27)
defined by
(α)
Shp (z, w) = w1 z1 , . . . , wp+1 z1 , . . . , (11.28)

w1 zα , . . . , wp+1 zα , zα+1 , . . . , zh
Pp+1
for z = (z1 , . . . , zh ) ∈ Ch , w = (w1 , . . . , wp+1 ) ∈ S p (1) with j=1 wj2 = 1.
The warping function of the CR-warped product Chα ×f S p (1) is
p
f = |z1 |2 + · · · + |zα |2 .
(α)
Definition 11.3. The embedding Shp defined by (11.28) is called a partial
Segre embedding. Also a CR-warped product defined via (11.25) is called a
partial Segre CR-warped product.
h
Remark 11.2. The map ψhp = Shp defined by
ψhp (z, x) = (z1 x1 , . . . , z1 xp , . . . , zh x1 , . . . , zh xp ) (11.29)
for z = (z1 , . . . , zh ) ∈ Ch∗ and x = (x1 , . . . , xp ) ∈ Ep∗ . is called the Euclidean

Segre map.
(α)
Remark 11.3. The map Shp defined in (11.27) and (11.28) can be defined
exactly the same way if Ch(α) in (11.27) were replaced by

Ch1,(α) = (z1 , . . . , zh ) ∈ Ch1 : |z1 |2 + · · · + |zα |2 6= 0 .
We also call such a map a partial Segre embedding.
11.4 A general inequality for CR-warped products
For CR-warped products in an arbitrary Kähler manifold, we have the

following result from [Chen (2001a)].
Theorem 11.3. Let M = N T ×f N ⊥ be a CR-warped product in a Kähler

manifold M̃ . Then we have
(1) The second fundamental form σ of M satisfies

||σ||2 ≥ 2p |∇(ln f )|2 , (11.30)
where ∇ ln f is the gradient of ln f and p is the dimension of N ⊥ .
(2) If the equality sign of (11.30) holds identically, then N T is a totally
geodesic submanifold and N ⊥ is a totally umbilical submanifold of M̃ .
Moreover, M is a minimal submanifold in M̃ .
(3) When M is anti-holomorphic and p > 1. The equality sign of (11.30)
holds identically if and only if N ⊥ is a totally umbilical submanifold of
M̃ .
(4) If M is a real hypersurface, then the equality sign of (11.30) holds
identically if and only if the characteristic vector field Jξ of M is a
principal vector field with zero as its principal curvature. Also, in this
case, the equality sign of (11.30) holds identically if and only if M is
a minimal hypersurface in M̃ .
Proof. From Lemma 11.1, we have

hσ(Z, JX), JZi = X ln f (11.31)
⊥
for any unit vector Z ∈ H . By applying (11.31) we have inequality (11.30)
immediately.
For any vector fields X ∈ H and Z, W ∈ H⊥ , Lemma 9.2 implies
h∇W Z, Xi = hJAJZ W, Xi = − hσ(JX, W ), JZi . (11.32)
Hence, by using Lemma 11.1(2) and (11.32), we find
h∇W Z, Xi = −(X ln f ) hZ, W i . (11.33)
On the other hand, if we denote by σ ⊥ the second fundamental form of
N in M = N T ×f N ⊥ , we get
⊥

⊥
σ (Z, W ), X = h∇W Z, Xi . (11.34)
Combining (11.33) and (11.34) yields
σ ⊥ (Z, W ) = − hZ, W i ∇ ln f. (11.35)
Now, assume that the equality case of (11.30) holds identically. Then
we obtain from (11.31) that
σ(H, H) = 0, σ(H⊥ , H⊥ ) = 0, σ(H, H⊥ ) ⊂ JH⊥ . (11.36)
Since N T is a totally geodesic submanifold in M , the first condition in
(11.36) implies that N T is totally geodesic in M̃ .
On the other hand, (11.35) shows that N ⊥ is totally umbilical in M .

Therefore the second condition in (11.36) implies that N ⊥ is also totally
umbilical in M̃ .
Moreover, it follows from (11.36) that M is minimal in M̃ . This proves
statement (2).
To prove statements (3) and (4) let us assume that M is an anti-
holomorphic CR-warped product in M̃ . Then we find from Lemma 11.1(1)
that
σ(H, H) = 0. (11.37)
⊥
If N is totally umbilical in M̃ , then there exists a normal vector field
Ĥ of N ⊥ in M̃ such that the second fundamental form σ̂ of N ⊥ in M̃
satisfies
σ̂(Z, W ) = hZ, W i Ĥ, (11.38)
⊥
for Z, W tangent to N . Since
σ̂(Z, W ) = σ ⊥ (Z, W ) + σ(Z, W ),
(11.38) implies that there is a normal vector field η such that
σ(Z, W ) = hZ, W i η. (11.39)
⊥
Therefore, for each unit vector W ∈ H and each unit vector Z ∈ H⊥
perpendicular to W , we have
hη, JW i = hσ(Z, Z), JW i
= hσ(Z, W ), JZi (11.40)
= hZ, W i hη, JZi = 0,
where we have applied Lemma 9.2. Since M is assumed to be anti-
holomorphic, (11.40) implies either p = 1 or
σ(H⊥ , H⊥ ) = 0. (11.41)
Therefore (11.31), (11.37) and (11.41) imply the equality case of (11.30)
holding whenever p > 1.
When p = 1, M is a real hypersurface of M̃ . In this case, the charac-
teristic vector field Jξ is a principal vector field with zero as its principal
curvature if and only if (11.41) holds. So, in this case we also have equal-
ity case of (11.30) if the characteristic vector field Jξ is a principal vector
field with zero as its principal curvature. Also, from the first condition
in (11.36), we also know that condition (11.41) holds if and only if M is
minimal in M̃ .
By applying statement (2), the converse is easy to verify.
For doubly warped product CR-submanifolds, we have the following

non-existence result from [Sahin (2007a)].
Theorem 11.4. There exist no doubly warped product CR-submanifolds
in a Kähler manifold M̃ in the form N T ×(f,b) N ⊥ such that N T is a
complex submanifold and N ⊥ is a totally real submanifold of M̃ which are
not (singly) warped product CR-submanifolds.
11.5 Twisted product CR-submanifolds
In this section we present some results on CR-submanifolds in a Kähler

manifold which are twisted products of complex submanifolds and totally
real submanifolds.
The following two results from [Chen (2000c)] are natural extension of
Theorem 11.1 and Theorem 11.3, respectively.
Theorem 11.5. If N ⊥ ×λ N T is a twisted product CR-submanifold of a
Kähler manifold M̃ such that N ⊥ is a totally real submanifold and N T is
a complex submanifold of M̃ , then M is a CR-product.
Theorem 11.6. Let N T ×λ N ⊥ be a twisted product CR-submanifold of a
Kähler manifold M̃ such that N ⊥ is a totally real submanifold and N T is
a complex submanifold of M̃ . Then we have
(1) The second fundamental form σ of M in M̃ satisfies
||σ||2 ≥ 2 p |∇T (ln λ)|2 , (11.42)
T T
where ∇ (ln λ) is the N -component of the gradient of ln λ and p is
the dimension of N ⊥ .
(2) If ||σ||2 = 2p |∇T ln λ|2 holds identically, then N T is a totally geodesic
submanifold and N ⊥ is a totally umbilical submanifold of M̃ .
(3) If M is anti-holomorphic in M̃ and dim N ⊥ > 1, then we have
||σ||2 = 2p |∇T ln λ|2 identically if and only if N T is a totally geodesic
submanifold and N ⊥ is a totally umbilical submanifold of M̃ .
Since the proofs of these two results are similar to the proofs of Theorem
11.1 and Theorem 11.3, so we omit their proofs.
Theorem 11.7. [Chen (2000c)] Let M = N T ×λ N ⊥ be a twisted product
CR-submanifold of a Kähler manifold M̃ such that N ⊥ is a totally real
submanifold and N T is a complex submanifold of M̃ . If M is mixed totally
geodesic, then we have
(1) The twisting function λ is a function on N ⊥ .

(2) N T ×Nλ⊥ is a CR-product, where Nλ⊥ denotes the manifold N ⊥ equipped
λ 2
with the metric gN ⊥ = λ gN ⊥ .
Proof. Let M = N T ×λ N ⊥ be a twisted product CR-submanifold of a

Kähler manifold M̃ such that N ⊥ is a totally real submanifold and N T is a
complex submanifold of M̃ . Since M̃ is Kählerian, ∇J˜ = 0. Thus, for any
⊥
vector fields X, Y ∈ H and Z ∈ H , we have
J∇X Z + Jσ(X, Z) = −AJZ X + DX JZ.
By taking the inner product of this equation with JY , we find
h∇X Z, Y i = − hAJZ X, JY i = − hσ(X, JY ), JZi . (11.43)
Since M = N T ×λ N ⊥ is a twisted product, N T is a totally geodesic
submanifold of M . Therefore, we also have h∇X Z, Y i = 0. Combining this
equation with (11.43), we obtain

σ(H, H), JH⊥ = 0, (11.44)
where H and H⊥ are the holomorphic distribution and the totally real
distribution of M , respectively.
On the other hand, since the Lie bracket [X, Z] = 0 for any vector field
X ∈ H and Z ∈ H⊥ , we have
∇X V = ∇V X. (11.45)
Thus, for any vector fields X, Y ∈ H and Z ∈ H⊥ , we get
Z hX, Y i = Z(λ2 (X, Y )) = 2λ(Zλ)(X, Y ) = 2(Z ln λ) hX, Y i , (11.46)
where ( , ) is the inner product of (N T , gN T ) such that
(X, Y ) = λ−2 hX, Y i .
Also, it follows from (11.46) and hX, Zi = 0 that
Z hX, Y i = h∇Z X, Y i + hX, ∇Z Y i
= h∇X Z, Y i + hX, ∇Y Zi (11.47)
= h∇X Z, Y i − h∇Y X, Zi .
If we denote by σ T and AT the second fundamental form and the shape
operator of N T in M , then from the formulas of Gauss and Weingarten
and from (11.47) we obtain

Z hX, Y i = −2 σ T (X, Y ), Z . (11.48)

σ T (X, Y ) = −∇⊥ (ln λ) hX, Y i , (11.49)
where ∇⊥ (ln λ) denotes the N ⊥ -component of the gradient ∇(ln λ) of ln λ.
By applying an argument similar to the proof of (11.50), we also know
that the second fundamental form σ ⊥ of N ⊥ in M is given by
σ ⊥ (Z, W ) = −∇T (ln λ) hZ, W i , (11.50)
for vector fields Z, W ∈ H⊥ . From (11.50) we get
h∇Z W, Xi = −X(ln λ) hZ, W i . (11.51)
On the other hand, from Lemma 9.2, we also have
hJAJW Z, Xi = h∇Z W, Xi . (11.52)
Hence, by combining (11.51) and (11.52) we have
hσ(JX, Z), JW i = − hJAJW Z, Xi = X(ln λ) hZ, W i (11.53)
for X ∈ H and Z, W ∈ H⊥ . Then we obtain from (11.53) that
hσ(JX, Z), JW i = − hJAJW Z, Xi = X(ln λ) hZ, W i (11.54)
for X in H and Z, W in H⊥ . Therefore, if M is mixed totally geodesic,
we have X(ln λ) = 0 for any vector X tangent to N T . Hence the twisting
function λ of the twisted product depends only on the second factor N ⊥ .
Hence λ is a function on N ⊥ . Clearly, in this case the twisted product
N T ×λ N ⊥ is isometric to the Riemannian product N T ×Nλ⊥ . Consequently,
λ ⊥
with respect to the metric gN ⊥ on N , N T × Nλ⊥ becomes a CR-product
of M̃ .
Now, we provide some examples of twisted product CR-submanifolds
which are not warped product CR-submanifolds.
Example 11.5. Let z : N T → Cm be a complex submanifold of a complex

Euclidean m-space Cm and w : N1⊥ → Cℓ be a totally real submanifold
such that the image of N T × N1⊥ under the product immersion ψ = (z, w)
does not contain the origin (0, 0) of Cm ⊕ Cℓ .
Let j : N2⊥ → S q−1 ⊂ Eq be an isometric immersion of a Riemannian
manifold N2⊥ into the unit hypersphere S q−1 of Eq centered at the origin.
Consider the map
φ = (z, w) ⊗ j : N T × N1⊥ × N2⊥ → (Cm ⊕ Cℓ ) ⊗ Eq (11.55)
defined by
φ(p1 , p2 , p3 ) = (z(p1 ), z(p2 )) ⊗ j(p3 ),
(11.56)
∀p1 ∈ N T , p2 ∈ N1⊥ , p3 ∈ N2⊥ .
On (Cm ⊕ Cℓ ) ⊗ Eq we define a complex structure J by
√
J((B, E) ⊗ F ) = (iB, iE) ⊗ F, i = −1, (11.57)
for any B ∈ Cm , E ∈ Cℓ and F ∈ Eq . Then (Cm ⊕ Cℓ ) ⊗ Eq becomes a
complex Euclidean (m + ℓ)q-space C(m+ℓ)q .
We put N ⊥ = N1⊥ × N2⊥ . We denote by |z| the distance function from
the origin of Cm to the position of N T in Cm via z; and by |w| the distance
ℓ ⊥ ℓ
function from the origin
p of C to the position of N1 in C via w. We define
a function λ by λ = |z|2 + |w|2 . Then λ > 0 is a differentiable function
on N T × N ⊥ , which depends on both N T and N ⊥ = N1⊥ × N2⊥ .
Let M denote the twisted product N T ×λ N ⊥ with twisting function λ.
Clearly, M is not a warped product.
For the twisted product N T ×λ N ⊥ in C(m+ℓ)q defined in Example 11.5,

we have the following.
Proposition 11.1. The map

φ = (z, w) ⊗ j : N T ×λ (N1⊥ × N2⊥ ) → C(m+ℓ)q (11.58)
defined by (11.56) satisfies the following properties:
(a) φ = (z, w) ⊗ j : N T ×λ (N1⊥ × N2⊥ ) → C(m+ℓ)q is an isometric immer-
sion.
(b) φ = (z, w) ⊗ j : N T ×λ (N1⊥ × N2⊥ ) → C(m+ℓ)q is a twisted product
CR-submanifold such that N T is a complex submanifold and N ⊥ is a
totally real submanifold of C(m+ℓ)q .
Proof. For vector fields U, V on N T × N1⊥ and vector fields Z, W tangent

to N2⊥ , we have
Xφ = X ⊗ j, Zφ = z ⊗ Z. (11.59)
It follows from (11.59) that the induced metric on N T ×N ⊥ via φ is given by
g = g1 + λ2 g2 , where g1 and g2 are the metrics on N T and N ⊥ = N1⊥ × N2⊥ ,
respectively, and with λ2 = |z|2 + |w|2 . Thus
φ = (z, w) ⊗ j : N T ×λ N ⊥ → C(m+ℓ)q
is an isometric immersion.
The differentiable function λ is a function which depends on both N T

and N1⊥ . Clearly, if dim N1⊥ > 0, the twisted product manifold N T ×λ N ⊥
is not a warped product.
Let X, Z1 and Z2 be vectors tangent to N T , N1⊥ and N2⊥ respectively.
Then we have
φ∗ (X) = Xφ = X ⊗ j, (11.60)
φ∗ (Z1 ) = Z1 φ = Z1 ⊗ j, (11.61)
φ∗ (Z2 ) = (z, w) ⊗ Z2 . (11.62)
Since z : N T → Cm is a complex submanifold, (11.50) implies J(φ∗ (X))

is tangent to M = N T ×λ N ⊥ . Moreover, since w : N1⊥ → Cℓ is a totally
real immersion, it follows from (11.50), (11.59) and (11.61) that J(φ∗ (Z1 ))
is normal to M .
Finally, from (11.50), (11.59) and (11.62), we also know that J(φ∗ (Z2 ))
is also normal to M . Consequently, these prove that (11.58) is a twisted
product CR-submanifold.
Proposition 11.1 shows that there do exist many twisted product CR-
submanifolds N T ×λ N ⊥ such that N T are complex submanifolds and
N ⊥ are totally real submanifolds. Moreover, such twisted product CR-
submanifolds are not warped product CR-submanifolds.
For doubly twisted product CR-submanifolds, we have the following
non-existence result from [Sahin (2007a)].
Theorem 11.8. There do not exist doubly twisted product CR-submanifold

in a Kähler manifold M̃ which are not (singly) twisted product CR-
submanifold in the form N T ×(f,b) N ⊥ such that N T is a complex sub-
manifold and N ⊥ is a totally real submanifold of M̃ .
11.6 Warped product submanifolds with a holomorphic

factor
The next non-existence result is an extension of Theorem 11.1 and Theorem

3.1 of [Sahin (2006a)].
Theorem 11.9. [Al-Solamy and Khan (2008b); Khan et al. (2008)] There
do not exist non-trivial warped product submanifolds N 0 ×f N T in a Kähler
manifold M̃ such that N T is a complex submanifold and N 0 is a proper
purely real submanifold of M̃ .
Proof. Let N 0 ×f N T be a warped product submanifold of a Kähler

manifold M̃ such that N T is a complex submanifold and N 0 is a proper
purely real submanifold of M̃ . Then Proposition 3.1(2) gives
∇X Z = ∇Z X = (Z ln f )X (11.63)
for X ∈ X(N T ) and Z ∈ X(N 0 ). Since M̃ is Kählerian, it follows from
(6.38), (11.63) and formulas of Gauss and Weingarten that
˜ JX JZ i = hJX, ∇JX P Zi − hσ(JX, JX), F Zi .
0 = h JX, ∇
Combining this with (11.63) yields
hσ(JX, JX), F Zi = (P Z ln f )|X|2 . (11.64)
Now, by applying Proposition 7.3(i), (6.42) and (11.63), we find
(P Z ln f )JX + (Z ln f )X = AF Z JX + tσ(JX, Z).
Taking the inner product of this with Y ∈ X(N T ) we obtain
(Z ln f ) hX, Y i + (P Z ln f ) hJX, Y i = hσ(JX, Y ), F Zi . (11.65)
Interchanging X and Y in (11.65) we also have
(Z ln f ) hX, Y i + (P Z ln f ) hX, JY i = hσ(X, JY ), F Zi . (11.66)
Adding (11.65) and (11.66) and using Proposition 8.1 give
(Z ln f ) hX, Y i = hσ(JX, Y ), F Zi . (11.67)
In particular, we have
hσ(JX, JX), F Zi = 0. (11.68)
P Z ln f = 0. (11.69)
0 0
for each Z tangent to N . Because N is assumed to be proper purely real,
we conclude from (11.69) that f is a constant. This is a contradiction.
Contrast to Theorem 11.9, the next example shows that there do exist
non-trivial warped product submanifolds of the form N T ×f N 0 in a Kähler
manifold M̃ , where N T is a complex submanifold and N 0 is a proper purely
real submanifold of M̃ .
Example 11.6. Consider the 4-dimensional submanifold N in the complex

Euclidean 5-space C5 defined by
φ = ((t + is) cos u, (t + is) cos v, (t + is) sin u, (t + is) sin v, u + iv)
for (t, s) 6= (0, 0) and u, v ∈ (0, π2 ). Let us put

X1 = (cos u, cos v, sin u, sin v, 0),
X2 = (i cos u, i cos v, i sin u, i sin v, 0),
Z1 = (−(t + is) sin u, 0, (t + is) cos u, 1),
Z2 = (0, −(t + is) sin v, 0, (t + is) cos v, u + i)
and we also put H = Span{X1 , X2 } and Hθ = Span{Z1 , Z1 }. Then H is a
holomorphic distribution and Hθ is a pointwise slant distribution with the
slant function

−1 1
θ = cos .
1 + t2 + s 2
Thus N is a pointwise semi-slant submanifold of C5 . Clearly, H and Hθ
are integrable. If N T denotes an integrable submanifold of H and N θ an
integral manifolds of Hθ , then the metric tensor g of N is
g = gN T + (1 + t2 + s2 )gN θ ,
with gN T = 2ds2 + 2dt2 and gN θ = du2 + dv 2 . Thus N is a non-trivial
T θ
p product submanifold of the form N θ ×f N with warping function
warped
2 2
f = (1 + t + s ). Since (t, s) 6= (0, 0), N is proper purely real.
11.7 Warped product hemi-slant submanifolds
Definition 11.4. A hemi-slant submanifold of a Kähler manifold is a

submanifold M with two orthogonal distributions H⊥ and Hθ such that
T M = H⊥ ⊕ Hθ , JH⊥ ⊂ T ⊥ M and Hθ is θ-slant.
Remark 11.4. Hemi-slant submanifolds were first defined in [Carriazo

(2000)] under the name of anti-slant submanifolds.
The next non-existence result was proved in [Sahin (2009a)] is a natural

extension of Theorem 11.1.
Theorem 11.10. There exist no warped product hemi-slant submanifolds

N ⊥ ×f N θ in a Kähler manifold M̃ such that N ⊥ is a totally real subman-
ifold and N θ is a proper slant submanifold of M̃ .
Proof. From ∇J˜ = 0 and (2.13), we have hAJZ T X, Xi = h ∇

˜ P X Z, JX i
θ ⊥
for X ∈ H and Z ∈ H . Then from (2.5) and (8.4) we get
hAJZ P X, Xi = h∇P X Z, P Xi + hσ(P X, Z), F Xi .
Using Proposition 3.1(2) we derive that

hAJZ P X, Xi = Z(ln f ) hP X, P Xi + hσ(P X, Z), F Xi .
Thus, from (2.14) and |P X|2 = (cos2 θ)|X|2 , we have
hσ(P X, X), JZi = Z(ln f )(cos2 θ) hX, Xi + hσ(P X, Z), F Xi . (11.70)
By substituting X by P X in (11.70) and using |F X|2 = (sin2 θ)|X|2 and
Lemma 11.2 we arrive at
hσ(X, P X), JZi = −(cos2 θ)Z(ln f ) hX, Xi + hσ(X, Z), F P Xi . (11.71)
On the other hand, from (2.13) we have
˜ Z F X, P X i .
hAF X Z, P Xi = − h ∇
˜ = 0 and (8.4) we get
Then using ∇J
˜ Z X, JP X i + h∇Z P X, P Xi .
hAF X Z, P Xi = h ∇
Using (2.5), (8.4), and Proposition 3.1(2) we obtain

hAF X Z, P Xi = ∇Z X, P 2 X + hσ(Z, X), F P Xi
+ Z(ln f ) hP X, P Xi .
Here, considering |P X| = (cos2 θ)|X|2 and Lemma 11.2, we arrive at
2
g(AF X W, T X) = − cos2 θ g(∇W X, X) + g(h(W, X), F T X)

+ cos2 θ W (ln f )g(X, X).
Then, using again Proposition 3.1(2) and (2.14), we derive
hσ(P X, Z), F Xi = hσ(Z, X), F P Xi . (11.72)
Thus from (11.70), (11.71) and (11.72) we conclude 2 cos2 θ Z(ln f ) = 0.
Since N θ is proper slant and Riemannian we obtain Z(ln f ) = 0, hence f
is constant.
In contrast, there do exist warped product hemi-slant submanifolds in
a Kähler manifold of the form N θ ×f N ⊥ , where N θ is a proper slant sub-
manifold and N ⊥ is a totally real submanifold. Clearly, such submanifolds
are spacial generic warped submanifolds.
The following two examples were given in [Sahin (2009a)].
Example 11.7. Consider a submanifold N in C3 defined by

ψ(x, y, z) = ((x + iky) cos z, (x + iky) sin z, y + ix), (11.73)
where k is constant 6= 0, 1. Clearly, the tangent bundle T M is spanned by

Z1 , Z2 and Z3 , where
Z1 = (cos z, sin z, i),

Z2 = (ik cos z, ik sin z, 1),
Z3 = (−(x + iky) sin z, (x + iky) cos z, 0).
Then H⊥ = Span{Z3 } is a totally real distribution and Hθ = Span{Z1 , Z2 }

is a slant distribution with the slant angle
!
−1 1−k
θ = cos p .
2(1 + k 2 )
It is direct to verify that Hθ is integrable. We denote the integral manifolds

of H⊥ and Hθ by N ⊥ and N θ , respectively. Then the metric tensor g of
N is given by g = gN θ + (x2 + k 2 y 2 )gN ⊥ . Thus N is a warped product
submanifold of C3 of the form N θ ×f N ⊥ with a non-constant warping
function.
The following is an example of mixed totally geodesic warped product

submanifold of the form: N θ ×f N ⊥ .
Example 11.8. Let N be a submanifold of C4 given by
χ(u, v, w) = (u + iv, cos v + i sin v, iu sin w, iu cos w),

θ ∈ 0, π2 , u 6= 0, v 6= 0.
Then the tangent bundle T N is given by
Z1 = (1, 0, i sin w, i cos w),

Z2 = (i, − sin v + i cos v, 0, 0),
Z3 = (0, 0, iu cos w, −iu sin w).
Then it is easy to verify that Hθ = Span{Z1 , Z2 } is a slant distribution

with the slant angle θ = π3 . It is also easy to get that H⊥ = Span{Z3 }
is an totally real distribution. Moreover, we can see that Hθ and H⊥ are
integrable.
Denote the integral manifolds of Hθ and H⊥ by N θ and N ⊥ , re-
spectively. Then the metric tensor of N is g = gN θ + u2 gN ⊥ . Thus
N = N θ ×f N ⊥ is a warped product hemi-slant submanifold of C4 with
warping function f = u.
On the other hand, it is easy to verify that the normal bundle T ⊥ N is

spanned by
W1 = (0, 0, sin w, cos w),
W2 = (−1, 0, i sin w, i cos w),
W3 = (i cos v, −i, 0, 0),
W4 = (i sin v, 1, 0, 0),
W5 = (0, 0, u cos w, u sin w).
Then using the Gauss formula, we have
σ(Z1 , Z2 ) = σ(Z1 , Z3 ) = σ(Z2 , Z3 ) = σ(Z1 , Z1 ) = 0,

sin v cos v
σ(Z2 , Z2 ) = W 3 − W4 ,
1 + cos2 v 1 + sin2 v
u
σ(Z3 , Z3 ) = − W2 .
2
Hence N is mixed geodesic, but it is neither totally geodesic nor totally
umbilical.
Another example was given in [Uddin et al. (2017)].
Example 11.9. Consider a submanifold M of C3 defined by

φ(u, v, w) = (v + iw) cos u, (v + iw) sin u, (11.74)

(i − 1)v + (1 + i)w .
It is easy to see that the tangent bundle T M of M is spanned by the
following vectors
v1 = (−(v + iw) sin u, (v + iw) cos u, 0),
v2 = (cos u, sin u, −1 + i), (11.75)
v3 = (i cos u, i sin u, 1 + i).
It is clear that Jv1 is orthogonal to T M . Thus D1 = Span{v1 } is a totally
real distribution.
D2 = Span{v2 , v3 } is a slant distribution
Also, it is easy to verify that
with slant angle θ = cos−1 31 . Hence the submanifold M defined by φ
given in (11.74) is a hemi-slant submanifold.
It is easy to verify that both distributions D1 and D2 are completely in-
tegrable. Let M⊥ and Mθ be integral manifolds of D1 and D2 , respectively.
Then it follows from (11.74) that the metric tensor of M is given by
g = gMθ + (v 2 + w2 )gM⊥ , (11.76)
where
gMθ = 3(dv 2 + dw2 ) and gM ⊥ = du2 (11.77)
are the metric tensors of Mθ and M⊥ , respectively. Consequently,
M = Mθ ×f M⊥
is a warped
√ product hemi-slant submanifold of C3 with warping function
f = v 2 + w2 and whose bi-slant angle satisfies θ 6= 0, π2 . Hence M is
warped product hemi-slant.
For warped product submanifolds N θ ×f N ⊥ , we also have the following

result from [Sahin (2009a)].
Theorem 11.11. Let N be an (m + n)-dimensional mixed geodesic warped

product submanifold N θ ×f N ⊥ in a Kähler manifold M̃ m+n such that N θ is
a proper slant submanifold and N ⊥ be a totally real submanifold of M̃ m+n .
Then we have:
(1) The second fundamental form σ of N satisfies
||σ||2 ≥ m (cot2 θ)|∇(ln f )|2 , m = dim N ⊥ . (11.78)
θ
(2) If the equality of (11.78) holds identically, then N is a totally geodesic
submanifold and N ⊥ is a totally umbilical submanifold of M̃ . Moreover,
M is never a minimal submanifold of M̃ .
11.8 Warped product semi-slant submanifolds
The following definition of semi-slant submanifolds was given in [Papaghiuc

(1994)].
Definition 11.5. A submanifold N of an almost Hermitian manifold M̃

is called semi-slant if the tangent bundle T N is the direct sum H ⊕ Hθ of
two orthogonal distributions H and Hθ , where H is holomorphic, i.e. H is
invariant with respect to the complex structure J of M̃ and Hθ is a θ-slant
distribution, i.e., the angle θ between JX and Hxθ is constant for any unit
vector X ∈ Hxθ and for any point x ∈ N .
Remark 11.5. Every CR-warped product N T ×f N ⊥ is a warped product

hemi-slant submanifold N θ ×f N ⊥ with the slant angle θ = 0. Thus warped
product hemi-slant submanifolds of the form N θ ×f N ⊥ are generalization
of CR-warped products.
Analogous to Lemma 7.2 we have the following (cf. [Sahin (2009a)]).
Lemma 11.2. Let N be a submanifold of a Kähler manifold M̃ . Then N

is a hemi-slant submanifold if and only if there exists a constant λ ∈ [−1, 0]
and a distribution Hθ on N such that
(i) Hθ = {X ∈ T N : P 2 X = λX}.
(ii) For any X ∈ T N perpendicular to Hθ , we have P X = 0.
Moreover, in this case λ = − cos2 θ for some real θ ∈ [0, π2 ].
Proof. Let N be a hemi-slant submanifold of M̃ . Then λ = − cos2 θ on

Hθ . By the definition of hemi-slant submanifold, (ii) is obvious.
⊥
Conversely (i) and (ii) imply T N = Hθ ⊕ Hθ . Since P (Hθ ) ⊂ Hθ from
⊥
(ii), we know that Hθ is an totally real distribution.
Contrast to CR-warped products N T ×f N ⊥ , we have the following
non-existence result for semi-slant warped product submanifolds.
Theorem 11.12. [Sahin (2006a)] There do not exist warped product sub-
manifolds of the form: N T ×f N θ in a Kähler manifold M̃ such that N T
is a complex submanifold and N θ is a proper slant submanifold of M̃ .
Proof. Assume that M̃ is a Kähler manifold which admits a warped

product submanifold N T ×f N θ such that N T is a complex submanifold of
M̃ and N θ is a proper slant submanifold. By using Proposition 3.1(2) and
hP Z, Zi = 0, we obtain for X ∈ H and Z ∈ Hθ that
h∇P Z X, Zi = X(ln f ) hP Z, Zi = 0,
θ
where H and H are the holomorphic and slant distribution. Since H and
Hθ are orthogonal, we have 0 = h∇P Z X, Zi = − h∇P Z Z, Xi . Thus we
˜ P Z JZ i = 0. By taking into account that ∇ is Levi-Civita
have h JX, ∇
connection, we obtain
˜ JX JX i = 0.
h JZ, ∇
˜ JX JX i = 0.
Using formulas of Gauss and Weingarten, we get h P ZF Z, ∇
Thus, using Gauss’ formula we have
hF Z, σ(JX, JX)i − h∇JX P Z, JXi = 0.
Thus we have hF Z, σ(JX, JX)i = P Z(ln f )|X|2 . Therefore, by polarization
identity we get
hF Z, σ(JX, JY )i = P Z(ln f ) hX, Y i (11.79)
for X, Y ∈ H and Z ∈ Hθ . Consequently, we find from formula (2.15) of

Weingarten, formula (2.16) and formula (8.4) that
˜ PZP Z + F Z i
0 = h JX, ∇
= − h ∇P Z JX, P Z i − h JX, AF Z P Z i
= h∇P Z JX, P Zi + hσ(P Z, JX), F Zi .
Hence, by using Proposition 3.1(2) and (8.4) for Z, W in Hθ , we get
JX(ln f ) cos2 θ hZ, Zi = − hσ(P Z, JX), F Zi (11.80)
θ
for X ∈ H and Z ∈ H . Substituting X by JX in (11.81) we obtain
X(ln f ) cos2 θ hZ, Zi = − hσ(P Z, X), F Zi . (11.81)
Also, by substituting Z by P Z in (11.81) and by using (7.22), we get

σ(P 2 Z, X), F P Z = −X(ln f ) cos2 θ|P Z|2 ,
cos2 θ hσ(Z, X), F P Zi = X(ln f ) cos4 θ|Z|2 .
Hence we have
hσ(Z, X), F P Zi = X(ln f ) cos2 θ|Z|2 (11.82)
for X ∈ H and Z ∈ Hθ . On the other hand, from Gauss’ formula we find
˜ X P Z, F W i
hσ(P Z, X), F W i = h ∇
for X ∈ H and Z, W ∈ Hθ . Hence we have
˜ XF W i .
hσ(P Z, X), F W i = − h P Z, ∇
Thus, by using (8.4) and that fact the M̃ is Kählerian, we derive that
˜ X JW − P W i
hσ(P Z, X), F W i = h P Z, ∇
= h JP Z, ∇˜ X W i + h P Z, ∇X P W i
˜ X W i + h F P Z, σ(X, W ) i + hP Z, ∇X P W i .
= − h P 2 Z, ∇
Thus, it follows from Proposition 3.1(2), P 2 Z = −(cos2 θ)Z and hP Z, P W i
= (cos2 θ) hZ, W i for Z, W ∈ Hθ that
hσ(P Z, X), F W i = − cos2 θX(ln f ) hZ, W i + hF P Z, σ(X, W )i
+ X(ln f ) cos2 θ hZ, W i .
Thus for Z = W we have
hσ(P Z, X), F Zi = hF P Z, σ(X, Z)i . (11.83)
So, we find from (11.81), (11.82) and (11.83) that X(ln f ) cos2 θ|Z|2 = 0.
Hence f is constant since N θ is assumed to be proper slant.
11.9 Warped product pointwise semi-slant submanifolds
A natural extension of semi-slant submanifolds are pointwise semi-slant

submanifolds defined and studied in [Sahin (2013d)].

is called pointwise semi-slant if the tangent bundle T N is the direct sum
H⊕Hθ of two orthogonal distributions H and Hθ , where H is a holomorphic
distribution and Hθ is a pointwise slant distribution, i.e., for any given point
x ∈ N the angle θ(x) between JX and Hxθ is independent of the choice of
the unit vector X ∈ Hxθ .
Example 11.10. Let N be a submanifold of C3 defined by

ϕ(t, s, u, v) = (t + is, u + i sin v, i cos v)
π π
for t, s, u ∈ R, 2 <v< 2. Then
X1 = (1, 0, 0), X2 = (i, 0, 0),
Z1 = (0, 1, 0), Z2 = (0, i cos v, −i sin v)
form an orthonormal frame of T N . Put H = Span{X1 , X2 } and Hθ =
Span{Z1 , Z2 }. It is direct to verify that H is a holomorphic distribution
and Hθ is a pointwise slant distribution with slant function θ = cos v.
Therefore, N is a pointwise semi-slant submanifold.
Proposition 11.2. Let M be a proper pointwise semi-slant submanifold of

a Kähler manifold. Then we have:
(a) The holomorphic distribution H is integrable if and only if
g(σ(X, JY ), F Z) = g(σ(JX, Y ), F Z)
for X, Y ∈ H and Z ∈ Hθ , where σ is the second fundamental form.
(b) The pointwise slant distribution Hθ is integrable if and only if
g(AF Z W − AF W Z, X) = g(∇Z (P W ) − ∇W (P Z), X)
for X ∈ H and Z, W ∈ Hθ .
Proof. This proposition is a special case of Propositions 8.1 and 8.2.
Proposition 11.3. Let M be a proper pointwise semi-slant submanifold of

a Kähler manifold. Then N is locally a Riemannian product N = N T × N θ
if and only if AF Hθ H = {0} holds, where N T is a complex submanifold and
N θ is a pointwise slant submanifold.
Proof. Follows immediately from Propositions 8.3 and 8.4.

Remark 11.6. Non-trivial warped product pointwise semi-slant submani-
folds of the form N T ×f N θ are natural extension of CR-warped products
defined in [Chen (2001a)]. Example 11.6 is such an example.
Indeed, every CR-warped product submanifold is a non-trivial warped
product pointwise semi-slant submanifold of the form M T ×f N θ with the
slant function θ = 0.
Theorem 11.13. [Sahin (2013d)] Let N = N T ×f N θ be a non-trivial
warped product pointwise semi-slant submanifold of a Kähler manifold
M̃ h+p , where N T is a complex submanifold with dimC N T = h and N θ
is a proper pointwise slant submanifold with dim N θ = p. Then we have
(1) The second fundamental form σ of N satisfies
||σ||2 ≥ 2p(csc2 θ + cot2 θ)|∇T (ln f )|2 . (11.84)
(2) If the equality of (11.84) holds identically, then N T is a totally geodesic
complex submanifold and N θ is a totally umbilical submanifold of
M̃ h+p . Moreover, N is a minimal submanifold of M̃ h+p .
11.10 Warped product pointwise bi-slant submanifolds
Pointwise bi-slant immersions are defined as follows (cf. [Chen and Uddin
(2016)]).
is called pointwise bi-slant if there exists a pair of orthogonal distributions
H1 and H2 of M such that the following three conditions hold.
(a) T N = H1 ⊕ H2 ;
(b) JH1 ⊥ H2 and JH2 ⊥ H1 ;
(c) Each distribution Hi is pointwise slant with slant function θi (i = 1, 2).
A pointwise bi-slant submanifold is a bi-slant submanifold if both slant
functions θ1 and θ2 are constant.
Analogous to CR-warped products, we define warped product pointwise
bi-slant submanifolds as follows.
Definition 11.8. A warped product N1 ×f N2 in an almost Hermitian
manifold M̃ is called a warped product pointwise bi-slant submanifold if
both factors N1 and N2 are pointwise slant submanifolds of M̃ .
A warped product pointwise bi-slant submanifold N1 ×f N2 is called

warped product bi-slant if both N1 and N2 are slant submanifolds.
Remark 11.7. For results on warped product pointwise bi-slant subman-

ifolds in Kähler manifolds, see [Chen and Uddin (2016)].
Remark 11.8. For a pointwise bi-slant submanifold N , if we assume θ1 = 0

and θ2 = π2 , then N is a CR-submanifold; if θ1 = 0 and θ2 6= 0, π2 , then
N is pointwise semi-slant; and if θ1 = π2 and θ2 6= 0, π2 , then N is called
pointwise hemi-slant.
Now, we give an example of pointwise bi-slant submanifold.
Example 11.11. Consider a submanifold N of C4 defined by

φ(u, v, w, s) = ((1 + i)u + iuv, (1 − i)w + isw).
If we put
V1 = (1 + i + iv, 0), V2 = (iu, 0),
V3 = (0, 1 − i + is), V4 = (0, iw).
Let us put H1 = Span{V1 , V2 } and H2 = Span{V3 , V4 }. It is direct to verify
that H1 and H2 satisfy both conditions (a) and (b) of Definition 11.7 as
well as condition (c) with slant functions
1 1
θ1 = p , θ2 = p .
1 + (1 + v)2 1 + (1 − s)2
Thus N is a pointwise bi-slant submanifold.
Next, we give some examples of bi-slant submanifolds.

π
Example 11.12. For two real numbers θ1 , θ2 satisfying 0 < θ1 , θ2 < 2, let
us consider the submanifold Mθ1 ,θ2 of C4 defined by
φθ1 ,θ2 (u, v, w, s) = (u + iv cos θ1 , iv sin θ1 , w + is cos θ2 , is sin θ2 ).
If we put
e1 = (1, 0, 0, 0), e2 = (i cos θ1 , i sin θ1 0, 0),
e3 = (0, 0, 1, 0), e4 = (0, 0, i cos θ2 , i sin θ2 ),
then the restriction of {e1 , e2 , e3 , e4 } to Mθ1 ,θ2 is an orthonormal frame of
the tangent bundle. Let us put D1 = Span{e1 , e2 } and D2 = Span{e3 , e4 }.
Then Mθ1 ,θ2 defined by φθ1 ,θ2 is a proper bi-slant submanifold in C4 with
θ1 , θ2 as the slant angles of D1 and D2 , respectively.
Example 11.11 and Example 11.12 illustrate that there exist pointwise
bi-slant submanifolds and bi-slant submanifolds in Kähler manifolds.
On the other hand, for warped product bi-slant submanifolds we have
the following classification result.
Theorem 11.14. [Uddin et al. (2017)] Let M = Mθ1 ×f Mθ2 be a warped

product bi-slant submanifold in a Kaehler manifold M̃ . Then one of the
following two cases must occurs:
(i) The warping function f is constant, i.e., M is a Riemannian product
of two slant submanifolds;
(ii) The slant angle of Mθ2 satisfies θ2 = π/2, i.e., M is a warped product
hemi-slant submanifold with Mθ2 being a totally real submanifold M⊥
of M̃ .
Remark 11.9. Examples 11.7 and 11.9 and Examples 11.11 and 11.12
illustrate that both case (1) and case (2) of Theorem 11.14 do occur.
Remark 11.10. Theorem 11.14 is a generalization of Theorems 11.10 and

11.12 of [Sahin (2006a)]. Moreover, Theorem 11.10 of [Sahin (2009a)] and
Theorem 11.1 of [Chen (2001a)] are special cases of Theorem 11.14 as well.
11.11 Warped products in locally conformal Kähler

manifolds
Let (M̃ , J, g) be an almost Hermitian manifold of dimension 2n, where J

denotes the almost complex structure and g the Hermitian metric. Then
(M̃ , J, g) is called a locally conformal Kähler manifold if for each point x of
M̃ there exists an open neighborhood U of x and a positive function fU on
U so that the local metric
gU = exp(−fU )g|U
is Kählerian (see, e.g. [Dragomir and Ornea (1998); Vilcu (2012)]).
If U = M̃ , then the manifold (M̃ , J, g) is said to be a globally conformal
Kähler manifold. Equivalently, (M̃ , J, g) is locally conformal Kähler if and
only if there exists a closed 1-form ω, globally defined on M̃ , such that
dΩ = ω ∧ Ω,
where Ω is the Kähler 2-form associated with (J, g), i.e.
Ω(X, Y ) = g(X, JY )
for X, Y ∈ Γ(T M̃ ).
The 1-form ω is called the Lee form and its metrically equivalent vector
field B̃ = ω ♯ is called Lee vector field, where ♯ means the rising of the indices
with respect to g, namely
g(X, B̃) = ω(X), (11.85)
for all X ∈ M̃ .
It is known that (M̃ , J, g) is globally conformal Kähler (respectively,
Kähler) if the Lee-form ω is exact (respectively ω = 0). It is also known
that the Levi-Civita connections ∇U of the local metrics gU glue up to a
globally defined torsion free linear connection ∇ˆ on M̃ given by
n o
∇ˆ XY = ∇˜ X Y − 1 ω(X)Y + ω(Y )X − g(X, Y )B̃ (11.86)
2
for X, Y ∈ X(M̃ ), where ∇ ˜ is the Levi-Civita connection of g. Moreover,
we have ∇g ˆ = ω ⊗ g and ∇J ˆ = 0. This connection ∇ ˆ is called the Weyl
connection of the locally conformal Kähler manifold M̃ .
Considering the anti-Lee form θ = ω ◦ J and the anti-Lee vector field
A = −JB, one can obtain a third equivalent definition in terms of the Levi-
Civita connection ∇ ˜ of the metric g (see [Dragomir and Ornea (1998)]).
Namely, (M̃ , J, g) is a locally conformal Kähler manifold if and only if the
following equation is satisfied:

˜ X J)Y = 1 θ(Y )X − ω(Y )JX − g(X, Y )A − Ω(X, Y )B̃
(∇ (11.87)
2
for any X, Y ∈ X(M̃ ).
A submanifold M of a locally conformal Kähler manifold (M̃ , J, g) is
called a CR-submanifold if there exists a differentiable distribution
H : x → Hx ⊂ Tx M
on M satisfying the following conditions:
(i) H is holomorphic, i.e. JHx =Hx , for each x ∈ M ;

(ii) the complementary orthogonal distribution H⊥ : x → Hx⊥ ⊂ Tx M is
totally real, i.e. JHx⊥ ⊂ Tx⊥ M for each x ∈ M .
If dim Hx⊥ = 0 (resp. dim Hx = 0), then the CR-submanifold is said to

be a holomorphic (resp. a totally real ) submanifold. A CR-submanifold is
called a proper if it is neither complex nor totally real.
For CR-submanifolds in a locally conformal Kähler manifold, we have
the following result from [Blair and Dragomir (2002)].
Theorem 11.15. Let N be a CR-submanifold of a locally conformal Kähler

manifold M̃ and let B be the restriction of the Lee vector field B̃ on N .
Then the following two statements are equivalent.
(1) The structure P defined by (6.38) is parallel, i.e. ∇P = 0.
(2) N is locally a Riemannian product N T × N ⊥ , where N T is a complex
submanifold and N ⊥ is a totally real submanifold of M̃ and either B =
0, i.e., N is normal to the Lee field of M̃ , or B 6= 0 and then B ∈ H
and dimC N T = 1.
For warped product CR-submanifolds in a locally conformal Kähler

manifold, we have the following.
Theorem 11.16. [Blair and Dragomir (2002)] Let N = N ⊥ ×f N T be a

warped product CR-submanifold of a locally conformal Kähler manifold M̃ .
Then we have:
(1) N T is totally umbilical in N with mean curvature |∇(ln f )| and with

d(ln f ) = 21 ω on H⊥ .
(2) Each local CR-submanifold, Ni⊥ =: N ⊥ ∩ Ui is a warped product
Ni⊥ ×αi exp(fi ) NiT ,
for αi > 0 and gi = exp(−fi )g⊥ + αi gT .
(3) If N is normal to B̃ or B ∈ H, then N is a CR-product and each fi is
constant on Ni⊥ = N ⊥ ∩ Ui .
For warped product CR-submanifolds of a locally Kähler manifold, we

also have the following result from [Bonanzinga and Matsumoto (2008)].
Proposition 11.4. In a warped product CR-submanifold N of a locally

conformal Kähler manifold M̃ , if the Lee vector field is normal to H⊥ , then
N is a CR-product.
Remark 11.11. Inequality (11.30) obtained in Theorem 11.3 has been

extended in [Bonanzinga and Matsumoto (2008)] to warped product CR-
submanifolds in a locally conformal Kähler manifold.
The same inequality was also extended in [Munteanu (2007b)] to doubly
warped product CR-submanifolds in a locally conformal Kähler manifold.
Those two generalized inequalities both involve the Lee vector field B̃ of
the ambient locally conformal Kähler manifold M̃ .
Chapter 12
CR-warped Products in Complex

Space Forms
12.1 CR-warped products
Theorem 11.3 shows that every non-trivial CR-warped product N T ×f N ⊥

in any Kähler manifold M̃ satisfies the following basic inequality
||σ||2 ≥ 2p|∇ ln f |2 , (12.1)
⊥
where p = dim N .
For simplicity we called the equality case:
||σ||2 = 2p|∇ ln f |2 , (12.2)
of the inequality (12.1) the basic equality of CR warped products.
In this section we discuss CR-warped products in complex space forms
which satisfy the basic equality identically.
Proposition 12.1. Let N T ×f N ⊥ be a non-trivial CR-warped product in

a complex space form M̃ (4c) of constant holomorphic sectional curvature
4c satisfying the basic equality: ||σ||2 = 2p|∇ ln f |2 . Then we have:
(a) N T is a totally geodesic complex submanifold of M̃ (4c). Hence N T is a
complex space form N h (4c) of constant holomorphic sectional curvature
4c.
(b) N ⊥ is a totally umbilical totally real submanifold of M̃ (4c). Hence N ⊥
is a real space form of constant sectional curvature ǫ with ǫ > c.
(c) When p = dim N ⊥ > 1, the warping function f satisfies
|∇f |2 = ǫ − cf 2 .
Proof. Under the hypothesis of this proposition, we have (11.36). Thus

statement (a) follows from the first equation in (11.36) and the fact that
N T is totally geodesic in N T ×f N ⊥ .
257
For statement (b), first we observe that the second equation in (11.36)
and the umbilicity of N ⊥ in N T ×f N ⊥ imply that N ⊥ is totally umbilical
in M̃ (4c). Hence, it follows from Gauss’ equation that N ⊥ is of constant
curvature, say ǫ ≥ c. It follows from (11.35) that ǫ = c occurs only when
the warping function is constant. Therefore we have statement (b).
⊥
Let RN denote the Riemann curvature tensor of N ⊥ . Then Proposition
3.2(5) gives
⊥
R(Z, W )V = RN (Z, W )V − |∇ ln f |2 {hW, V i Z − hZ, V i W }, (12.3)
for vectors Z, W, V ∈ T N ⊥ . Therefore, statement (c) follows from (11.36),
(12.3), Gauss’ equation, and statement (b).
The next result from [Chen (2001b)] shows that in order to study CR-
warped products satisfying the basic equality in a complex space form, it
is sufficient to consider anti-holomorphic submanifolds.
Proposition 12.2. Let M = N T ×f N ⊥ be a CR-warped product in a

complete simply-connected complex space form M̃ m (4c). If M satisfies the
basic equality, then it lies in a complex totally geodesic M̃ h+p (4c) ⊂ M̃ m (4c)
as an anti-holomorphic submanifold.
Proof. Assume that M = N T ×f N ⊥ is a CR-warped product in M̃ m (4c).

As usual, we denote the holomorphic and totally real distribution of M by
H and H⊥ , respectively. Then we find from Lemma 11.1 that
hσ(Z, JX), JZi = hσ(JX, Z), JZi = (X ln f )|Z|2 (12.4)
for X ∈ H and Z ∈ H⊥ . Thus if the basic equality ||σ||2 = 2p |∇(ln f )|2
holds, then we have
σ(H, H) = 0, σ(H⊥ , H⊥ ) = 0, σ(H, H⊥ ) ⊂ JH⊥ . (12.5)
Also, it follows from (12.5) that, at any point x ∈ M , the first normal space
νx = Im σx satisfies νx ⊂ JHx⊥ . Moreover, from Lemma 11.1(4) and (12.5),
we also have
DX JH⊥ ⊂ JH ⊥ . (12.6)
On the other hand, we obtain from (12.5) that
DZ JW = J∇Z W + AJW Z
for Z, W ∈ H . Thus DZ JH⊥ ⊂ JH⊥ . Combining this with (12.6) shows
⊥
that JH⊥ is a parallel subbundle of T ⊥ M with respect to the normal

connection. Since each first normal space νx = Im σx of M is contained
CR-warped Products in Complex Space Forms 259
in JH⊥ and JH⊥ is a parallel normal subbundle, we conclude that M is

contained in an (h + p)-dimensional totally geodesic complex submanifold
M̃ h+p (4c) of M̃ m (4c) by reduction theorem.
Because the codimension of M in M̃ h+p (4c) is equal to p = dim N ⊥ , M
is an anti-holomorphic submanifold of M̃ h+p (4c).
12.2 A PDE system associated with the basic equality
In order to classify CR-warped products in a complex space form which

satisfy the basic equality (12.2), we need the following proposition from
[Chen (2001a)].
Proposition 12.3. The non-constant solutions φ = φ(x1 , y1 , . . . , xh , yh ) of

the following system of partial differential equations:
∂2φ ∂φ ∂φ ∂φ ∂φ
= − , (12.7)
∂xj ∂xk ∂yj ∂yk ∂xj ∂xk
∂2φ ∂φ ∂φ ∂φ ∂φ
=− − , (12.8)
∂xj ∂yk ∂yj ∂xk ∂xj ∂yk
∂2φ ∂φ ∂φ ∂φ ∂φ
= − , (12.9)
∂yj ∂yk ∂xj ∂xk ∂yj ∂yk
j, k = 1, . . . , h,
are the functions given by
1
φ= ln hα, zi2 + hiα, zi2 , (12.10)
2
where z = (x1 + iy1 , . . . , xh + iyh ), α is a nonzero constant vector in Cn
and h , i is the standard Euclidean inner product on Ch .
Proof. From (12.8) we get

∂ ∂φ ∂φ
ln = −2 . (12.11)
∂x1 ∂y1 ∂x1
After solving (12.11) we obtain
∂φ
= e−2φ ψ, (12.12)
∂y1
for some function ψ = ψ(y1 , x2 , y2 , . . . , xh , yh ). Therefore by solving (12.12)
we obtain
1
φ = ln η(x1 , x2 , y2 , . . . , xh , yh ) + µ(y1 , x2 , y2 , . . . , xh , yh ) , (12.13)
2
Ry
for some function η = η(x1 , x2 , y2 , . . . , xh , yh ), where µ = 2 1 ψdy1 .
From (12.13) we find
ηx1 µy 1
φx1 = , φy1 = , (12.14)
2(η + µ) 2(η + µ)
ηx x (η + µ) − ηx21
φx1 x1 = 1 1 . (12.15)
2(η + µ)2
By applying (12.7), (12.14) and (12.15) we derive
2(η + µ)ηx1 x1 = ηx21 + µ2y1 . (12.16)
Similarly, from (12.6) with j = k = 1 and (12.13), we also have
2(η + µ)µy1 y1 = ηx21 + µ2y1 . (12.17)
By combining (12.16) and (12.17) we find ηx1 x1 = µy1 y1 . Since η and µ are
independent of y1 and x1 respectively, we find
ηx1 x1 = µy1 y1 = 2F (x2 , y2 , . . . , xh , yh ), (12.18)
for some positive function F = F (x2 , y2 , . . . , xh , yh ). Hence, after solving
(12.18), we obtain
η = F (x2 , . . . , yh )x21 + G(x2 , . . . , yh )x1 + H(x2 , . . . , yh ),
(12.19)
µ = F (x2 , . . . , yh )y12 + K(x2 , . . . , yh )y1 + L(x2 , . . . , yh ),
for some functions G, H, K, L of 2n − 2 variables.
Substituting (12.19) into (12.16) gives
4F (H + L) = G2 + K 2 .
Hence, by (12.19), we get
1
η+µ= (2F x1 + G)2 + (2F y1 + K)2 .
4F
Combining this with (12.13) yields
1 n o
φ = ln (ax1 + β)2 + (ay1 + δ)2 , (12.20)
2
where
√ G K
a(x2 , . . . , yh ) = F > 0, β(x2 , . . . , yh ) = √ , δ(x2 , . . . , yh ) = √ .
2 F 2 F
a(ax1 + β)
φx1 = ,
(ax1 + β)2 + (ay1 + δ)2
a(ay1 + δ)
φy1 = ,
(ax1 + β)2 + (ay1 + δ)2
(12.21)
(ax1 + β)(axj x1 + βxj ) + (ay1 + δ)(axj y1 + δxj )
φxj = ,
(ax1 + β)2 + (ay1 + δ)2
(ax1 + β)(ayj x1 + βyj ) + (ay1 + δ)(ayj y1 + δyj )
φyj = ,
(ax1 + β)2 + (ay1 + δ)2
for j = 2, . . . , h.
Hence, by applying (12.7) for φx1 xj , we obtain
(axj (2ax1 + β) + aβxj )((ax1 + β)2 + (ay1 + δ)2 )

− a(ax1 + β)[(ax1 + β)(axj x1 + βxj ) + (ay1 + δ)(axj y1 + δxj )] (12.22)
= a(ay1 + δ)[(ax1 + β)(ayj x1 + βyj ) + (ay1 + δ)(ayj y1 + δyj )].
By comparing the coefficients of x31 and y13 in (12.12) we find
∂a ∂a
= = 0, j = 2, . . . , h,
∂xj ∂yj
which show that a is constant. Therefore (12.21) and (12.22) yield
(ax1 + β)βxj + (ay1 + δ)δxj
φxj = ,
(ax1 + β)2 + (ay1 + δ)2
(12.23)
(ax1 + β)βyj + (ay1 + δ)δyj
φyj = .
(ax1 + β)2 + (ay1 + δ)2
Since a is constant, (12.23) implies
βxj βyk + δxj δyk + (ax1 + β)βxj yk + (ay1 + δ)δxj yk
φxj yk =
(ax1 + β)2 + (ay1 + δ)2
(12.24)
[(ax1 + β)βxj + (ay1 + δ)δxj ] [(ax1 + β)βyk + (ay1 + δ)δyk ]
−2 ,
[(ax1 + β)2 + (ay1 + δ)2 ]2
for 2 ≤ j, k ≤ h. Therefore, by using (12.8) with j = k, (12.23) and (12.24),
we obtain
βxj βyj + δxj δyj + (ax1 + β)βxj yj + (ay1 + δ)δxj yj = 0. (12.25)
Thus we find
βxj yj = δxj yj = 0, j = 2, . . . , h,
after comparing the coefficients of x31 and y13 in (12.25), respectively.
Similarly, by using (12.23) and by comparing the coefficients of x31 and
3
y1 in other equations from (12.7)-(12.9), we may also obtain
βx j x j = βx k y j = βx j y k = δ x j x j
= δxk yj = δxj yk = 0, 2 ≤ j, k ≤ h.
Hence there exist constants a3 , . . . , a2n , b3 , . . . , b2h such that
β = a3 x2 + a4 y2 + · · · + a2h−1 xh + a2n yh ,
(12.26)
δ = b3 x2 + b4 y2 + · · · + b2h−1 xh + b2h yh .

1
φ = ln (ax1 + a3 x2 + a4 y2 + · · · + a2h yh )2
2 (12.27)

+ (ay1 + b3 x2 + · · · + b2h yh )2 .
Finally, from (12.7) with j = 1 and (12.27), we obtain a2k−1 = b2k and
a2k = −b2k−1 . Thus we obtain from (12.27) that
1
φ = ln (ax1 + a3 x2 + a4 y2 + · · · + a2h−1 xn + a2h yh )2
2 (12.28)

+ (ay1 − a4 x2 + a3 y2 − · · · − a2h xh + a2h−1 yh )2 .
If we put
α = (a, a3 + ia4 , . . . , a2h−1 + ia2h ), z = (x1 + iy1 , . . . , xn + iyh ),
then (12.28) becomes (12.10) with a1 = a and a2 = 0. Consequently, after
applying a suitable change of variable on z1 , we obtain the proposition.
12.3 CR-warped products in Cm satisfying basic equality
Now, we present the complete classification of all non-trivial CR-warped

products in Cm satisfying the basic equality (12.2). By Proposition 12.3
we only need to consider anti-holomorphic CR-warped products.
As before we put h = dimC N T and p = dim N ⊥ .
Theorem 12.1. An anti-holomorphic CR-warped product M = N T ×f N ⊥

in Ch+p satisfies ||σ||2 = 2p|∇(ln f )|2 if and only if
(1) N T is an open portion of a complex Euclidean h-space Ch ,
(2) N ⊥ is an open portion of the unit p-sphere S p , and
(3) M is locally a partial Segre CR-warped product with f = |z1 | and up to
rigid motion the immersion x : N T ×f N ⊥ → Ch+p is given by
x(z, w) = ψhp (z, w), z ∈ Ch , w ∈ S p (1), (12.29)
where ψhp is the Euclidean Segre map defined by (11.29).
Proof. Assume that x : N T ×f N ⊥ → Ch+p is a CR-warped product

satisfying the basic equality ||σ||2 = 2p|∇(ln f )|2 . Then, from Lemma 11.1
and Theorem 11.4 we have
(a) N T is a totally geodesic complex submanifold of Cm ;
(b) N ⊥ is a totally umbilical totally real submanifold of Cm ;
(c) For any X ∈ H and Z ∈ H⊥ , we have

σ(JX, Z) = (X ln f )JZ; (12.30)
(d) The second fundamental form σ satisfies
σ(H, H) = σ(H⊥ , H⊥ ) = 0. (12.31)
Applying Lemma 11.1(4) and (12.30), we find
DX JZ = J∇X Z for any X ∈ H and Z ∈ H⊥ . (12.32)
T m T
Since N is a totally geodesic complex submanifold of C , N is an open
portion of a linear complex subspace Ch of Cm . Also, because N ⊥ is a
totally umbilical totally real submanifold of Cm , N ⊥ is an open portion of
an ordinary hypersphere S p lying in a totally real (p + 1)-subspace Ep+1 of
Cm (cf. [9]). Without loss of generality, we may assume that the radius of
S p is one.
Let z = (z1 , . . . , zh ) be a natural complex coordinate system on Ch . If
we put zj = xj + iyj for j = 1, . . . , h, then the standard metric on Ch is
given by
g0 = dx21 + dy12 + · · · + dx2h + dyh2 .
On S p we consider a spherical coordinate system {u1 , . . . , up } such that
the metric tensor g1 on S p is given by
g1 = du21 + cos2 u1 du22 + · · · + cos2 u1 · · · cos2 up−1 du2p .
The warped product metric on N T ×f N ⊥ is then given by
h
X
g= (dx2k + dyk2 ) + f 2 du21 + cos2 u1 du22 + · · ·
k=1
(12.33)
2 2

+ cos u1 · · · cos up−1 du2p .
Now, from (12.33) and a straightforward computation we know that the
Levi-Civita connection on N T ×f N ⊥ satisfies
∂ ∂ ∂
∇ ∂ =∇ ∂ =∇ ∂ = 0, j, k = 1, . . . , h, (12.34)
∂xk
∂xj ∂xj ∂y
k ∂yj ∂y
k
∂ fxj ∂
∇ ∂ = , j = 1, . . . , h; t = 1, . . . , p, (12.35)
∂xj ∂u f ∂ut
t
∂ fy ∂
∇ ∂ = j , j = 1, . . . , h; t = 1, . . . , p, (12.36)
∂yj ∂u f ∂ut
t
∂ ∂
∇ ∂ = − tan us , 1 ≤ s < t ≤ p, (12.37)
∂us ∂u ∂u
t t
h
∂
t−1
Y X
∂ ∂
∇ ∂ =− cos2 us f fxk + f fyk
∂ut ∂ut s=1
∂xk ∂yk
k=1
t−1 t−1
! (12.38)
X sin 2uq Y ∂
+ cos2 us , t = 1, . . . , p.
q=1
2 s=q+1 ∂uq
By applying (12.34)-(12.38) we know that the Riemann curvature tensor

of N T ×f N ⊥ satisfies

∂ ∂ ∂ ∂2φ ∂φ ∂φ ∂
R , = +
∂xj ∂ut ∂xk ∂xj ∂xk ∂xj ∂xk ∂ut
2
∂ ∂ ∂ ∂ φ ∂φ ∂φ ∂
R , = + (12.39)
∂xj ∂ut ∂yk ∂xj ∂yk ∂xj ∂yk ∂ut
2
∂ ∂ ∂ ∂ φ ∂φ ∂φ ∂
R , = + .
∂yj ∂ut ∂yk ∂yj ∂yk ∂yj ∂yk ∂ut
From (12.30) we have

∂ ∂ ∂φ ∂
σ , =− J ,
∂xj ∂ut ∂yj ∂ut
(12.40)
∂ ∂ ∂φ ∂
σ , = J ,
∂yj ∂ut ∂xj ∂ut
for j = 1, . . . , h; t = 1, . . . , p, where φ = ln f .
Hence, by applying Gauss’ equation, (12.39), and (12.40), we obtain the
PDE system (12.7)-(12.9). Therefore, after applying Proposition 12.2, we
conclude that there exists a constant vector
α = (a1 , a3 + ia4 , . . . , a2h−1 + ia2h )
h
in C such that
1 n o
ln hα, zi2 + hiα, zi2 ,
φ= (12.41)
2
where z = (x1 + iy1 , . . . , xh + iyh ).
Since φ = ln f , (12.41) implies
q
2 2 (12.42)
f = hα, zi + hiα, zi , |∇f | = |α|.
After applying the following rotations:
q
(a2j−1 xj + a2j yj , −a2j xj + a2j−1 yj ) 7→ a22j−1 + a22j (xj , yj ),
for j = 2, . . . , h, we obtain from (12.42) that
p
f = (a1 x1 + · · · + ah xh )2 + (a1 y1 + · · · + ah yh )2 . (12.43)
Without loss of generality, we may assume a1 6= 0.

From Gauss’ formula, (12.31), (12.32), (12.34)-(12.38), and (12.43), we
know that the immersion x : N T ×f N ⊥ → Cm satisfies
xxj xk = xxj yk = xyj yk = 0, j, k = 1, . . . , h, (12.44)
aj
xxj ut = 2 {a1 (x1 − iy1 ) + · · · + ah (xh − iyh )}xut (12.45)
f
aj
xyj ut = 2 {a1 (y1 + ix1 ) + · · · + ah (yh + ixh )}xut (12.46)
f
xus ut = −(tan us )xut , 1 ≤ s < t ≤ p, (12.47)
t−1
Y h n
X
xut ut = − cos2 us ak (a1 x1 + · · · + ah xh )xxk (12.48)
s=1 k=1
o
+ ak (a1 y1 + · · · + ah yh )xyk
t−1 t−1
!
X sin 2uq Y 2
+ cos us xuq .
q=1
2 s=q+1
Solving (12.44) gives
h
X h
X
x= Âk (u1 , . . . , up )xk + B̂ k (u1 , . . . , up )yk + C(u1 , . . . , up ) (12.49)
k=1 k=1
for some functions Â , . . . , Â , B̂ 1 , . . . , B̂ h , C. By substituting (12.49) into

1 h
(12.45) we find
{(a1 x1 + · · · + ah xh )2 + (a1 y1 + · · · + ah yh )2 }Âjut

= aj a1 (x1 − iy1 ) + · · · + ah (xh − iyh )
( h h
) (12.50)
X X
k k
× Âut xk + B̂ut yk + Cut ,
k=1 k=1
for j = 1, . . . , h; t = 1, . . . , p. Therefore, after comparing the coefficients of
x21 , xk yℓ and x1 in (12.50) respectively, we derive
a21 Âjut = a1 aj Â1ut , aj ak B̂uℓ t = iaj aℓ Âkut , Cu1 = · · · = Cup = 0 (12.51)
for j, k, ℓ = 1, . . . , h.
The last equation in (12.51) shows that C is a constant vector in Cm .
Hence we may choose C = 0 by applying a suitable translation on Cm if
necessary. Since a1 6= 0, the first two equations in (12.51) with j = k = 1
imply that
aj
Âjut = Â1ut , B̂uj t = i Âjut , j = 1, . . . , h, t = 1, . . . , p. (12.52)
a1

∂φ 1
Âj = aj Â + γ̂j , B̂ j = i Âj + δ̂ j , (12.53)
a1
for some constant vectors γ̂j , δ̂j , j = 1, . . . , h.
Equations (12.49) and (12.53) imply that there exist a Cm -valued func-
tion A(u1 , . . . , uk ) and some vectors γj , δj , j = 1, . . . , h, in Cm such that
h
X h
X
x = A(u1 , . . . , up ) ak (xk + iyk ) + (γk xk + δk yk ). (12.54)
k=1 k=1
Now divide the proof into two cases.
Case (a): p = 1. In this case, after substituting (12.54) into (12.48) with
t = 1, we find
h
X
Au1 u1 ak (xk + iyk )
k=1
h n
X (12.55)
=− ak (a1 x1 + · · · + ah xh )(ak A + γk )
k=1
o
+ ak (a1 y1 + · · · + ah yh )(iak A + δk )
which implies that
Au1 u1 + A = 0, (12.56)
h
X h
X
a k γk = ak δk = 0, (12.57)
k=1 k=1
since A is a function depending on u1 .

After solving (12.56) we find
A = c1 cos u1 + c2 sin u1 , (12.58)
for some vectors c1 , c2 in Cm .
After combining (12.54), (12.57) and (12.58) we derive that
x(x1 , y1 , . . . , xh , yh , u1 )
h
X
= (c1 cos u1 + c2 sin u1 ) ak (xk + iyk )
k=1 (12.59)
h
X
+ (γk xk + δk yk ),
k=1
where γk , δk , k = 1, . . . , h, are vectors in Cm satisfying condition (12.57).

If we choose the following initial conditions:
xx1 (1, 0, . . . , 0) = (1, 0, . . . , 0, 0, . . . , 0),
···
h-th
z}|{
xxh (1, 0, . . . , 0) = (0, . . . , 0, 1 , 0, . . . , 0),
xy1 (1, 0, . . . , 0) = (i, 0, . . . , 0, 0, . . . , 0),
(12.60)
···
h-th
z}|{
xyh (1, 0, . . . , 0) = (0, . . . , 0, i , 0, . . . , 0),
(h+1)-th
z}|{
xu1 (1, 0, . . . , 0) = (0, . . . , 0, 0, a1 , 0, . . . , 0),
then we obtain from (12.59) and (12.60) that
a1 c1 + γ1 = (1, 0, . . . , 0), · · ·
h-th
z}|{
ah c1 + γh = (0, . . . , 0, 1 , 0, . . . , 0),
ia1 c1 + δ1 = (i, 0, . . . , 0), . . .
(12.61)
h-th
z}|{
iah c1 + δh = (0, . . . , 0, i , 0, . . . , 0),
(h+1)-th
z}|{
c2 = (0, . . . , 0, 1 , 0, . . .).
Hence we find
1
c1 = (a1 , . . . , ah , 0, . . . , 0),
b
(h+1)-th
z}|{
c2 = (0, . . . , 0, 1 , 0, . . .)
1
γ1 = (b − a21 , −a1 a2 , . . . , −a1 ah , 0, . . . , 0),
b
···
(12.62)
1
γh = (−a1 ah , . . . , −ah−1 ah , b − a2h , 0, . . . , 0),
b
i
δ1 = (b − a21 , −a1 a2 , . . . , −a1 ah , 0, . . . , 0),
b
···
i
δh = (−a1 ah , . . . , −ah−1 ah , b − a2h , 0, . . . , 0),
b
Ph
where b = j=1 a2j .
Now, we obtain from (12.9) that
a1
h
X
x = x1 + iy1 + (cos u1 − 1) ak (xk + iyk ), . . . ,
b
k=1
X h
ah
xh + iyh + (cos u1 − 1) ak (xk + iyk ), (12.63)
b
k=1
h
X
sin u1 ak (xk + iyk ) .
k=1

1
hxx1 , xx1 i = 1 + 1− sin2 u1 .
a21 (12.64)
b
From (12.33) with (12.64) we get b = 1. So, we find (12.29) with p = 1.
Case (b): p > 1. In this case, we get |∇f |2 = 1 from Proposition 12.1.
Hence we obtain a21 + a22 + · · · + a2h = 1 from (12.42).
By substituting (12.4) into (12.47), we find
Aus ut = −(tan us )Aut , 1 ≤ s < t ≤ p. (12.65)
Solving (12.65) for t = p and s = 1 gives
A(u1 , . . . , up ) = D1 (u2 , . . . , up ) cos u1 + E1 (u1 , . . . , up−1 ) (12.66)
for some functions D and E. Hence, we derive from (12.4) that
h
X
x = D1 (u2 , . . . , up ) cos u1 + E1 (u1 , . . . , up−1 ) ak (xk + iyk )
k=1
h
X
+ (γk xk + δk yk )
k=1
By substituting this into (12.47) with s = 2, t = p, we find
D1 (u2 , . . . , up ) = D2 (u3 , . . . , up ) cos u2 + E2 (u2 , . . . , up−1 ).
Continuing such procedure (p − 1)-times we get
n
x = Dp−1 (up ) cos u1 · · · cos up−1 + E1 (u1 , . . . , up−1 )
+ E2 (u2 , . . . , up−1 ) cos u1 + · · ·
oX
h
+ Ep−1 (up−1 ) cos u1 · · · cos up−2 ak (xk + iyk ) (12.67)
k=1
h
X
+ (γk xk + δk yk ).
k=1
Therefore, by substituting (12.67) into (12.48) with t = 1 and by applying

a21 + a22 + · · · + a2h = 1, we find
h
X h
X
∂ 2 E1
+ E1 = 0, a k γ k = ak δk = 0. (12.68)
∂u21
k=1 k=1
Solving the first equation in (12.68) gives
E1 = F̂1 (u2 , . . . , up−1 ) cos u1 + F2 (u2 , . . . , up−1 ) sin u1 .
Hence we obtain from (12.70) that
n
x = Dp−1 (up ) cos u1 · · · cos up−1 + F1 (u2 , . . . , up−1 ) cos u1
+ F2 (u2 , . . . , up−1 ) sin u1 + E3 (u3 , . . . , up−1 ) cos u1 + · · ·
oX
h
+ Ep−1 (up−1 ) cos u1 · · · cos up−2 ak (xk + iyk ) (12.69)
k=1
h
X
+ (γk xk + δk yk ),
k=1
where F1 = F̂1 + E2 . Now, by substituting (12.69) into (12.47) with s = 1
and 1 < t < p, we find ∂F2 /∂ut = 0. Hence F2 is a constant vector, say c2 .
Continuingn such procedure sufficiently many times, we obtain
x = c1 cos u1 · · · cos up + c2 sin u1 + c3 sin u2 cos u1 + · · ·
oX
h
+ cp+1 sin up cos u1 · · · cos up−1 ak (xk + iyk )
(12.70)
k=1
h
X
+ (γk xk + δk yk )
k=1
for some vectors c1 , . . . , cp+1 in Cm , where γk , δk are vectors satisfying the
second condition in (12.68).
Consequently, after choosing suitable initial conditions, we obtain from
(12.70) that
Xh h
X
x(z, w) = z1 + (w0 − 1)a1 aj zj , · · · zh + a(w0 − 1)ah aj z j ,
j=1 j=1
h h
!
X X
w1 aj zj , . . . , wp aj z j ,
j=1 j=1
z = (z1 , . . . , zh ) ∈ Ch , w = (w0 , . . . , wp ) ∈ S p ⊂ Ep+1 .

If we choose a = (1, 0, . . . , 0) ∈ S h−1 , then the immersion reduces to (12.29).
The converse can be verified by a direct long computation.
12.4 CR-warped products in CP m and CH m
In this section we provide the classification of non-trivial CR-warped prod-

uct submanifolds in CP m and CH m satisfying the basic equality (12.2). In
order to do so, we present the following general construction method.
Case (1): CP m (4). Let C∗ = C − {0} and Cm+1

∗ = Cm+1 − {0}. Consider
the action of C∗ on Cm+1
∗ defined by
λ · (z0 , . . . , zm ) = (λz0 , . . . , λzm ). (12.71)
The set of equivalent classes obtained from this action is denoted by CP m .

Let π(z) denote the equivalent class contains z. Then π : Cm+1∗ → CP m is
m
a surjection. It is known that the CP admits a complex structure induced
from the complex structure on Cm+1 and a Kähler metric g with constant
holomorphic sectional curvature 4.
Assume that ψ : M → CP m (4) is an isometric immersion. Then M̆ =
π (M ) is a C∗ -bundle over M and the lift ψ̆ : π −1 (M ) → Cm+1
−1
∗ of ψ is
an isometric immersion satisfying π ◦ ψ̆ = ψ ◦ π. Then M̆ is invariant under
the C∗ -action. Conversely, if ψ̆ : N̆ → Cm+1 ∗ is an isometric immersion
invariant under the C∗ -action, then there is a unique isometric immersion
ψ : π(N̆ ) → CP m (4) satisfying π ◦ ψ̆ = ψ ◦ π.
There is an alternate way to view the lift ψ̆ : π −1 (N ) → Cm+1 ∗ via
the Hopf fibration as follows: Let S 2m+1 (1) denote the unit hypersphere of
Cm+1 centered at the origin and let U (1) = {λ ∈ C : λλ̄ = 1}. Then we
have a U (1)-action on S 2m+1 (1) defined by z 7→ λz. At z ∈ S 2m+1 (1) ⊂
Cm+1 , the vector V = iz is tangent to the flow of the action. The quotient
space S 2m+1 (1)/ ∼, obtained from the U (1)-action, is CP m (4). The almost
complex structure J on CP m (4) is induced from the complex structure J
on Cm+1 via the Hopf fibration: ϕ : S 2m+1 (1) → CP m (4).
If ψ : M → CP m (4) is an isometric immersion, then M̂ = ϕ−1 (M )
is a principal circle bundle over M with totally geodesic fibers. The lift
ψ̂ : M̂ → S 2m+1 (1) of ψ is an isometric immersion satisfying ϕ ◦ ψ̂ = ψ ◦ ϕ.
Conversely, if ψ : M̂ → S 2m+1 (1) is an isometric immersion invariant under
the U (1)-action, then there is a unique isometric immersion ψϕ : ϕ(M̂ ) →
CP m (4) satisfying ϕ ◦ ψ̂ϕ = ψϕ ◦ ϕ.
Since V = iz generates the vertical subspaces of the Hopf fibration, we
have an orthogonal decomposition:
Tz S 2m+1 (1) = (Tϕ(z) CP m )∗ ⊕ Span {V }. (12.72)

For each vector X tangent to CP m (4), we denote by X ∗ a horizontal lift

of X via the Hopf fibration ϕ. The horizontal lift X ∗ and X have the same
length, since the Hopf fibration is a Riemannian submersion.
For an isometric immersion ψ : M → CP m (4), M̆ = π −1 (M ) is diffeo-
morphic to R∗ × M̂ where R∗ = R−{0} and M̂ = ϕ−1 (M ). The immersion
ψ̆ : M̆ → Cm+1
∗ is related to the immersion ψ̂ : M̂ → S 2m+1 (1) by
ψ̆(t, q) = tψ̂(q), t ∈ R∗ , q ∈ M̂ . (12.73)
M̆ is the cone over M̂ with the origin of Cm+1 as its vertex. The metric
tensor ğ of M̆ and the metric tensor ĝ of M̂ are related by
ğ = t2 ĝ + dt2 . (12.74)
Case (2): CH m (−4). Consider the complex number (m + 1)-space Cm+1
1
endowed with the pseudo-Euclidean metric
m
X
g0 = −dz0 dz̄0 + dzj dz̄j .
j=1
Put Cm+1
1∗ = Cm+1
1 − {0}. Consider the C∗ -action on Cm+1
1∗ by
λ · (z0 , . . . , zm ) = (λz0 , . . . , λzm ).
The set of equivalent classes obtained from this action is denoted by CH m .
Just like CP m , there is an alternate way to view CH m as follows: Let
H12m+1 (−1) = {z = (z1 , z2 , . . . , zm+1 ) ∈ Cm+1
1 : hz, zi = −1},
where h , i is the inner product on Cm+1 1 induced from the pseudo-
Euclidean metric g0 . H12m+1 (−1) is known as the anti-de Sitter spacetime.
We put

Tz′ = z ∈ Cm+1
1 : Re hu, zi = Re hu, izi = 0 .
Then we have an U (1)-action on H12m+1 defined by z 7→ λz. At each
point z ∈ H12m+1 (−1), the vector iz is tangent to the flow of the action.
The orbit lies in the negative definite plane spanned by z and iz. The
quotient space H12m+1 (−1)/ ∼ under the U (1)-action is exactly the complex
hyperbolic space CH m with constant holomorphic sectional curvature −4.
The complex structure J on CH m is induced from the canonical complex
structure J on Cm+1
1 via the Riemannian submersion:
ϕ : H12m+1 (−1) → CH m (−4),
which has totally geodesic fibers. This submersion is called the hyperbolic
Hopf fibration.
Assume ψ : M → CH m (−4) is an isometric immersion. Then M̆ =

ϕ (M ) is a C∗ -bundle over M and the lift ψ̆ : M̆ → Cm+1
−1
1∗ ⊂ Cm+1 of ψ
is the isometric immersion satisfying π ◦ ψ̆ = ψ ◦ π. M̆ is invariant under
the action of C∗ . Conversely, if ψ̆ : M̆ → Cm+1
1∗ is an isometric immersion
invariant under the C∗ -action, then there is an isometric immersion ψ :
π(M̆ ) → CH m (−4) satisfying π ◦ ψ̆ = ψ ◦ π.
For an isometric immersion ψ : M → CH m (−4), M̆ = π −1 (M ) is
diffeomorphic to R∗ × M̂ , where M̂ = ϕ−1 (M ). The immersion ψ̆ : M̆ →
Cm+1
1∗ is related to ψ̂ : M̂ → H12m+1 (−1) by
ψ̆(t, q) = tψ̂(q), t ∈ R∗ , q ∈ M̂ . (12.75)
Assume that ψ : M → CP m (4) is an isometric immersion. Let ψ̂ : M̂ →

2m+1
S (1) and ψ̆ : M̆ → Cm+1∗ be the lifts of ψ via ϕ : S 2m+1 (1) → CP m (4)
and π : Cm+1
∗ → CP m (4), respectively, as described above.
˘ ˆ
Let ∇, ∇ and ∇ denote the Levi-Civita connections on M̆ , M̂ and M
respectively, and let σ̂ be the second fundamental form of M̂ in S 2m+1 (1).
Then we have
ˆ X ∗ Y ∗ = (∇X Y )∗ − hP X, Y i V,
∇ (12.76)
ˆ V X∗ = ∇
∇ ˆ X ∗ V = (P X)∗ , (12.77)
ˆ V V = 0,
∇ (12.78)
∗ ∗ ∗ ∗ ∗
σ̂(X , Y ) = (σ(X, Y )) , σ̂(X , V ) = (F X) , σ̂(V, V ) = 0 (12.79)
for vector fields X, Y tangent to M , where P X and F X are the tangential
and the normal components of JX, respectively.
For a vector U tangent to M̂ ⊂ S 2m+1 (1) ⊂ Cm+1∗ , we extend U to a
vector field in Cm+1
∗ , also denoted by U, by parallel translation along the
rays of the cone M̆ over M̂ . Then we obtain that
1
σ̆(U, W )(t, q) = σ̂(U, W )(q), t ∈ R∗ , q ∈ M̂ , (12.80)
t
∂ ∂ ∂
σ̆ U, = σ̆ , =0 (12.81)
∂t ∂t ∂t
for U, W tangent to M̂ , where σ̆ is the second fundamental form of ψ̆.
Proposition 12.4. Let N = N T ×f N ⊥ be a CR-warped product in

CP m (4). Then we have
(1) N̆ = π −1 (N ) is isometric to N̆ T ×tf˘ N ⊥ , where N̆ T = π −1 (N T ) and
f˘ = f ◦ π, π : Cm+1 → CP m (4).
∗
(2) N̆ is isometrically immersed in Cm+1 ∗ as a CR-warped product such

that N̆ T is as a complex submanifold and N ⊥ as a totally real subman-
ifold.
(3) N T ×f N ⊥ satisfies ||σ||2 = 2p|∇(ln f )|2 if and only if N̆ T ×tf˘ N ⊥
˘
satisfies the corresponding equality ||σ̆||2 = 2p|∇(ln tf˘)|2 .
Proof. Assume that N T ×f N ⊥ is a CR-warped product in CP m (4).

Denote by H and H⊥ the holomorphic and the totally real distributions
of N , respectively. Let Ĥ be the distribution on N̂ spanned by H∗ and
V = iz, where
H∗ = {X ∗ , X ∈ D}
and X ∗ a horizontal lift of X. Since H is integrable, (12.76)-(12.78) implies
that the distribution Ĥ is also integrable. From (12.76)-(12.78), we also
know that each leaf of Ĥ is a totally geodesic submanifold of N̂ .
Let Ĥ⊥ = {Z ∗ ∈ T N̂ : Z ∈ H⊥ } which is the orthogonal complementary
distribution of Ĥ in T N̂ . For any Z, W ∈ H⊥ , (12.76) gives
ˆ Z ∗ W ∗ = (∇Z W )∗ .
∇ (12.82)
Since H⊥ is integrable, (12.82) implies that Ĥ⊥ is an integrable distribution.
On the other hand, it is known from Lemma 11.1(2) that, for any vector
field X ∈ H and Z, W ∈ H⊥ , we have
h∇Z W, Xi = −(X ln f ) hZ, W i . (12.83)
Thus, by (12.82), (12.83), h(∇Z W )∗ , V i = 0, and the fact that the Hopf
fibration is a Riemannian submersion, we obtain
ˆ Z ∗ W ∗ , X ∗ i = −(X ln f ) hZ ∗ , W ∗ i ,
h∇
(12.84)
ˆ Z ∗ W ∗ , V i = 0.
h∇
Thus each leaf of Ĥ⊥ is an extrinsic sphere in N̂ , i.e., a totally umbilical
submanifold with parallel mean curvature vector. Therefore, according to
a result of [Hiepko (1979)], N̂ is a warped product N̂ T ×fˆ N∗⊥ , where N̂ T is
a leaf of Ĥ, N ⊥ a horizontal lift of N ⊥ and fˆ the warping function. From
∗
the definitions of Ĥ, N̂ T and ϕ, we may choose N̂ T to be ϕ−1 (N T ).
Because the Hopf fibration ϕ : S 2m+1 → CP m (4) is a Riemannian sub-
mersion, dϕ preserves the length of vectors normal to fibres. In particular,
dϕ preserves the inner product of any two vectors in F ⊥ . Therefore, the
warping function fˆ of N̂ T ×fˆ N∗⊥ is given by f ◦ ϕ. Consequently, we obtain
N̂ = N̂ T ×f ◦ϕ N∗⊥ , where N∗⊥ is a horizontal lift of N ⊥ via ϕ. Since N̆ is the
punctured cone over N̂ with 0 as its vertex, N̆ is nothing but N̆ T ×tf˘ N̆ ⊥ ,

where N̆ T = π −1 (N T ), f˘ = f ◦ π, and N̆ ⊥ is a horizontal lift of N ⊥ via π.
Because N̆ ⊥ is isometric to N ⊥ , N̆ is thus isometric to N̆ T ×tf˘ N ⊥ . This
proves statement (1).
Since N T is a holomorphic submanifold of CP m (4), we know from the
definition of π that N̆ T = π −1 (N T ) is a complex submanifold of Cm+1 ∗ .
Because N ⊥ is a totally real submanifold of CP m (4) and N̆ ⊥ is a horizontal
lift of N ⊥ via π, N̆ ⊥ is a totally real submanifold of Cm+1
∗ . Therefore, N̆
m+1
is isometrically immersed in C∗ as a CR-warped product. Thus we have
statement (2).
To prove statement (3), assume that N = N T ×f N ⊥ is a CR-warped
product in CP m (4) satisfying the basic equality (12.2). Then, according
to Proposition 12.2, N is contained in a totally geodesic linear complex
subspace CP h+p (4) of CP m (4) as an anti-holomorphic submanifold.
Without loss of generality, we may put m = h + p. From the proof of
Proposition 3.1, we have
σ(H, H) = 0, σ(H⊥ , H⊥ ) = 0, (12.85)
σ(X, Z) = −((JX) ln f )JZ (12.86)
⊥
for any X ∈ H and any unit vector Z ∈ H .
∂
Let H̆ denote the distribution on N̆ spanned by Ĥ and ∂t and let H̆⊥
⊥
denote the orthogonal distribution of H̆ in T N̆ . Then H̆ is spanned by
vectors in Cm+1
∗ obtained from Ĥ⊥ by parallel translation along rays of the
cone N̆ over N̂ .
From (12.79), (12.80) and the second condition in (12.85) we derive that
σ̆(H̆⊥ , H̆⊥ ) = 0. (12.87)
Also, from (12.79)-(12.81) and the first condition in (12.85), we have
σ̆(H̆, H̆) = 0. (12.88)
Applying (12.79)-(12.81) yields
1
σ̆(X ∗ , Z ∗ ) = − (σ(X, Z))∗ , (12.89)
t
1
σ̆(V, Z ∗ ) = JZ ∗ , (12.90)
t
∂
σ̆ , Z∗ = 0 (12.91)
∂t
for X ∈ H̆ and Z ∈ H⊥ . Therefore, by applying (12.75) and (12.86)-(12.91),
we find
1 + |∇(ln f )|2
||σ̆||2 = 2p . (12.92)
t2
On the other hand, from V (tf˘) = t(V fˆ) = 0 and ∂t ∂

(tf˘) = 1/t, it is
direct to verify that the gradient of ln(tf˘) on N̆ satisfies
1
˘
|∇(ln tf˘)|2 = 2 (1 + |∇(ln f )|2 ).
t
˘
This N̆ T ×tf˘ N ⊥ satisfies the corresponding equality ||σ̆||2 = 2p|∇(ln tf˘)|2 .
Conversely, if N = N T ×f N ⊥ is a CR-warped product in CP m (4)
˘ ln(tf˘)|2 , then we
such that N̆ T ×tf˘ N ⊥ satisfies the equality: ||σ̆||2 = 2p|∇
obtain
σ̆(H̆, H̆) = σ̆(H̆⊥ , H̆⊥ ) = 0, (12.93)

σ̆(U, Z ∗ ) = −((JU ) ln(tf˘))JZ ∗ (12.94)
for any U ∈ D̆ and Z ∗ ∈ D̆⊥ . Hence we find (12.85) from (12.79), (12.80)
and (12.93). Moreover, from (12.79), (12.80), (12.94), and f˘ = f ◦ π, we
obtain (12.86). Consequently, the basic equality ||σ||2 = 2p|∇(ln f )|2 must
holds for N in CP m (4).
The following theorem from [Chen (2001b)] determines all CR-warped

products in CP m (4) satisfying the basic equality (12.2).
Theorem 12.2. A non-trivial CR-warped product N T ×f N ⊥ in CP m (4)

satisfies the basic equality ||σ||2 = 2p|∇(ln f )|2 if and only if we have
(1) N T is an open portion of complex Euclidean h-space Ch ,

(2) N ⊥ is an open portion of a unit p-sphere S p , and
(3) up to rigid motions, the immersion x of N T ×f N ⊥ into CP m (4) is
the composition π ◦ x̆, where π is the projection Cm+1 ∗ → CP m (4) and

x̆(z, w) = ψh+1 p (z, w), 0, . . . , 0
for z = (z1 , z1 , . . . , zh+1 ) ∈ Ch+1 and w = (w1 , . . . , wp+1 ) ∈ S p (1),

where ψh+1 p is the Euclidean Segre map defined in (11.29).
Proof. Follows from Theorem 12.1 and Propositions 12.2 and 12.4.
Similarly, for CR-warped products in CH m (−4) satisfying the basic

equality (12.2), we have the following result from [Chen (2001b)].
Theorem 12.3. A CR-warped product N T ×f N ⊥ in CH m (−4) satisfies

the basic equality ||σ||2 = 2p|∇(ln f )|2 if and only if one of the following
two cases occurs:
(1) N T is an open portion of complex Euclidean h-space Ch , N ⊥ is an open

portion of a unit p-sphere S p and, up to rigid motions, the immersion is
the composition π ◦ x̆, where π is the projection π : Cm+1
1∗ → CH m (−4)
and
Xh Xh
x̆(z, w) = z0 + a0 (1 − w0 ) aj zj , z1 + a1 (w0 − 1) aj z j , . . . ,
j=0 j=0
h h h
!
X X X
zh + ah (w0 − 1) aj zj , w1 aj zj , . . . , wp aj zj , 0, . . . , 0
j=0 j=0 j=0
for some real numbers a0 , . . . , ah satisfying a20 − a21 − · · · − a2h = −1,

where z = (z0 , . . . , zh ) ∈ Ch+1
1 and w = (w0 , . . . , wp ) ∈ S p ⊂ Ep+1 .
(2) p = 1, N T is an open portion of Ch and, up to rigid motions, the
immersion is the composition π ◦ x̆, where
h
X h
X
x̆(z, t) = z0 + a0 (cosh t − 1) aj zj , z1 + a1 (1 − cosh t) aj z j ,
j=0 j=0
h h
!
X X
. . . , zh + ah (1 − cosh t) aj zj , sinh t aj zj , 0, . . . , 0
j=0 j=0
for some real numbers a0 , a1 . . . , ah+1 satisfying a20 − a21 − · · · − a2h = 1.
Proof. This can be done in a way similar to Theorem 12.2.
12.5 CR-warped products with compact holomorphic

factor
In this section we discuss CR-warped product submanifolds in CP m with

compact holomorphic factor N T . For the proof of the main results of this
section, we need the following two lemmas.
Lemma 12.1. Let N T ×f N ⊥ be a CR-warped product in CP m (4). Then

hσ(X, Y ), σ(V, Z)i = hσν (X, Y ), σν (V, Z)i = 0 (12.95)
⊥
for X, Y, V ∈ H and Z ∈ H .
Proof. Let N T ×f N ⊥ be a CR-warped product in CP m (4) and let

X, Y, V be vector fields in H and Z in H⊥ . Since N T is totally geodesic in
N T ×f N ⊥ , we obtain from Proposition 6.2 that
R(X, Y ; V, Z) = R̃(X, Y ; V, Z) = 0. (12.96)
Thus, by applying Gauss’ equation, we get

hσ(X, V ), σ(Y, Z)i = hσ(X, Z), σ(Y, V )i . (12.97)
Also, from Lemma 11.1(1), we find
σ(X, V ) = σν (X, V ) (12.98)
for any vector fields X, V ∈ H. By applying (12.97), (12.98), and Lemma
11.1(1), we obtain
hσ(X, V ), σ(JX, Z)i
= hσ(X, Z), σ(JX, V )i by(12.97)
= − hJσ(X, Z), σ(X, V )i by Lemma 11.1(4)
(12.99)
= − hJσ(X, Z), σν (X, V )i by (12.97)
= − hσ(JX, Z), σν (X, V )i by Lemma 11.1(7)
= − hσ(JX, Z), σ(X, V )i by (12.98).
Therefore we get
hσ(X, V ), σ(JX, Z)i = hσ(X, Z), σ(X, JV )i = 0 (12.100)
for any vector fields X, V ∈ H and Z ∈ H⊥ .
Therefore we have hσ(X, Z), σ(X, V )i = 0 for vector fields X, V ∈ H
and Z ∈ H⊥ . Hence by applying polarization we obtain
hσ(X, Z), σ(Y, V )i + hσ(Y, Z), σ(X, V )i = 0. (12.101)
hσ(X, Z), σ(Y, V )i = 0 (12.102)
for X, Y, V ∈ H and Z ∈ H⊥ .
Now, the equation hσν (X, Y ), σν (V, Z)i = 0 follows from (12.102) and
Lemma 11.1(1). This completes the proof of the lemma.
Lemma 12.2. Let N T ×f N ⊥ be a CR-warped product in CP m (4). Then

we have

|σν (X, Z)|2 = |X|2 + HCln f (X, X) |Z|2 , (12.103)
for any X ∈ H and Z ∈ H⊥ , where
1 ln f 1
HCln f (X, Y ) =
H (X, X) + H ln f (JX, JX), (12.104)
2 2
is called the complex Hessian of ln f .
Proof. Assume that N T ×f N ⊥ is a CR-warped product in CP m (4). For

vector fields X, Y ∈ H and Z, W ∈ H⊥ , equation (6.5) gives
R̃(X, JX; JZ, Z) = 2|X|2 |Z|2 , (12.105)
since we have X, Y, JX ⊥ Z, W and JZ ⊥ W .
On the other hand, for X ∈ H and Z ∈ H⊥ , the equation of Codazzi
implies that
R̃(X, JX, JZ, Z)
= hDJX σ(X, Z) − σ(∇JX X, Z) − σ(X, ∇JX Z), JZi (12.106)
− hDX σ(JX, Z) − σ(∇X JX, Z) − σ(JX, ∇X Z), JZi .
Since N T is totally geodesic in N T ×f N ⊥ , ∇X Z and ∇JX Z lie in
⊥
H and ∇X JX and ∇JX X lie in H. Therefore, by applying (12.105) and
statements (1), (2) and (3) of Lemma 11.1, we get
2|X|2 |Z|2 = − JX(|Z|2 JX ln f ) − hσ(X, Z), DJX JZi
− X(|Z|2 X ln f ) + hσ(JX, Z), DX JZi
(12.107)
+ {(∇JX JX) ln f + (∇X X) ln f + (X ln f )2
+ ((JX ln f ))2 }|Z|2 .
Since X|Z|2 = 2(X ln f )|Z|2 by Lemma 11.1(2), we find
JX(|Z|2 JX ln f ) + X(|Z|2 X ln f )
(12.108)
= {(JX)2 ln f + X 2 ln f + 2(JX ln f )2 + 2(X ln f )2 }|Z|2 .
Because CP m (4) is Kählerian, we have
J∇X Z + Jσ(X, Z) = −AJZ X + DX JZ. (12.109)
Applying (12.109) and statements (2) and (3) of Lemma 11.1, we find
hσ(JX, Z), DX JZi = hσ(JX, Z), J∇X Zi
+ hσ(JX, Z), Jσ(X, Z)i (12.110)
2 2
= (X ln f ) |Z| + hσ(JX, Z), Jσ(X, Z)i
for vector fields X ∈ H and Z ∈ H⊥ .
On the other hand, if σν (X, Z) denotes the ν-component of σ(X, Z),
then (9.4) gives
hσ(JX, Z), Jσ(X, Z)i = hσ(JX, Z), Jσν (X, Z)i

(12.111)
= AJσν (X,Z) JX, Z = Aσν (X,Z) X, Z = |σν (X, Z)|2 .
hσ(JX, Z), DX JZi = (X ln f )2 |Z|2 + |σν (X, Z)|2 . (12.112)
Similarly, we also have

hσ(X, Z), DJX JZi = −(JX ln f )2 |Z|2 − |σν (X, Z)|2 (12.113)
T m
by Lemma 11.1(6). Since N is a complex submanifold of CP (4) and
N T is a totally geodesic submanifold of N T ×f N ⊥ , we get
J∇JX X = ∇JX JX, J∇X JX = −∇X X. (12.114)
Combining (12.107), (12.108), (12.112), (12.113) and (12.114) gives
2|X|2 |Z|2 = {(∇X X + ∇JX JX) ln f − X 2 ln f
(12.115)
− (JX)2 ln f }|Z|2 + 2|σν (X, Z)|2 .
Therefore we obtain
n o
|σν (X, Z)|2 = |X|2 + HCln f (X, X) |Z|2 , (12.116)
which is (12.103).
Theorem 12.4. Let N T ×f N ⊥ be a CR-warped product in the complex

projective m-space CP m (4) of constant holomorphic sectional curvature 4.
If N T is compact, then m ≥ h + p + hp.
Proof. Assume that {X1 , . . . , X2h } and {Z1 , . . . , Zp } are orthonormal

bases of H and H⊥ , respectively. Since the complex Hessian HCln f , the
inner product h , i and the second fundamental form σ are bilinear, thus
by applying polarization and (12.116), we obtain

hσν (Xi , Zs ), σν (Xj , Zs )i = δij + HCln f (Xi , Xj ) δst . (12.117)
If N T is compact, then the function ln f has an absolute minimum
value at some point, say at u ∈ N T . At this critical point the complex
Hessian HCln f is non-negative definite. Hence each σν (Xi , Zs ) is nonzero at
u by (12.116). Moreover, since HCln f is clearly self-adjoint, there exists an
orthonormal basis X1 , . . . , X2h at u ∈ N T such that HCln f at u is diagonal-
ized with respect to X1 , . . . , X2h . Therefore it follows from (12.117) that
the following vectors at u:
σν (Xi , Zt ), i = 1, . . . , 2h; t = 1, . . . , p,
are mutually orthogonal nonzero vectors in the complex subspace νu of
the normal space at u. Therefore the complex rank of the complex vector
bundle ν is at least hp. Consequently, we get m ≥ h + p + hp. This prove
the theorem.
Theorem 12.5. If N T ×f N ⊥ is a CR-warped product in CP h+p+hp (4)

with compact N T , then N T is holomorphically isometric to CP h (4).
Proof. Assume N T ×f N ⊥ is a CR-warped product in CP h+p+hp (4)

with compact holomorphic factor N T . Let {Z1 , . . . , Zp } be a basis of H⊥
and let {X1 , . . . , X2h } be a basis of H which diagonalizes the complex Hes-
sian HCln f at the point u on which ln f attains its absolute minimum. Then
we known from above that
σν (Xi , Zt ), i = 1, . . . , 2h; t = 1, . . . , p,
are mutually orthogonal nonzero vectors in νu . Therefore, by applying
continuity we conclude that there exists an open neighborhood U of u such
that, restricted to U ×f N ⊥ , the following vector fields
{σν (Xj , Zs ), j = 1, . . . , 2h; p = 1, . . . , p}
span the complex vector bundle ν.
On the other hand, it follows from Lemma 12.1 and Lemma 11.1(1)
that, for any X, Y ∈ H, σ(X, Y ) is a normal vector field in ν which is
perpendicular to each σν (Xj , Zs ). Thus, we have σ(X, Y ) = 0 identically
on U ×f N ⊥ . Therefore, by using the fact that N T is totally geodesic in
N T ×f N ⊥ , we conclude that for any fixed q ∈ N ⊥ the non-empty open
subset U × {q} is immersed as a totally geodesic complex subspace CP h (4)
of CP h+p+hp (4). Hence U × {q} is immersed as an open portion of a linear
complex submanifold CP p (4) of CP h+p+hp (4).
On the other hand, since N T × {q} is a complex submanifold of the
Kähler manifold CP h+p+hp , N T × {q} is real analytic. Hence the whole
complex submanifold N T × {q} must be the linear complex subspace
CP p (4). Consequently, N T is holomorphic isometric to CP h (4).
Theorem 12.6. For any CR-warped product N T ×f N ⊥ in CP m (4) with

compact N T and any q ∈ N ⊥ , we have
Z
||σ||2 dVT ≥ 4hp vol(N T ), (12.118)
N T ×{q}
where ||σ|| is the norm of the second fundamental form, dVT is the volume
element of N T and vol(N T ) is the volume of N T .
The equality sign of (12.118) holds identically if and only if we have:
(a) The warping function f is constant.
(b) (N T , gN T ) is holomorphically isometric to CP h (4) and it is isometri-
cally immersed in CP m (4) as a totally geodesic complex submanifold.
(c) (N ⊥ , f 2 gN ⊥ ) is isometric to an open portion of the real projective p-
space RP p (1) of constant sectional curvature one and it is isometrically
immersed in CP m (4) as a totally geodesic totally real submanifold.
(d) N T ×f N ⊥ is immersed linearly fully in a linear complex subspace

CP h+p+hp (4) of CP m (4); and moreover, the immersion is rigid.
Proof. Assume N T ×f N ⊥ is a CR-warped product in CP m (4). Then

Lemma 12.2 yields
|σν (X, Z)|2 = |X|2 |Z|2
1 (12.119)
+ H ln f (X, X) + H ln f (JX, JX) |Z|2
2
for X ∈ H and Z ∈ H⊥ . On the other hand, from Lemma 11.1(3), we find

|σJH⊥ (JX, Z)|2 = − σJH⊥ (JX, Z), J 2 σJH⊥ (JX, Z)
= (X ln f ) hσJH⊥ (JX, Z), JZi (12.120)
2 2
= (X ln f ) |Z| .
|σJH⊥ (X, Z)|2 = ((JX) ln f )2 |Z|2 . (12.121)
Combining (12.119), (12.120) and (12.121) gives
||σ||2 ≥ 2p {|∇ ln f |2 − ∆(ln f ) + 2h} (12.122)
with equality sign holding if and only if we have
σ(H, H) = σ(H⊥ , H⊥ ) = 0, (12.123)
i.e., N T ×f N ⊥ is both N T and N ⊥ -totally geodesic.
When N T is compact, (12.122) and Hopf’s lemma imply that
Z Z
2
||σ|| dVT ≥ 2p |∇ ln f |2 dVT + 4hp vol(N T ) (12.124)
N T ×{q} NT
for each q ∈ N ⊥ .
Obviously, inequality (12.124) implies (12.118), with the equality sign
holding if and only if we have
(a) f is constant and
(b) ||σ||2 = 4hp holds identically.
When f is constant, the warped product is the Riemannian product of
(N T , gN T ) and (N ⊥ , f 2 gN ⊥ ). Therefore, N T ×f N ⊥ can be regarded as a
CR-product in CP m (4). Consequently, we may applying Theorem 6.6 to
conclude that
(1) N T is the complex projective h-space CP h (4) immersed as a totally
geodesic complex submanifold.
(2) (N ⊥ , f 2 fN ⊥ ) is an open portion of RP p (1) immersed as a totally real

totally geodesic submanifold.
(3) N T ×f N ⊥ is immersed linearly fully in a linear complex subspace
CP h+p+hp (4) of CP m (4); moreover, the immersion is rigid.
Conversely, if φ : N T ×f N ⊥ → CP m (4) is a CR-warped product in
CP m (4) which satisfies conditions (a)-(d), then the CR-warped product is
both N T -totally geodesic and N ⊥ -totally geodesic.
On the other hand, since φ is a CR-product by statement (a), we find
from Lemma 12.2 and Lemma 11.1(3) that |σ(X, Z)|2 = |X|2 |Z|2 . Thus we
have ||σ||2 = 4hp identically. Consequently, the equality case of (12.118)
holds identically.
Theorem 12.7. Let N T ×f N ⊥ be a CR-warped product with compact
N T in CP m (4). If the warping function f is non-constant, then, for each
q ∈ N ⊥ , we have
Z Z
||σ||2 dVT ≥ 2pλ1 (ln f )2 dVT + 4hp vol(N T ), (12.125)
N T ×{q} NT
T
where dVT , λ1 and vol(N ) are the volume element, the first positive eigen-
value of the Laplacian ∆ and the volume of N T , respectively.
Moreover, the equality sign of (12.125) holds identically if and only if
we have
(a) ∆ ln f = λ1 ln f .
(b) The CR-warped product is both N T -totally geodesic and N ⊥ -totally
geodesic.
Proof. Assume that f is non-constant. Then the minimum principal on
λ1 yields (cf. page 187 of [Berger et al. (1971)])
Z Z
|∇ ln f |2 dVT ≥ λ1 (ln f )2 dVT , (12.126)
NT NT
with equality holding if and only if ∆ ln f = λ1 ln f holds. So, by combining
(12.124) and (12.126), we obtain inequality (12.125).
Clearly, from the discussion above, we know that the equality sign of
(12.125) holds identically if and only if we have
(a) ∆ ln f = λ1 ln f and
(b) the warped product is both N T and N ⊥ -totally geodesic.
Remark 12.1. The results given in this section were obtained in [Chen
(2004c)].
Chapter 13
More on CR-warped Products in

Complex Space Forms
13.1 Another optimal inequality for CR-warped products
For CR-warped product in a complex space form, we also have the following
optimal inequality from [Chen (2003a)].
Theorem 13.1. Let φ : N T ×f N ⊥ → M̃ m (4c) be a CR-warped product

in a complex space form of constant holomorphic sectional curvature 4c. If
we put h = dimC N T and p = dim N ⊥ , then we have

||σ||2 ≥ 2p |∇(ln f )|2 − ∆(ln f ) + 4hpc. (13.1)
Proof. Let N T ×f N ⊥ be a CR-warped product in a complex space form

M̃ m (4c). Then the equation of Codazzi implies
R̃(X, JX, JZ, Z) = hDJX σ(X, Z), JZi − hσ(∇JX X, Z), JZi
− hσ(X, ∇JX Z), JZi − hDX σ(JX, Z), JZi (13.2)
+ hσ(∇X JX, Z), JZi + hσ(JX, ∇X Z), JZi
for all vector fields X ∈ H and Z ∈ H⊥ . Since N T is totally geodesic in
N T ×f N ⊥ , ∇X Z and ∇JX Z lie in H⊥ and ∇X JX and ∇JX X lie in H.
Therefore, by applying Lemma 11.1, we get
2|X|2 |Z|2 c = − JX(|Z|2 JX ln f ) − hσ(X, Z), DJX JZi
− X(|Z|2 X ln f ) + hσ(JX, Z), DX JZi
(13.3)
+ {(J∇JX X) ln f − (J∇X JX) ln f }|Z|2
+ {(X ln f )2 + ((JX ln f ))2 }|Z|2 .
By applying Lemma 11.1 we find
JX(|Z|2 JX ln f ) + X(|Z|2 X ln f )
(13.4)
= {(JX)2 ln f + X 2 ln f + 2(JX ln f )2 + 2(X ln f )2 }|Z|2 .
283
Since M̃ m (4c) is Kählerian, we have

J∇X Z + Jσ(X, Z) = −AJZ X + DX JZ. (13.5)
By using (13.5) and Lemma 11.1, we find
hσ(JX, Z), DX JZi = hσ(JX, Z), J∇X Zi
+ hσ(JX, Z), Jσ(X, Z)i (13.6)
2 2
= (X ln f ) |Z| + hσ(JX, Z), Jσ(X, Z)i
for vector fields X ∈ H and Z ∈ H⊥ .
On the other hand, if we denote the ν-component of σ(X, Z) by
σν (X, Z), then, by applying Lemma 11.1(3), we also have
hσ(JX, Z), Jσ(X, Z)i = hσ(JX, Z), Jσν (X, Z)i

= AJσν (X,Z) JX, Z

(13.7)
= Aσν (X,Z) X, Z
= |σν (X, Z)|2 .
Hence, after combining (13.6) and (13.7) we derive that
hσ(JX, Z), DX JZi = (X ln f )2 |Z|2 + |σν (X, Z)|2 . (13.8)
hσ(X, Z), DJX JZi = −(JX ln f )2 |Z|2 − |σν (X, Z)|2 . (13.9)
T T
Since N is a complex submanifold of a Kähler manifold and N is totally
geodesic in N T ×f N ⊥ , we find
J∇JX X = ∇JX JX, J∇X JX = −∇X X. (13.10)
Now, combining (13.3), (13.4) and (13.8)-(13.10) gives
2|X|2 |Z|2 c = {(∇X X + ∇JX JX) ln f − X 2 ln f
(13.11)
− (JX)2 ln f }|Z|2 + 2 |σν (X, Z)|2 .
Let us assume that {X1 , . . . , X2h } is an orthonormal frame of N T and
{Z1 , . . . , Zp } an orthonormal frame on N ⊥ . Then (13.11) yields
p
2h X
X
2 |σν (Xj , Zt )|2 = 4hpc + 2p ∆(ln f ). (13.12)
j=1 t=1
On the other hand, Lemma 11.1(3) implies

p
2h X
X
|σJH⊥ (Xj , Zt )|2 = p |∇ ln f |2 , (13.13)
j=1 t=1
More on CR-warped Products in Complex Space Forms 285
where σJH⊥ (Xj , Zt ) denotes the JH⊥ -component of σ(Xj , Zt ). Thus, by

combining (13.12) and (13.13) we obtain

2 ||σ(H, H⊥ )||2 = 2p |∇ ln f |2 − ∆(ln f ) + 2hc , (13.14)
where
p
2h X
X
||σ(H, H⊥ )||2 = |σ(Xj , Zt )|2 . (13.15)
j=1 t=1
The desired inequality (13.1) now follows from (13.14).
Next, we present a class of CR-warped products in a complex Euclidean
space which satisfy the equality case of the inequality (13.1).
Let Ch∗ = Ch − {0} and j : S p → Ep+1 be the inclusion of the unit
hypersphere S p centered at the origin into Ep+1 .
For a natural number α ≤ h and a vector X tangent to Cα ∗ at a point
z ∈ Cα∗ , we decompose X as
X = Xz|| + Xz⊥ ,
||
where Xz is parallel to z and Xz⊥ is perpendicular to z.
For three natural numbers h, p, α satisfying α ≤ h, we have a partial
(α)
Segre embedding Shp : Ch∗ × S p → Cαp+h defined in (11.27).
(α)
Theorem 13.2. For 1 ≤ α ≤ h and p ≥ 1, Shp : Ch∗ × S p → Cαp+h
defined by (11.27) satisfies the following properties:
(α)
(1) Shp : Ch∗ ×f S p → Cαp+h is an isometric immersion which gives rise
qP
α
to a CR-warped product with warping function: f = j=1 zj z̄j .
(α)
(2) The second fundamental form σ of Shp satisfies

||σ||2 = 2p |∇(ln f )|2 − ∆(ln f ) . (13.16)
Proof. For tangent vector fields X of Ch∗ and Z of S p , we derive from
(11.27) that
(α)
XShp = (X (1) ⊗ j, Xα+1 , . . . , Xh ), (13.17)
(α)
ZShp = (z (1) ⊗ Z, 0, . . . , 0), (13.18)
where
X (1) ⊗ j = (w0 X1 , . . . , wp X1 , . . . , w0 Xα , . . . , wp Xα ), (13.19)
(1)
z ⊗ Z = (Z0 z1 , . . . , Zp z1 , . . . , Z0 zα , . . . , Zp zα ), (13.20)
(1) (2)
X = (X1 , . . . , Xα ), X = (Xα+1 , . . . , Xh ), (13.21)
(1) (2) (1)
X = (X ,X ), Z = (Z0 , . . . , Zp ), z = (z1 , . . . , zα ). (13.22)
From (13.17) and (13.18) we know that the tangent space of Ch∗ × S p at
a point (z, w) is spanned by vectors given by (13.17) and (13.18). Since S p
is the unit hypersphere centered at the origin, it follows from (13.17) and
(α)
(13.18) that the induced metric on Ch∗ × S p via Shp is the warped product
qP
α
metric g = g0 + f 2 g1 with warping function f = j=1 zj z̄j , where g0 and
g1 are the metrics of Ch∗ and S p , respectively.
It follows from (13.17) that Ch∗ is immersed as a complex submanifold
of Cαp+h . From (13.17) and (13.18) we also know that S p is immersed as
a totally real submanifold of Cαp+h . Hence we have statement (1).
Applying (11.27) and (13.17)-(13.22) gives
(α)
XY Shp = ∇ ˜ X (1) Y (1) ⊗ j, ∇˜ X (2) Y (2) , (13.23)
(α)
ZW §hp = z (1) ⊗ ∇ ˜ Z W, 0, . . . , 0 , (13.24)
(α)
XZShp = X (1) ⊗ Z, 0, . . . , 0 , (13.25)
˜ is the
for vector fields X, Y tangent to Ch∗ and Z, W tangent to S p , where ∇
Levi-Civita connection for Euclidean space as well as for complex Euclidean
space. From (13.17)-(13.18) and (13.23)-(13.25), we find
(1)
σ(X, Y ) = σ(Z, W ) = 0, σ(X, Z) = Xz(1) ⊥ ⊗ Z, 0, . . . , 0 (13.26)
for vector fields X, Y tangent to Ch∗ and Z, W tangent to S p . Hence the
squared norm of the second fundamental form is given by
Xα
2p(2α − 1)
||σ||2 = , f 2
= zj z̄j . (13.27)
f2 j=1
On the other hand, it is straightforward to verify that

1 2(1 − α)
|∇(ln f )|2 = 2 , ∆(ln f ) = . (13.28)
f f2
Now, by combining (13.27) and (13.28) we obtain statement (2). This
13.2 CR-warped products in Cm satisfying the equality
Theorem 13.3. Let φ : N T ×f N ⊥ → Cm be a CR-warped product in

complex Euclidean m-space Cm . Then we have
(1) The second fundamental form σ of φ satisfies

||σ||2 ≥ 2p |∇(ln f )|2 − ∆(ln f ) . (13.29)
(2) If the CR-warped product satisfies the equality case of (13.29), then
(2.a) N T is an open portion of Ch∗ ;
(2.b) N ⊥ is an open portion of S p ;
(2.c) There exist a natural number α ≤ h and a complex coordinate system
. . , zh } on Ch∗ such that the warping function f is given by
{z1 , .q
Pα
f= j=1 zj z̄j ;
(2.d) Up to rigid motions of Cm , the immersion φ of the CR-warped
product in Cm is given by

(α)
φ(z, w) = Shp (z, w), 0, . . . , 0 ,
where z = (z1 , . . . , zh ) ∈ Ch∗ , w = (w1 , . . . , wp+1 ) ∈ S p (1) ⊂ Ep+1 ,

(α)
and Shp is a partial Segre embedding defined by (11.27).
Proof. Let φ : N T ×f N ⊥ → Cm be a CR-warped product in Cm . Then

inequality (13.1) reduces to inequality (13.29).
Now, let us assume the CR-warped product satisfying the equality case
of (13.29) identically. Then (13.14) and the equality case of (13.29) give
σ(H, H) = 0, σ(H⊥ , H⊥ ) = 0. (13.30)
Since N T is totally geodesic in N T ×f N ⊥ , the first equation in (13.30) and
the totally geodesy of N T in N T ×f N ⊥ imply that N T is isometrically
immersed as a totally geodesic complex submanifold of Cm . Hence N T is
an open portion of a complex Euclidean h-space Ch . For any vector fields
X ∈ H and Z, W ∈ H⊥ , Lemma 7.2 implies
h∇W Z, Xi = hJAJZ W, Xi = − hσ(JX, W ), JZi . (13.31)
Hence, by applying Lemma 11.1(2) and (13.31), we find
h∇W Z, Xi = −(X ln f ) hZ, W i . (13.32)
⊥
On the other hand, if we denote

⊥by σ the second
fundamental form of
⊥ T ⊥
N in M = N ×f N , we have σ (Z, W ), X = h∇W Z, Xi . Combining
this with (13.32) yields
σ ⊥ (Z, W ) = − hZ, W i ∇ ln f. (13.33)
Hence, by applying (13.33) and the second equation of (13.30), we see that
N ⊥ is immersed as a totally umbilical submanifold of Cm . Therefore N ⊥
is an open portion of an ordinary p-sphere S p (or of R when p = 1).
If p ≥ 2, we may assume S p is of radius one, by rescaling the warping
function f if necessary. Thus in this case N T ×f N ⊥ is an open portion of
Ch ×f S p (or of Ch ×f R when p = 1). Consequently, we may choose a

complex Euclidean coordinate system {z1 , . . . , zh } on Ch and a coordinate
system {u1 , . . . , up } on S p (or on R if p = 1) such that the metric tensor
on N T ×f N ⊥ is given by
h
X
g= dzj dz̄j + f 2 du21 + cos2 u1 du22 + · · ·
j=1 (13.34)
2 2

+ cos u1 · · · cos up−1 du2p
with zj = xj + iyj .
It follows from equation (13.34) and a straightforward computation that
the Levi-Civita connection on N T ×f N ⊥ satisfies
∂φ ∂φ ∂φ
∇ ∂φ = ∇ ∂φ =∇ ∂ = 0, j, k = 1, . . . , h, (13.35)
∂xj∂xk ∂xj ∂y
k ∂yj ∂y
k
∂ fx ∂
∇ ∂ = j , j = 1, . . . , h; t = 1, . . . , p, (13.36)
∂xj ∂u f ∂ut
t
∂ fy ∂
∇ ∂ = j , j = 1, . . . , h; t = 1, . . . , p, (13.37)
∂yj ∂u f ∂ut
t
∂ ∂φ
∇ ∂ = − tan us , 1 ≤ s < t ≤ p, (13.38)
∂us ∂u ∂ut
t
h
∂φ
t−1
Y X
∂ 2 ∂φ
∇ ∂ =− cos us f fxk + f fyk (13.39)
∂ut ∂u ∂xk ∂yk
t s=1 k=1
t−1 t−1
!
X sin 2uq Y ∂φ
+ cos2 us , t = 1, . . . , p.
q=1
2 s=q+1
∂u q
From (13.30), (13.35), (13.38) and (13.39), we know that φ satisfies

φzj zk = φzj z̄k = φz̄j z̄k = 0, j, k = 1, . . . , h, (13.40)
φus ut = − tan us φut , 1 ≤ s < t ≤ p, (13.41)
t−1
Y h
X
φut ut = − cos2 us f fxk φxk + f fyk φyk (13.42)
s=1 k=1
t−1 t−1
!
X sin 2uq Y
2
+ cos us φuq , t = 1, . . . , p,
q=1
2 s=q+1
∂2 φ
where φzj z̄k = ∂zj ∂ z̄k , . . . ,
etc., and

∂φ 1 ∂ ∂ ∂ 1 ∂ ∂
= −i , = +i . (13.43)
∂zj 2 ∂xj ∂yj ∂ z̄j 2 ∂xj ∂yj
By solving (13.40) we get

h
X
φ(z1 , . . . , zh , u1 , . . . , up ) = Aj (u1 , . . . , up )zj + B(u1 , . . . , up ) (13.44)
j=1
for Cm -valued functions A1 , . . . , Ah , B. From (13.42) with t = 1, we find

h
!
1 X ∂f 2 ∂f 2
φu1 u1 = − φx + φy . (13.45)
2 ∂xk k ∂yk k
k=1
Substituting (13.44) into (13.45) yields
h
X Xh
∂ 2 Aj ∂2B ∂f 2
2 zj + 2 =− Aj . (13.46)
j=1
∂u1 ∂u1 j=1
∂ z̄j
Ph ∂f 2
Case (1): j=1 ( ∂ z̄j )Aj is independent of z1 , . . . , zh . In this case, (13.46)
implies that
∂ 2 Aj
= 0, j = 1, . . . , h, (13.47)
∂u21
Xh
∂2B ∂f 2
2 = − Aj . (13.48)
∂u1 j=1
∂ z̄j
After solving (13.47) we have
Aj = Dj (u2 , . . . , up )u1 + Ej (u2 , . . . , up ), j = 1, . . . , h, (13.49)
for vector functions Dj and Ej . Applying (13.44) and (13.49) yields

φzj , φzj = |Dj |2 u21 + 2 hDj , Ej i u1 + |Ej |2 .

On the other hand, (13.34) gives φzj , φzj = 1 which is independent
of u1 . Therefore we obtain D1 = · · · = Dh = 0. Consequently (13.26) is
reduced to
Aj (u1 , . . . , up ) = Ej (u2 , . . . , up ), j = 1, . . . , h, (13.50)
From (13.48) and (13.50), we find
h
1 X ∂f 2
B=− Ej (u2 , . . . , up )u21 + F (u2 , . . . , up )u1
2 j=1 ∂ z̄j (13.51)
+ G(u2 , . . . , up )
for some vector functions F, G. Thus we may derive from (13.44), (13.50)
and (13.51) that
h
!
X 1 ∂f 2 2
φ= Ej zj − u + F u1 + G. (13.52)
j=1
2 ∂ z̄j 1
After substituting (13.52) into (13.41) with s = 1 and 1 < t ≤ p we shall

obtain that
h
1 X ∂f 2 ∂Ej ∂F
u1 − (13.53)
2 j=1 ∂ z̄j ∂ut ∂ut
( h h
)
X ∂Ej 1 X ∂f 2 ∂Ej 2 ∂F ∂G
= tan u1 zj − u + u1 + .
j=1
∂ut 2 j=1 ∂ z̄j ∂ut 1 ∂ut ∂ut
Since Ej , F, G and ∂f 2 /∂ z̄j are independent of u1 , equation (13.53) implies

∂Ej ∂F ∂G
= = =0
∂ut ∂ut ∂ut
for j = 1, . . . , h and t = 2, . . . , p. Therefore E1 , . . . , Eh , F, G are constant
vectors in Cm . From (13.52) we also have
h
X ∂f 2
φu1 = − Ej u1 + F. (13.54)
j=1
∂ z̄j
On the other hand, after applying (13.34), we obtain hφu1 φu1 i = f 2

which is a non-constant function independent of u1 . Hence (13.31) implies
Ph 2 2 2
j=1 (∂f /∂ z̄j )Ej = 0. Thus f = |F | is constant which contradicts to
properness of the CR-warped product.
Ph ∂ 2 f 2
Case (2): j=1 ( ∂ z̄j )Aj depends on z1 , . . . , zh . In this case, by taking the
derivative of (13.46) with respect to ∂/∂zj , we find
h
X ∂2f 2
∂ 2 Aj
= − Ak , j = 1, . . . , h. (13.55)
∂u21 ∂zj ∂ z̄k
k=1
On the other hand, by (13.44), we get φzj = Aj (u1 , . . . , up ). Hence
A1 , . . . , Ah form an orthonormal frame by (13.34). Since ∂ 2 Aj /∂u21 and
A1 , . . . , Ah are independent of z1 , . . . , zh , we conclude from (13.55) that
∂ 2 f 2 /∂zk ∂ z̄j , j, k = 1, . . . , h, are constant. Thus we may put
∂2f 2
= γj k̄ , j, k = 1, . . . , h, (13.56)
∂zj ∂ z̄k
for some constants γj k̄ . Solving (13.56) yields
h
X
f 2 (z1 , . . . , zh ) = γj k̄ zj z̄k + H + K (13.57)
j,k=1
for some functions H, K satisfying

∂H ∂K
= = 0, j = 1, . . . , h. (13.58)
∂ z̄j ∂zj
It follows from (13.56) that (γj k̄ ) is Hermitian, i.e., γ̄j k̄ = γkj̄ . Hence
the Spectral Theorem in Matrix Theory shows that there exists a unitary
matrix which diagonalizes (γj k̄ ). So there is a suitable complex Euclidean
coordinate system {z1 , . . . , zh } on Ch such that (13.57) reduces to
h
X
2
f = bj zj z̄j + H(z1 , . . . , zh ) + K(z1 , . . . , zh ). (13.59)
j=1
Since f is a real-valued function, we may put
H = X + iY, K = U − iY, (13.60)
for some real-valued functions X, Y, U . From (13.58) and (13.60), we obtain
the following Cauchy-Riemann equations:
∂X ∂Y ∂Y ∂X
=− , = ,
∂xj ∂yj ∂xj ∂yj
(13.61)
∂U ∂Y ∂Y ∂U
= , =− .
∂xj ∂yj ∂xj ∂yj
From (13.61) we find that H + K = X + U is constant, say δ. Hence (13.59)
reduces to
Xh
f2 = bj zj z̄j + δ.
j=1
We may put δ = 0 by applying a suitable translation on Cm . Thus
Xh
f2 = a2j zj z̄j , (13.62)
j=1
for some real numbers a1 , . . . , ah ≥ 0, since we have f > 0.
Now, by combining (13.46) and (13.62) we find
∂ 2 Aj
= −a2j Aj , j = 1, . . . , h, (13.63)
∂u21
∂2B
= 0. (13.64)
∂u21
Since f > 0, there exists at least one aj greater than zero. Without loss of
generality, we may assume
a1 , . . . , aα > 0, aα+1 = · · · = ah = 0. (13.65)
for some 1 ≤ α ≤ h. From (13.63), (13.64) and (13.65), we obtain
Aj = Dj (u2 , . . . , up ) cos(aj u1 ) + Ej (u2 , . . . , up ) sin(aj u1 ), (13.66)
Ak = Dk (u2 , . . . , up )u1 + Ek (u2 , . . . , up ), (13.67)
B = F (u2 , . . . , up )u1 + G(u2 , . . . , up ) (13.68)
for j = 1, . . . , α, and k = α + 1, . . . , h. Substituting (13.66), (13.67) and

(13.68) into (13.44) gives
α
X
φ= Dj (u2 , . . . , up ) cos(aj u1 ) + Ej (u2 , . . . , up ) sin(aj u1 ) zj
j=1
h
X
+ Dk (u2 , . . . , up )u1 + Ek (u2 , . . . , up ) zk (13.69)
k=α+1
+ F (u2 , . . . , up )u1 + G(u2 , . . . , up ).
By differentiating (13.69) with respect to zk , we obtain φzk = Dk u1 + Ek

for k = α + 1, . . . , h. Thus
hφzk , φzk i = |Dk |2 u21 + 2 hDk , Ek i + |Ek |2 .
By comparing this with (13.34) we obtain Dα+1 = · · · = Dh = 0. Therefore

(13.45) becomes
α
X
φ= Dj (u2 , . . . , up ) cos(aj u1 ) + Ej (u2 , . . . , up ) sin(aj u1 ) zj
j=1
h
(13.70)
X
+ Ek (u2 , . . . , up )zk + F (u2 , . . . , up )u1 + G(u2 , . . . , up ).
k=α+1
From (13.41) with s = 1, t > 1 and (13.46), we find

Xα ∂D ∂Ej ∂F
j
aj sin(aj u1 ) − cos(aj u1 ) zj +
j=1
∂ut ∂ut ∂ut
( α
X ∂Dj ∂Ej
= tan u1 cos(aj u1 ) + sin(aj u1 ) zj (13.71)
j=1
∂ut ∂ut
h
)
X ∂Ek ∂F ∂G
+ zk + u1 +
∂ut ∂ut ∂ut
k=α+1
which implies ∂Ek /∂ut = ∂F /∂ut = ∂G/∂ut = 0 for k = α + 1 . . . , h and

t = 2, . . . , p. Hence Eα+1 , . . . , Eh , F and G are constant vectors.
Equation (13.71) also implies
∂Dj ∂Ej
aj sin(aj u1 ) − aj cos(aj u1 )
∂ut ∂ut
(13.72)
∂Dj ∂Ej
= tan u1 cos(aj u1 ) + sin(aj u1 )
∂ut ∂ut
for j = 1, . . . , α, which are equivalent to

∂Dj
(aj − 1) sin((aj + 1)u1 ) − (aj + 1) sin((aj − 1)u1 ) (13.73)
∂ut
∂Ej
= (aj − 1) cos((aj + 1)u1 ) + (aj + 1) cos((aj − 1)u1 )
∂ut
for j = 1, . . . , α. By letting u1 = 0, we get ∂Ej /∂ut = 0. Thus E1 , . . . , Eα
are constant vectors. Consequently, we obtain from (13.70) that
α
X
φ= Dj (u2 , . . . , up ) cos(aj u1 ) + Ej sin(aj u1 ) zj
j=1
h
(13.74)
X
+ Ek zk + F u1 + G,
k=α+1
where E1 , . . . , Eh , F, G are constant vectors. From (13.74) we get

φxj = Dj cos(aj u1 ) + Ej sin(aj u1 ), j = 1, . . . , α, (13.75)
φyj = iDj cos(aj u1 ) + iEj sin(aj u1 ), j = 1, . . . , α, (13.76)
φxk = Ek , k = α + 1, . . . , h, (13.77)
φyk = iEk , k = α + 1, . . . , h, (13.78)
Xα

φu1 = aj Ej cos(aj u1 ) − Dj sin(aj u1 ) zj + F. (13.79)
j=1
By applying (13.34) and (13.75), we find

2δjℓ = hDj , Dℓ i cos((aj + aℓ )u1 ) + cos((aj − aℓ )u1 ))
+ hEj , Eℓ i cos((aj − aℓ )u1 ) − cos((aj + aℓ )u1 )) (13.80)
+ hDj , Eℓ i sin((aj + aℓ )u1 ) − sin((aj − aℓ )u1 ))
+ hDℓ , Ej i sin((aj + aℓ )u1 ) + sin((aj − aℓ )u1 ))
for j, ℓ = 1, . . . , α.
Since cos((aj −aℓ )u1 ), cos((aj +aℓ )u1 ), sin((aj +aℓ )u1 ) are independent
functions, (13.80) implies hDj , Eℓ i + hDℓ , Ej i = 0 for j, ℓ = 1, . . . , α. By
setting u1 = 0, (13.80) also yields hDj , Dℓ i = δjℓ . Hence, by combining
these with (13.80), we have hEj , Eℓ i = δjℓ . Consequently, we obtain
hDj , Dℓ i = hEj , Eℓ i = δjℓ , hDj , Eℓ i + hEj , Dℓ i = 0, 1 ≤ j, ℓ ≤ α. (13.81)
Similarly, by differentiating (13.80) with respect to u1 , we find
aℓ hDj , Eℓ i + aj hDℓ , Ej i = 0, j, ℓ = 1, . . . , α. (13.82)
Also, it follows from (13.34), (13.75) and (13.76) that

hDj , iDℓ i = hEj , iEℓ i = δjℓ , hDj , iEℓ i + hEj , iDℓ i = 0, (13.83)
aℓ hDj , iEℓ i + aj hDℓ , iEj i = 0, j, ℓ = 1, . . . , α. (13.84)
From (13.34) and (13.75)-(13.78), we also have
hEk , Dj i = hEk , Ej i = hEk , iDj i = hEk , iEj i = 0 (13.85)
for j = 1, . . . , α; k = α + 1, . . . , h.
Equations (13.34), (13.62), (13.79), (13.81) and (13.82) imply
X α α
X α
X
a2j zj z̄j = a2j zj z̄j + 2 aj hEj cos(aj u1 )zj , F i
j=1 j=1 j=1
α
X
−2 aj hDj sin(aj u1 )zj , F i + |F |2 .
j=1
Thus we get F = 0. Therefore (13.74) reduces to
Xα
φ= (Dj (u2 , . . . , up ) cos(aj u1 ) + Ej sin(aj u1 ))zj
j=1
h
(13.86)
X
+ Ek zk + G,
k=α+1
where E1 , . . . , Eh , G are constant vectors. By using (13.73) we know that
either Dj is a constant vector or aj = 1. Without loss of generality, we may
assume that a1 , . . . , ar 6= 1 and ar+1 = · · · = aα = 1. Hence D1 , . . . , Dr are
constant vectors. Therefore (13.86) reduces to
Xr
φ= (Dj cos(aj u1 ) + Ej sin(aj u1 ))zj
j=1
α h
(13.87)
X X
+ (Dj (u2 , . . . , up ) cos u1 + Ej sin u1 )zj + Ek zk + G,
j=r+1 k=α+1
where D1 , . . . , Dr , E1 , . . . , Eh and G are constant vectors satisfying (13.57)-
(13.61). Substituting (13.39) and (13.63) into (13.42) with t = 2 yields
Xα
∂ 2 Dj
cos u1 zj
j=r+1
∂u22
α
X
= − cos2 u1 aj Dj cos(aj u1 ) + Ej sin(aj u1 ) zj (13.88)
j=1
α
X
− sin u1 cos u1 aj Dj sin(aj u1 ) − Ej cos(aj u1 ) zj ,
j=1
where ar+1 = · · · = aα = 1.
If r > 1, then (13.88) implies

cos u1 Dj cos(aj u1 ) + Ej sin(aj u1 )
(13.89)
+ sin u1 Dj sin(aj u1 ) − Ej cos(aj u1 ) = 0
for j = 1, . . . , r. Since a1 , . . . , ar 6= 1, (13.89) implies
D1 = · · · = Dr = E1 = · · · = Er = 0
which is a contradiction. Hence a1 = · · · = aα = 1. Thus we find from
(13.88) that ∂ 2 Dj /∂u22 = −Dj holds for j = 1, . . . , α. After solving these
equations we get
Dj = Fj (u3 , . . . , up ) cos u2 + Gj (u3 , . . . , up ) sin u2 .
Consequently, (13.86) becomes
Xα

φ= Fj (u3 , . . . , up ) cos u1 cos u2
j=1

+ Gj (u3 , . . . , up ) cos u1 sin u2 + Ej sin u1 zj (13.90)
h
X
+ Ek zk + G.
k=α+1
By substituting (13.90) into (13.41) with s = 2 and t > 2, we conclude that
G1 , . . . , Gα are constant vectors.
Continuing these procedures sufficiently many times, we find
α
( p
X j
Y
φ= c1 cos ut + c2j sin u1 + c3j sin u2 cos u1
j=1 t=1
p−1
) (13.91)
Y h
X
j
+ · · · + cp+1 sin up cos ut zj + Ek zk + G,
t=1 k=α+1
where ctj , Ek and G are constant vectors in Cm .

Since N T ×f N ⊥ is a CR-warped product in Cm , we may choose the
following initial conditions:
φ(1, 0, . . . , 0) = (1, 0, . . . , 0, . . . , 0),
φz1 (1, 0, . . . , 0) = (1, 0, . . . , 0, . . . , 0),
(p + 2)-th
z}|{
φz2 (1, 0, . . . , 0) = (0, 0, . . . , 0, 1 , 0, . . . , 0),
······ (13.92)
(αp − p + α)-th
z}|{
φzα (1, 0, . . . , 0) = (0, . . . , 0, 1 , 0, . . . , 0),
(1 + αp + α)-th
z}|{
φzα+1 (1, 0, . . . , 0) = (0, . . . , 0, 1 , 0, . . . , 0),
······
(αp + h)-th
z}|{
φzh (1, 0, . . . , 0) = (0, . . . , 0, 1 , 0, . . . , 0),
(p + 3)-th (1 + αp − p + α)-th
z}|{ z}|{
φu1 (1, 0, . . . , 0) = (0, 1, . . . , 0, 1 , 0, . . . , 0, 1 , 0, . . . , 0),
······
(p + 1)-th α(p + 1)-th
z}|{ z}|{
φup (1, 0, . . . , 0) = (0, . . . , 0, 1 , 0, . . . , 0, 1 , 0, . . . , 0).
Now, by applying (13.91) and (13.92) we derive that

φ = w0 z1 , . . . , wp z1 , . . . , w0 zα , . . . , wp zα ,
(13.93)
zα+1 , . . . , zh , 0, . . . , 0 ,
where
p
Y
w0 = cos ut , w1 = sin u1 , w2 = sin u2 cos u1 , . . . ,
t=1
p−1
Y
wp+1 = sin up cos ut .
t=1
Since φ is an immersion, (13.93) implies that N T is contained in Ch∗ . This
13.3 CR-warped products in CP m satisfying the equality
The next theorem from [Chen (2003a)] classifies all CR-warped product
submanifolds in the complex projective m-space CP m (4) satisfying the
equality case of inequality (13.1) with c = 1.
Theorem 13.4. Let φ : N T ×f N ⊥ → CP m (4) be a CR-warped product.

Then we have:

||σ||2 ≥ 2p |∇(ln f )|2 − ∆(ln f )} + 4hp. (13.94)
(2) The CR-warped product satisfies the equality case of (13.94) if and only
if the following statements hold.
(2.a) N T is an open portion of complex projective h-space CP h (4).

(2.b) N ⊥ is an open portion of the unit p-sphere S p .
(2.c) There exists a natural number α ≤ h such that, M is a partial Segre
CR-warped product such that φ = π ◦ φ̆, where π : Cm+1 ∗ → CP m (4)
is the natural projection and
(α)
φ̆(z, w) = Shp (z, w), 0, . . . , 0 , z ∈ Ch+1
∗ , w ∈ S p ⊂ Ep+1
(α)
and Shp is a partial Segre embedding defined by (11.27).
Proof. Let φ : M = N T ×f N ⊥ → CP m (4) be a CR-warped product in

CP m (4). Then we obtain inequality (13.94) as a special case of (13.1) with
c = 1. As before, we denote by H and H⊥ the holomorphic and the totally
real distributions of Nt ×f N ⊥ , respectively.
Let M̆ and M̂ be the corresponding manifolds induced from M vis
˘ ∇
Hopf’s fibration as described in Section 12.4. Let ∇, ˆ and ∇ denote the
Levi-Civita connections on M̆ , M̂ and M respectively. Denote by σ̂ the sec-
ond fundamental form of the lift φ̂ : M̂ → S 2m+1 of φ via Hopf’s fibration.
Then we have
ˆ X ∗ Y ∗ = (∇X Y )∗ − hP X, Y i V,
∇ (13.95)
ˆ V X∗ = ∇
∇ ˆ X ∗ V = (P X)∗ , (13.96)
ˆ V V = 0,
∇ (13.97)
∗ ∗ ∗ ∗ ∗
σ̂(X , Y ) = (σ(X, Y )) , σ̂(X , V ) = (F X) , σ̂(V, V ) = 0, (13.98)
for vector fields X, Y tangent to M , where P X and F X are the tangential
and the normal components of JX, respectively.
For a vector U tangent to M̂ ⊂ S 2m+1 ⊂ Cm+1 ∗ , we extend U to a
vector field, also denoted by U , in Cm+1
∗ by parallel translation along the
rays of the cone M̆ over M̂ . We obtain from (12.73) that
1
σ̆(U, W )(t, q) = σ̂(U, W )(q), t ∈ R∗ , q ∈ M̂ , (13.99)
t
∂ ∂ ∂
σ̆ U, = σ̆ , = 0, (13.100)
∂t ∂t ∂t
for U, W tangent to M̂ , where σ̆ denotes the second fundamental form of
the lift φ̆ : M̆ → Cm+1∗ of φ via π. Let Ĥ denote the distribution on
M̂ = ϕ−1 (M ) spanned by H∗ = {X ∗ , X ∈ H} and V = iz, where X ∗ is
a horizontal lift of X via ϕ. Since H is integrable, (13.95)-(13.97) implies
that the distribution Ĥ is also integrable. From (13.95)-(13.97), we know
that each leaf of Ĥ is a totally geodesic submanifold of M̂ .
Let Ĥ⊥ = {Z ∗ ∈ T M̂ : Z ∈ H⊥ }, i.e., Ĥ⊥ is the orthogonal comple-

mentary distribution of Ĥ in T M̂. For vector fields Z, W in H⊥ , (13.95)
implies
ˆ Z ∗ W ∗ = (∇Z W )∗ .
∇ (13.101)
Since H⊥ is integrable, (13.101) implies that Ĥ⊥ is also an integrable dis-
tribution.
On the other hand, (13.32) gives
h∇W Z, Xi = −(X ln f ) hZ, W i (13.102)
for vector field X ∈ H and Z, W ∈ H⊥ . Hence, by (13.101), (13.102),
h(∇Z W )∗ , V i = 0, and the fact that the Hopf fibration is a Riemannian
submersion, we obtain
h∇ ˆ Z ∗ W ∗ , V i = 0.
ˆ Z ∗ W ∗ , X ∗ i = −(X ln f ) hZ ∗ , W ∗ i , h ∇ (13.103)
Thus each leaf of Ĥ⊥ is an extrinsic sphere in M̂ . Therefore, by applying
a result of [Hiepko (1979)], we know that M̂ is also a warped product
N̂ T ×fˆ N∗⊥ , where N̂ T is a leaf of Ĥ, N∗⊥ a horizontal lift of N ⊥ and fˆ the
warping function.
From the definitions of Ĥ, N̂ T and ϕ, we may choose N̂ T to be ϕ−1 (N T ).
Since the Hopf fibration ϕ : S 2m+1 → CP m (4) is a Riemannian submersion,
dϕ preserves the length of vectors normal to fibres. Therefore, the warping
function fˆ of N̂ T ×fˆ N∗⊥ is given by f ◦ ϕ.
Since M̆ is the punctured cone over M̂ with 0 as its vertex, M̆ is nothing
but N̆ T ×tf˘ N̆ ⊥ , where N̆ T = π −1 (N T ), f˘ = f ◦ π, and N̆ ⊥ is a horizontal
lift of N ⊥ via π. Because N̆ ⊥ is isometric to N ⊥ , M̆ is thus isometric to
N̆ T ×tf˘ N ⊥ . Now, it follows from our constructions that N̆ T = π −1 (N T )
is a complex submanifold of Cm+1
∗ and N̆ ⊥ is a totally real submanifold
in C∗ . Consequently, M̆ is isometrically immersed in Cm+1
m+1
∗ as a CR-
warped product.
Now, suppose that φ : M = N T ×f N ⊥ → CP m (4) satisfies the equality
case of (13.94). Then we obtain from (13.14) and (13.1) that
σ(H, H) = 0, σ(H⊥ , H⊥ ) = 0. (13.104)
∂
Let H̆ be the distribution on M̆ spanned by Ĥ and and H̆⊥ the
∂t
⊥
orthogonal distribution of H̆ in T M̆. Then H̆ is spanned by vectors in
Cm+1
∗ obtained from Ĥ⊥ by parallel translation along rays of the cone M̆
over M̂ . Thus, from (13.98), (13.99) and the second equation of (13.104),
we obtain
σ̆(H̆⊥ , H̆⊥ ) = 0. (13.105)
Also, after applying (13.98)-(13.100) and the first equation in (13.104), we

obtain
σ̆(H̆, H̆) = 0. (13.106)
Hence it follows from (13.14) that π −1 (M ) = N̆ T ×tf˘ N ⊥ satisfies the
corresponding equality:
n o
˘
||σ̆||2 = 2p |∇(ln tf˘)|2 − ∆(ln
˘ tf˘)
in Cm+1
∗ . Therefore Theorem 13.3 implies that, up to rigid motions, the
immersion of M̆ is the φ̆ defined defined in (2.c) of this theorem for some
natural number α satisfying α ≤ h. Consequently, up to rigid motions, φ
is given by the composition π ◦ φ̆.
Conversely, it is straightforward to verify that the immersion φ̆ is a
CR-warped product Ch+1 ∗ ×f S p in Cm+1 which is invariant under the C∗ -
action. Therefore the projection π ◦ φ̆ of φ̆ under π : Cm+1∗ → CP m (4)
m
defines a submanifold M in CP (4).
It is direct to verify that M is a CR-warped product CP h (4) ×f˜ S p in
CP m (4) for some suitable warping function f˜. Moreover, it follows from
(13.98) that the CR-warped product M satisfies the condition (13.104).
Consequently, after applying (13.14), we conclude that M = π(Ch+1 ∗ ×f S p )
satisfies the equality case of (13.94) identically. This completes the proof
of the theorem.
13.4 CR-warped products in CH m satisfying the equality
The next theorem from [Chen (2003a)] classifies all CR-warped product
submanifolds in the complex hyperbolic m-space CH m (−4) which satisfy
the equality case of (13.1) with c = −1.
Theorem 13.5. Let φ : N T ×f N ⊥ → CH m (−4) be a CR-warped product.

Then we have:

||σ||2 ≥ 2p |∇(ln f )|2 − ∆(ln f )} − 4hp. (13.107)
(2) The CR-warped product satisfies the equality case of (13.107) if and
only if the following three statements hold.
(2.a) N T is an open portion of complex hyperbolic h-space CH h (−4);
(2.b) N ⊥ is an open portion of unit p-sphere S p (or R, when p = 1); and
(2.c) φ is the composition π ◦ φ̆, where π : Cm+1 1∗ → CH m (−4) is the

natural projection and φ̆ is either congruent to
(α)
φ̆(z, w) = Shp (z, w), 0, . . . , 0 ,
or congruent to
φ̆(z, u) = z0 cosh u, z0 sinh u, z1 cos u, z1 sin u, . . . ,

zα cos u, zα sin u, zα+1 , . . . , zh , 0, . . . , 0 ,
for some integer α ∈ {1, . . . , h}, where z ∈ Ch+1 p
1∗ , w ∈ S (1) ⊂ E
p+1
and u ∈ R.
Proof. This can be done in a similar way as Theorem 13.4.
13.5 Irreducibility of real hypersurfaces in non-flat complex

space forms
A Riemannian manifold M is called reducible if it is locally a Riemannian

product of two or more Riemannian manifolds of positive dimension.
The main result of this section is the following.
Theorem 13.6. [Chen and Maeda (2003)] Every real hypersurface of a

non-flat complex space form is irreducible.
To prove this theorem we need several lemmas.
Lemma 13.1. Let N1 × N2 be a real hypersurface of a non-flat complex

space form M̃ n (4c) with dim N1 = n1 and dim N2 = n2 . Then exactly one
of the following two cases occur:
(1) n1 = 1 or n2 = 1.
(2) n ≥ 3 and either (n1 , n2 ) = (n, n − 1) or (n1 , n2 ) = (n − 1, n).
Moreover, Case (2) occurs only when, restricted to some open dense subset
of M , N1 and N2 are both purely real submanifolds of M̃ n (4c).
Proof. Assume that M = N1 × N2 is a real hypersurface of M̃ n (4c).

Then n1 + n2 = 2n − 1. Thus either n1 ≥ n or n2 ≥ n. If n = 2, we have
(n1 , n2 ) = (1, 2) or (n1 , n2 ) = (2, 1). Hence case (1) occurs. So, without
loss of generality, we may assume that n ≥ 3 and n1 > n ≥ 3.
For a given point x = (x1 , x2 ) ∈ N1 × N2 , if we put
Dxj j = Txj Nj ∩ J(Txj Nj ) and U j = {x ∈ M : dim Dxj j > 0 }, j = 1, 2.
then the Dimension Formula implies U 1 = M . Thus there is a non-empty

connected open subset U of M on which the dimension of D1 is a positive
constant. We shall work on U instead of M .
Clearly, the restriction of D1 on U is a distribution on U × N2 . We
denote this distribution also by D1 .
Let X be a unit vector field in D1 and Z a unit vector field in T N2 .
Then we have ∇Z X = ∇Z (JX) = 0. Thus, by the formulas of Gauss and
Weingarten, we obtain
σ(JX, Z) = ∇ ˜ Z (JX) = J ∇
˜ Z X = Jσ(X, Z). (13.108)
Since Jσ(X, Z) is tangent to M , (13.108) gives
σ(X, Z) = σ(JX, Z) = 0, X ∈ D1 , Z ∈ T N2 . (13.109)
Since the sectional curvature of M satisfies K(X, Z) = 0, Gauss’ formula
and (13.109) give
0 = K̃(X, Z) + hσ(X, X), σ(Z, Z)i , (13.110)
where K̃(X, Z) is the sectional curvature of the plane section X ∧ Z on
M̃ n (4c). Since X, Z are orthonormal vectors spanning a totally real plane,
(6.5) and (13.110) give
λ(X)µ(Z) = −c 6= 0, (13.111)
where λ(X) and µ(Z) are defined by σ(X, X) = λ(X)ξ and σ(Z, Z) =
µ(Z)ξ with ξ being a unit normal vector field of M .
It follows from (13.111) that λ(X) and µ(Z) are independent of the
choice of X and Z, respectively. Thus
σ(X, X) = λ hX, Xi ξ, σ(Z, Z) = µ hZ, Zi ξ, (13.112)
1
for X ∈ D and Z ∈ T N2 . Since N2 is totally geodesic in M , the sec-
ond equation in (13.112) implies that N2 is totally umbilical in M̃ n (4c).
Therefore, by applying Theorem 1 of [Chen and Ogiue (1974b)], we obtain
either
(a) n2 = 1 or
(b) n2 ≥ 2 and N2 is a real space form immersed in M̃ n (4c) as a totally
real submanifold whose mean curvature vector H2 is perpendicular to
J(T N2 ).
If case (b) occurs, then H2 is parallel to ξ according to (13.112), which
implies J(T N2 ) ⊂ T N1 . Hence
∇Z JW + σ(Z, JW ) = J∇Z W + Jσ(Z, W ) (13.113)
for Z, W in T N2 . Since J∇Z W ∈ J(T N2 ) ⊂ T N1 , (13.113) yields

σ(Z, JW ) = 0, Z, W ∈ T N2 . (13.114)
Therefore, by applying the equation of Gauss, we get
0 = K(Z, JW ) = K̃(Z, JW ) + hσ(Z, Z), σ(JW, JW )i (13.115)
for any unit vectors Z, W tangent to N2 . Since Z, JW span a totally real
plane for orthonormal vectors Z, W in T N2 , (13.112) and (13.115) imply
λµ = hσ(Z, Z), σ(JW, JW )i = −c 6= 0. (13.116)
λµ = hσ(Z, Z), σ(JZ, JZ)i = −K̃(Z, JZ) = −4c. (13.117)
Combining (13.116) and (13.117) gives c = 0, which is a contradiction.
Thus, if n1 > n, we must have n2 = 1. Consequently, exactly one of case
(1) or case (2) occurs.
Next, suppose that case (2) occurs. Let us assume that D1 contains a
non-empty open subset V of M . Then, by applying the same argument as
case (i) to V instead of U , we conclude that n2 = 1 which is a contradiction.
Similarly, D2 does not contain any non-empty open subset of M . Hence,
when case (2) occurs, then, restricted to some open dense subset of M , N1
and N2 are both purely real. This completes the proof of lemma.
Now, we consider the case (1) and case (2) of Lemma 13.1 separately.
Case (1): dim N2 = 1. First, assume that Jξ ∈ T N1 . In this case, we
may choose an orthonormal frame e1 , . . . , e2n−1 on M in such a way that
e1 , . . . , e2n−2 are tangent to N1 , e2n−1 tangent to N2 and e1 = Jξ, e3 =
Je2 , . . . , e2n−3 = Je2n−4 , e2n−1 = Je2n−2 . Clearly, the distribution D1 is
spanned by e3 , . . . , e2n−3 .
By applying the equations of Gauss and Weingarten’s formulas and also
∇e2n−1 e2n−1 = ∇e2n−1 e2n−2 = 0, we find
˜ e2n−1 e2n−1 = Jσ(e2n−1 , e2n−1 )
−σ(e2n−1 , e2n−2 ) = J ∇
which implies that σ(e2n−1 , e2n−2 ) = σ(e2n−1 , e2n−1 ) = 0. Hence, by using
Gauss’ equation, we find 0 = K(e2n−1 , e2n−2 ) = 4c which is a contradiction.
Therefore we find Jξ ∈ / T N1 .
Next, let us assume Jξ ∈ T N2 . Then N1 is a complex submanifold and
N2 is a totally real submanifold of M̃ n (4c). In this case, M is a proper
CR-product. But this is impossible according to Remark 9.2. Hence, we
also have Jξ ∈/ T N2 . Consequently, there exist a function α such that
Jξ = cos αe2n−2 + sin αe2n−1 , sin α cos α 6= 0, (13.118)
where e2n−2 is a unit vector tangent to N1 and e2n−1 a unit vector tangent
to N2 . It follows from (13.118) that
Je2n−1 = − sin αξ − cos αe2n−3 (13.119)
for some unit vector e2n−3 ∈ T N1 with he2n−3 , e2n−2 i = 0.
Since hJe2n−3 , ξi = − he2n−3 , Jξi = 0, (13.118) gives
Je2n−3 = − sin αe2n−2 + cos αe2n−1 . (13.120)
Using (13.118) and (13.119) , we find
Je2n−2 = − cos αξ − sin αe2n−3 . (13.121)
Clearly, e2n−3 , e2n−2 , e2n−1 , ξ span a complex 2-plane Hx at each point
x ∈ M and Dx1 is the orthogonal complementary subspace of Hx .
Since N2 is totally geodesic in M , (13.120) and the formulas of Gauss
and Weingarten imply
Jσ(V, e2n−1 ) = J ∇ ˜ V e2n−1 = ∇˜ V Je2n−1
= −(cos α){(V α)ξ + ∇V e2n−3 + σ(V, e2n−3 )} (13.122)
+ sin α{(V α)e2n−3 + AV },
for V ∈ T M , where A = Aξ is the shape operator. Using (13.122), we get
(V α) = − hAV, e2n−3 i , V ∈ T M. (13.123)
Also, by taking the inner product of (13.122) with e2n−2 , we get
h∇V e2n−3 , e2n−2 i = tan α hAV, e2n−2 i − hAV, e2n−1 i , V ∈ T M. (13.124)
Moreover, by taking the inner product of (13.122) with X ∈ D1 , we find
h∇V e2n−3 , Xi = tan α hAV, Xi , X ∈ D1 , V ∈ T M. (13.125)
In particular, if V = e2n−1 , (13.124) and (13.125) reduce to
σ(e2n−1 , e2n−1 ) = (tan α)σ(e2n−1 , e2n−2 ), σ(e2n−1 , X) = 0,
for X ∈ D1 .
We summary the above results as the following.
Lemma 13.2. Let N1 × N2 be a real hypersurface of a non-flat complex
space form M̃ n (4c). If dim N2 = 1 and n ≥ 3, then we have
V α = − hAV, e2n−3 i , (i)
h∇V e2n−3 , e2n−2 i = tan α hAV, e2n−2 i − hAV, e2n−1 i , (ii)
h∇V e2n−3 , Xi = tan α hAV, Xi , (iii)
σ(e2n−1 , e2n−1 ) = (tan α)σ(e2n−1 , e2n−2 ), (iv)
σ(e2n−1 , X) = 0 (v)
1
for X ∈ D and V ∈ T M.
For simplicity, we put σAB = hAeA , eB i. Using (i) we find

(∇¯ e2n−2 σ)(e2n−3 , e2n−3 ) = −(e2n−2 e2n−3 α)ξ
2n−2
X j (13.126)
−2 ω2n−3 (e2n−2 )σj 2n−3 ξ.
j=1
From Lemma 13.2(i), we also have

e2n−2 e2n−3 α − e2n−2 e2n−3 α
2n−2
X j j (13.127)
= {ω2n−2 (e2n−3 ) − ω2n−3 (e2n−2 )}σj 2n−3 .
j=1
By applying the equation of Codazzi and (13.126)-(13.127), we obtain

2n−2
X 2n−2
X
j j
ω2n−3 (e2n−3 )σj 2n−2 = ω2n−3 (e2n−2 )σj 2n−3 . (13.128)
j=1 j=1
On the other hand, from Lemma 13.2(iii), we have

j
ω2n−3 (e2n−2 ) = (tan α)σj 2n−2 ,
j
(13.129)
ω2n−3 (e2n−3 ) = (tan α)σj 2n−3 .
Substituting these into (13.128), we get
2n−2 2n−2
ω2n−3 (e2n−3 )σ2n−2 2n−2 = ω2n−3 (e2n−2 )σ2n−3 2n−2 . (13.130)
We find from Lemma 13.2(ii) and (13.130) that
σ2n−3 2n−1 σ2n−2 2n−2 = σ2n−2 2n−1 σ2n−3 2n−2 . (13.131)
We also have from Lemma 13.2(i) that
¯ e2n−1 σ)(e2n−3 , e2n−3 ) = −(e2n−1 e2n−3 α)ξ,
(∇ (13.132)
¯ e2n−3 σ)(e2n−1 , e2n−3 ) = −(e2n−3 e2n−1 α)ξ
(∇ (13.133)
2n−2
X j
− ω2n−3 (e2n−3 )σj 2n−1 ξ.
j=1
Moreover, from (6.5), (13.119) and (13.120) we find

⊥
R̃(e2n−1 , e2n−3 )e2n−3 = 0. (13.134)
Also, we have e2n−1 e2n−3 α − e2n−1 e2n−3 α = [e2n−1 , e2n−3 ]α = 0. Thus, by
the equation of Codazzi and (13.132)-(13.134), we obtain
2n−2
X j
ω2n−3 (e2n−3 )σj 2n−1 = 0. (13.135)
j=1
Hence, by Lemma 13.2(v) and (13.135) we get

2n−2
ω2n−3 (e2n−3 )σ2n−2 2n−1 = 0,
which is equivalent to
σ2n−2 2n−1 {σ2n−3 2n−1 − (tan α)σ2n−3 2n−2 } = 0, (13.136)
by Lemma 13.2(ii). It follows from (i) and (v) of Lemma 13.2 that
¯ e2n−1 σ)(e2n−3 , e2n−2 ) = −(e2n−1 e2n−2 α)ξ,
(∇ (13.137)
¯ e2n−2 σ)(e2n−1 , e2n−3 ) = −(e2n−2 e2n−1 α)ξ
(∇ (13.138)
2n−2
− ω2n−3 (e2n−2 )σ2n−2 2n−1 ξ.
From (6.5), (13.119)-(13.121) we find
⊥
R̃(e2n−1 , e2n−2 )e2n−3 = −c ξ. (13.139)
Since e2n−1 e2n−2 α = e2n−1 e2n−3 α, (13.137)-(13.139) and the equation of
Codazzi give
2n−2
ϕω2n−3 (e2n−2 ) = −c 6= 0, ϕ = σ2n−2 2n−1 . (13.140)
From (13.136), (13.137) and (13.140) we find σ2n−22n−1 6= 0 and
2n−2
ω2n−3 (e2n−3 ) = 0, σ2n−3 2n−1 = (tan α)σ2n−3 2n−2 . (13.141)
By substituting the second equation of (13.141) into (13.130), we find

σ2n−3 2n−2 ϕ − (tan α)σ2n−2 2n−2 = 0. (13.142)
Since (13.140) gives
2n−3
ϕ − (tan α)σ2n−2 2n−2 = ω2n−2 (e2n−2 ) 6= 0,
we find σ2n−3 2n−2 = 0 from (13.142). Hence (13.141) yields σ2n−3 2n−1 = 0.
Hence, by Lemma 13.2(i), we obtain e2n−1 α = e2n−2 α = 0. Consequently,
we derive
σ2n−3 2n−2 = σ2n−3 2n−1 = e2n−1 α = e2n−2 α = 0. (13.143)
Also, we find from (13.137), (13.143) and Lemma 13.2(i) that
¯ e2n−1 σ)(e2n−3 , e2n−2 ) = 0.
(∇ (13.144)
On the other hand, from statements (i) and (v) of Lemma 13.2, (13.141)
and (13.143), we have
¯ e2n−3 σ)(e2n−1 , e2n−2 ) = (e2n−3 ϕ)ξ.
(∇ (13.145)
Also, from (6.5), (13.119)-(13.121) we find

(R̃(e2n−1 , e2n−3 )e2n−2 )⊥ = c(1 − 3 cos2 α)ξ. (13.146)
Hence, by applying the equation of Codazzi and (13.144)-(13.146), we get
e2n−3 ϕ = c(3 cos2 α − 1). (13.147)
Also, we find rom (2.23) that
¯ e2n−3 σ)(e2n−1 , e2n−1 ) = (e2n−3 σ2n−12n−1 )ξ.
(∇ (13.148)
On the other hand, from Lemma 13.2(i) and (13.143), we have
¯ e2n−1 σ)(e2n−3 , e2n−1 ) = 0.
(∇ (13.149)
From (6.5), (13.119)-(13.121) we find
(R̃(e2n−3 , e2n−1 )e2n−1 )⊥ = 3c sin α cos αξ. (13.150)
Hence, by using the equation of Codazzi and (13.148)-(13.150), we get
e2n−3 σ2n−12n−1 = 3c sin α cos α. (13.151)
We obtain Lemma 13.2(iv), (13.147) and (13.151) that
3c sin α cos α = e2n−3 (ϕ tan α)
= ϕ(sec2 α)e2n−3 α + c(3 cos2 α − 1) tan α
which implies e2n−3 α = (c/ϕ) sin α cos α. Thus, by applying Lemma 13.2(i)
we find
c
σ2n−3 2n−3 = − sin 2α.
2ϕ
Hence, by using Lemma 13.2(iv), (6.5), (13.119), (13.143), Gauss’ equation
and also the last equation, we derive that
0 = K(e2n−3 , e2n−1 ) = 4c cos2 α.
But this is a contradiction. Hence case (1) cannot occur.
Case (2): n ≥ 3, n1 = n and n2 = n − 1. We know from Lemma 13.1 that,
restricted to some open dense subset Û of M , N1 and N2 are purely real
submanifolds of M̃ n (4c). We shall only work on Û to derive a contradiction.
Without loss of generality, we may just simply assume that Û = M .
Since dim N1 = n and N1 is a purely real submanifold of M̃ n (4c), Jξ
cannot be tangent to N2 at every point x ∈ M = N1 × N2 . Thus
Jξ = cos αe1 + sin αen+1 , cos α 6= 0 (13.152)
for some unit vectors e1 ∈ T N1 , en+1 ∈ T N2 .
Let H = T M ∩ J(T M ) denote the maximal complex subbundle of T M .

Then H is the orthogonal complementary subbundle of the complex line
bundle spanned by ξ, Jξ. Put
Hj = H ∩ T Nj , j = 1, 2. (13.153)
1
Since n ≥ 3, n1 = n and n2 = n − 1, we have rank (H ) = n − 1 and
rank (H2 ) = n − 1 or n − 2 according to sin α = 0 or sin α 6= 0, respectively.
We need the following.
Lemma 13.3. In Case (2) we have the following.
σ(Z, JX) = 0, Z ∈ T N2 , X ∈ H1 , (a)
2
σ(Y, JW ) = 0, Y ∈ T N1 , W ∈ H , (b)
2
hAV, JW i = sin α h∇V en+1 , W i , W ∈ H , (c)
for any V ∈ T M .
Proof. For vector fields X in H1 and Z in T N2 , the formulas of Gauss
and Weingarten give
∇Z JX + σ(Z, JX) = Jσ(Z, X),
which implies formula (a). Similarly, we have formula (b).
For vector V ∈ T M , we have
−JAV = J ∇ ˜V ξ = ∇
˜ V Jξ.
Thus we obtain formula (c) from (13.152).
Let X ∈ T N1 , Z ∈ T N2 and U, V ∈ T M . Then we obtain from Gauss’
equation that
0 = R̃(Z, U, V, X) + hσ(Z, X), σ(U, V )i − hσ(Z, V ), σ(U, X)i . (13.154)
From (6.5) we have
R̃(Z, U, V, X) = c {− hZ, V i hU, Xi + hJU, V i hJZ, Xi
(13.155)
+ hZ, JV i hJU, Xi + 2 hZ, JU i hJV, Xi}.
It follows from (a) and (b) of Lemma 13.3, (13.152) and (13.155) that
Lemma 13.4. In Case (2), we have the following:
hσ(Z, X), σ(V, JY )i = c {hX, V i hZ, JY i − hV, Y i hJZ, Xi (d)
+ 2 hX, Y i hZ, JV i}
for X, V ∈ T N1 , Z ∈ T N2 and Y ∈ H1 ;
hσ(X, Z), σ(W, JP )i = c {hZ, W i hX, JP i − hW, P i hJX, Zi (e)
+ 2 hZ, P i hX, JW i}
for X ∈ T N1 , Z, W ∈ T N2 and P ∈ H2 .
Case (2-a): Jξ = e1 ∈ T N1 . In this case, we find sin α = 0 from (13.152)

and
H = {V ∈ T M : hX, e1 i = 0}, H2 = T N2 . (13.156)
Hence we obtain from Lemma 13.3(c) that
σ(V, JW ) = 0, V ∈ T M, W ∈ T N2 . (13.157)
If σ(Z, e1 ) = 0 for all Z ∈ T N2 , then from the equation of Gauss we have
hσ(Z, Z), σ(e1 , e1 )i = −c (13.158)
for any unit Z ∈ T N2 .
From (13.158) we obtain σ11 6= 0 and σ(Z, Z) = σ(W, W ) for any unit
Z, W ∈ T N2 . Since N2 is totally geodesic in M , this implies that N2 is
totally umbilical in M̃ n (4c). Because dim N2 ≥ 2, Theorem 1 of [Chen and
Ogiue (1974b)] implies that N2 is a totally real submanifold in M̃ n (4c) such
that ξ is perpendicular to J(T N2 ). Hence, by applying (13.154) and the
equation of Gauss, we obtain
hσ(Z, Z), σ(JZ, JZ)i = −4c,
(13.159)
hσ(W, W ), σ(JZ, JZ)i = −c,
for orthonormal vectors Z, W ∈ T N2 . Clearly, this is impossible, since
c 6= 0 and σ(Z, Z) = σ(W, W ). Hence, σ(Z, Jξ) 6= 0 for some Z ∈ T N2 .
Therefore, by applying Lemma 13.4(e), we obtain
σ(V, JY ) = 0 for V ∈ H, Y ∈ H1 . (13.160)
Let e2 be an unit vector in H1 . Then there exist a θ ∈ R and unit
vectors e3 ∈ H1 , en+1 ∈ T N2 with he2 , e3 i = 0 such that
Je2 = cos θe3 + sin θen+1 , sin θ 6= 0. (13.161)
When n = 3, (13.161) gives hJe2 , en+1 i = hJe2 , e1 i = 0. Thus (13.161)
implies that
Je3 = − cos θe2 + sin θ η, sin θ 6= 0, (13.162)
where η = en+2 is a unit vector in T N2 with hen+2 , en+1 i = 0.
When n ≥ 4, (13.161) implies
Je3 = − cos θe2 + sin θ η, (13.163)
where η = cos γe4 + sin γen+2 , γ ∈ R with sin γ 6= 0, e4 is a unit vector
in H1 with he4 , e2 i = he4 , e3 i = 0 and en+2 is a unit vector in T N2 with
hen+2 , en+1 i = 0.
From (13.161), (13.162) and (13.163), we get

Jη = − sin θe3 + cos θen+1 , (13.164)
Jen+1 = − sin θe2 − cos θ η. (13.165)
Applying (13.160) with V = ej , j ∈ {2, . . . , n} and Y = e3 and (13.162)-
(13.163), we have
cos θσ2j − sin θ(cos γσ4j + sin γσjn+2 ) = 0. (13.166)
Notice that cos γ = 0 and sin γ = 1 when n = 3.
On the other hand, from Lemma 13.3(b) with Y = ej , j ∈ {2, . . . , n}
and W = en+1 , we find
sin θσ2j + cos θ(cos γσ4j + sin γσjn+2 ) = 0. (13.167)
Combining (13.166) and (13.167) gives σ22 = · · · = σ2n = 0. Also, from
(13.160) with V = Y = e2 and (13.161), we get
cos θσ23 + sin θσ2n+1 = 0.
Thus σ2n+1 = 0. Now, by applying Gauss’ equation again, we obtain
0 = K(e2 , en+1 ) = K̃(e2 , en+1 ) = c(1 + 3 sin2 θ)
which is a contradiction.
Case (2-b): Jξ = cos αe1 + sin αen+1 and sin α cos α 6= 0. Since Je1 is
perpendicular to e1 and en+1 , there exist γ ∈ R, unit vector e2 ∈ H1 and
unit vector en+2 ∈ H2 such that
Je1 = − cos αξ + sin α η,
(13.168)
η = cos γe2 + sin γen+2 .
Jη = − sin αe1 + cos αen+1 ,
(13.169)
Jen+1 = − sin αξ − cos α η.
Clearly, ξ, e1 , en+1 , η span a complex vector subbundle L of rank 2. It is
easy to verify that ζ = − sin γe2 + cos γen+2 is a unit vector perpendicular
to L. Moreover, it is easy to see that
H1 = {X ∈ T N1 : hX, e1 i = 0},
(13.170)
H2 = {Z ∈ T N2 : hZ, en+1 i = 0}.
Assume sin γ = 0. Then we may choose e2 such that
Je1 = − cos αξ + sin αe2 ,
Je2 = − sin αe1 + cos αen+1 , (13.171)
Jen+1 = − sin αξ − cos αe2 .
We get, from Lemma 13.3(a) with X = e2 and Z = en+1 , en+2 , and

(13.171), that
σn+1n+1 = (tan α)σ1n+1 ,
(13.172)
σn+1n+2 = (tan α)σ1n+2 .
Also from Lemma 13.4(d) , we get
σ1n+1 σ(e1 , Je2 ) 6= 0,
(13.173)
σ2n+1 σ(e1 , Je2 ) = σ1n+2 σ(e1 , Je2 ) = 0
which imply σ2n+1 = σ1n+2 = 0. Hence, by applying (13.172), we obtain
σn+1n+2 = 0. Therefore, by applying Gauss’ equation, we find
0 = R̃(e1 , en+2 , en+1 , en+2 ) = σ1n+1 σn+2n+2 .
If σ1n+1 = 0, then (13.172) gives σn+1n+1 = 0. So, by the equation of
Gauss, we get 0 = K(e1 , en+1 ) = K̃(e1 , en+1 ) = c, which is a contradiction.
Similarly, if σn+2n+2 = 0, then the equation of Gauss gives
0 = K(e1 , en+2 ) = K̃(e1 , en+2 ) = c,
which is also a contradiction. Consequently, we have sin γ 6= 0.
Next, we assume cos γ = 0. Then we may choose en+2 such that
Je1 = − cos αξ + sin αen+2 ,
Jen+1 = − sin αξ − cos αen+2 , (13.174)
Jen+2 = − sin αe1 + cos αen+1 .
Using Lemma 13.3(b) with Y = ej and W = en+2 and (13.174), we get
σjn+1 = (tan α)σ1j , j = 1, . . . , n. (13.175)
Also from Lemma 13.4(e) and (13.174), we obtain
σ1n+1 σ(en+1 , Jen+2 ) 6= 0,
(13.176)
σ2n+1 σ(en+1 , Jen+2 ) = σ1n+2 σ(en+1 , Jen+2 ) = 0,
which imply σ2n+1 = σ1n+2 = 0. Hence, by applying (13.175), we get
σ12 = 0. Thus, by the equation of Gauss, we find
0 = R̃(en+1 , e2 , e1 , e2 ) = σ1n+1 σ22 .
If σ1n+1 = 0, then (13.175) yields σ11 = 0. Therefore, by the equation of
Gauss, we get
0 = K(e1 , en+1 ) = K̃(e1 , en+1 ) = c
which is a contradiction.
Similarly, if σ22 = 0, then by σ2n+1 = 0 and the equation of Gauss we

get 0 = K̃(e2 , en+1 ) = c, which is also a contradiction. Therefore we have
cos γ 6= 0. Consequently, we obtain sin γ cos γ sin α cos α 6= 0 in Case (2-b).
Case (2-b-i): n = 3. In this case, for each unit vector e3 in H1 perpendicular
to e2 and e3 is perpendicular to both L and ζ. Since e3 , ζ are orthonormal
vectors which span the orthogonal complementary holomorphic distribution
L⊥ of L, we may thus choose e3 such that
Je3 = − sin γe2 + cos γe5 , cos γ 6= 0. (13.177)
Hence, we also have
Je5 = − sin γ sin αe1 − cos γe3 + sin γ cos αe4 . (13.178)
From (13.168), (13.169), and (13.178) we get
Je2 = − cos γ sin αe1 + sin γe3 + cos γ cos αe4 . (13.179)
Applying Lemma 13.3(a), (13.177) and (13.179), we have
(sin α)σ1t − (tan γ)σ3t − (cos α)σ4t = 0, (13.180)
σ55 = −(tan γ)σ25 . (13.181)
Similarly, we derive from Lemma 13.3(b) with W = e5 and (13.178)
that
(tan γ sin α)σ1j + σj3 − (tan γ cos α)σj4 = 0, j = 1, 2, 3. (13.182)
Also, we find from 13.4(d) that

σit hσ(ej , Jek ), ξi = c δij het , Jek i − δjk hei , Jet i
(13.183)
+ 2δik het , Jej i
for i, j = 1, 2, 3; k = 2, 3; t = 4, 5. Moreover, we obtain from (13.168),
(13.169), (13.177)-(13.179) and (13.183), that
σ14 σ(e1 , Je2 ) = c cos γ cos αξ 6= 0,
σ25 σ(e1 , Je2 ) = 2c sin γ sin αξ 6= 0,
(13.184)
σ14 σ(e1 , Je3 ) = σ14 σ(e2 , Je3 ) = 0,
σ15 σ(e1 , Je3 ) = σ24 σ(e2 , Je3 ) = σ35 σ(e1 , Je3 ) = 0.
From the first two equations of (13.184), we get σ14 , σ25 6= 0 and
σ25 = 2(tan γ tan α)σ14 . (13.185)
Moreover, from the remaining equations of (13.184) we get
σ15 = σ24 = σ34 = σ35 = 0, (13.186)
σ(e1 , Je3 ) = σ(e2 , Je3 ) = 0. (13.187)
Now, by applying σ(e1 , Je3 ) = 0, (13.186) and (13.182) with j = 2, we may

obtain that
σ12 = σ23 = 0. (13.188)
Using σ(e2 , Je3 ) = 0, we find
σ25 = (tan γ)σ22 . (13.189)
By (13.189) and the equation of Gauss, we get
2
0 = K(e2 , e5 ) = c + σ22 σ55 − σ25 .
2
Therefore, after applying (13.181) and (13.189), we derive that 2σ25 = c.
Consequently, we obtain from (13.181), (13.189) and the second equation
of (13.184) that
r r r
c c c
σ25 = , σ22 = cot γ, σ55 = − tan γ, c > 0. (13.190)
2 2 2
From (13.185) and (13.190) we find
√
c
σ14 = √ cot γ cot α. (13.191)
2 2
Also, by applying σ35 = 0, (13.190), and the equation of Gauss for K(e3 , e5 ),
we find
√
σ33 = 2c(1 + 3 cos2 γ) cot γ. (13.192)
It follows from (13.179) with j = 3 and (13.186) that
σ13 = −(cot γ csc α)σ33 .
Hence, by (13.192) we obtain
√
σ13 = − 2c(1 + 3 cos2 γ) cot2 γ csc α. (13.193)
Applying (13.179) and the first equation in (13.184), we get
σ14 (sin γσ13 + cos γ cos ασ14 − cos γ sin ασ11 )
(13.194)
= c cos2 γ cos α.
Combining (13.194) with (13.182) with j = 1, we find
σ14 (− sin ασ11 + cos ασ14 ) = c cos γ cos α. (13.195)
Substituting (13.191) into (13.195), we obtain
√
c
σ11 = √ (cot2 α cot γ − 8 sin γ cos γ). (13.196)
2 2
Substituting (13.191), (13.193) and (13.196) into (13.182) with j = 1 gives

(1 + 3 cos2 γ) cot2 γ + 2 sin2 α sin2 γ c = 0
which is a contradiction. Consequently, we obtain that every real hyper-
surface in a non-flat space form M̃ n (4c) is irreducible if n ≤ 3.
Case (2-b-ii): n ≥ 4. In this case, we have
Je1 = − cos αξ + sin α(cos γe2 + sin γen+2 ), (13.197)
Jen+1 = − sin αξ − cos α(cos γe2 + sin γen+2 ), (13.198)
where sin α cos α sin γ cos γ 6= 0 and e2 ∈ H1 , en+2 ∈ H2 . Moreover, at each
point x ∈ M , the vectors ξ, e1 , en+1 , η = cos γe2 + sin γen+2 span a complex
2-plane Lx ⊂ Tx M̃ n (4c).
Since Je2 is perpendicular to ξ, e2 , en+2 , we derive from (13.197) and
(13.198) that
Je2 = − cos γ sin αe1 + sin γ cos δe3
(13.199)
+ cos γ cos αen+1 + sin γ sin δen+3
for some δ ∈ R, unit vector e3 ∈ H1 with he2 , e3 i = 0, and unit vector
en+3 ∈ H2 with hen+2 , en+3 i = 0.
From (13.197)-(13.199) we get
Jen+2 = − sin γ sin αe1 − cos γ cos δe3
(13.200)
+ sin γ cos αen+1 − cos γ sin δen+3 .
If sin δ = 0, then (13.199) and (13.200) reduce to (13.179) and (13.178),
respectively. In this case, we also have Je3 = − sin γe2 + cos γen+2 with
cos γ 6= 0 from (13.197), (13.198), and (13.199). Hence, in this case the
exact same argument as Case (2-b-i) yields a contradiction. Thus we have
sin δ 6= 0.
If cos δ = 0, then (13.197)-(13.200) reduce to
Je2 = − cos γ sin αe1 + cos γ cos αen+1 + sin γen+3 , (13.201)
Jen+2 = − sin γ sin αe1 + sin γ cos αen+1 − cos γen+3 . (13.202)
Hence, by (13.197), (13.198), and (13.201), we find
Jen+3 = − sin γe2 + cos γen+2 . (13.203)
Using (13.197), (13.198), and Lemma 13.4(d) with X = ej ∈ T N1 ,
V = e1 , Y = e2 , and Z = en+1 , en+2 , we find
σjn+1 σ(e1 , Je2 ) = cδ1j cos γ cos α ξ,
(13.204)
σjn+3 σ(e1 , Je2 ) = cδ1j sin γ ξ, j = 1, 2.
It follows from (13.204) that σ(e1 , Je2 ), σ2 n+1 , σ2 n+3 6= 0. Hence we derive
from the equation of Gauss that
0 = R̃(e2 , en+2 ; en+1 , en+3 ).
On the other hand, from (6.5), (13.200), (13.201), and (13.202) we get
R̃(e2 , en+2 ; en+1 , en+3 ) = c cos α 6= 0,
which is a contradiction. Hence we must have sin δ cos δ 6= 0, too.
Finally, from (13.197), (13.199), and Lemma 13.4(d), we get
σjn+2 σ(e1 , Je2 ) = 2cδ2j sin γ sin α ξ,
(13.205)
σjn+3 σ(e1 , Je2 ) = cδ1j sin γ sin α ξ
for j = 1, 2, 3. From (13.205) we find σ3n+2 = σ3n+3 = 0. Hence, by Gauss’
equation, we get
0 = R̃(e3 , en+2 ; en+3 , en+2 ).
On the other hand, from (6.5), (13.198), and (13.200), we get
R̃(e3 , en+2 ; en+3 , en+2 ) = 3c cos2 γ cos δ sin δ 6= 0,
which is a contradiction. Therefore case (2-b) is impossible as well. Hence
the real hypersurface must be irreducible. This completes the proof of the
theorem.
13.6 Warped product real hypersurfaces
Recall that a contact manifold is an odd-dimensional manifold M 2n+1 with

a 1-form η such that η ∧ (dη)n 6= 0. A curve γ = γ(t) in a contact manifold
is called a Legendre curve if η(β ′ (t)) = 0 along β.
Let S 2n+1 (c) denote the hypersphere in Cn+1 with curvature c centered
at the origin. Then S 2n+1 (c) is a contact manifold endowed with a canonical
contact structure which is the dual 1-form of Jξ, where J is the complex
structure and ξ the unit normal vector on S 2n+1 (c).
Legendre curves are known to play an important role in the study of
contact manifolds, e.g. a diffeomorphism of a contact manifold is a contact
transformation if and only if it maps Legendre curves to Legendre curves.
Let Cn+1 denote the complex Euclidean (n + 1)-space endowed with
Pn+1
metric g = j=1 dzj dz̄j , zj = xj + iyj .
We put
n o
S 2n+1 (c) = (z1 , . . . , zn+1 ) ∈ Cn+1 : hz, zi = c−1 > 0 ,
where h , i denotes the inner product induced from the metric.

The following lemma from [Chen (1997a)] provide a simple relationship
between Legendre curves and a second order differential equation (see also
[Castro and Chen (2006)]).
Lemma 13.5. Let c be a positive number and z = (z1 , z2 ) : I → S 3 (c) ⊂ C2

be a unit speed curve where I is either an open interval or a circle. If
z : I → C2 satisfies
z ′′ (t) − iλγ(t)z ′ (t) + cz(t) = 0 (13.206)
for some nonzero real-valued function λ on I, then it defines a Legendre

curve in S 3 (c).
Conversely, if z defines a Legendre curve in S 3 (c), it satisfies differential
equation (13.206) for some real-valued function λ.
Proof. Suppose z : I → S 3 (c) ⊂ C2 is a unit speed curve which satisfies

(13.206). Then, by taking the derivative of hiz ′ , izi = 0 and by applying
hz, zi = 1c , hz ′ , z ′ i = 1 and equation (13.206), we have hz ′ , izi λ = 0.
Put U = {x ∈ I : λ(x) 6= 0}. If U = I, then hz ′ , izi = 0 identically
on I. Thus z defines a Legendre curve in S 3 (c). Suppose U 6= I, then we
′
have λ = 0 on the open subset I − U . Since hz ′ , izi = 0 on I − U , the
′
continuity implies hz , izi = 0 identically. Therefore z defines a Legendre
curve in S 3 (c).
Conversely, if z defines a Legendre curve in S 3 (c), then hz ′ , izi = 0
identically. Hence we have hz ′′ , izi = 0. Because z, iz, z ′ and iz ′ form an
orthogonal frame field along the Legendre curve, we have
z ′′ (x) = iλ(x)z ′ (x) + k(x)z(x)
for some real-valued functions λ and k.

On the other hand, from hz, zi = c−1 and hz ′ , z ′ i = 1, we also have
′′
hz , zi = −1. Hence we find k(x) = −c which implies (13.206). This proves
the lemma.
Remark 13.1. For any nonzero function λ(t) there exists a (unit speed)
Legendre curve z = z(t) in S 3 (c) satisfying equation (13.206). Such a
Legendre curve is unique if one imposes the initial conditions: z(0) = z0 ∈
S 3 (c) and z ′ (0) = u for some unit vector tangent to S 3 (c).
Similar to Lemma 13.5, we also have the following result from [Chen
(1997a)].
Lemma 13.6. Let c be a negative number and z = (z1 , z2 ) : I → H13 (c) ⊂

C21 be a unit speed curve where I is an open interval. If z : I → C21 satisfies
differential equation
z ′′ (x) = iλz ′ (x) − cz (13.207)
for some nonzero real-valued function λ on I, then it defines a Legendre
curve in H13 (c).
Conversely, if z defines a Legendre curve in H13 (c), then it satisfies
differential equation (13.207) for some real-valued function λ.
Proof. This can be done in the same way as Lemma 13.5.

Contrast to Theorem 13.6, there do exist many warped product real
hypersurfaces in complex space forms.
Theorem 13.7. [Chen (2002c)] Let a be a positive number and γ(t) =

(Γ1 (t), Γ2 (t)) be a unit speed Legendre curve γ : I → S 3 (a2 ) ⊂ C2 defined
on an open interval I. Then

x(z1 , . . . , zn , t) = aΓ1 (t)z1 , aΓ2 (t)z1 , z2 , . . . , zn , z1 6= 0 (13.208)
defines a real hypersurface which is the warped product Cn∗∗ ×a|z1 | I of a
complex n-plane and I, where Cn∗∗ = {(z1 , . . . , zn ) : z1 6= 0}.
Conversely, up to rigid motions of Cn+1 , every real hypersurface in
n+1
C which is the warped product N ×f I of a complex hypersurface N and
an open interval I is either obtained in the way described above or given by
the direct product Cn × γ ⊂ Cn × C1 of Cn and a real curve γ in C1 .
Proof. Let N ×f I be a real hypersurface in Cn+1 which is the warped

product N ×f I of a complex hypersurface N of Cn+1 and an open interval
I. Without loss of generality, we may assume that I contains 0. Then we
obtain from Lemma 11.1 that
σ(H, H) = 0. (13.209)
Since N is totally geodesic in N ×f I, (13.209) implies that N is immersed
as a totally geodesic complex submanifold in Cn+1 . Hence, N is holomor-
phically isometric to a complex Euclidean n-space Cn .
Let z = (z1 , . . . , zn ) be a natural complex coordinate system on the
complex Euclidean n-space Cn . If we put zj = xj + iyj , then the warped
product metric on N ×f I is given by
h
X
g= (dx2k + dyk2 ) + f 2 dt2 . (13.210)
k=1
From (13.210) and a straightforward computation we know that the

Riemannian connection on N ×f I satisfies
∂ ∂ ∂
∇ ∂ =∇ ∂ =∇ ∂ = 0, j, k = 1, . . . , n, (13.211)
∂xk
∂xj ∂xj ∂y
k ∂yj ∂y
k
∂ fxj ∂
∇ ∂ = , j = 1, . . . , n, (13.212)
∂xj ∂t f ∂t
∂ fyj ∂
∇ ∂ = , j = 1, . . . , n, (13.213)
∂yj ∂t f ∂t
Xn
∂ ∂ ∂
∇∂ = −f fxk + fyk , (13.214)
∂t ∂t ∂xk ∂yk
k=1
where fxj = ∂f /∂xj and fyj = ∂f /∂yj .

By applying (13.211)-(13.214) we know that the Riemann curvature
tensor of satisfies

∂ ∂ ∂ ∂2φ ∂φ ∂φ ∂
R , = +
∂xj ∂t ∂xk ∂xj ∂xk ∂xj ∂xk ∂t

∂ ∂ ∂ ∂2φ ∂φ ∂φ ∂
R , = + (13.215)
∂xj ∂t ∂yk ∂xj ∂yk ∂xj ∂yk ∂t
2
∂ ∂ ∂ ∂ φ ∂φ ∂φ ∂
R , = + ,
∂yj ∂t ∂yk ∂yj ∂yk ∂yj ∂yk ∂t
for j, k = 1, . . . , n, where φ = ln f . Now, we obtain from Lemma 11.1(3)
that
∂ ∂ ∂φ ∂
σ , =− J ,
∂xj ∂t ∂yj ∂t
∂ ∂ (13.216)
∂φ ∂
σ , = J , j = 1, . . . , n.
∂yj ∂t ∂xj ∂t
Let us put
∂ ∂ ∂
σ , = λJ , (13.217)
∂t ∂t ∂t
for some function λ = λ(z1 , . . . , zn , t). By applying the equation of Gauss,
(13.215), and (13.216), we obtain
∂2φ ∂φ ∂φ ∂φ ∂φ
= − ,
∂xj ∂xk ∂yj ∂yk ∂xj ∂xk
∂2φ ∂φ ∂φ ∂φ ∂φ
=− − , (13.218)
∂xj ∂yk ∂yj ∂xk ∂xj ∂yk
∂2φ ∂φ ∂φ ∂φ ∂φ
= − ,
∂yj ∂yk ∂xj ∂xk ∂yj ∂yk
for j, k = 1, . . . , n.
Clearly, every constant function φ is a solution of the PDE system
(13.218). However, if φ is constant, the warping function f is constant.
In this case the hypersurface is a CR-product. Thus, by applying Theorem
4.6 of [Chen (1981b)], we know that the real hypersurface is locally the
product of Cn and a curve in C1 .
It follows from Proposition 12.3 that the solution of system (13.218) is
given by
1
φ = ln f, f = {hα, zi2 + hiα, zi2 } (13.219)
2
for some vector α ∈ Cn+1 . So, after choosing a suitable Euclidean complex
coordinates on Cn+1 , we may obtain
α = (b1 + ib2 , 0, . . . , 0).
With respect to this vector α, we have
f = {(b1 x1 + b2 y1 )2 + (−b2 x1 + b1 y1 )2 }1/2 ,
which can also be expressed as
q q
f = a x21 + y12 , a = b21 + b22 . (13.220)
From Gauss’ formula, (13.209), (13.211)-(13.214), (13.216) and (13.220),
we know that the immersion x of N ×f I in Cm satisfies
xxj xk = xxj yk = xyj yk = 0, j, k = 1, . . . , n, (13.221)
x1 − iy1 y1 + ix1
xx1 t = 2 xt , xy1 t = 2 xt , (13.222)
x1 + y12 x1 + y12
xxj t = xyj t = 0, j = 2, . . . , n, (13.223)
2
xtt = −a (x1 xx1 + y1 xy1 ) + iλxt . (13.224)
It is straightforward to verify from (13.221)-(13.224) that the immersion
x satisfies
xxj tt = xttxj , xyj tt = xttyj , j = 1, . . . , n
hold if and only if ∂λ/∂xj = ∂λ/∂yj = 0 and xyj = ixxj for j = 1, . . . , n.
Hence λ = λ(t) is a function of t. Therefore (13.224) reduces to
xtt = −a2 z1 xx1 + iλ(t)xt . (13.225)
n
X n
X
k
x= Â (t)xk + B̂ k (t)yk + C(t) (13.226)
k=1 k=1
for some functions Â1 , . . . , Ân , B̂ 1 , . . . , B̂ n , C of t. Substituting (13.226)

into (13.222) gives
( n n
)
X X
2 2 1 k k
(x1 + y1 )Ât = (x1 − iy1 ) Ât xk + B̂t yk + Ct , (13.227)
k=1 k=1
for j = 1, . . . , n, which implies
Âkt = B̂tk = Ct = 0, k = 2, . . . , n, (13.228)
B̂t1 = i Â1t . (13.229)
2 n 2 n
Condition (13.228) implies that Â , . . . , Â , B̂ , . . . , B̂ and C are con-
stant vectors. We may choose C = 0 by applying a suitable translation if
necessary. Solving (13.229) shows that there is a vector δ1 so that
B̂ 1 = iÂ1 + iδ1 . (13.230)
If we put
Â1 = aγ + δ1 , Âk = βk , B̂ k = iδk , k = 2, . . . , n, (13.231)
then we derive from (13.226), (13.230) and (13.231) that
n
X
x(z1 , . . . , zn , t) = aγ(t)z1 + (βk xk + iδk yk ), (13.232)
k=1
where β1 = δ1 .
Now, by using xyj = ixxj and (13.232), we find δk = βk , k = 1, . . . , n.
Therefore (13.232) gives
n
X
x(z1 , . . . , zn , t) = aγ(t)z1 + βk z k . (13.233)
k=1
Substituting (13.233) into (13.225) yields β1 = δ1 = 0 and
γ ′′ (t) − iλ(t)γ ′ (t) + a2 γ(t) = 0. (13.234)
If we choose the initial conditions:
(k+1)-th
z}|{
xxk (1, 0, . . . , 0) = (0, . . . , 0, 1 , 0, . . . , 0), (13.235)
xt (1, 0, . . . , 0) = (a, 0, . . . , 0), k = 1, . . . , n,
γ(0) = (0, a−1 , 0, . . . , 0),
γ ′ (0) = (1, 0, . . . , 0),
β2 = (0, 0, 1, 0, . . . , 0), (13.236)
···
βn = (0, . . . , 0, 1).
So, combining (13.233) and (13.236) yields
x(z1 , . . . , zn , t) = aγ(t)z1 + (0, 0, z2 , . . . , zn ). (13.237)

Since γ(t) is a solution (13.234), γ(t) can be expressed as
γ(t) = c1 A1 (t) + c2 A2 (t),
where c1 and c2 are constant vectors in Cn+1 and A1 (t), A2 (t) are two
independent solutions of (13.234). Thus the image of γ(t) must lie in the
complex plane, say C2 , spanned by c1 and c2 .
Clearly, C2 is the complex plane defined by
{(w1 , w2 , 0, . . . , 0) : w1 , w2 ∈ C}.
Hence, if we denote the curve γ by γ(t) = (Γ1 (t), Γ2 (t)), then the hyper-
surface is given by
x(z1 , . . . , zn , t) = (aΓ(t)z1 , aΓ(t)z1 , z2 , . . . , zn ). (13.238)

From (13.237) we get hxt , xt i = f 2 |γ ′ (t)|2 . Comparing this with the warped
metric (13.234) of the hypersurface yields |γ ′ (t)| = 1. Hence γ(t) is of unit
speed, so we have hγ ′ (t), γ ′′ (t)i = 0.
Now, by taking the inner product of γ ′ (t) with (13.234), we obtain
hγ ′ (t), γ(t)i = 0. Thus γ(t) has constant length. Therefore, by applying the
first condition on γ(0) in (13.236), we obtain |γ(t)| = a−1 . Consequently,
γ defines a unit speed curve in S 3 (a2 ) so that γ : I → S 3 (a2 ) ⊂ C2 .
Case (a): λ = 0. In this case, the solution of (13.234) is given by γ(t) =
c1 cos(at) + c2 sin(at). Thus, if we choose the initial conditions:
γ(0) = (0, a−1 , 0, . . . , 0), γ ′ (0) = (1, 0, . . . , 0),

then we get
c1 = (0, a−1 , 0, . . . , 0), c2 = (a−1 , 0, . . . , 0).

Hence γ(t) = a−1 sin(at), a−1 cos(at), 0, . . . , 0 is a unit speed Legendre
curve in S 3 (a2 ).
Case (b): λ(t) 6= 0. In this case, the curve γ = γ(t) satisfies the differential
equation (13.234) with λ(t) 6= 0. Thus Lemma 13.6 implies that γ = γ(t)
is a unit speed Legendre curve in S 3 (a2 ).
Conversely, since γ(t) = (Γ1 (t), Γ2 (t)) is a unit speed Legendre curve in
3 2
S (a ), it is easy to verify that the real hypersurface defined by (13.208) is
the warped product Cn∗ ×a|z1 | I of Cn∗ and an open interval I.
Remark 13.2. The real hypersurface of Cn+1 defined by (13.208) is a

non-complete ruled hypersurface. Moreover, a direct computation shows
that the squared mean curvature H 2 and scalar curvature τ of the real
hypersurface are given respectively by
λ2 1
H2 = , τ =− (13.239)
(2n + 1)2 a2 |z1 |2 |z1 |2
where λ is equal to the curvature of the unit speed Legendre curve γ(t) =
(Γ1 (t), Γ2 (t)) in S 3 (a2 ).
Also, it follows from (13.239) that the hypersurface has non-constant
scalar curvature. Moreover, it has non-constant mean curvature unless the
Legendre curve γ is a geodesic in S 3 (a2 ).
Example 13.1. If λ is constant, the unique Legendre curve obtained from

the solution of the differential equation (13.234), which also satisfies the
initial conditions: γ(0) = (a−1 , 0), γ ′ (0) = (0, 1) is given by
i i λt !
e 2 λt bt iλ bt e2 bt
γ(t) = cos − sin , sin (13.240)
a 2 b 2 b 2
√
with b = 4a2 + λ2 . Thus it follows from (13.236) and (13.240) that the
corresponding warped product hypersurface is given by
i i !
e 2 λt bt iλ bt e 2 λt bt
x= cos − sin z1 , sin z1 , z2 , . . . , zn .
a 2 b 2 b 2
The following theorem from [Chen (2002c)] classifies all warped products
real hypersurfaces of the form N ×f I in complex projective spaces.
Theorem 13.8. Let a be a positive number and γ(t) = (Γ1 (t), Γ2 (t)) be a
unit speed Legendre curve γ : I → S 3 (a2 ) ⊂ C2 on an open interval I. If
x : S∗2n+1 × I → Cn+2 is the map defined by
n
X
x(z0 , . . . , zn , t) = aΓ1 (t)z0 , aΓ2 (t)z0 , z1 , . . . , zn , zk z̄k = 1, (13.241)
k=0
then we have:
(1) x induces an isometric immersion ψ : S∗2n+1 ×a|z0 | I → S 2n+3 .
(2) The image ψ(S∗2n+1 ×a|z0 | I) in S 2n+3 is invariant under the action of
U (1).
(3) the projection ψπ : π(S∗2n+1 ×a|z0 | I) → CP n+1 of ψ via π is a warped
product hypersurface CP0n ×a|z0 | I in CP n+1 .
Conversely, if a real hypersurface in CP n+1 is a warped product N ×f I

of a complex hypersurface N of CP n+1 and an open interval I, then, up to
rigid motions, it is locally obtained in the way described above.
Proof. Statement (1) is easy to verify, since γ(t) = (Γ1 (t)Γ2 (t)) is a unit
speed Legendre curve in S 3 (a2 ). Statement (ii) follows from (13.241) and
the definition of the U (1)-action.
Since γ = γ(t) is a Legendre curve in S 3 (a2 ), for each z = (z0 , . . . , zn )
P
with zk z̄k = 1 (13.241) implies that the curve Γz defined by

Γz (t) = aΓ1 (t)z0 , aΓ2 (t)z0 , z1 , . . . , zn
is horizontal in S 2n+3 . So, π : Γz → π(Γz ) is isometric. It is clear that the
restriction of π on S∗2n+1 is also a Riemannian submersion. Therefore the
projection of
ψπ : ψ(S∗2n+1 ×a|z0 | I) → CP n+1 (4)
of ψ : S∗2n+1 ×a|z0 | I → S 2n+3 is a warped product real hypersurface
CP n+1 (4) ×a|z0 | I into CP n+1 (4).
Conversely, assume that M = N ×f I is a warped product hypersurface
of CP n+1 (4), where N is a complex hypersurface of CP n+1 (4) and I is
an open interval. Then, according to Theorem 13.6, the warping function
f is non-constant.
Let H1 and H2 denote the distributions on M spanned by vectors tan-
gent to the N and I, respectively. Trivially, H1 and H2 are integrable
distributions. From Lemma 11.1(1) we know that the second fundamen-
tal form σ of M in CP n+1 (4) satisfies σ(H1 , H1 ) = 0. Since N is totally
geodesic in N ×f I, N is thus totally geodesic in CP n+1 (4). Hence, N is
holomorphically isometric to an open part of a CP n+1 (4).
Let ∇ˆ and ∇ denote the Riemannian connections of M̂ and M respec-
tively. And let σ̂ denote the second fundamental form of M̂ in S 2n+3 . Then
we have
ˆ X ∗ Y ∗ = (∇X Y )∗ − hP X, Y i V,
∇ (13.242)
ˆ V X∗ = ∇
∇ ˆ X ∗ V = (P X)∗ , (13.243)
ˆ V V = 0,
∇ (13.244)
∗ ∗ ∗
σ̂(X , Y ) = (σ(X, Y )) , (13.245)
∗ ∗
σ̂(X , V ) = (F X) , (13.246)
σ̂(V, V ) = 0, (13.247)
for vector fields X, Y tangent to M , where P X and F X denote the tan-

gential and the normal components of JX, respectively.
Let Ĥ1 denote the distribution on M̂ = π −1 (M ) spanned by H1∗ and
V = iz, where H1∗ = {X ∗ : X ∈ H1 }. Since H1 is integrable, (13.242)-
(13.244) implies that Ĥ1 is also integrable.
Also, from (13.245)-(13.247) we know that each leaf of Ĥ1 is totally
geodesic in S 2n+3 . Thus, each leaf of Ĥ is isometric to an open portion of
the unit sphere S 2n+1 .
Clearly, H2∗ = {Z ∗ ∈ T M̂ : Z ∈ H2 } is the orthogonal complementary
distribution of Ĥ1 in T M̂ . Since H2∗ is of rank one, H2∗ is also integrable
and P H2 = {0}. Thus, it follows from (13.242) that
ˆ Z ∗ W ∗ = (∇Z W )∗ , Z, W ∈ H2 .
∇ (13.248)
On the other hand, for any vector field X in H1 , and Z, W in H2 , we
have
h∇Z W, Xi = −(X ln f ) hZ, W i . (13.249)
Applying (13.248), (13.249), (∇Z W )∗ ⊥ V , and the fact that the Hopf
fibration is a Riemannian submersion, we obtain
h∇ ˆ Z∗ W ∗, V i = 0
ˆ Z ∗ W ∗ , X ∗ i = −(X ln f ) hZ ∗ , W ∗ i , h ∇
which implies that each integral curve of Ĥ2 is a circle in M̂ , i.e., a order
two Frenet curve with constant curvature in M̂ . Thus, a result of [Hiepko
(1979)] implies that locally M̂ is a warped product S 2n+1 ×fˆI with warping
function fˆ.
Let M̆ denote the punctured cone over M̂ with 0 as its vertex defined
by

M̆ = tw ∈ Cn+2 : w ∈ M̂ = S 2n+1 ×fˆ I ⊂ S 2n+3 ⊂ Cn+2 , t > 0 .
Since the tangent vector field ∂/∂t on M̆ is parallel to the position vector
field of M̆ in Cn+2 and V is tangent to the first component of S 2n+1 ×fˆ I,
we see that locally M̆ is the warped product Cn+1 ×tfˆ I, where Cn+1 is a
complex hyperplane of Cn+2 . Since the warping function is non-constant,
Theorem 4.1 implies that, up to rigid motions, M̆ is given by

x(z1 , . . . , zn , t) = aΓ1 (t)z1 , aΓ2 (t)z1 , z2 , . . . , zn , z1 6= 0,
for some positive number a and a unit speed Legendre curve γ(t) =
(Γ1 (t), Γ2 (t)) in S 3 (a2 ). Consequently, up to rigid motions, the warped
product hypersurface in CP n+1 (4) is the projection of ψ given by (13.241)
via the Hopf fibration.
Remark 13.3. The real hypersurface M of CP n+1 (4) defined by (13.241)

is a non-complete ruled hypersurface. A direct computation shows that
the squared mean curvature H 2 and the scalar curvature τ of M are given
respectively by
λ2 1
H2 = , τ =− ,
(2n + 1)2 a2 |z0 |2 |z0 |2
where λ is the curvature of the Legendre curve γ(t) = (Γ1 (t), Γ2 (t)) in
S 3 (a2 ). Therefore, the hypersurface M has non-constant mean curvature
in CP n+1 (4) unless the Legendre curve γ is a geodesic of S 3 (a2 ).
Moreover, it is straightforward to verify that the real hypersurface has
non-constant scalar curvature.
The next theorem from [Chen (2002c)] classifies all warped products
hypersurfaces of the form N ×f I in complex hyperbolic spaces.
Theorem 13.9. Let a be a positive number and γ(t) = (Γ1 (t), Γ2 (t)) be a
2n+1
unit speed Legendre curve γ : I → S 3 (a2 ) ⊂ C2 . If y : H1∗ × I → Cn+2
1
be the map defined by
y(z0 , . . . , zn , t) = (z0 , . . . , zn−1 , aΓ1 (t)zn , aΓ2 (t)zn ),
X n
(13.250)
z0 z̄0 − zk z̄k = 1,
k=1
then we have
2n+1
(1) y induces an isometric immersion ψ : H1∗ ×a|zn | I → H12n+3 .
2n+1 2n+3
(2) The image ψ(H1∗ ×a|zn | I) in H1 is invariant under the U (1)-
action.
2n+1
(3) the projection ψπ : π(H1∗ ×a|zn | I) → CH n+1 of ψ via π is a warped
product hypersurface CH∗ ×a|zn | I in CH n+1 .
n
Conversely, if a real hypersurface in CH n+1 is a warped product N ×f I

of a complex hypersurface N and an open interval I, then, up to rigid
motions, it is locally obtained in the way described above.
Proof. This can be proved in a similar way as Theorem 13.8.
Remark 13.4. A direct computation shows that the real hypersurface of
CH n+1 (−4) obtained from (13.250) is a non-complete ruled hypersurface
which has non-constant scalar curvature τ = −|zn |−2 and it also has non-
constant mean curvature, unless the Legendre curve γ is a geodesic in
S 3 (a2 ).
Chapter 14
δ-invariants, Submersions and

Warped Products
Curvatures invariants play key roles in differential geometry as well as in

physics. For instance, the magnitude of a force required to move an object
at constant speed is, according to Newton’s law, a constant multiple of the
curvature of the trajectory. The motion of a body in a gravitational field
is determined by the curvatures of spacetime according to A. Einstein.
One of the most fundamental problems in the theory of submanifolds is
the immersibility of a Riemannian manifold in a Euclidean space (or, more
generally, in a space form). According to the famous embedding theorem
of J. F. Nash, every Riemannian manifold can be isometrically embedded
in some Euclidean spaces with sufficiently high codimension [Nash (1956)].
The Nash embedding theorem was aimed for in the hope that if Rieman-
nian manifolds could always be regarded as Riemannian submanifolds, this
would then yield the opportunity to use extrinsic help. Till when observed
in [Gromov (1985)], this hope had not been materialized however.
There were several reasons why it is so difficult to apply the Nash theo-
rem. One reason is that it requires very large codimension for a Riemannian
manifold to admit an isometric embedding in Euclidean spaces in general.
On the other hand, submanifolds of higher codimension are very difficult to
understand. Another difficulty to apply Nash’s theorem is that there did
not exist general optimal relationships between the known intrinsic invari-
ants and main extrinsic invariants for arbitrary Riemannian submanifolds
of Euclidean spaces. In order to overcome those difficulties, one needs
to introduce new types of Riemannian invariants, different in nature from
“classical” invariants. One also needs to establish general optimal relation-
ships between the main extrinsic invariants with the new intrinsic invariants
on the submanifolds. These were the author’s motivation to introduce his
δ-invariants in the 1990s.
325
14.1 δ-invariants
For a Riemannian n-manifold M , let K(π) denote the sectional curvature

associated with a plane section π ⊂ Tx M, x ∈ M . For an orthonormal
basis e1 , . . . , en of Tx M , the scalar curvature τ at x is defined to be
X
τ (x) = K(ei ∧ ej ).
i<j
Let L be a subspace of Tx M of dimension r ≥ 2 and {e1 , . . . , er } an

orthonormal basis of L. We define the scalar curvature τ (L) of L by
X
τ (L) = K(eα ∧ eβ ), 1 ≤ α, β ≤ r.
α<β
If L is a 2-plane section, τ (L) is the sectional curvature K(L) of L. Ge-

ometrically, τ (L) is nothing but the scalar curvature of the image expx (L)
under the exponential map at x.
Let M be an n-dimensional Riemannian manifold. For an integer k ≥ 0,
denote by S(n, k) the finite set consisting of k-tuples (n1 , . . . , nk ) of integers
≥ 2 satisfying n1 < n and n1 + · · · + nk ≤ n. Denote by S(n) the set of
(unordered) k-tuples with k ≥ 0 for a fixed positive integer n.
The cardinal number #S(n) of S(n) is equal to p(n) − 1, where p(n)
denotes the number of partition of n which increases quite rapidly with n.
For instance, for
n = 2, 3, 4, 5, 6, 7, 8, 9, 10, . . ., 20, . . . , 50, . . . ,
100, . . . , 200,
the cardinal number #S(n) are given respectively by
1, 2, 4, 6, 10, 14, 21, 29, 41, . . ., 626, . . . , 204 225, . . . ,
190 569 291, . . ., 3 972 999 029 387.
For each (n1 , . . . , nk ) ∈ S(n), the invariant δ(n1 , . . . , nk ) was defined in
[Chen (1998a, 2000a)] by
δ(n1 , . . . , nk )(x) = τ (x) − inf{τ (L1 ) + · · · + τ (Lk )}, (14.1)
where L1 , . . . , Lk run over all k mutually orthogonal subspaces of Tx M such
that dim Lj = nj , j = 1, . . . , k. In particular, we have
(1) δ(∅) = τ (k = 0, the trivial δ-invariant),
(2) δ(2) = τ − inf K, where K is the sectional curvature,
(3) δ(n − 1)(x) = max Ric(x).
δ-invariants, Submersions and Warped Products 327
Remark 14.1. The δ-curvatures, also known as Chen invariants, defined

above are very different in nature from the “classical” scalar and Ricci
curvatures; simply due to the fact that both scalar and Ricci curvatures
are the “total sum” of sectional curvatures on a Riemannian manifold. In
contrast, the δ-curvature invariants are obtained from the scalar curvature
by deleting a certain amount of sectional curvatures.
14.2 An inequality for submanifolds in real space forms
Let n be a natural number ≥ 2 and let n1 , . . . , nk be k natural numbers.

Then (n1 , . . . , nk ) is called a partition of n if n1 + · · · + nk = n.
Lemma 14.1. Let a1 , . . . , an be real numbers and k be an integer satisfying

2 ≤ k ≤ n − 1. Then, for any partition (n1 , . . . , nk ) of n, we have
X X
ai1 aj1 + ai2 aj2 + · · ·
1≤i1 <j1 ≤n1 n1 +1≤i2 <j2 ≤n1 +n2
X
+ aik ajk (14.2)
n1 ···+nk−1 +1≤i1 <j1 ≤n
1
(a1 + · · · + an )2 − k(a21 + · · · + a2n ) ,
≥
2k
with the equality holding if and only if
a1 + · · · + an1 = · · · = an1 +···+nk−1 +1 + · · · + an . (14.3)
Proof. Under the hypothesis we find

(
X X
2k ai1 aj1 + ai2 aj2 + · · ·
1≤i1 <j1 ≤n1 n1 +1≤i1 <j1 ≤n1 +n2
) X
n 2 n
X X
+ aik ajk − aα +k a2α
n1 +···+nk−1 +1≤i1 <j1 ≤n α=1 α=1
(
X X
= 2k ai1 aj1 + ai2 aj2 + · · ·
1≤i1 <j1 ≤n1 n1 +1≤i1 <j1 ≤n1 +n2
) n
X X X
+ aik ajk + (k−1) a2α − 2 aα aβ
n1 +···+nk−1 +1≤i1 <j1 ≤n α=1 1≤α<β≤n
( 2
X X
= ai1 − ai2
1≤ai1 ≤n1 n1 +1≤ai2 ≤n1 +n2
( )2
X X
+ ai1 − ai3 + ···
1≤ai1 ≤n1 n1 +n2 +1≤ai3 ≤n1 +n2 +n3
( )2
X X
+ aik−1 − aik
n1 +···+nk−2 +1≤ai1 ≤n1 +···+nk−1 n1 +···+nk−1 +1≤aik ≤n
≥ 0,
with equality holding if and only if (14.3) holds.
An immediate consequence of Lemma 14.1 is the following.
Corollary 14.1. Let a1 , . . . , an , η be n + 1 real numbers such that

n
!2 n
!
X X
ai = (n − 1) η + a2i . (14.4)
i=1 i=1
Then 2a1 a2 ≥ η, with equality holding if and only if

a1 + a2 = a3 = · · · = an .
Proof. Let a1 , . . . , an be real numbers. Then (14.4) is equivalent to

(a1 + · · · + an )2
η= − (a21 + · · · + a2n ).
n−1
By choosing k = n − 1, n1 = 2 and n2 = · · · = nk = 1, (14.2) becomes
2a1 a2 ≥ η, with the equality holding if and only if a1 + a2 = a3 = · · · = an
holds according to Lemma 14.1.
For each (n1 , . . . , nk ) ∈ S(n), we put

k
1 1X
a(n1 , . . . , nk ) = n(n − 1) − nj (nj − 1) (14.5)
2 2 j=1
P
n2 (n + k − 1 − j nj )
b(n1 , . . . , nk ) = P . (14.6)
2(n + k − j nj )
We have the following universal inequality for submanifolds.
Theorem 14.1. [Chen (1998a, 2000a)] For any n-dimensional submanifold

N of a Riemannian m-manifold Rm (c) of constant curvature c and for any
k-tuple (n1 , . . . , nk ) ∈ S(n), we have
δ(n1 , . . . , nk ) ≤ b(n1 , . . . , nk )H 2 + a(n1 , . . . , nk )c, (14.7)
where H 2 is the squared mean curvature of N in Rm (c).
The equality case of inequality (14.7) holds at a point x ∈ N if and only

if, there exists an orthonormal basis e1 , . . . , em at x, such that the shape
operators of N in Rm (c) at x take the following form:
 r 
A1 ... 0
 .. .. .. 
Aer =
 . . . 0 
, r = n + 1, . . . , m, (14.8)
0 ... Ark 
0 µr I
where Arj are symmetric submatrices satisfying
Tr (Ar1 ) = · · · = Tr (Ark ) = µr . (14.9)
Proof. If k = 0, this was done in [Chen (1996a)]. Thus we assume that

k ≥ 1. From the equation of Gauss we know that the scalar curvature and
the squared mean curvature H 2 of N satisfy
2τ = n2 H 2 − ||σ||2 + n(n − 1)c (14.10)
where ||σ||2 is the squared norm of the second fundamental form σ.
Let (n1 , . . . , nk ) ∈ S(n). Put
P
n2 (n + k − 1 − nj ) 2
η = 2τ − n(n − 1)c − P H . (14.11)
n + k − nj
Then from (14.10) and (14.11) we find
X
n2 H 2 = γ η + ||σ||2 , γ = n + k − nj . (14.12)
Let L1 , . . . , Lk be mutually orthogonal subspaces of Tx N with dim Lj =
nj , j = 1, . . . , k. By choosing an orthonormal basis e1 , . . . , em at x such
that
Lj = Span {en1 +···+nj−1 +1 , . . . , en1 +···+nj }, j = 1, . . . , k
and en+1 is in the direction of the mean curvature vector, we obtain from
(14.12) that
!2  
X n X n X Xm Xn
ai = γ η + (ai )2 + n+1 2
(σij ) + r 2
(σij ) , (14.13)
i=1 i=1 i6=j r=n+2 i,j=1
n+1 P
where ai = σii ,i = 1, . . . , n, and γ = n + k − nj .
We put
∆1 = {1, . . . , n1 },
...
∆k = {n1 + · · · + nk−1 + 1, . . . , n1 + · · · + nk }.
Equation (14.13) is equivalent to

!2 
γ+1
X γ+1
X X m
X n
X
āi = γ η + (āi )2 + n+1 2
(σij ) + r 2
(σij )
i=1 i=1 i6=j r=n+2 i,j=1
 (14.14)
X X X
− aα1 aβ1 − aα2 aβ2 − · · · aαk aβk  ,
2≤α1 6=β1 ≤n1 α2 6=β2 αk 6=βk
for α2 , β2 ∈ ∆2 , . . . , αk , βk ∈ ∆k , where
ā1 = a1 , ā2 = a2 + · · · + an1 ,
ā3 = an1 +1 + · · · + an1 +n2 ,
...
āk+1 = an1 +···+nk−1 +1 + · · · + an1 ···+nk ,
āk+2 = an1 ···+nk +1 ,
...
āγ+1 = an .
Applying Corollary 14.1 to (14.14) yields
X X X
aα1 aβ1 + aα2 aβ2 + · · · + aαk aβk
α1 <β1 α2 <β2 αk <βk
X m n (14.15)
η n+1 2 1 X X r 2
≥ + (σij ) + (σ ) ,
2 i<j 2 r=n+2 i,j=1 ij
for αj , βj ∈ ∆j , j = 1, . . . , k. On the other hand, Gauss’ equation yield

nj (nj − 1) X m X
τ (Lj ) = c+ σαr j αj σβr j βj − (σαr j βj )2 ,
2 r=n+1 αj <βj (14.16)
αj , βj ∈ ∆j , j = 1, . . . , k.
Combining (14.15) and (14.16), we get
k
η X nj (nj − 1)
τ (L1 ) + · · · + τ (Lk ) ≥ + c
2 j=1 2
1 X X X r 2
m m k
1 X X
r
+ (σαβ )2 + σαj αj (14.17)
2 r=n+1 2 r=n+2 j=1
/ 2
(α,β)∈∆ αj ∈∆j
k
η X nj (nj − 1)
≥ + c,
2 j=1 2
where
∆ = ∆1 ∪ · · · ∪ ∆k , ∆2 = (∆1 × ∆1 ) ∪ · · · ∪ (∆k × ∆k ).
From (14.1), (14.11), and (14.17), we obtain (14.7). If the equality in
(14.5) holds at a point p, then the inequalities in (14.15) and (14.17) are
actually equalities at x. In this case, after applying Corollary 14.1, (14.14),
(14.15), (14.16) and (14.17), we obtain (14.8) and (14.9).
The converse can be verified by a direct computation.
Since δ(n − 1)(p) = maxu∈Tp1 (M) Ric(u), Theorem 14.1 implies
Corollary 14.2. For any n-dimensional submanifold M of Rm (c) with
n ≥ 3, the Ricci curvature Ric of M satisfies
n2 2
Ric(u) ≤ H + (n − 1)c
4
for any unit vector u ∈ T 1 M , where T 1 M denotes the unit tangent bundle
of M .
It follows immediately from the equation of Gauss that a necessary
condition for a Riemannian manifold to admit a minimal immersion in a
Euclidean space is “Ric ≤ 0”. For many years before the invention of δ-
invariants, this was the only known Riemannian obstruction for a general
Riemannian manifold to admit a minimal immersion into a Euclidean space
with arbitrary codimension.
An immediate application of Theorem 14.1 is the following theorem
which provides many new obstructions to minimal immersions.
Theorem 14.2. [Chen (1998a, 2000a)] Let N be a Riemannian n-
manifold. If there exist a point p and a k-tuple (n1 , . . . , nk ) ∈ S(n) with
δ(n1 , . . . , nk )(p) > 0, then N never admits a minimal immersion into any
Riemannian manifold with non-positive sectional curvature. In particular,
N never admits a minimal immersion into any Euclidean space for any
codimension.
Remark 14.2. There exist many Riemannian manifolds with Ricci tensor
Ric ≤ 0, but with some positive δ-invariants. For instance, the following
example was constructed in [Suceavă (2001)].
Let ε > 0 and put U = {(x, y, u, v) ∈ R4 : y > 2ε and v > 0}. Denote by
N 4 the Riemannian 4-manifold (U, g) with the warped product metric
1 2ε tan−1 y
g = 2 (dx2 + dy 2 ) + (du2 + dv 2 ).
y v2
Then, for sufficiently small ε > 0, N 4 satisfies Ric < 0 and δ(2, 2) > 0.
An important special case of Theorem 14.1 is the following.
Corollary 14.3. Let N be an n-dimensional submanifold of a Riemannian

manifold Rm (c) of constant sectional curvature c. Then we have
n2 (n − 2) 2 1
δ(2) ≤ H + (n + 1)(n − 2)c. (14.18)
2(n − 1) 2
The equality case of inequality (14.18) holds at a point x ∈ N if and only
if, there exists an orthonormal basis e1 , . . . , em at x, such that the shape
operators of N in Rm (c) at x take the following form:
 r 
A 0
Aer = , r = n + 1, . . . , m, (14.19)
0 µr I
where Ar is a symmetric 2 × 2 submatrix satisfying Tr (Ar ) = µr .
Inequality (14.18) is also known as Chen’s basic inequality and the

equality case of (14.18) is known as Chen’s basic equality in [Dajczer and
Florit (1998)].
Definition 14.1. An isometric immersion φ : N → Rm (c) of a Riemannian

manifold into Rm (c) is called a δ(n1 , . . . , nk )-ideal immersion if it satisfies
the equality case of (14.7) identically for a k-tuple (n1 , . . . , nk ) ∈ S(n).
Remark 14.3. Roughly speaking, ideal immersions are the immersions

which receive the least possible amount of tension at each point form its
ambient space. In this sense, ideal immersions of a Riemannian manifold
are the best immersions among all possible isometric immersions of the
manifold.
Remark 14.4. The δ-invariants have many applications to several areas

in mathematics (see [Chen (2011b)] for details).
14.3 Inequalities for submanifolds in complex space forms
We have the following optimal result from [Chen (1998a, 2000a)].
Theorem 14.3. For any n-dimensional totally real submanifold N of a

complex space form M̃ m (4c) of constant holomorphic sectional curvature
4c and for any k-tuple (n1 , . . . , nk ) ∈ S(n), we have
δ(n1 , . . . , nk ) ≤ b(n1 , . . . , nk )H 2 + a(n1 , . . . , nk )c, (14.20)
where H 2 is the squared mean curvature of N in Rm (c).

The equality case of inequality (14.20) holds at a point x ∈ N if and
only if, there exists an orthonormal basis e1 , . . . , e2m at x, such that the
shape operators of N in M̃ m (4c) at x take the following form:
 
Ar1 . . . 0
 .. . . .. 
Aer =
 . . . 0 
, r = n + 1, . . . , 2m, (14.21)
 0 . . . Ark 
0 µr I
where Arj are symmetric submatrices satisfying
Tr (Ar1 ) = · · · = Tr (Ark ) = µr ; (14.22)
Proof. This theorem follows from the fact that the proof of Theorem
14.1 is based only on the equation of Gauss and the Gauss equation for
submanifolds of a real space form Rm (c) is the same as the Gauss equation
of a totally real submanifold in a complex space form M̃ m (4c).
Let N be a Lagrangian submanifold of a complex space form M̃ n (4c).
Let p ∈ N and V be a d-dimensional subspace of Tp N . Denote by πV :
Tp N → V the orthogonal projection. For each vector v ∈ V , we define a
symmetric endomorphism AVJv on V by AVJv = πV ◦ AJv , where AJv is the
shape operator at Jv.
In order to prove the next theorem, we need the following two lemmas
from [Chen (2000d)].
Lemma 14.2. Let N be a Lagrangian submanifold of a complex space form

M̃ n (4c) and V be a d-dimensional subspace of Tp N at some point p ∈ N .
Then there exists an orthonormal basis {ε1 , . . . , εd } of V such that
AVJε1 εi = λi εi , i = 1, . . . , d, (14.23)
where λ1 , . . . , λd satisfy λ1 ≥ 2λj , j = 2, . . . , d. Moreover, we have
(a) λ1 > 0 unless AVJv = 0 for all v ∈ V and
(b) if λ1 > 0, then the multiplicity of λ1 is one and λ1 > λj for j = 2, . . . , d.
Proof. Since N is a Lagrangian submanifold, we have

hσ(u, v), Jwi = hσ(w, v), Jui = hσ(v, u), Jwi , u, v, w ∈ V.
If AVJv= 0 for all v ∈ V , there is nothing to prove. So, we may assume
AVJv 6=0 for some v ∈ V , which implies hσ(u, u), Jui =
6 0 for some u ∈ V
by polarization.
We define a function fx by
fx : V1 → R : u 7→ fx (u) = hσ(u, u), Jui , (14.24)
where V1 is the unit hypersphere of V consisting of all unit vectors in V .
Since V1 is a compact set, there exists a vector v in V1 such that fx attains
an absolute maximum at v. We have fx (v) > 0.
By taking the directional derivative of fx with respect to w, we also
have hσ(v, v), Jwi = 0 for all w ∈ V orthogonal to v. So, by (2.14), we
know that v is an eigenvector of the symmetric operator AVJv . If we put
ε1 = v and choose ε2 , . . . , εn so that {ε1 , . . . , εn } is an orthonormal basis of
V and each εi is an eigenvector of AVJε1 with eigenvalue λi , then we obtain
(14.23).
Since fx attains an absolute maximum at ε1 , the function fi , i ∈
{2, . . . , d}, defined by fi (t) = fx (cos t ε1 + sin t εi ) has a relative maximum
at t = 0. So, after computing fi′′ (0), we get λ1 ≥ 2λi .
Since fx (v) > 0, we get λ1 > 0. Hence we find λ1 > λi , for i ≥ 2. In
particular, this implies that the eigenspace of AJ ē1 with eigenvalue λ1 is
1-dimensional. This proves the lemma.
Now, assume N is a Lagrangian submanifold of a complex space form
M̃ n (4c) satisfying the equality case of (14.20) identically for some k-tuple
(n1 , . . . , nk ) ∈ S(n). Thus, according to Theorem 14.3, there exists an
orthonormal basis {e1 , . . . , en } of Tx N at each x ∈ N such that, for each
normal vector ξ at x, the shape operator Aξ with respect to {e1 , . . . , en }
takes the form:
 
Aξ1 . . . 0
 .. . . ..
 .
 . .
Aξ =  0 . . . Aξ
0 , (14.25)
 k 
 
0 µξ I
where Aξj is a symmetric nj × nj submatrix so that
Tr (Aξ1 ) = · · · = Tr (Aξk ) = µξ . (14.26)

With respect to the orthonormal basis {e1 , . . . , en } given above, we put
Li = Span {eα : α ∈ Ii },
X
TrLi σ = σ(eα , eα ), (14.27)
α∈Ii
where Ii = {n1 + · · · + ni−1 + 1, . . . , n1 + · · · + ni } for i = 1, . . . , k.

Lemma 14.3. If N is a Lagrangian submanifold of a complex space form

M̃ n (4c) which satisfies (14.20) at a point x ∈ N , then there exists an
orthonormal basis {e1 , . . . , en } of Tx N such that the second fundamental
form of N satisfies
σ(Li , Li ) ⊂ J(Li ), σ(Li , Lj ) = 0, (14.28)
σ(X, en1 +···+nk +1 ) = · · · = σ(X, en ) = 0, (14.29)
TrLi σ = 0, (14.30)
for 1 ≤ i 6= j ≤ k and X ∈ Tx N .
Proof. If hσ(u, v), Jwi = 0 for all tangent vectors u, v and w at x, then
we get σ = 0 by polarization. In this case there is nothing to prove. So
we may assume that this is not the case. By applying Lemma 14.2 to
V = Tx M n , we know that there exists an orthonormal basis {ε1 , . . . , εn } of
V such that
AJε1 εi = λi εi , i = 1, . . . , n, (14.31)
where λ1 , . . . , λn satisfy λ1 ≥ 2λj , j = 2, . . . , n. Moreover, we have
(a) λ1 > 0 unless AJv = 0 for all v ∈ V and
(b) if λ1 > 0, then the multiplicity of λ1 is one.
Assume that N satisfies equality (14.20) at a point x. Then Theorem
14.3 implies that there exists an orthonormal basis {e1 , . . . , en } of Tx N such
that, for each normal vector ξ at x, the shape operator Aξ takes the form
of (14.25) so that condition (14.26) holds.
Define the subspaces L1 , . . . , Lk of Tx M n by (14.27) and Lk+1 by
Lk+1 = Span {en1 +···+nk +1 , . . . , en }.
Obviously, we have Tp N = L1 ⊕ · · · ⊕ Lk+1 .
Now, we claim that ε1 must lies in one of L1 , . . . , Lk . This can be seen
as follows. First we put
ε1 = v1 + · · · + vk + vk+1 , (14.32)
where v1 ∈ L1 , . . . , vk+1 ∈ Lk+1 . Then, we have
λ1 ε1 = AJε1 ε1 = AJε1 v1 + · · · + AJε1 vk + AJε1 vk+1 . (14.33)
From (14.25) we have AJε1 vi ∈ Li . Thus (14.32) and (14.33) imply
AJε1 v1 = λ1 v1 , . . . , AJε1 vk+1 = λ1 vk+1 . (14.34)
Since the eigenspace of AJε1 with eigenvalue λ1 is 1-dimensional, (14.34)
implies that exactly one of v1 , . . . , vk+1 is nonzero.
If ε1 = vk+1 , then n1 + · · · + nk = n − 1, since λ1 has multiplicity one.

Hence, in this case, ε1 = ±en . Therefore, it follows from Lemma 6.2(1) and
AJε1 εi = λi εi , we get
±AJε1 en = AJεi ε1 = AJε1 εi = λi εi ⊥ en = ±ε1
for i = 2, . . . , n. Thus we obtain λ2 = · · · = λn = 0 by applying (14.25).
Hence, by (14.26) we obtain λ1 = 0 which is a contradiction. Consequently,
ε1 must belongs to one of L1 , . . . , Lk which proves the claim.
Without loss of generality we may now assume ε1 = e1 by applying the
claim. Moreover, since AJe1 takes the form of (14.25), we may also assume
ε2 = e2 , . . . , εn = en . In other words, we may choose the eigenvectors
ε2 , . . . , εn of AJe1 to be compatible with the decomposition arisen from
(14.25). Therefore, by applying (2.14), (14.25) and Lemma 6.2(1), we have
hσ(Xi , Yi ), JXj i = hσ(Xi , Xj ), JYi i = 0 (14.35)
for vectors Xi , Yi ∈ Li , Xj ∈ Lj , 1 ≤ i 6= j ≤ k.
If n1 + · · · + nk = n, then k ≥ 2 by the definition of S(n). Thus, (14.25),
(14.26) and (14.35) imply (14.28) and (14.29) and (14.30).
If n1 + · · · + nk < n, it follows from (14.25) that, for any j 6= t and any
t ∈ {n1 + · · · + nk + 1, . . . , n}, we have
hAJet ei , ej i = hAJei et , ej i = 0. (14.36)
By applying (14.36) and (14.26) with ξ = Jet , we obtain hAJet et , et i = 0.
Hence we have AJet = 0 for each t ∈ {n1 + · · · + nk + 1, . . . , n}. Combining
this with (14.35), we also obtain (14.28), (14.29) and (14.30).
It is known that there exist many non-minimal submanifold of a real
space form Rm (c) satisfying the equality case of (14.7). In contrast, we
have the following theorem from for Lagrangian submanifolds.
Theorem 14.4. [Chen (2000d)] Let N be a Lagrangian submanifold of a

complex space form M̃ n (4c). If N satisfies the equality case of (14.20) at
some point x ∈ N , then the mean curvature vector of N vanishes at x.
Proof. Follows immediately from Lemma 14.3.
14.4 Improved inequalities for Lagrangian submanifolds
Theorem 14.4 suggests that inequality (14.20) given in Theorem 14.3 is not
optimal, i.e., that the coefficient of H 2 can be replaced by a smaller value.
In this respect, the following two theorems were proved.
Theorem 14.5. [Chen et al. (2013)] Let N be a Lagrangian submanifold

of a complex space form M̃ n (4c). If n1 , . . . , nk are k integers satisfying
2 ≤ n1 ≤ · · · ≤ nk ≤ n − 1 and n1 + · · · + nk < n, then, at any point of N ,
we have
Pk Pk
1
n2 n − i=1 ni + 3k − 1 − 6 i=1 2+n i
δ(n1 , . . . , nk ) ≤ Pk Pk H2
1
2 n − i=1 ni + 3k + 2 − 6 i=1 2+ni
k
!
1 X
+ n(n − 1) − ni (ni − 1) c.
2 i=1
Assume that the equality sign of the inequality holds at a point x ∈ N .
Then with the choice of basis and the notations introduced at the beginning
of this section, one has
A
(1) σBC = 0 if A, B, C are mutually different and not all in the same ∆i
with i ∈ {1, . . . , k},
X
αi
(2) σααji αj = σrr = σβαiiβi = 0 for i 6= j,
βi ∈∆i
r r
(3) σrr = 3σss = (ni + 2)σαr i αi for r 6= s.
Remark 14.5. For δ(2), Theorem 14.5 is due to [Oprea (2007)].
Theorem 14.6. [Chen et al. (2013)] Let N be a Lagrangian submanifold

of a complex space form M̃ n (4c). Let n1 , . . . , nk be integers satisfying 2 ≤
n1 ≤ · · · ≤ nk ≤ n − 1 and n1 + · · · + nk = n. Then, at any point of M n ,
the following holds:
Pk
n2 k − 1 − 2 i=2 ni1+2
δ(n1 , . . . , nk ) ≤ P H2
2 k − 2 ki=2 ni1+2
k
!
1 X
+ n(n − 1) − ni (ni − 1) c.
2 i=1
Assume that the equality sign of the inequality holds at a point x ∈ N .
Then with the choice of basis and the notations introduced at the beginning
of this section, we have
(a) σαAi αj = 0 if i 6= j and A 6= αi , αj ,
P β β
(b) if nj 6= min{n1 , . . . , nk }, we have αj ∈∆j σαjj αj = 0 and σαji αi = 0,
P β β
(c) if nj = min{n1 , . . . , nk }, we have αj ∈∆j σαjj αj = (ni + 2)σαji αi ,
for any i 6= j and αi ∈ ∆i .
14.5 CR-warped products and δ-invariants
Let N be a CR-submanifold of a Kähler manifold. Let H and H⊥ denote

the complex and the totally real distributions of N respectively as before.
The CR δ-invariant δ(H) of N was defined in [Chen (2012)] by
δ(H)(x) = τ (x) − τ (Hx ), (14.37)
where τ (x) and τ (Hx ) denote the scalar curvature of N at x ∈ N and the
scalar curvature of Hx ⊂ Tx N , respectively.
Let N = N T ×f N ⊥ be a CR-warped product in the complex space
form M̃ h+p (4c) with h = dimC N T and p = dim N ⊥ and let us choose a
local orthonormal frame {e1 , . . . , e2h+p } on N such that
e1 , . . . , eh , eh+1 = Je1 , . . . , e2h = Jeh
are in H and e2h+1 , . . . , e2h+p are in H⊥ . Put
r
σAB = hσ(eA , eB ), Jer i .
In the following, we shall use the following convention on the range of
indices unless mentioned otherwise:
i, j, k = 1, . . . , 2h; α, β, γ = 1, . . . , h,
r, s, t = 2h + 1, . . . , 2h + p; A, B, C = 1, . . . , 2h + p.
From Lemma 6.2(2) we have
X 2h
∆f
= K(ej ∧ er ), r ∈ {2h + 1, . . . , 2h + p}. (14.38)
f j=1
The next theorem provides an optimal inequality for CR-warped sub-

manifolds involving the CR δ-invariant δ(H).
Theorem 14.7. [Chen (2012)] Let N = N T ×f N ⊥ be a CR-warped

product in a complex space form M̃ h+p (4c) with h = dimC N T ≥ 1 and
p = dim N ⊥ ≥ 2. Then we have

2(p + 2) p∆f p(p − 1)c
H2 ≥ δ(H) − − , (14.39)
(2h + p)2 (p − 1) f 2
where ∆f is the Laplacian of f on N T and H 2 the squared mean curvature.
The equality sign of (14.39) holds at a point x ∈ N if and only if there
exists an orthonormal basis {e2h+1 , . . . , en } of Hx⊥ such that the coefficients
of the second fundamental σ with respect to {e2h+1 , . . . , en } satisfy
r r
σrr = 3σss , for 2h + 1 ≤ r 6= s ≤ 2h + p,
r
(14.40)
σst = 0, for distinct r, s, t ∈ {2h + 1, . . . , 2h + p}.
Proof. Assume that N = N T ×f N ⊥ is a CR-warped product in the

complex space form M̃ h+p (4c) with h = dimC N T ≥ 1 and p = dim N ⊥ ≥
2. Let us choose an orthonormal frame {e1 , . . . , e2h+p } on N as above.
Then it follows from Gauss’ equation, (14.37), and (14.38) that
X X
δ(H) = K(ei , er ) + K(er , es )
i,r r<s
p∆f X
= + hσ(er , er ), σ(es , es )i (14.41)
f r<s
X p(p − 1)
− |σ(er , es )|2 + c.
r<s
2
On the other hand, we obtain from Lemma 6.2(1) and JH⊥ = T ⊥ N

that
X X n2 2 1
hσ(er , er ), σ(es , es )i − |σ(er , es )|2 = H − ||σ⊥ ||2 (14.42)
r<s r<s
2 2
P
with n = 2h + p, where ||σ⊥ ||2 = r,s |σ(er , es )|2 is the squared norm of σ
restricted to H⊥ .
p∆f p(p − 1) n2 2 1
δ(H) = + c+ H − ||σ⊥ ||2 . (14.43)
f 2 2 2
Thus we obtain

2 2(p + 2) p∆f
2
n H + − δ(H) + p(p + 2)c
p−1 f
(14.44)
3n2 2 p + 2
= H + ||σ⊥ ||2 .
1−p p−1
r s t
From Lemma 6.1(2) we find σst = σrt = σrs . Now, we derive from (14.44)
and Lemma 6.2(1) that

2(p + 2) p∆f
n2 H 2 + − δ(H) + p(p + 2)c
p−1 f
XX 2 3(p + 1) X 6(p + 2) X r 2
r r 2
= σss + (σss ) + (σ )
r s
p−1 p − 1 r<s<t st
r6=s
(14.45)
2(p + 2) X X r r
+ σss σtt
p − 1 r s<t
X 3(p + 1) X r 2 6(p + 2) X r 2
r 2
= (σrr ) + (σss ) + (σ )
r
p−1 p − 1 r<s<t st
r6=s
6 XX r r
− σ σ
p − 1 r s<t ss tt
6(p + 2) X r 2 3 XX r r 2
= (σst ) + (σss − σtt )
p − 1 r<s<t p−1 s<t r6=s,t
1 X r r 2
+ (σrr − 3σss )
p−1
s6=r
≥ 0.
Hence inequality (14.39) follows from (14.45). Moreover, it is easy to verify
that the equality sign of (14.39) holds if and only if (14.40) holds. This
proves the theorem.
CR-warped products in Ch+p satisfying the equality case of inequality
(14.39) have been completely classified as follows.
Theorem 14.8. [Chen (2012)] If ψ : N T ×f N ⊥ → Ch+p is a CR-warped
product in Ch+p with h = dimC N T ≥ 1 and p = dim N ⊥ ≥ 2, then

2 2(p + 2) p∆f
H ≥ δ(H) − . (14.46)
(2h + p)2 (p − 1) f
The equality sign of (14.46) holds identically if and only if, up to dila-
tions and rigid motions of Ch+p , one of the following three cases occurs:
(a) The CR-warped product is an open part of the CR-product Ch × W p ⊂

Ch × Cp , where W p is the Whitney p-sphere in Cp ;
(b) N T is an open part of Ch , N ⊥ is an open part of the unit p-sphere S p ,
f = |z1 | and ψ is the minimal immersion defined by

z1 w1 , . . . , z1 wp+1 , z2 , . . . , zh ,
where z = (z1 , . . . , zh ) ∈ Ch and w = (w1 , . . . , wp+1 ) ∈ S p ⊂ Ep+1 ;
(c) f = |z1 |, N T is an open part of Ch , N ⊥ is the warped product of a
curve and an open part of S p−1 with warping function
√ √ !
c2 − 1 c2 − 1
ϕ= √ cn c t, √ , c > 1,
2 2c
where cn is a Jacobi’s elliptic function and ψ is a non-minimal immer-
sion defined by
R
ϕ(ϕ′ +ikϕ2 )
dt
R R
ik ϕdt ik ϕdt
z1 e ϕ 2 −1
, z1 ϕe w1 , . . . , z1 ϕe wp , z2 , . . . , zh ,
√
with k = 21 c4 − 1, z = (z1 , . . . , zh ) ∈ Ch , and (w1 , . . . , wp ) ∈ S p−1 (1)
⊂ Ep .
14.6 Anti-holomorphic submanifolds with p ≥ 2
For a CR-submanifold N of a Kähler manifold, the two partial mean cur-

vature vectors HH and HH⊥ of N are defined respectively by
2h 2h+p
1 X 1 X
HH = σ(ei , ei ) HH⊥ = σ(er , er ). (14.47)
2h i=1 p
r=2h+1
An anti-holomorphic submanifold N of a Kähler manifold M̃ is called

H-minimal (or H⊥ -minimal ) if HH = 0 (or HH⊥ = 0) holds identical.
For anti-holomorphic submanifolds with p = rank H⊥ ≥ 2, we have the
following optimal inequality.
Theorem 14.9. [Al-Solamy et al. (2014)] Let N be an anti-holomorphic
submanifold of a complex space form M̃ h+p (4c) with h = rankC H ≥ 1 and
p = rank H⊥ ≥ 2. Then we have
(p − 1)(2h + p)2 2 p
δ(H) ≤ H + (4h + p − 1)c. (14.48)
2(p + 2) 2
The equality sign of inequality (14.48) holds identically if and only if the
following three conditions are satisfied:
(a) N is H-minimal, i.e., HH = 0.
(b) N is mixed totally geodesic.
(c) there exist an orthonormal frame {e2h+1 , . . . , en } of H⊥ such that the
second fundamental σ of N satisfies
(
r r
σrr = 3σss , for 2h + 1 ≤ r 6= s ≤ 2h + p,
r
(14.49)
Proof. Let N be an anti-holomorphic submanifold of M̃ h+p (4c). Let us

choose an orthonormal frame {e1 , . . . , e2h+p } on N as above. It follows
from the equation of Gauss and the definition of CR δ-invariant that
X2h 2h+p
X X 1
δ(H) = K(ei , er ) + K(er , es )
i=1
2
r=2h+1 2h+1≤r6=s≤2h+p
2h
X 2h+p
X 2h 2h+p
X X
= hσ(ei , ei ), σ(er , er )i − |σ(ei , er )|2
i=1 r=2h+1 i=1 r=2h+1 (14.50)
2h+p
X 2h+p
X
1 1
+ hσ(er , er ), σ(es , es )i − |σ(er , es )|2
2 2
r,s=2h+1 r,s=2h+1
p
+ (4h + p − 1)c.
2
On the other hand, we have

2h 2h+p
X X 2h+p
X 1
hσ(ei , ei ), σ(er , er )i + hσ(er , er ), σ(es , es )i
i=1 r=2h+1
2
r,s=2h+1
2h+p
X 1 (14.51)
− |σ(er , es )|2
2
r,s=2h+1
(2h + p)2 2 1
= H − 2h2 |HH |2 − ||σH⊥ ||2 ,
2 2
where ||σH⊥ ||2 is defined by
2h+p
X
||σ⊥ ||2 = |σ(er , es )|2 . (14.52)
r,s=2h+1
(2h + p)2 2 p
δ(H) = H + (4h + p − 1)c − 2h2 |HH |2
2 2
X2h 2h+p
X (14.53)
1
− |σ(ei , er )|2 − ||σH⊥ ||2 .
i=1
2
r=2h+1
It follows from statement (2) of Lemma 6.1 that the coefficients of the
second fundamental form σ satisfy
r s t
σst = σrt = σrs . (14.54)
Hence we find from (14.47), (14.52) and (14.54) that
2h+p 2h+p
!2
X X
2 2 2 r
(p + 2)||σH⊥ || − 3p |HH⊥ | = (p − 1) σss
r=2h+1 s=2h+1
X
r 2
+ 3(p + 1)(σss )
2h+1≤r6=s≤2h+p
X
r 2
+ 6(p + 2)(σst )
2h+1≤r<s<t≤2h+p
2h+p
X X
r r
+ 2(p + 2)σss σtt (14.55)
r=2h+1 2h+1≤s<t≤2h+p
2h+p
X X
r 2 r 2
= (p − 1)(σrr ) + 3(p + 1)(σss )
r=2h+1 2h+1≤r6=s≤2h+p
X
r 2
+ 6(p + 2)(σst )
2h+1≤r<s<t≤2h+p
2h+p
X X
r r
− 6σss σtt
r=2h+1 2h+1≤s<t≤2h+p
X X
r 2 r r 2
= 6(p + 2)(σst ) + (σrr − 3σss )
2h+1≤r<s<t≤2h+p 2h+1≤s6=r≤2h+p
X X
r r 2
+ 3(σss − σtt )
r6=s,t 2h+1≤s<t≤2h+p
≥ 0.
Thus we get
3p2
||σH⊥ ||2 ≥ |H ⊥ |2 , (14.56)
p+2 H
with equality holding if and only if
r r
σrr = 3σss , for 2h + 1 ≤ r 6= s ≤ 2h + p,
r
(14.57)
Now, by combining (14.53) and (14.56), we obtain
(2h + p)2 2 p
H + (4h + p − 1)c − δ(H)
2 2
X2h 2h+p
X 3p2
≥ 2h2 |HH |2 + |σ(ei , er )|2 + |HH⊥ |2
i=1
2(p + 2)
r=2h+1
2h 2h+p
X X
= 2h2 |HH |2 + |σ(ei , er )|2
i=1 r=2h+1
(
3
+ (2h + p)2 H 2 − 4h2 |HH |2
2(p + 2)
) (14.58)
X2h 2h+p
X
−2 |σ(ei , er )|2
i=1 r=2h+1
2
3(2h + p) 2 2h2 (p
− 1)
= H + |HH |2
2(p + 2) p+2
2h 2h+p
p−1X X
+ |σ(ei , er )|2
p + 2 i=1
r=2h+1
3(2h + p)2 2
≥ H .
2(p + 2)
It is obvious that the equality of the last inequality in (14.58) holds if and
only if N is H-minimal and mixed totally geodesic. Consequently, we have
inequality (14.48) from (14.58). It is direct to verify that the equality sign
of (14.48) holds identically if and only if conditions (a), (b) and (c) of
Theorem 14.9 are satisfied.
14.7 Anti-holomorphic submanifolds satisfying the equality
Example 14.1. Consider the product immersion:

φ : Ch × S p (1) → Ch ⊕ Cp = Ch+p
defined by
φ(z, x) = (z, w(x)), ∀z ∈ Ch , ∀x ∈ S p (1). (14.59)
It is direct to verify that φ is an anti-holomorphic isometric immersion
which satisfies the equality sign of (14.48) identically.
Anti-holomorphic submanifolds satisfying the equality case of inequality
(14.48) were classified by the next two theorems.
submanifold of a complex space form M̃ h+p (4c) with h = rankC H ≥ 1 and
p = rank H⊥ ≥ 2. If N satisfies the equality case of (14.48) identically
and if the holomorphic distribution H is integrable, then c = 0 so that
M̃ h+p (4c) = Ch+p . Moreover, we have either
(i) N is a totally geodesic anti-holomorphic submanifold of Ch+p or,
(ii) up to dilations and rigid motions of Ch+p , N is given by an open
portion of the following product immersion:
φ : Ch × S p (1) → Ch+p ; (z, x) 7→ (z, w(x)), z ∈ Ch , x ∈ S p (1),
where w : S p (1) → Cp is the Whitney p-sphere.
Proof. Assume that N is an anti-holomorphic submanifold of a complex
space form M̃ h+p (4c) with h = rankC H ≥ 1 and p = rank H⊥ ≥ 2. If N
satisfies the equality case of (14.48) and if the holomorphic distribution H
is integrable, then it follows from Theorem 14.9 that N is mixed foliate.
Hence Lemma 6.3 implies that c = 0. Thus, according to Theorem 9.13, N
is a CR-product. Therefore, N is locally a CR-product given by
Ch × N ⊥ ⊂ Ch × Cp ,
where Ch is a complex Euclidean h-subspace and N ⊥ is a Lagrangian sub-
manifold of Cp .
Consequently, condition (c) of Theorem 14.9 implies that N ⊥ is a La-
grangian H-umbilical submanifold in Cp whose second fundamental form
satisfying
σ(e2h+1 , e2h+1 ) = 3λJe2h+1 , σ(e2h+1 , es ) = λJes ,
σ(e2h+2 , e2h+2 ) = · · · = σ(e2h+p , e2h+p ) = λJe2h+1 , (14.60)
σ(er , es ) = 0, 2h + 2 ≤ r 6= s ≤ 2h + p,
for some suitable function λ with respect to some suitable orthonormal local
frame field {e2h+1 , . . . , e2h+p } of T N ⊥ .
If λ = 0, then N ⊥ is an open portion of a totally geodesic totally real
p-plane in Cp . Hence, in this case N is a totally geodesic anti-holomorphic
submanifold.
If λ 6= 0, it follows from (14.60) that, up to dilations and rigid motions,
N ⊥ is an open part of the Whitney p-sphere in Cp (cf. [Borrelli et al.
(1995); Chen (2011b)]). Consequently, up to dilations and rigid motions
of Ch+p , the anti-holomorphic submanifold is locally given by the product
immersion:
φ : Ch × S p (1) → Ch+p ; (z, x) 7→ (z, w(x)), (14.61)
for z ∈ Ch and x ∈ S p (1), where w : S p (1) → Cp is the Whitney p-sphere.
submanifold in a complex space form M̃ 1+p (4c) with h = dimC H = 1 and
p = dim H⊥ ≥ 2. Then we have
(p − 1)(p + 2)2 2 p
δ(H) ≤ H + (p + 3)c. (14.62)
2(p + 2) 2
The equality case of (14.62) holds identically if and only if c = 0 and
either
(i) N is a totally geodesic anti-holomorphic submanifold of Ch+p or,

(ii) up to dilations and rigid motions, N is given by an open portion of the
following product immersion:
φ : C × S p (1) → C1+p ; (z, x) 7→ (z, w(x)), z ∈ C, x ∈ S p (1),
where w : S p (1) → Cp is the Whitney p-sphere.
Proof. Let N be an anti-holomorphic submanifold in a complex space

form M̃ 1+p (4c). Then we have inequality (14.60) from inequality (14.48).
Assume that N satisfies the equality case of (14.62) identically. Then
Theorem 14.9 implies that N satisfies conditions (a), (b) and (c) of Theorem
14.9. By condition (a), N is H-minimal. Thus we find
σ(Je1 , Je1 ) = −σ(e1 , e1 ) (14.63)
for any unit vector e1 ∈ H.
It is direct to verify from (14.63) and polarization that the second fun-
damental form satisfies
σ(X, JY ) = σ(JX, Y ), ∀X, Y ∈ H.
Thus, by Lemma 6.2(1), we conclude that H is integrable. Consequently,
we obtain Theorem 14.11 from Theorem 14.10.
14.8 An optimal inequality for real hypersurfaces
Anti-holomorphic submanifolds with rank H⊥ = 1 are nothing but real

hypersurfaces. A real hypersurface N of a Kähler manifold M̃ is called a
Hopf hypersurface if Jξ is a principal curvature vector, i.e., an eigenvector
of the shape operator Aξ , where ξ is a unit normal vector of N .
In the following, a Hopf hypersurface is called a special Hopf hypersur-
face if the vector field Jξ is an eigenvector field of Aξ with eigenvalue 0,
i.e., Aξ (Jξ) = 0.
For real hypersurfaces, we have the following.
Theorem 14.12. [Al-Solamy et al. (2014)] If N is a real hypersurface of

a complex space form M̃ h+1 (4c), then its Ricci tensor Ric satisfies
(2h + 1)2 2
Ric(Jξ, Jξ) ≤ H + 2hc, (14.64)
2
where ξ is a unit normal vector field of N in M̃ h+1 (4c).
The equality sign of inequality (14.64) holds identically if and only if N
is a minimal special Hopf hypersurface.
Proof. Let N be a real hypersurface of a complex space form M̃ h+1 (4c).

Then it follows from the definition of δ(H) that
δ(H) = Ric(Jξ, Jξ). (14.65)
Let us choose an orthonormal frame
{e1 , . . . , eh , eh+1 = Je1 , . . . , e2h = Jeh }
for the holomorphic distribution H and choose e2h+1 to be a unit vector
field in H⊥ .
We put
σa,b = hσ(ea , eb ), Je2h+1 i , a, b = 1, . . . , 2h + 1. (14.66)
Define the connection forms by
2h
X
∇X ei = ωij (X)ej + ωi2h+1 (X)e2h+1 ,
j=1
(14.67)
2h
X j
∇X e2h+1 = ω2h+1 (X)ej ,
j=1
for i = 1, . . . , 2h. Then it follows from (14.37) and Gauss’ equation that
2h
X 2h
X
δ(H) = σi,i σ2h+1,2h+1 − (σi,2h+1 )2 + 2hc. (14.68)
i=1 i=1
On the other hand, we have

2h
X (2h + 1)2 2 1
σi,i σ2h+1,2h+1 = H − (σ2h+1,2h+1 )2 − 2h2 |HH |2 . (14.69)
i=1
2 2
By combining (14.68) and (14.69) we obtain
(2h + 1)2 2 1
δ(H) = H + 2hc − 2h2 |HH |2 − (σ2h+1,2h+1 )2
2 2
2h
X
− (σi,2h+1 )2 (14.70)
i=1
(2h + 1)2 2
≤ H + 2hc.
2
Now, it follows from (14.70) and Lemma 6.2(2) that the equality sign of
the inequality (14.64) holds identically if and only if the following two
statements hold:
(i) N is a special Hopf hypersurface and
(ii) N is H-minimal in M̃ h+1 (4c).
Obviously, conditions (i) and (ii) imply that N is a minimal real hyper-
surface of M̃ h+1 (4c).
The following results from [Al-Solamy et al. (2014)] are consequences of
Theorem 14.12.
Corollary 14.4. Let N be a real hypersurface of a complex space form

M̃ h+1 (4c). If N satisfies the equality case of (14.64) identically, then the
holomorphic distribution of N is non-integrable, unless c = 0 and N is
totally geodesic.
Proof. Under the hypothesis of this corollary, if N satisfies the equality

case of (14.64) identically and if the holomorphic distribution H is inte-
grable, then Theorem 14.12 implies that N is mixed foliate. Therefore it
follows from Proposition 9.1, Corollary 9.3 and Theorem 9.14 that c = 0
and N is a CR-product of a complex h-subspace in Ch and an open portion
of line in C. Consequently, N must be totally geodesic.
Theorem 14.13. If N is a real hypersurface of a complex space form

M̃ 2 (4c), then we have
9
Ric(Jξ, Jξ) ≤ H 2 + 2c. (14.71)
2
The equality sign of inequality (14.71) holds identically if and only if

c = 0 and N is totally geodesic.
Theorem 14.14. Let N be a real hypersurface of the complex Euclidean
3-space C3 . Then we have
25 2
Ric(Jξ, Jξ) ≤ H . (14.72)
2
If the equality case of inequality (14.72) holds identically, then N is a
totally real 3-ruled minimal submanifold of C3 .
Theorem 14.15. If N is a real hypersurface of the complex projective 3-
space CP 3 (4), then we have
25 2
Ric(Jξ, Jξ) ≤ H + 4. (14.73)
2
locally there exists an orthonormal frame {e1 , e2 , e3 = Je1 , e4 = Je2 , e5 }
such that
σ(e1 , e1 ) = λξ, σ(e2 , e2 ) = −λξ,
1 1
σ(e3 , e3 ) = ξ, σ(e4 , e4 ) = − ξ,
λ λ
σ(ea , eb ) = 0 otherwise,
where λ is a nowhere zero function.
Theorem 14.16. If N is a real hypersurface of the complex hyperbolic 3-
space CH 3 (−4), then we have
25 2
Ric(Jξ, Jξ) ≤ H − 4. (14.74)
2
locally there exists an orthonormal frame {e1 , e2 , e3 = Je1 , e4 = Je2 , e5 } on
N such that
σ(e1 , e1 ) = λξ, σ(e2 , e2 ) = −λξ,
1 1
σ(e3 , e3 ) = − ξ, σ(e4 , e4 ) = ξ,
λ λ
σ(ea , eb ) = 0 otherwise,
where λ is a nowhere zero function.
Corollary 14.5. A real hypersurface of CP 3 (4) (resp., of CH 3 (−4)) sat-
isfying the equality case of (14.73) (resp., the equality case of (14.74)) is
δ(2, 2)-ideal.
For the proofs of the above results, see [Al-Solamy et al. (2014)].
14.9 Another optimal inequality involving a δ-invariant
Let π : M → B be a Riemannian submersion with totally geodesic fibers

and let N be a Riemannian n-manifold isometrically immersed in the Rie-
mannian manifold B. If we denote the pre-image π −1 (N ) of N in M by
Ñ , then π̃ : Ñ → N is also a Riemannian submersion with totally geodesic
fibers, where π̃ is the restriction π|Ñ . For a horizontal 2-plane Px ⊂ Tx Ñ ,
let P̄x be the (m−b+2)-subspace spanned by Px and the vertical space Vx .
The submersion δ-invariant δ H on N e is defined by [Alegre et al. (2014)]:
δ H (x) = τÑ (x) − inf τÑ (P̄x ), (14.75)
P̄x
where P̄x runs over (m − b + 2)–subspaces associated with all horizontal

2-planes Px at x ∈ Ñ . Obviously, we have δÑ (r) ≥ δ H with r = 2 + m − b.
Lemma 14.4. [Alegre et al. (2014)] Let π : M → B be a Riemannian

submersion with totally geodesic fibers. Then the scalar curvature τM of M
and the scalar curvature τB of B satisfy
τM = τB + Ăπ − 3Åπ + τF , (14.76)
where τF is the scalar curvature of fibers.
Proof. Let π : M → B be a Riemannian submersion with totally geodesic

fibers. For orthonormal basic horizontal vector fields X1 , . . . , Xb and for
orthonormal vertical vector fields Vb+1 , . . . , Vm on M , we find from Lemma
5.3 that
KM (Xi ∧ Vα ) = |AXi Vα |2 . (14.77)
Also, it follows from Corollary 5.1 and (5.7) that the scalar curvature τ (H)
of the horizontal space satisfies
τ (H) = τB − 3Åπ . (14.78)
Since π has totally geodesic fibers, the scalar curvature τF equals to the
scalar curvature τ (V) of the vertical distribution. Consequently, we obtain
(14.76) from (14.77) and (14.78).
Lemma 14.5. Let π : M → B be a Riemannian submersion with totally

geodesic fibers and let N be a submanifold of B. Then, for orthonormal
vectors e1 , e2 at π(x) ∈ N , x ∈ Ñ , we have
τÑ (x) − τÑ (P̄x ) = τN − KN (e1 , e2 ) − 3(Åπ̃ − |Aē1ē2 |2 )
2 m−b
X X (14.79)
+ Ăπ̃ − KÑ (ēi , vα ),
i=1 α=1
where ē1 , ē2 are horizontal vectors at x, {v1 , . . . , vm−b } is an orthonormal

basis of the vertical space Vx , and P̄x is the subspace spanned by ē1 , ē2 and
the vertical space Vx .
Proof. Under the hypothesis of the lemma, π̃ : Ñ = π −1 (N ) → N is a

Riemannian submersion with totally geodesic fibers. Thus it follows from
Lemma 14.4 that
τÑ = τN + Ăπ̃ − 3Åπ̃ + τF . (14.80)
Let x ∈ Ñ and e1 , e2 orthonormal vectors at π(x) ∈ N . Denote by ē1 , ē2
the horizontal lifts of e1 , e2 at x ∈ Ñ .
As before let P̄x denote the subspace of Tx Ñ spanned by ē1 , ē2 and Vx .
Then we have
2 m−b
X X
τÑ (P̄x ) = τF + KÑ (ē1 , ē2 ) + KÑ (ēi , vα ). (14.81)
i=1 α=1
From Corollary 4.1 we find

KÑ (ē1 , ē2 ) = KN (e1 , e2 ) − 3 |Aē1ē2 |2 . (14.82)
Therefore, after combining (14.80), (14.81), (14.82) and Lemma 5.5(2), we
obtain (14.79).
Let N be a Riemannian submanifold of a Kähler manifold. For each
X ∈ T N we put
JX = P X + F X, (14.83)
respectively. It follows from J 2 = −I and (14.83) that
hP X, Y i = − hX, P Y i (14.84)
for X, Y tangent to N .
Let ψ be a 2-plane section of Tx̄ N , x̄ ∈ N , spanned by two orthonormal
vectors e1 , e2 ∈ Tx̄ N . We put
Θ(ψ) = hP e1 , e2 i2 . (14.85)
If {e1 , . . . , en } is an orthonormal frame of N , then the squared norms ||P ||2
and ||F ||2 of P and F are defined respectively by
n
X n
X
||P ||2 = |P ei |2 , ||F ||2 = |F ei |2 . (14.86)
i=1 i=1
We need the following lemma from [Chen (1996b)].

Lemma 14.6. Let φ : N → CP m (4) be an isometric immersion from a

Riemannian n-manifold N into the complex projective m-space CP m (4).
Then, for any 2-plane section ψ ⊂ Ty N, y ∈ N , we have
n2 (n − 2) 2 (n + 1)(n − 2)
τN − KN (ψ) ≤ H +
2(n − 1) 2 (14.87)
3 2
+ ||P || − 3Θ(ψ).
2
The equality of inequality (14.87) holds at a point y ∈ N if and only if
there exists an orthonormal basis e1 , . . . , em at y such that
(i) ψ = Span{e1 , e2 } and
(ii) the shape operator A at y satisfies
 
Bs 0
Aes = , s = n + 1, . . . , 2m, (14.88)
0 µs I
where I is an identity (n − 2) × (n − 2)-submatrix and Bs are symmetric
2 × 2 submatrices with µs = Tr Bs , s = n + 1, . . . , 2m.
An application of Lemma 14.5 is the following.
Theorem 14.17. [Alegre et al. (2014)] Let π : S 2m+1 → CP m (4) be the

Hopf fibration and N an n-dimensional submanifold of CP m (4). Then
n2 (n − 2) 2 1
δH ≤ H + ||P ||2 + (n2 − n − 2), (14.89)
2(n − 1) 2
where H 2 is the squared mean curvature of N in CP m (4).
The equality sign of (14.89) holds identically if and only if there is an
orthonormal frame e1 , . . . , em such that
(a) the shape operator A of N in CP m (4) satisfies
 
Bs 0
Aes = , s = n + 1, . . . , 2m, (14.90)
0 µs I
where I is an identity (n − 2) × (n − 2) matrix and Bs are symmetric 2 × 2
submatrices satisfying µs = Tr Bs , s = n + 1, . . . , 2m, and
(b) P e1 = P e2 = 0.
Proof. Let π : S 2m+1 (1) → CP m (4) be the Hopf fibration and put ξ = iz
ˆ and ∇
as before. Denote by ∇ ˇ the Levi-Civita connections of S 2m+1 (1) and
CP (4), respectively. For vector fields X, Y tangent to CP m (4), we have
m
ˆ X̄ Ȳ = ∇
∇ ˇ X Y − hJX, Y i ξ, (14.91)
ˆ X̄ ξ = ∇
∇ ˆ ξ X̄ = JX. (14.92)
Let N be an n-dimensional submanifold of CP m (4). Denote by Ñ the

pre-image of N via the Hopf fibration π : S 2m+1 (1) → CP m (4). Let Py be
a 2-plane section of a tangent space Ty N of N spanned by two orthonormal
vectors e1 , e2 .
As before, we denote by P̄x the 3-plane spanned by ξx and the horizontal
lifts ē1 , ē2 of e1 , e2 at a point x with π(x) = y. Then it follows from Lemma
14.5 that
τÑ (x) − τÑ (P̄x ) = τN −KN (e1 , e2 )−3Åπ̃ +3|Aē1ē2 |2
2
X (14.93)
+ Ăπ̃ − KÑ (ēi , ξ),
i=1
where ξ = iz is the characteristic vector field of the Sasakian space form

S 2m+1 (1).
If η is a normal vector field of N in CP m (4), then by using ξ = iz we
find
∇ ˇ X η − hF X, ηi ξ.
ˆ X̄ η̄ = ∇ (14.94)
Hence Weingarten’s formula yields
Âξ̄ X̄ = Aξ X + hF X, ηi ξ, D̂X̄ η̄ = DX η, (14.95)
m 2m+1
where A, Â are the shape operators of N in CP (4) and Ñ in S (1),
respectively, and D and D̂ are the corresponding normal connections.
From (14.91) we get
σ̂(X̄, Ȳ ) = σ(X, Y ), (14.96)
where σ̂ is the second fundamental form of Ñ in S 2m+1 (1).
By using (14.91) we find
˜ X̄ Ȳ = ∇X Y − hJX, Y i ξ,
∇ (14.97)
where ∇ ˜ and ∇ are the Levi-Civita connections of Ñ and N , respectively.
Also, it follows from (14.92) that
˜ X̄ ξ = ∇
σ̂(X̄, ξ) = F X, ∇ ˜ ξ X̄ = F X (14.98)
for X ∈ T N . Moreover, since Hopf’s fibration has totally geodesic fibers,
we get
σ̂(ξ, ξ) = 0. (14.99)
Now, it follows from (14.98), (14.99) and Gauss’ equation that
KÑ (X̄, ξ) = 1 − |F X|2 (14.100)
for each unit tangent vector X of N .

By applying (5.5), (5.7) and (14.100) we find
Ăπ̃ = n − ||F ||2 . (14.101)
For an orthonormal frame {e1 , . . . , en }, we find from ξ = iz and Lemma
6.2(1) that
2Aēi ēj = V[ēi , ēj ]
˘ ēi ēj − ∇
= h∇ ˘ ēj ēi , ξ i ξ
˘ ēj z i ξ − h ēj , i∇
= h ēi , i∇ ˘ ēi z i ξ (14.102)
= 2 hēi , iēj i ξ
= 2 hei , P ej i ξ,
where ∇ ˘ is the Levi-Civita connection of Cm+1 . Now, by combining (5.7)
and (14.102) we obtain
1
Åπ̃ = ||P ||2 . (14.103)
2
After applying (14.93), (14.100), (14.102) and (14.103), we obtain
3
τÑ (x) − τÑ (P̄x ) = τN − KN (e1 , e2 ) − ||P ||2 + 3 hP e1 , e2 i2
2
X2 (14.104)
2 2
+ n − 2 − ||F || + |F ei | .
i=1
2 2 2
Since |P X| + |F X| = |X| , we derive from Lemma 14.6 and (14.104)
that
τÑ (x) − τÑ (P̄x )
X2
n2 (n − 2) 2 n2 + n − 6
≤ H + − ||F ||2 + |F ei |2
2(n − 1) 2 i=1
X2
n2 (n − 2) 2 n2 − n − 6
= H + + ||P ||2 + |F ei |2
2(n − 1) 2 i=1
X2
n2 (n − 2) 2 n2 − n − 2
= H + + ||P ||2 − |P ei |2
2(n − 1) 2 i=1
n2 (n − 2) 2 n2 − n − 2
= H + + ||P ||2 ,
2(n − 1) 2
which gives inequality (14.89). Consequently, we conclude from Lemma
14.6 and (14.95) that the equality sign of (14.89) holds identically if and
only if there exists an orthonormal frame e1 , . . . , em such that statements
(a) and (b) of the theorem hold.
The next result from [Alegre et al. (2014)] provides the necessary and
sufficient condition for π −1 (N ) to satisfy the equality case of the inequality
(14.89), where N is a CR-submanifold of CP m (4).
Theorem 14.18. Let N be a CR-submanifold of the complex projective

m-space CP m (4). Then N satisfies the equality case of (14.89) identically
if and only if N is a totally real submanifold satisfying
n2 (n − 2) 2 1 2
δ(2) = H + (n − n − 2) (14.105)
2(n − 1) 2
identically.
For the proof of this theorem, see [Alegre et al. (2014)].
Theorem 14.19. If N is a totally real totally geodesic submanifold of

CP m (4), then the equality sign of (14.89) holds identically.
Proof. Let N be a totally real totally geodesic submanifold of CP m (4)

with dim N = n ≥ 3. In view of (14.100) we have
n(n − 1)
τÑ = , τÑ (P̄x ) = 1.
2
Thus we have δ H = 21 (n2 − n − 2). Hence we obtain the equality sign of
(14.89) identically due to |H| = P = 0.
The next result provides ample examples of totally real submanifolds of
CP m (4) which satisfy the equality case of inequality (14.89).
Theorem 14.20. There exist many non-totally geodesic totally real sub-
manifolds of CP m (4) which satisfy the equality case of inequality (14.89)
identically.
Proof. Let N be an n-dimensional submanifold of a unit m-sphere S m (1)

satisfying the equality
n2 (n − 2) 2 1 2
δ(2) = H + (n − n − 2). (14.106)
2(n − 1) 2
Then N can be isometrically immersed as a totally real submanifold of
CP m (4) satisfying the equality case of (14.89) via the following standard
isometric immersion:
2 to 1 totally geodesic
S m (1) −−−−−→ RP m (1) −−−−−−−−−−→ CP m (4). (14.107)
covering totally real
Since N in S m (1) satisfies equality (14.106), the shape operator of N in

m
S (1) satisfies statement (a) of Theorem 14.17, the shape operator of N in
CP m (4) satisfies (a) as well. Because N is totally real in CP m (4), it also

satisfies statement (b) of Theorem 14.17. It is known that there exist ample
submanifolds in spheres which satisfy equality (14.106) identically. Conse-
quently, there exist many non-totally geodesic, totally real submanifolds
of CP m (4) which satisfy the equality case of inequality (14.89) identically
according to Theorem 14.17.
Remark 14.6. For further results on CR-submanifolds in Kähler mani-

folds related to δ-invariants, in particular for CR-submanifolds in complex
hyperbolic spaces, see [Chen (1998a, 2011b); Chen and Vrancken (1999,
2002b); Sasahara (1999, 2014)].
14.10 Examples of δ(2)-ideal warped product submanifolds
The question when an immersion can be decomposed into a warped prod-

uct immersion is solved for immersions into Euclidean spaces, spheres or
hyperbolic spaces in [Nölker (1996)]. The main result from [Nölker (1996)]
can be stated as follows.
Let
f : N1 ×f2 N1 × · · · ×fk Nk → Rm (c)
be an isometric immersion into a space Rm (c) of constant curvature c. If
the second fundamental form σ of φ satisfies σ(Xi , Xj ) = 0 for all vector
fields Xi and Xj tangent to Ni and Nj respectively for i 6= j, then locally
φ is a warped product immersion.
The problem of how Rm (c) can be decomposed into a warped product
is solved in [Hiepko (1979)] (see also [Dillen and Nölker (1999)]) for the
statement.
If c = 1, k = 2, and N0 is 1-dimensional, then every warped product
decomposition of S m (1) ⊂ Em+1 can be obtained as follows:
Let Ek+1 ⊕ Eℓ+1 be an orthogonal direct sum decomposition of Em+1
(m = k + ℓ − 1), and E2 any 2-dimensional linear space. Let (x, y) be
Euclidean coordinates on E2 . Consider a quarter unit circle
1

S+ (1) = (x, y) ∈ E2 : x2 + y 2 = 1, x > 0, y > 0 .
Let S k (1) and S ℓ (1) be the unit hyperspheres of Ek+1 and Eℓ+1 centered
at the origin. Then the mapping
1
ψ : S+ (1) ×x S k (1) ×y S ℓ (1) → S m (1);
((x, y), p, q) 7→ ψ((x, y), p, q) = xp + yq
is an isometry onto an open dense subset of S m (1). Hence this mapping

provides a warped product representation of S m (1)
Let us consider the decomposition given above with m = 2n+1, E2n+2 =
n+1
C , k = n + 3 and ℓ = n − 3. We also consider the parametrization
1
(cos t, sin t) of S+ (1), t ∈ (0, π2 ). Let ψ : M 2 → S n+3 (1) be an isometric
immersion. Then we consider the warped product immersion

F : 0, π2 ×cos t M 2 ×sin t S n−3 (1) → S 2n+1 (1) (14.108)
defined by
F (t, p, q) = (cos t)ψ(p) + (sin t)q.
It is direct to verify that F is a minimal isometric immersion if ψ is.
We now investigate under which conditions F is C-totally real with
respective to the standard Sasakian structure on S 2n+1 (1).
If X1 and X2 are vector field tangent to M 2 and S n−3 (1) respectively,
then we find

∂
F∗ = −(sin t)ψ(p) + (cos t)q,
∂t
F∗ (X1 ) = (cos t)ψ∗ (X1 ),
F∗ (X2 ) = (sin t)X2 .
From this it is easy to verify that F is C-totally real if and only if
hJψ(p), qi = 0, (14.109)
hJψ(p), ψ∗ (X1 )i = 0, (14.110)
hJψ∗ (X1 ), qi = 0, (14.111)
hJψ(p), X2 i = 0, (14.112)
hJq, X2 i = 0. (14.113)
2n+1
Now (14.110) says that ψ has to be C-totally real in S (1); (14.113) says
n−3 2n+1
that S (1) is C-totally real in S (1). On the other hand, (14.111) and
(14.112) are consequences of (14.109).
Suppose that ψ : M 2 → S n+3 (1) is linearly full in some S r (1). Then
(14.109) implies that Jq is orthogonal to this S r (1) for every q ∈ S n−3 (1).
Moreover, from (14.113) we obtain that Jq lies in S n+3 (1). A simple argu-
ment on dimensions implies that n − 2 + r + 1 ≤ n + 4, or still r ≤ 5.
Let En−2 be the linear space containing S n−3 (1). Then the orthogonal
direct sum
J(En−2 ) ⊕ En−2
is a complex subspace Cn−2 ; and S r (1) is contained in S 5 (1), the unit

hypersphere of the orthogonal complement C3 of Cn−2 .
For a Riemannian submanifold M with second fundamental form σ, the
subspace
Nx = {X ∈ Tx M : σ(X, Y ) = 0 for all Y ∈ Tx M }
of the normal space Tx⊥ M is called the relative nullity space at x ∈ M .
We now are in a position to prove the following proposition.
Theorem 14.21. [Chen et al. (1994)] Let S 2n+1 (1) be the unit hypersphere
of Cn+1 and consider the orthogonal decomposition
Cn+1 = C3 ⊕ J(En−2 ) ⊕ En−2 .
Let f : M 2 → S 5 (1) ⊂ C3 be a minimal, isometric, C-totally real immer-
sion and consider the hypersphere S n−3 (1) in En−2 . Then we have:
(1) F : (0, π2 ) ×cos t M 2 ×sin t S n−3 (1) → S 2n+1 (1) defined by
F (t, p, q) 7→ (cos t)ψ(p) + (sin t)q
is a minimal δ(2)-ideal C-totally real immersion.
(2) If ψ has no totally geodesic points, then the dimension of Nx is exactly
n − 2 for each point x.
(3) If we extend F to a map

F̃ : − π2 , π2 × M 2 × S n−3 (1) → S 2n+1 (1)
defined by
F̃ (t, p, q) = (cos t)ψ(p) + (sin t)q.
Then F̃ fails to be immersive at t = 0, but the image of F̃ is an im-
mersed minimal C-totally real submanifold. If ψ is not totally geodesic,
then this image can not be extended further.
Proof. The first part of the theorem follows from the discussion above.
F is δ(2)-ideal because the dimension of N is at least n − 2. In case f has
no totally geodesic points, we have

N = T 0, π2 ⊕ T S n−3 (1) and N ⊥ = T M 2 .
Obviously the conditions (1) and (2) are satisfied in this case.
In order to prove the second part, we introduce a map
n−2
γ : − π2 , π2 × M 2 × S n−3 (1) → S+ × M2
defined by γ(t, p, q) 7→ ((cos t, (sin t)q), p), where

n−2
S+ = {x ∈ En−1 : |x| = 1 and x1 > 0}.
Clearly there exists a map
n−2
y : S+ × M 2 → S 2n+1 (1)
such that y(γ) = F̃ . Now γ(0, p, q) = ((1, 0), p) and it is easy to check that
y is immersive at ((1, 0), p).
We remark that immersion y is one of the examples constructed in [Chen
et al. (1995a)], and that is was not really necessary to extend to (− π2 , π2 );
including 0 would be sufficient. This means that we include the points ψ(p),
which are not obtained by F . Note that all these points are projected by
the Hopf fibration into a totally geodesic CP 2 of CP n .
For proving that we cannot extend the image of F further, we only have
to check the points where t = π/2. An easy computation yields that the
sectional curvature K ′ of the warped product of a plane tangent to M 2 is
given by
K ′ = 1 + (K − 1) sec2 t,
where K is the Gauss curvature of M 2 .
If t goes to π/2, then all points (t, p, q) are mapped into q, so if the
immersed submanifold were to be extendable, K ′ should converge to a finite
value independent of p. Clearly this is not the case, unless K is identically
equal to 1 on M 2 , meaning that f is totally geodesic.
Note that the points q are projected by the Hopf fibration into a totally
geodesic RP n−3 of CP n .
Remark 14.7. It was proved in [Chen et al. (1994)] that if

ψ : M → CP n (4)
is a Lagrangian immersion satisfying the following three conditions:
(1) the equality case of (14.20) identically,
(2) the dimension of the relative nullity space N on M is constant, and
(3) the orthogonal distribution N ⊥ of N is integrable,
then either ψ is totally geodesic or ψ has no totally geodesic point and, up
to holomorphic transformation, φ(M ) is contained in the image of the Hopf
fibration
π : S 2n+1 (1) → CP n (4)
of the image of one of the immersions described in Theorem 14.21.
Chapter 15
Warped Products in Nearly Kähler

Manifolds
15.1 Nearly Kähler manifolds
An almost Hermitian manifold M̃ with metric tensor g and almost complex

structure J is called nearly Kähler if we have [Gray (1970)]
˜ X J)X = 0, ∀X ∈ T M,
(∇ (15.1)
where ∇ ˜ is the Levi-Civita connection of M̃ . Nearly Kähler manifolds are
exactly almost Tachibana manifolds studied in [Tachibana (1959)]. Every
Kähler manifold is nearly Kähler. The converse is not true. Non-Kählerian
nearly Kähler manifolds are called strict nearly Kähler manifolds.
A curve in an almost Hermitian manifold is said to be holomorphically
planar if the holomorphically sections determined by its tangent vectors are
parallel along the curve (cf. [Yano (1965)]). It follows from the definitions
that nearly Kähler manifolds are exactly almost Hermitian manifolds whose
geodesics are holomorphically planar.
Nearly Kähler manifolds can be characterized by the following geometric
property as well: Let γ be a piecewise differentiable loop at x ∈ M̃ and let
π be the almost complex section of the tangent space Tx M̃ which contains
γ ′ (0). Denote by τγ the parallel translation along γ. Then there exists ψ ∈
U (n) such that τγ |π = ψ|π , where U (n) is regarded as the structure group
of the tangent bundle of M . Conversely, any almost Hermitian manifold
with this property is a nearly Kähler manifold [Gray (1970)].
It was showed in [Gray (1970)] that a compact nearly Kähler manifold of
positive holomorphic sectional curvature is simply-connected; generalizing
a well known result of [Tsukamoto (1957)]. He also proved the following:
(a) A nearly Kähler manifold M̃ is Kählerian (resp. Einstein) if dimR M̃ =

4 (resp. if dimR M̃ = 6).
359
(b) A complete simply-connected nearly Kähler manifold with dimR M̃ = 8

is the direct product M1 × M2 of 2-dimensional Kähler manifold M1
and a 6-dimensional nearly Kähler manifold M2 .
(c) {X ∈ Tx M̃ : (∇˜ X J)Y = 0, ∀Y ∈ Tx M̃ , x ∈ M̃ } defines an integrable
distribution on a nearly Kähler manifold M , whose leaves are Kähler.
(d) If a strict nearly Kähler manifold M̃ has sufficiently large sectional
curvature (or holomorphic pinching), then H 2 (M̃ ; R) = 0.
The best known example of nearly Kähler manifolds, but not Kählerian
manifold, is S 6 with the nearly Kählerian structure induced from the vector
cross product on the space of purely imaginary Cayley numbers O (see
§15.2).
It was proved that any 6-dimensional strict nearly Kähler manifold is
an Einstein manifold and it has vanishing first Chern class [Nagy (2002)]
(in particular, this implies spin). It was shown in [Nagy (2002)] that any
complete, strict nearly Kähler manifold is locally a Riemannian product of
homogeneous nearly Kähler spaces, twistor spaces over Kähler manifolds
and 6-dimensional nearly Kähler manifolds. Moreover, P.-A. Nagy proved
in [Nagy (2002)] the following properties for a strict and complete nearly
Kähler manifold M :
(1) If M̃ is not Einstein, then the canonical Hermitian connection on M̃

has reduced holonomy.
(2) M̃ has positive Ricci curvature; hence M̃ is a compact manifold with
a finite fundamental group.
(3) The scalar curvature is a strictly positive constant.
(4) If M̃ is simply-connected, then M is a Riemannian product M1 × M2
of a Kähler manifold M1 and a strict nearly Kähler manifold M2 .
The only known 6-dimensional strict nearly Kähler manifolds are:

S 6 , S 3 × S 3 , Sp(2)/(SU (2) × U (1)),
SU (3)/(U (1) × U (1)).
In fact, these are the only homogeneous nearly Kähler manifolds in dimen-
sion six [Butruille (2005)].
More general examples of nearly Kähler manifolds are the homogeneous
spaces G/K, where G is a compact semisimple Lie group and K is the fixed
point set of an automorphism of G of order 3 [Wolf and Gray (1968)]. It
was also proved in [Butruille (2005)] that homogeneous and strictly nearly
Kähler manifolds are 3-symmetric.
Warped Products in Nearly Kähler Manifolds 361
Nearly Kähler manifolds are related to the existence of a Killing spinor.

Indeed, it was proved in [Friedrich and Grunewald (1985)] that a Rieman-
nian 6-manifold admits a Riemannian Killing spinor if and only if it is nearly
Kähler. Furthermore, nearly Kähler manifolds provide a natural example of
almost-Hermitian manifold admitting a Hermitian connection with totally
skew symmetric torsion. From this point of view, nearly Kähler manifolds
are of interest in string theory (see [Friedrich and Ivanov (2002)]).
15.2 Nearly Kähler structure on S 6
The unit 6-sphere S 6 can be considered as a homogeneous space G2 /SU (3),

where G2 is the group of automorphisms of the Cayley numbers O. From
this representation, one can define a nearly Kählerian structure on S 6 by
making use of the vector cross product of O.
Now, we give a brief explanation of how this standard nearly Kähler
structure on S 6 arises in a natural manner from Cayley multiplication.
The multiplication of the Cayley algebra O may be used to define a vector
cross-product on the purely imaginary Cayley numbers R7 using the formula
u × v = 21 (uv − vu), (15.2)
while the standard inner product on R7 is given by
(u, v) = − 21 (uv + vu). (15.3)
It is now elementary to show that
u × (v × w) + (u × v) × w
(15.4)
= 2(u, w)v − (u, v)w − (w, v)u,
and that the triple scalar product hu × v, wi is skew symmetric in u, v, w.
Conversely, Cayley multiplication on O is given in terms of the vector
cross product and the inner product by
(r + u)(s + v) = rs + rv + su − (u, v) + (u × v) (15.5)
for r, s ∈ Re O, u, v ∈ Im O. In view of (15.2), (15.3) and (15.5), it is
clear that the group G2 of automorphisms of O is precisely the group of
isometries of R7 which preserve the vector cross-product.
An ordered orthonormal basis {e1 , . . . , e7 } of R7 is called canonical if
e3 = e1 × e2 , e5 = e1 × e4 ,
(15.6)
e6 = e2 × e4 , e7 = e3 × e4 .
For example, the standard basis of R7 is canonical. Moreover, (E7 , ×) is

generated by e1 , e2 , e4 subject to the relations
ei × (ej × ek ) + (ei × ej ) × ek = 2δik ej − δij ek − δjk ei . (15.7)
For two given canonical bases {e1 , . . . , e7 }, {f1 , . . . , f7 } of R7 , there exists
a unique g ∈ G2 such that gei = fi ; thus g is uniquely determined by
ge1 , ge2 , ge4 .
For any G2 -frame, we have the following multiplication table.
× e1 e2 e3 e4 e5 e6 e7
e1 0 e3 −e2 e5 −e4 −e7 e6
e2 −e3 0 e1 e6 e7 −e4 −e5
e3 e2 −e1 0 e7 −e6 e5 −e4
e4 −e5 −e6 −e7 0 e1 e2 e3
e5 e4 −e7 e6 −e1 0 −e3 e2
e6 e7 e4 −e5 −e2 e3 0 −e1
e7 −e6 e5 e4 −e3 −e2 e1 0
It follows from the above that G2 acts transitively on S 6 and that the
stabilizer of the point (1, 0, . . . , 0) is SU (3). Thus we have S 6 = G2 /SU (3).
Moreover, it follows that G2 , a connected subgroup of SO(7) of dimension
14, is the group of automorphisms of the nearly Kähler structure J.
Let J be the automorphism of T S 6 defined by
Ju = x × u, u ∈ Tx S 6 , x ∈ S 6 . (15.8)
Then J is an almost complex structure on S 6 , which is not integrable; but
in fact J is a nearly Kähler structure on S 6 .
The fundamental 2-form Ω of (S 6 , g, J) is clearly not closed, but it
satisfy
Ω ∧ dΩ = 0.
Moreover, it was shown in [Calabi and Gluck (1993)] that J is the best
almost complex structure on S 6 from variational point of view.
Define a skew-symmetric (2, 1)-tensor field G on S 6 by
˜ X J)(Y ),
G(X, Y ) = (∇ (15.9)
where ∇˜ is the Levi-Civita connection of S 6 .

It is direct to show that
G(X, Y ) = X × Y − g(x × X, Y )x. (15.10)
It is also known that the (2, 1)-tensor G satisfies the following properties
[Sekigawa (1983)]:
G(X, JY ) = −JG(X, Y ), (15.11)
˜
(∇G)(X, Y, Z) = g(Y, JZ)X + g(X, Z)JY − g(X, Y )JZ, (15.12)
g(G(X, Y ), Z) + g(G(X, Z), Y ) = 0, (15.13)
g(G(X, Y ), G(Z, W )) = g(X, Z)g(Y, W ) − g(X, W )g(Z, Y ) (15.14)
+ g(JX, Z)g(Y, JW ) − g(JX, W )g(Y, JZ).
15.3 Complex submanifolds of nearly Kähler manifolds
A submanifold M of an almost Hermitian manifold M̃ is called a complex

submanifold if its tangent bundle T M is invariant under the action of the
almost complex structure J of M̃ .
Proposition 15.1. [Gray (1969)] Let M̃ be a nearly Kähler manifold with

almost complex structure J. If M is a complex submanifold of M , then we
have:
(1) M is nearly Kählerian;

(2) σ(X, JX) = Jσ(X, X) for all X ∈ T M ;
(3) M is an austere submanifold. In particular, it minimal in M̃ .
Proof. For X ∈ T M we have

˜ X J)X = (∇X J)X + σ(X, JX) − Jσ(X, X),
0 = (∇
where ∇ is the Levi-Civita connection of M . Hence we have statement (1)
and statement (2).
Statement (3) is an easy consequence of statement (2).
The following result from [Gray (1969)] is well-known.
Proposition 15.2. Every nearly Kähler 4-manifold is a Kähler manifold.
Proof. Assume that M̃ be a 4-dimensional nearly Kähler manifold. Let

us choose an orthonormal frame field on an open subset of M̃ to be of the
form {X, JX, Y, JY }. Clearly, (∇X J)Y is perpendicular to X and Y . It is
also perpendicular to JX and JY since (∇X J)Y = J(∇X J)(JY ). Hence
we obtain (∇X J)Y = 0 for all X, Y ∈ T M̃ . Consequently, M̃ is a Kähler
manifold.
Definition 15.1. A complex submanifold M of an almost Hermitian man-

ifold M̃ is called a σ-submanifold if its second fundamental form satisfies
σ(JX, Y ) = Jσ(X, Y )
for X, Y ∈ T M.
For complex submanifolds of a nearly Kähler manifold we have the fol-

lowing result from [Vanhecke (1977)].
Proposition 15.3. Every complex submanifold of a nearly Kähler manifold

is a σ-submanifold.
Proof. Let M be a complex submanifold of a nearly Kähler manifold M̃ .

It follows from the definition of nearly Kähler manifold that
˜ X J)Y + (∇
(∇ ˜ Y J)X = 0, X, Y ∈ T M. (15.15)
˜ we have
From the definition of ∇J,
˜ X JY = J ∇
∇ ˜ X Y + (∇˜ X J)Y = 0,
(15.16)
˜ JX J 2 Y = J ∇
∇ ˜ JX JY + (∇˜ JX J)JY = 0.
Thus, by using (15.15) we get

˜ X JY − ∇
∇ ˜ JX Y = J(∇
˜ XY + ∇
˜ JX JY ). (15.17)
It follows from (15.17) and Gauss’ formula that
∇X JY − ∇JX Y = J(∇X Y + ∇JX JY ) (15.18)
and further
σ(X, JY ) − σ(JX, Y ) = Jσ(X, Y ) + Jσ(JX, JY ). (15.19)
The right-hand side of (15.19) is symmetric in X and Y due to the symme-
try of σ and the left-hand side is skew-symmetric what is easily seen with
(2.9). Therefore we have
σ(X, JY ) = σ(JX, Y ). (15.20)
It follows from (15.16) and the definition of nearly Kähler manifold that
˜ X JY + ∇
∇ ˜ Y JX = J(∇
˜ XY + ∇
˜ Y X) (15.21)
and so we get
σ(X, JY ) + σ(JX, Y ) = 2Jσ(X, Y ). (15.22)
Now, the proposition follows immediately from (15.20) and (15.22).
For complex submanifolds of the nearly Kähler S 6 we have
Theorem 15.1. [Gray (1969)] With respect to the usual nearly Kähler
structure, S 6 has no 4-dimensional complex submanifolds.
Proof. Suppose that M is a 4-dimensional complex submanifold of S 6 .

Assume that there exists x ∈ M and a unit vector X ∈ Tx M such that
σ(X, X) 6= 0. Then we have σ(X, JX) = Jσ(X, X) 6= 0. We may assume
that X is chosen so that |σ(X, X)|2 is maximum.
If g(X, Y ) = g(JX, Y ) = 0, then
g(σ(X, X), σ(X, Y )) = 0.
Also, we find g(Jσ(X,
√ X), σ(X, Y )) = 0 because Jσ(X, X) = σ(U, U ) with
U = (X + JY )/ 2.
Similarly, σ(X, JY ) is perpendicular to both σ(X, X) and Jσ(X, X).
Since σ(X, X) and Jσ(X, X) span Tx⊥ M , we have σ(X, Y ) = σ(X, JY ) = 0.
Let K̃ denote the constant sectional curvature of the curvature operator
of S 6 . Then if |Y | = 1, we have
K̄ = g(R̃(X, Y )Y, X) − g(R̃(X, Y )JY, JX)
˜ X J)Y |2
= |(∇
= |(∇X J)Y + σ(X, JY ) − Jσ(X, Y )|2
= 0.
This is a contradiction, thus σ(X, X) = 0 for all X ∈ Tx M . Therefore
M is totally geodesic in S 6 , and hence it must be an open submanifold of
a 4-sphere with constant curvature one, and also must be Kählerian. We
then have 0 = |(∇X J)Y |2 = 1, which is impossible.
Let ψ : M → S 6 be an isometric immersion of a surface M into S 6 ⊂ E7 .
Definition 15.2. The ellipse of curvature at a point x ∈ M is defined as

the image of the unit circle of Tx M under the second fundamental σ, i.e.,
Ex = {σ(u, u) | u ∈ Tx M, |u| = 1},
where σ is the second fundamental form.
The third fundamental form III of ψ is defined by
˜ X σ(Y, Z))⊥ , X, Y, Z ∈ T M,
III(X, Y, Z) = (∇
where ∇˜ is the flat connection on E7 and (∇ ˜ X σ(Y, Z))⊥ is the normal
component of ∇ ˜ X σ(Y, Z) orthogonal to the first normal subspace of ψ.
The image of the unit circle of Tx M under III is called the ellipse of second
curvature.
Definition 15.3. A minimal surface M in S 6 is called superminimal if,

at each point x ∈ M , the ellipse of curvature and the ellipse of second
curvature are either circles or points.
Almost complex curves in S 6 have been studied among others in [Bolton
et al. (2007); Byrant (1982); Sekigawa (1983)]. Bryant’s results states that
superminimal complex curves can be completely described by means of a
Weierstrass’ formula and that curves of every genus can occur.
It was proved in [Sekigawa (1983)] that if the Gauss curvature G of a
complex curve in S 6 is constant, then G = 0, 61 or 1.
Almost complex curves in S 6 were classified into four classes in [Bolton
et al. (2007)]; namely,
(a) linearly full in S 6 and superminimal;

(b) linearly full in S 6 but not superminimal;
(c) linearly full in a total geodesic S 5 ⊂ S 6 necessarily not superminimal;
(d) totally geodesic.
15.4 Lagrangian submanifolds of nearly Kähler manifolds
Let M be a Lagrangian submanifold of a nearly Kähler manifold M̃ . Then

from (2.5), (2.13) and (15.9), we have
DX JY = J∇X Y + G(X, Y ), (15.23)
AJY X = −Jσ(X, Y ), (15.24)
for X, Y ∈ T M.
For Lagrangian submanifolds in S 6 we have the following well-known
result of [Ejiri (1981)].
Theorem 15.2. Every Lagrangian submanifold of the nearly Kähler S 6 is
orientable and minimal.
Combining Theorem 13.3 and Theorem 15.2 gives the following.
Theorem 15.3. Every Lagrangian submanifold of the nearly Kähler S 6
satisfies δ(2) ≤ 2.
The following is an immediate consequence of Theorem 15.3 (Theorem
19.4 of [Chen (2011b)]).
Corollary 15.1. A Lagrangian submanifold of the nearly Kähler S 6 is δ(2)-
ideal if and only if δ(2) = 2.
The existence and uniqueness theorems for Lagrangian submanifolds in

the nearly Kähler S 6 are the following [Ejiri (1981)] (see also [Chen et al.
(1995b)]).
Theorem 15.4. Let x1 , x2 : M → S 6 be two Lagrangian immersions of

a Riemannian manifold M with second fundamental forms σ 1 and σ 2 and
with the same orientation (induced by JG). If

1

σ (X, Y ), Jx1⋆ Z = σ 2 (X, Y ), Jx2⋆ Z (15.25)
for all vector fields X, Y, Z tangent to M , then there exists an isometry Ψ
of S 6 such that x1 = x2 ◦ Ψ.
Proof. From (15.23) and (15.24) we see that the normal connections and
the shape operators coincide. Thus, by applying the standard uniqueness
result for minimal immersions in real space forms we get the theorem.
Theorem 15.5. Let M be a 3-dimensional simply connected (hence it is

oriented) Riemannian manifold. Let ∧ be the skew symmetric operator
∧ : TM × TM → TM
assigning to two linearly
q independent vectors X and Y the unique vector
X ∧ Y with length hX, Xi hY, Y i − hX, Y i2 which is orthogonal to both X
and Y such that {X, Y, X ∧ Y } form a positively oriented basis.
Let α be a symmetric bilinear T M -valued form on M satisfying
(1) Tr α = 0,
(2) hα(X, Y ), Zi is totally symmetric,
(3) (∇α)(X, Y, Z) + X ∧ α(Y, Z) is totally symmetric,
(4) R(X, Y )Z = hY, Zi X − hX, Zi Y + α(α(Y, Z), X) − α(α(X, Z), Y ).
Then there exists a Lagrangian immersion x : M → S 6 such that the second
fundamental form σ satisfies
σ(X, Y ) = Jα(X, Y )
such that G(X, Y ) = J(X ∧ Y ).
Proof. We first derive some general properties of the skew-symmetric

operator ∧ on M . First, by taking a positively oriented basis, it is easy to
check that
hX ∧ Y, Zi + hX ∧ Z, Y i = 0,
hY3 ∧ Y1 , Y3 ∧ Y2 i = 0,
for arbitrary tangent vector fields X,Y and Z and for orthonormal vector
fields Y1 , Y2 and Y3 .
Next, we take X and Y orthonormal. Then we have
h∇Z (X ∧ Y ), Xi = − hX ∧ Y, ∇Z Xi
= h(∇Z X) ∧ Y, Xi
= h(∇Z X) ∧ Y + X ∧ (∇Z Y ), Xi
h∇Z (X ∧ Y ), Y i = − hX ∧ Y, ∇Z Y i
= hX ∧ ∇Z Y, Y i
= h(∇Z X) ∧ Y + X ∧ (∇Z Y ), Y i
h∇Z (X ∧ Y ), (X ∧ Y )i = 0
h(∇Z X) ∧ Y + X ∧ (∇Z Y ), X ∧ Y i = 0.
So, we have
∇Z (X ∧ Y ) = (∇Z X) ∧ Y + X ∧ (∇Z Y ). (15.26)
A straightforward computation shows that (15.26) remains valid if X and
Y are arbitrary tangent vector fields.
Now, we take N M = T M as bundle over M . We define a connection
⊥
∇ on N M by
∇⊥
X J0 Y = J0 ∇X Y + J0 X ∧ Y,
for X and Y tangent to M , where J0 is the identification between T M and
N M . We define the second fundamental form σ by
σ : T M × T M → N M : (X, Y ) 7→ σ(X, Y ) = J0 α(X, Y ),
and the Weingarten operator A by AJ0 X Y = α(X, Y ) for X, Y ∈ T M .
A straightforward computation shows that the Gauss, Codazzi and Ricci
equations are satisfied and hence, by the existence and uniqueness theorem
for immersions into a real space form, there exists an isometric immersion
x : M 3 → S 6 with second fundamental form σ, normal connection ∇⊥ and
Weingarten operator A.
Let us now prove that x is a Lagrangian immersion. In order to do so,
we consider S 6 as a hypersphere in E7 . Let p ∈ M . We define a vector
cross product on E7 by
p × X = J0 X,
p × J0 X = −X,
X × Y = J0 X ∧ Y, (15.27)
J0 X × Y = (X ∧ Y ) − hX, Y i p,
J0 X × J0 Y = −X ∧ Y.
If we denote the Levi-Civita connection of E7 by D, we have

nDY (p × X) − (DY p) × X − p × DY X
= DY (J0 X) − Y × X − J0 DY X
= ∇⊥
Y J0 X − AJ0 X Y + J0 X ∧ Y
(15.28)
− J0 (∇Y X) − J0 σ(X, Y )
= J0 ∇Y X + J0 Y ∧ X + J0 X ∧ Y − J0 ∇Y X
= 0.
Computing the other equations in a similar way, we obtain that × is parallel.
The uniqueness of the vector cross product (see [Calabi (1958)]) then shows
that after applying a suitable element of SO(7), M is Lagrangian in S 6 ,
and that the almost complex structure of S 6 , applied to vectors tangent to
M , coincides with J0 .
Remark 15.1. Theorem 15.5 was first given in a preprint of [Ejiri (1981)].
For minimal totally real surfaces in the nearly Kähler S 6 we have the fol-
lowing.
Theorem 15.6. [Dillen et al. (1988)] If M is a minimal totally real surface

of the nearly Kähler S 6 and if M is homeomorphic to a sphere, then M is
totally geodesic.
For Lagrangian submanifolds in the nearly Kähler S 6 we have the fol-

lowing pinching theorem.
Theorem 15.7. [Dillen et al. (1987)] Let M be a compact Lagrangian sub-

manifold of the nearly Kähler S 6 . If all sectional curvatures K satisfy
1
K > 16 , then M is totally geodesic.
Theorem 15.8. [Dillen et al. (1990)] Let M be a compact Lagrangian sub-

manifold of the nearly Kähler S 6 . If all sectional curvatures K satisfy
21 1
16 > K ≥ 16 , then M is totally geodesic or M has constant sectional
1
curvature 16 .
Theorem 15.9. [Zhong (2001)] Let M be a compact Lagrangian subman-

ifold of the nearly Kähler S 6 . If the second fundamental form σ satisfies
||σ||2 < 52 , then M is totally geodesic.
Theorem 15.10. [Antić et al. (2005)] Let M be a compact Lagrangian

submanifold of the nearly Kähler S 6 . Then all Ricci curvatures Ric satisfy
Ric(v) ≥ 34 if and only if M is totally geodesic.
15.5 CR-submanifolds in nearly Kähler manifolds
The Nijenhuis tensor NJ of an almost complex manifold (M̃ , J) is defined

by
NJ (U, V ) = [JU, JV ] − [U, V ] − J[JU, V ] − J[U, JV ] (15.29)
for any U, V ∈ T M .
From (15.29) we have
NJ (JU, V ) = −JNJ (U, V ) = NJ (U, JV ). (15.30)
The well-known Newlander-Nirenberg theorem states that an almost
complex structure J on an almost complex manifold M̃ is integrable if and
only if the Nijenhuis tensor of J vanishes identically. In this case, M̃ is a
complex manifold [Newlander and Nirenberg (1957)].
The Nijenhuis tensor of a nearly Kähler manifold M̃ can be written in
the following simpler form.
Proposition 15.4. [Tachibana (1959)] The Nijenhuis tensor of the almost

complex structure J on a nearly Kähler manifold M̃ is given by
NJ (U, V ) = 4J(∇˜ V J)U (15.31)
for U, V ∈ T M .
The following result gives the integrability condition for the holomorphic
distribution of a CR-submanifold in nearly Kähler manifold.
Proposition 15.5. [Sato (1981, 1982)] Let M be a CR-submanifold of a

nearly Kähler manifold M̃ . Then the holomorphic distribution H of M is
integrable if and only if the following two conditions hold:
(1) The second fundamental form satisfies σ(JX, Y ) = σ(X, JY )
(2) The Nijenhuis tensor satisfies NJ (X, Y ) ∈ H,
for X, Y ∈ H.
Proof. From (15.31) we have

hNJ (U, V ), W i = − hNJ (U, W ), V i . (15.32)
It follows from (15.32) and Gauss’ formula that
∇X (JY ) − ∇Y (JX) + σ(X, JY ) − σ(Y, JX)
1 (15.33)
= JNJ (X, Y ) + J[X, Y ].
2
The proposition now follows from (15.29) and (15.33).
The next result gives integrability condition for the totally distribution
of a CR-submanifold in nearly Kähler manifold.
Proposition 15.6. [Sato (1981, 1982)] Let M be a CR-submanifold of a

nearly Kähler manifold. Then the totally real distribution H⊥ is integrable
if and only if we have
hσ(X, W ), JZi = hσ(X, Z), JW i (15.34)
⊥
for X ∈ H and Z, W ∈ H .
Proof. Let M be a CR-submanifold of a nearly Kähler manifold M̃ .

Assume that the totally real distribution H⊥ is integrable. From (15.23)
and the formulas of Gauss and Weingarten we get
∇˜ X JZ = −AJZ X + DX JZ, (15.35)
1
˜ X JZ = JNJ (X, Z) + J(∇X Z + σ(X, Z)).
∇ (15.36)
4
∇˜ JX JZ = ∇Z X + J(∇JX Z + ∇Z JX) (15.37)
+ σ(X, Z) + 2Jσ(Z, JX).
From (15.35), (15.36) and (15.37) we derive that
hAJZ JX, W i + hJ∇JX Z, W i + 2 hJσ(JX, Z), W i
(15.38)
+ h∇Z X, W i + hJ∇Z JX, W i = 0,
and hence we have
hσ(JX, W ), JZi = 2 hσ(JX, Z), JW i + hX, ∇Z W i . (15.39)
Similarly, we find
hσ(JX, Z), JW i = 2 hσ(JX, W ), JZi + hX, ∇W Zi . (15.40)
Now, the proposition follows from (15.39) and (15.40). This completes the
proof of the proposition.
The following is an easy consequence of Proposition 15.6.
Corollary 15.2. [Sato (1981, 1982)] Let M be a CR-submanifold of a

nearly Kähler manifold. If the totally real distribution H⊥ is integrable,
then the leaves of H⊥ are totally geodesic in M if and only if we have the
following
hσ(X, W ), JZi = 0
holds for X ∈ H and Z, W ∈ H⊥ .
Proof. Follows immediately from (15.33) and Proposition 15.5.

15.6 Warped products in nearly Kähler manifolds
The following result from [Khan et al. (2007); Sahin and Günes (2008);
Al-Luhaibi et al. (2009)] is analogous to Theorem 11.1 for warped product
CR-submanifolds of a Kähler manifold.
Theorem 15.11. If N ⊥ ×f N T is a warped product CR-submanifold in a

nearly Kähler manifold M̃ such that N ⊥ is a totally real submanifold and
N T is a complex submanifold of M̃ , then N ⊥ ×f N T is non-proper. Hence
the warping function f is constant.
Proof. Under the hypothesis, we have h∇X Z, Xi = h ∇ ˜ X Z, X i for any

X ∈ H and Z ∈ H⊥ . Combining this with Proposition 3.1(2) gives
˜ X Z, X i = X(ln f ) hX, Xi .
h∇ (15.41)
Since H⊥ ⊥ H, we have hX, Zi = 0. Hence we find from (15.41) that
˜ X X i = −X(ln f ) hX, Xi .
h Z, ∇
˜ X X i = −Z(ln f ) hX, Xi . So we obtain
Thus, using (15.1), we get h JZ, J ∇
˜ X JX − (∇
Z(ln f ) hX, Xi = − h JZ, ∇ ˜ X J)X i .
˜ X JX i = −Z(ln f ) hX, Xi . Hence, by
So we derive from (15.1) that h JZ, ∇
Gauss’ formula we find
hσ(X, JX), JZi = −Z(ln f ) hX, Xi . (15.42)
From Weingarten’s formula we get
˜ X JZ, JX i = h JZ, ∇
hAJZ X, JXi = − h ∇ ˜ X JX i .
˜ is a torsion free connection, we derive
Since ∇
˜ JX X i= h JZ, ∇
hAJZ X, JXi= h JZ, [X, JX] + ∇ ˜ JX X i = h JZ, ∇
˜ JX J 2 X i ,
which implies
˜ JX J)JX + J ∇
hAJZ X, JXi = − h JZ, (∇ ˜ JX JX i .
Hence, from (15.1) and hX, Y i = hJX, JY i, we find

˜ JX JX i = h ∇
hAJZ X, JXi = − h Z, ∇ ˜ JX Z, JX i = h∇JX Z, JXi .
Therefore we obtain from Proposition 3.1(2) that

hσ(X, JX), JZi = Z(ln f ) hJX, JXi = Z(ln f ) hX, Xi . (15.43)
Now, by combining (15.42) and (15.43) we find Z(ln f ) = 0, which implies
that f is constant. Consequently, N ⊥ ×f N T is non-proper.
The next result from [Sahin and Günes (2008); Al-Luhaibi et al. (2009)]
is analogous to Theorem 11.3.
Theorem 15.12. Let M = N T ×f N ⊥ be a CR-warped product in a nearly

Kähler manifold M̃ . Then we have:
||σ||2 ≥ 2p |∇(ln f )|2 , (15.44)
where p is the dimension of N ⊥ .

(2) If the equality sign of (15.44) holds identically, then N T is a totally
Moreover, M is a minimal submanifold in M̃ .
(3) When M is anti-holomorphic and p > 1. The equality sign of (15.44)
holds identically if and only if N ⊥ is a totally umbilical submanifold of
M̃ .
(4) If M is a real hypersurface, then the equality sign of (15.69) holds
identically if and only if the characteristic vector field Jξ of M is a
principal vector field with zero as its principal curvature, where ξ a
unit normal vector field of M . Also, in this case, the equality sign of
(15.44) holds identically if and only if M is a minimal hypersurface.
Proof. This can be done in a similar way as Theorem 11.3 for CR-warped
products in Kähler manifolds.
Remark 15.2. It was proved in [Khan et al. (2009)] that there does not
exist a proper doubly warped product CR-submanifold f1 M T ×f2 M ⊥ in a
nearly Kähler manifold such that M T is a complex submanifold and M ⊥
is a totally real submanifold.
For semi-slant warped product submanifolds of a nearly Kähler mani-

folds, we have the following.
Proposition 15.7. [Khan and Khan (2009)] Let N T ×f N θ be a semi-slant

warped product of a nearly Kähler manifold M̃ with N T and N θ complex
and θ-slant submanifolds of M̃ , respectively. Then the second fundamental
form σ of M satisfies

1
||σ||2 ≥ 2q csc2 θ 1 + cos4 θ |∇(ln f )|2 , (15.45)
9
where q = dim M θ .
For hemi-slant warped product submanifolds of a nearly Kähler mani-

folds, we have the following result from [Al-Solamy and Khan (2014a)].
Proposition 15.8. Let M = N θ ×f N ⊥ be a hemi-slant warped product of a
nearly Kähler manifold with N θ and N ⊥ slant and totally real submanifolds
of M̃ , respectively. Denote the canonical slant and totally real distributions
of M by Hθ and H⊥ , respectively.
If the mean curvature vector Hx at x ∈ M lies in the maximal complex
subspace µx of the normal space Tx⊥ M for each x ∈ M , then we have
||σ||2 ≥ 4q cos2 θ |∇(ln f )|2 , q = dim N ⊥ . (15.46)
θ θ ⊥ ⊥ θ ⊥ ⊥
(2) If σ(H , H ) = σ(H , H ) = {0} and σ(P H , H ) ⊥ F H ⊕ F (T M )
hold, then the equality sign of (15.46) holds identically
For a submanifold M of a nearly Kähler manifold M̃ , we put
(∇˜ U J)V = PU V + QU V, U, V ∈ T M, (15.47)
where PU V and QU V denote the tangential and the normal components
˜ U J)V , respectively.
of (∇
For hemi-slant warped product submanifolds of a nearly Kähler mani-
folds, we also have the following results from [Uddin and Chi (2011a)].
Proposition 15.9. A warped product submanifold N ⊥ ×f N θ of a nearly
Kähler manifold M̃ is a Riemannian product of a proper slant submanifold
N θ and a totally real submanifold N ⊥ if and only if we have PX (P X) ∈
T N θ for any X ∈ T N θ .
Proposition 15.10. A warped product submanifold N θ ×f N ⊥ of a nearly
Kähler manifold M̃ is a Riemannian product of a proper slant submanifold
N θ and a totally real submanifold N ⊥ if and only if we have
hσ(X, Z)F Zi = hσ(Z, Z), F Xi (15.48)
θ ⊥
for any X ∈ T N and Z ∈ T N .
Proposition 15.11. There does not exist any mixed totally geodesic warped
product submanifold M = N θ ×f N ⊥ in a nearly Kähler manifold such that
σ(T M, H⊥ ) ⊂ µ, where µ is the maximal complex subbundle of the normal
bundle of M .
Remark 15.3. For further results on semi-slant and semi-slant submani-
folds in a nearly Kähler manifold, see also [Khan and Khan (2007); Lone
and Shahid (2016)].
15.7 Examples of warped product CR-submanifolds in

nearly Kähler S 6
The following is an example of warped product CR-submanifold in the

nearly Kähler S 6 whose holomorphic and totally real distributions are both
integrable.
Example 15.1. [Sekigawa (1983)] Put

S 2 = y = (y2 , y4 , y6 ) ∈ E3 : y22 + y42 + y62 = 1
be a unit 2-sphere in E3 and let S 1 = {z = eit : t ∈ R} be a unit circle in
the complex plane C.
Consider the map ψ : S 2 × S 1 → S 5 ⊂ S 6 defined by
ψ(y, z) = 0, y2 cos t, −y2 sin t, y4 cos 2t,
(15.49)
y4 sin 2t, y6 cos t, y6 sin t ,
with y = (y2 , y4 , y6 ) ∈ S 2 , t ∈ R.
It is direct to verify that the map ψ gives rise to an isometric immersion
from the warped product S 2 ×f S 1 into S 6p , where f is the function on S 2
given by the restriction of F (y2 , y4 , y6 ) = 1 + 3y42 to S 2 . The map ψ is
not embedding since

ψ (y2 , y4 , y6 ), eit = ψ (−y2 , y4 , −y6 ), ei(t+π)
for any (y2 , y4 , y6 ) ∈ S 2 and t ∈ R.

Let σ ′ denote the second fundamental form of M = S 2 ×f S 1 in E7 and
let σ be the second fundamental form of M = S 2 ×f S 1 in S 6 . Then we
obtain
σ ′ (X, Y ) = σ(X, Y ) − hX, Y i x (15.50)
for any X, Y ∈ T M , where x denotes the position vector of M in E7 .
For each t ∈ R, we find
n
St2 = (x1 , . . . , x7 ) ∈ S 6 : (sin t)x2 + (cos t)x2 = 0,
o
(sin 2t)x4 − (cos 2t)x5 = 0, (sin t)x6 + (cos t)x7 = 0 .
Hence each St2 is a totally geodesic surface of S 6 .

Now, we put

∂ ∂
Xα = ψ∗ , Z = ψ∗ , α = 1, 2, (15.51)
∂uα ∂t
where u = (u1 , u2 ) is a local coordinate system of S 2 . From (15.50), (15.51)

and the table in section 14.2, we have
Jx X α =

∂y6 ∂y4 ∂y4 ∂y6
0, y4 − y6 cos t, y6 − y4 sin t,
∂uα ∂uα ∂uα ∂uα

∂y2 ∂y6 ∂y2 ∂y6
y6 − y2 cos 2t, y6 − y2 sin 2t,
! (15.52)
∂y4 ∂y2 ∂y2 ∂y4
y2 − y4 cos t, y4 − y2 sin t ,

Jx Z = − y22 + 2y42 − y62 , 3y4 y6 sin t, 3y4 y6 cos t, 0, 0, −3y2y4 sin t,

− 3y2 y4 cos t .
If we put H = Span{X1 , X2 } and H⊥ = Span{Z}, then it follows from

(15.66) that ψ : S 2 ×f S 1 → S 6 is a proper CR-submanifold with holomor-
phic distribution H and the totally real distribution H⊥ .
Further, it is direct to verify that the distributions H and H⊥ are both
integrable. Moreover, it is straightforward to verify that ψ : S 2 ×f S 1 → S 6
is a minimal immersion in the nearly Kähler S 6 .
Example 15.1 was extended in [Hashimoto and Mashimo (1999)] to the

following.

Example 15.2. Let S 2 = y = (y2 , y4 , y6 ) ∈ E3 : y22 + y42 + y62 = 1 . For a
given triple r = (r1 , r2 , r3 ) of real numbers such that r1 + r2 + r3 = 0 and
r1 r2 r3 6= 0, let us consider the map ψr : S 2 × R → S 5 ⊂ S 6 defined by

ψr (x1 , x2 , x3 , t) = x1 cos(r1 t), x2 cos(r2 t), x3 cos(r3 t), 0,
(15.53)
x1 sin(r1 t), x2 sin(r2 t), x3 sin(r3 t) ,
where x21 + x22 + x23 = 1 and t ∈ R. It is direct to verify that ψr gives

rise to an isometric immersion from the warped product S 2 ×f R into S 6 .
Moreover, the induced warped product metric is given by

g = π1∗ g0 + r1 x1 + r2 x2 + r3 x3 )2 π2∗ dt2 , (15.54)
where π1 : S 2 ×f R → S 2 , π2 : S 2 ×f R → R are natural projections and
g0 is the standard metric on the unit 2-sphere S 2 . Similar to Example 15.1,
for each triple r = (r1 , r2 , r3 ) the map ψr gives rise to a CR-immersion into
the nearly Kähler S 6 .
Remark 15.4. It was shown in [Hashimoto and Mashimo (1999)] that, up

to G2 , there exists a unique 3-dimensional CR-submanifold in the nearly
Kähler S 6 which is obtained as an orbit of reducible representations of
SU (2) on E7 .
As orbits of the subgroup which corresponds to the irreducible repre-
sentation of SU (2) on E7 , Hashimoto and Mashimo obtained 2-parameter
family of 3-dimensional CR-submanifolds in S 6 .
15.8 Non-existence of CR-products in nearly Kähler S 6
A submanifold of an almost Hermitian manifold M̃ is called a CR-product

if it is the Riemannian product N T × N ⊥ of a complex submanifold N T
and a totally real submanifold N ⊥ of M̃ (cf. Definition 9.3). As before,
we denote the holomorphic distribution and the totally real distribution of
N T × N ⊥ by H and H⊥ , respectively.
The following result from [Sekigawa (1983)] shows that CR-products in
the nearly Kähler S 6 are either complex or totally real.
Theorem 15.13. There does not exist any proper CR-product in the nearly
Kähler S 6 .
Proof. Assume M = N T × N ⊥ is a CR-product in the nearly Kähler S 6 .

Then N T is a complex submanifold of S 6 . Thus dimR N T = 2 by Theorem
15.1. Let σ T and σ denote the second fundamental forms of N T and M in
S 6 . Then we have
σ T (X, Y ) = σ(X, Y ) (15.55)
for X, Y ∈ T N T due to the fact that N T is totally geodesic in M .
Since N T is a complex submanifold of S 6 , Proposition 15.3 implies that
M is a σ-submanifold of S 6 , i.e., we have
T
σ T (JX, Y ) = Jσ(X, Y ), X, Y ∈ T N T . (15.56)

σ(JX, Y ) = Jσ(X, Y ). (15.57)
Since N T × N ⊥ is a Riemannian product, we have
h∇U X, Zi = − hX, ∇U Zi = 0 (15.58)
for vector fields U ∈ T M , X ∈ H and Z ∈ H⊥ .
The Gauss equation and (15.58) imply that

hσ(X, X), σ(Z, Z)i − hσ(X, Z), σ(X, Z)i = −1, (15.59)
hσ(X, JX), σ(Z, Z)i − hσ(X, Z), σ(JX, Z)i = 0, (15.60)
hσ(X, JX), σ(X, Z)i − hσ(X, X), σ(JX, Z)i = 0, (15.61)
for unit vectors X ∈ H and Z ∈ H⊥ . By applying (15.57) we find
hσ(X, Y ), JZi = − hσ(JX, Y ), Zi = 0 (15.62)
for X, Y ∈ H and Z ∈ H⊥ .
By using (15.25) and Gauss’ formula we get
˜ Z JX = 1 JNJ (Z, X) + J(∇Z X + σ(Z, X)),

∇ (15.63)
4
˜
∇Z JX = ∇Z JX + σ(Z, JX). (15.64)
Since ∇Z X ∈ H, it follows from (15.63) and (15.64) that
1
σ(Z, JX) = JNJ (Z, X) + Jσ(Z, X). (15.65)
4
Let us fix a unit vector Z ∈ Hx⊥ and let E be a unit vector in Hx satisfying
hσ(E, E), σ(Z, Z)i = max {hσ(X, X), σ(Z, Z)i}. (15.66)
|X|=1,X∈Hx
Then we have
hσ(E, JE), σ(Z, Z)i = 0. (15.67)
From (15.60) and (15.67) we get
hσ(E, Z), σ(JE, Z)i = 0. (15.68)
Now, we claim that σ(E, Z) 6= 0. Suppose that σ(E, Z) = 0. Then it
follows from (15.59) that
hσ(E, E), σ(Z, Z)i = −1. (15.69)
From (15.59) and (15.66) we obtain
hσ(E, Z), σ(E, Z)i = max {hσ(X, Z), σ(X, Z)i}.
|X|=1,X∈Hx
Thus we find hσ(JE, Z), σ(JE, Z)i = 0, and hence from (15.57) and (15.59)
it follows that
−1 = hσ(JE, JE), σ(Z, Z)i = − hσ(E, E), σ(Z, Z)i ,
which contradicts (15.69). Hence we must have σ(E, Z) 6= 0.
From Corollary 15.2 we know that σ(E, Z) is perpendicular to JH⊥ .

Therefore we obtain dim M = 3. Hence we have
dim M = 3 and σ(E, Z) 6= 0. (15.70)
By applying (15.68), (15.70) and Corollary 15.1, we may put
σ(JE, Z) = aJσ(E, Z) (15.71)
for some a ∈ R. Using (15.57), (15.61) and (15.71) we get
(a + 1) hσ(E, E), Jσ(E, Z)i = (a + 1) hσ(E, E), σ(E, Z)i = 0. (15.72)
Case (a): a 6= −1. From (15.62), (15.72) and Corollary 15.1, we get
σ(X, Y ) = 0, X, Y ∈ H. (15.73)
From (15.25), (15.57), (15.62), (15.73), the formulas of Gauss and Wein-
garten and the fact that M is the Riemannian product of N T and N ⊥ , we
obtain

¯ Z σ)(X, Y ), JZ = 0.
(∇ (15.74)
Thus, from Corollary 15.2, taking account of (15.25), (15.74) and equations
of Gauss and Codazzi, we get
1
hJNJ (X, Z), σ(Y, Z)i + hJσ(X, Z), σ(Y, Z)i = 0.
4
Hence, after using (15.24) we find
1
hNJ (JX, Z), σ(Y, Z)i + hJσ(X, Z), σ(Y, Z)i = 0. (15.75)
4
Putting Y = JX in (15.75) gives
1
hNJ (JX, Z), σ(JX, Z)i + hJσ(X, Z), σ(JX, Z)i = 0. (15.76)
4
On the other hand, it follows from (15.65) that
1
hJσ(X, Z), σ(JX, Z)i = − hσ(X, Z), NJ (X, Z)i
4
+ hσ(X, Z), σ(X, Z)i ,
which implies
1
hJσ(X, Z), Jσ(X, Z)i = − hσ(JX, Z), NJ (JX, Z)i
4 (15.77)
+ hσ(X, Z), σ(X, Z)i .
2 hσ(JX, Z), Jσ(X, Z)i = hσ(JX, Z), σ(JX, Z)i . (15.78)
Thus we also have

2 hσ(JX, Z), Jσ(X, Z)i = hσ(X, Z), σ(X, Z)i . (15.79)
From (15.71), (15.78) and (15.79) we get
(a2 − 1) hσ(E, Z), σ(E, Z)i = 0,
(2a − 1) hσ(E, Z), σ(E, Z)i = 0.
But this is impossible according to (15.70).
Case (b): a = −1. From (15.57), (15.59) and (15.71), we get
hσ(E, E), σ(Z, Z)i − hσ(E, Z), σ(E, Z)i = −1,
hσ(E, E), σ(Z, Z)i + hσ(X, Z), σ(JX, Z)i = 1,
for unit vectors X ∈ H and Z ∈ H⊥ . Thus we obtain
hσ(E, E), σ(Z, Z)i = 0, hσ(E, Z), σ(E, Z)i = 1. (15.80)
Now, from (15.67), (15.68), (15.71), (15.80), we get
hσ(X, Y ), σ(Z, Z)i = 0, hσ(X, Z), σ(X, Z)i = 1 (15.81)
for unit vectors X, Y ∈ H and Z ∈ H⊥ .
From (15.62), taking account of (15.24), (15.57), (15.81), formulas of
Gauss and Weingarten and the fact that M is a Riemannian product of
N T and N ⊥ , we get the same equation as (15.74). Hence, by the similar
arguments to the first step in Case (a), we may also deduce a contradiction.
15.9 A special class of warped product submanifolds in

nearly Kähler S 6
Let S 5 be a totally geodesic hypersurface of S 6 . Denote by N the unit

vector orthogonal to the hyperplane containing S 5 . We parametrize the
1
half unit circle S+ = {x ∈ E2 : |x| = 1 and x1 > 0} by
1
− π2 , π2 → S+ ; t 7→ (cos t, sin t).
Consider the following warped product representation of S 6 given by
1
φ : S+ ×cos t S 5 → S 6 ; (t, p) 7→ (sin t)N + (cos t)p. (15.82)
5 5
Let ψ : M → S be an immersion of a surface into S . Consider the
associated warped product immersion
1
ϕ : − π2 , π2 ×cos t M → S+ ×cos t S 5 ⊂ S 6 (15.83)
defined by
ϕ(t, p) = (sin t)N + cos(t)ψ(p).
The next proposition from [Chen et al. (1995a)] shows that the warped
product immersion ψ satisfies some nice properties.
Proposition 15.12. Let ψ : M → S 6 be a minimal non-totally geodesic

totally real surface in the nearly Khler S 6 whose ellipse of curvature is a
circle. Then M is contained in a unique totally geodesic S 5 and the warped
product immersion ψ defined by (15.83) is Lagrangian.
Proof. Let X be a vector field tangent to M . Under the hypothesis of

the proposition, we have

∂ ∂
ψ∗ = cos(t)N − sin(t)ψ(p), Jψ∗ = N × ψ(p),
∂t ∂t
ψ∗(X) = cos(t)ψ∗ (X),
Jψ∗(X) = cos(t) sin(t)N × ψ∗ (X) + sin2 (t)Jψ∗ (X).
From these it is easy to check that ψ is Lagrangian if and only if
hN × ψ∗ (X), ψ∗ (Y )i = 0 (15.84)
hN × ψ(p), ψ∗ (X)i = 0 (15.85)
hJψ∗ (X), ψ∗ (Y )i = 0, (15.86)
hold for all tangent vector fields X and Y to M . (15.86) simply says that ψ
has to be totally real; from (15.10), we obtain that (15.84) is equivalent to
hG(ψ∗ (X), ψ∗ (Y )), N i = 0; (15.85) is equivalent to hJψ∗ (X), N i = 0. We
have reduced the condition that ψ is Lagrangian to conditions depending
only on ψ. From now on, for simplicity, we identify M with ψ(M ), so we
do not write ψ∗ if there is no confusion.
Differentiating (15.85) gives us
hN × Y, Xi + hN × p, σ(X, Y )i = 0. (15.87)
Since the first term in (15.87) is skew symmetric and the second is sym-
metric, both terms have to be zero. Therefore (15.85) implies (15.84). Now
take any orthonormal basis {e1 , e2 } of Tp M . From the properties of G, we
obtain that G(e1 , e2 ) is orthogonal to e1 , e2 , Je1 , Je2 and p; (15.84) implies
that G(e1 , e2 ) is also orthogonal to N .
On the other hand, (15.85) implies that N ×p is orthogonal to e1 and e2 ;
clearly N × p is orthogonal to p and N , and a straightforward calculation
shows that N × p is orthogonal to Je1 and Je2 . Thus
G(e1 , e2 ) = ±p × N.
After changing the sign of e1 , we can make sure that e1 × e2 = JN if

necessary. This implies that
e1 × N = −Je2 , e2 × N = Je1 .
Note that the normal space of M in S 5 at p is spanned by Je1 , Je2 and
JN .
Differentiating (15.84), we obtain
hN × σ(Y, Z), Xi + hN × Y, σ(X, Z)i = 0, (15.88)
for all tangent vector fields X, Y and Z to M .
Putting X = e1 and Y = e2 , we obtain that
0 = hJe2 , σ(e2 , Z)i + hJe1 , σ(e1 , Z)i
= hσ(e2 , e2 ), JZi + hσ(e1 , e1 ), JZi ,
such that the mean curvature vector H of M in S 5 is orthogonal to Je1
and Je2 ; from (15.87) we obtain that H is orthogonal to JN . Therefore H
can only be zero. Then (15.87) immediately implies that σ is of the form
σ(e1 , e1 ) = αJe1 + βJe2 ,
σ(e1 , e2 ) = βJe1 − αJe2 ,
σ(e2 , e2 ) = −αJe1 − βJe2
such that the ellipse of curvature at p is a circle (possibly a point).
Corollary 15.1 states that a Lagrangian submanifold of the nearly Kähler
S 6 is δ(2)-ideal if and only if it satisfies δ(2) = 2.
The next theorem from [Chen et al. (1995a)] shows that the warped
product Lagrangian immersion ϕ defined by (15.83) is always δ(2)-ideal.
Theorem 15.14. Let ψ : M → S 6 be a minimal (non-totally geodesic)
totally real immersion in the nearly Kähler S 6 whose ellipse of curvature is
a circle. Then M is linearly full in a totally geodesic S 5 .
Moreover, let N be a unit vector perpendicular to this S 5 . Then

ϕ : − π2 , π2 ×cos t M → S 6 , (t, p) 7→ (sin t)N + (cos t)ψ(p)
is a δ(2)-ideal Lagrangian warped product immersion.
Conversely, every non-totally geodesic δ(2)-ideal Lagrangian immersion
of M into S 6 satisfying
(1) the dimension of the relative nullity subspace Nx of ϕ is independent
of the choice of x on M 3 =: − π2 , π2 ×cos t M ,
(2) the orthogonal distribution N ⊥ of N is an integrable distribution,
can be obtained locally as a warped product immersion in this way.
Chapter 16
Warped Products in Para-Kähler

Manifolds
Para-complex numbers were introduced as a generalization of complex num-

bers in [Graves (1845)]. J. T. Graves applied para-complex numbers to solve
some questions in number theory.
Para-Kähler geometry is the geometry related to the algebra of para-
complex numbers. Properties of para-Kähler manifolds were first studied in
[Rashevsky (1948)], in which P. K. Rashevsky considered a neutral metric
of signature (n, n) defined from a potential function on a locally product 2n-
manifold. He called such manifolds stratified space. Para-Kähler manifolds
were explicitly defined in [Rozenfeld (1949)]. B. A. Rozenfeld compared
Rashevsky’s definition with Kähler’s definition in the complex case and
established the analogy between Kähler and para-Kähler manifolds
16.1 Para-Kähler manifolds
First we give the following.

Definition 16.1. An almost para-Hermitian manifold M̃ is a manifold en-
dowed with an almost product structure P =
6 ±I and a pseudo-Riemannian
metric g such that
P 2 = I, g(Pv, Pw) = −g(v, w) (16.1)
for vectors v, w ∈ Tx M̃ , x ∈ M̃ , where I is the identity map.
The fundamental 2-form Ω of an almost para-Hermitian manifold M̃ is
defined by
Ω(X, Y ) = g(X, PY ), X, Y ∈ T M̃ .
Definition 16.1 implies that the dimension of an almost para-Hermitian
manifold is always even and the metric is neutral.
383
Definition 16.2. An almost para-Hermitian manifold (M̃ , P, g) is called

para-Kähler if it satisfies
˜ =0
∇P (16.2)
˜ is the Levi-Civita connection of M̃ .
identically, where ∇
Definition 16.3. Two para-Kähler manifolds (M̃ , P, g) and (M̃ ′ , P ′ , g ′ )

are called P-isometric if there exists an isometric ψ : M̃ → M̃ ′ such that
ψ∗ ◦ P = P ′ ◦ ψ∗ .
The simplest example of para-Kähler manifold is the para-Kähler n-

n
plane (E2n
n , P, g0 ), simply denoted by PK , which consists of the pseudo-
Euclidean 2n-space E2n n equipped with the standard flat neutral metric
Xn Xn
g0 = − dx2j + dyj2 ,
j=1 j=1
and the almost product structure
Xn ∂ Xn ∂
P= ⊗ dxj + ⊗ dyj .
j=1 ∂yj j=1 ∂xj
There exist many other para-Kähler manifolds; e.g., a homogeneous

manifold M = G/H of a semisimple Lie group G admits an invariant para-
Kähler structure (g, P) if and only if it is a covering of the adjoint orbit
AdG h of a semisimple element h (cf. [Hou et al. (1997)]).
g
Lemma 16.1. The Riemann curvature tensor R̃ and the Ricci tensor Ric
of a para-Kähler manifold satisfy
R̃(X, Y ) ◦ P = P ◦ R̃(X, Y ); (16.3)
R̃(PX, PY ) = −R̃(X, Y ), R̃(X, PY ) = −R̃(PX, Y ); (16.4)
g
Ric(PX, g
PY ) = −Ric(X, Y ). (16.5)
Proof. The proof of (16.3) is the same as that of (16.3). (16.4) follows
from (1.10), (16.1), (16.3) and P 2 = I. (16.5) follows from (16.4).
From (16.1) we find
g(Pv, w) + g(v, Pw) = 0, v, w ∈ Tp M̃ , p ∈ M̃ . (16.6)
Thus g(v, Pv) = 0. When {v, Pv} determines a non-degenerate plane, the
sectional curvature H P (v) = K(v ∧ Pv) is called the para-holomorphic
sectional curvature of the P-plane section spanned by v.
Definition 16.4. A para-Kähler space form is a para-Kähler manifold of

constant para-holomorphic sectional curvature.
Warped Products in Para-Kähler Manifolds 385
Proposition 16.1. Let M̃ be a para-Kähler manifold of constant para-

holomorphic sectional curvature c. Then the Riemann curvature tensor of
M̃ satisfies
c
R̃(X, Y ; Z, W ) = {g(X, W )g(Y, Z) − g(X, Z)g(Y, W )
4
+ g(PX, W )g(Y, PZ) − g(X, PZ)g(PY, W ) (16.7)
− 2g(X, PY )g(PZ, W )}.
Proof. This can be done in the same way as Proposition 6.2.
16.2 Non-flat para-Kähler space forms
The model of a nonflat para-Kähler space form was constructed in [Gadea

et al. (1989)] as follows: Let B be R2 with the product
(a, b)(a′ , b′ ) = (aa′ , bb′ ),
then B is a commutative algebra.
If we define the conjugate w̄ of an element w = (a, b) ∈ B by w̄ = (b, a),
then an element w ∈ B is called real if w = w̄, and is invertible if ww̄ 6= 0.
Let us put
B+ = {(a, b) ∈ B : a > 0, b > 0}.
Then B+ is a Lie group. Let
Xr
B0r+1 = {z = (zj ) ∈ B r+1 : hz, z̄i > 0}, hz, z̄i = zj z̄j .
j=0
Denote by gl(B; r + 1) the algebra of (r + 1)× (r + 1)-matrices with elements
in B. Then
gl(B, r + 1) = gl(R; r + 1) × gl(R; r + 1).
We have the Lie group

U (B; r + 1) = {Z ∈ gl(B; r + 1) : Zz, Z̄ z̄ = hz, z̄i ∀z ∈ B r+1 }.
Let P r (B) be the quotient of B0r+1 under the equivalence given by
(zj ) = (qzj ), q ∈ B+ . Then if π : B0r+1 → P r (B) is the natural projection,
we can identify π(z) with the unique element w = qz such that
hw, w̄i = 1, hw, wi = hw̄, w̄i ,
where q = (a, b) ∈ B+ . Indeed, if z = (zj ) = ((uj , vj )), we have
hw, w̄i = (ab hu, vi, ab hu, vi), hw, wi = (a2 hu, ui, b2 hv, vi),
hw̄, w̄i = (b2 hv, vi, a2 hu, ui)
with
1 1
hv, vi 4 hu, ui 4
a= 1 1 , b = 1 1 .
hu, ui 4 hu, vi 2 hv, vi 4 hu, vi 2
Thus
P r (B) ∼
= {(u, v) ∈ Rr+1 × Rr+1 : hu, ui = hv, vi , hu, vi = 1}. (16.8)
Since Z(qz) = qZ(z) for all Z ∈ U (B, r + 1), z ∈ B0r+1 ,
q ∈ B+ , the action
of U (B; r + 1) pass to the quotient P r (B).
For a nonzero real number c, consider on B0r+1 the tensor field
( r r
2 X X
g̃c = duj ⊗ dvj + dvj ⊗ duj
c hu, vi
j=0 j=0
r
)
1 X
− ui vj (dvi ⊗ duj + duj ⊗ dvi ) .
hu, vi
i,j=0
Then g̃c is invariant by U (B, r + 1). If ι : P r (B) → B0r+1 is the inclusion,

we have (ι ◦ π)∗ g̃c = g̃c . Hence the tensor field gc = ι∗ g̃c is a pseudo-
Riemannian metric on P r (B), which is also invariant by U (B; r + 1).
For P r (B) we have the coordinate charts (Uj± , ϕj ), where
Uj+ = {(u, v) ∈ P r (B) : uj , vj > 0},
Uj− = {(u, v) ∈ P r (B) : uj , vj < 0},
ĉj
u0 u ur v0 vj
c vr
ϕj (u, v) = ,..., ,..., ; ,..., ,..., ,
uj uj uj vj vj vj
and b· denotes the missing term.
If we call (xk , yk ) to the coordinates of any one of these charts, say
xk = uk /u0 , yk = vk /v0 , then
X X
2
gc = dxj ⊗ dyj + dyj ⊗ dxj
c(1 + hx, yi) j j
X (16.9)
1
− xi yj (dyi ⊗ dxj + dxj ⊗ dyi ) .
1 + hx, yi i,j
Also, we have the almost-product structure on B0r+1 given by

X ∂ X ∂
P̂ = ⊗ duj − ⊗ dvj ,
j
∂uj
j
∂vj
and it defines an almost-product structure P on P r (B) by the relation

π∗ ◦ P̂ = P ◦ π∗ . Thus in the same chart P is given by
X ∂ X ∂
P= ⊗ dxj − ⊗ dyj . (16.10)
j
∂x j
j
∂yj
The P r (B), together with the metric (16.9) and the almost product struc-
ture P given by (16.10), is a para-Kähler manifold of constant para-
holomorphic sectional curvature c 6= 0.
Moreover, if r > 1, then P r (B) is complete, connected, simply connected
para-Kähler space form. To obtain a complete simply-connected model of
para-Kähler space form for r = 1, it is enough to extend the above structure
on P 1 (B) = S 1 ×R to its universal covering, which is R2 . We simply denote
the triple (P r (B), P, gc ) by Pcr (B).
Remark 16.1. Two complete simply-connected para-Kähler space forms

of equal para-holomorphic sectional curvature are P-isometric [Gadea et al.
(1989)].
16.3 Invariant submanifolds of para-Kähler manifolds
Let M be a (non-degenerate) submanifold of a para-Kähler manifold M̃ .

For any vector field X tangent to M , we put
PX = P X + F X, (16.11)
where P X and F X are the tangential and the normal components of PX
with respect to the following orthogonal decomposition:
Tx M̃ = Tx M ⊕ Tx⊥ M, x ∈ M.
Then P is an endomorphism of the tangent bundle T M and F is a normal
bundle valued 1-form on M .
Similarly, for a normal vector field ξ, we put
Pξ = tξ + f ξ, (16.12)
where tξ and f ξ are the tangential and the normal part of Pξ, respectively.
Definition 16.5. A pseudo-Riemannian submanifold M of a para-Kähler

manifold (M̃ , P, g) is called an invariant submanifold if each of its tangent
spaces is invariant under the action of P, i.e., we have P(Tx M ) = Tx M for
any x ∈ M .
Example 16.1. For two given integers r, p ≥ 1, we consider the map

φr,p : Pcr (B) × Pcp (B) → Rr+p+rp+1 × Rr+p+rp+1
defined by
Pcr (B) × Pcp (B) ∋ (((uj , vj )), ((ûα , v̂α ))) 7→ ((uj ûα , vj v̂α )).
It is obvious that conditions

g(u, u) = g(v, v), g(û, û) = g(v̂, v̂), g(u, v) = g(û, v̂) = 1
imply
X p
r X p
r X
X p
r X
X
u2j û2α = vj2 v̂α2 , uj vj ûα v̂α = 1.
j=0 α=0 j=0 α=0 j=0 α=0
Therefore, according to (16.8), φr,p gives rise to a map
ϕr,p : Pcr (B) × Pcp (B) → Pcr+h+rp (B). (16.13)
It is direct to verify that ϕr,p gives rise to an invariant submanifolds of
Pcr+h+rp (B).
Lemma 16.2. An invariant submanifold M of a para-Kähler manifold M̃
is para-Kählerian with respect to its induced structures. Further, its second
fundamental form σ satisfies
σ(PX, Y ) = σ(X, PY ) = Pσ(X, Y ), X, Y ∈ T M. (16.14)
Proof. If M is an invariant submanifold of a para-Kähler manifold M̃ ,

then (16.1) implies that M is an almost para-Hermitian manifold with re-
spect to the induced metric g ′ and the induced para-structure, also denoted
by P. For any vector fields X, Y tangent to M we have
˜ X (PY ) = ∇X (PY ) + σ(X, PY )
∇ (16.15)
˜
where ∇ and ∇ denote the Levi-Civita connections of M̃ and M , respec-
tively.
˜ = 0. Thus we get
On the other hand, since M is para-Kählerian, ∇P
˜ X (PY ) = P∇X Y + Pσ(X, Y ).
∇ (16.16)
Comparing (16.15) and (16.16) gives
∇P = 0, σ(X, PY ) = Pσ(X, Y ).
Thus M is para-Kählerian; and by symmetry of σ we also have (16.14).
Proposition 16.2. Every invariant submanifold of a para-Kähler manifold
is austere; in particular it is minimal.
Proof. If M is an invariant submanifold of a para-Kähler manifold M̃ ,
then, by (16.1) and Lemma 16.2, we may choose a local orthonormal frame
{e1 , . . . , en , e1∗ , . . . , en∗ }
on M such that ei∗ = Pei . Hence we have
ǫi∗ = g(ei∗ , ei∗ ) = −g(ei , ei ) = −ǫi .
On the other hand, it follows from (16.14) that σ(ei∗ , ei∗ ) = σ(ei , ei ). Thus
M is an austere submanifold.
Proposition 16.3. An invariant submanifold M of a para-Kähler space

2m
form Mm (c) has constant para-holomorphic sectional curvature c if and
only if M is null-isotropic.
Proof. Let M be an invariant submanifold of a para-Kähler space form

2m
Mm (c). For any vector in the unit tangent bundle T 1 M , it follows from
Gauss’ equation, Proposition 16.1 and Proposition 16.2 that the para-
holomorphic sectional curvature H P (v) of v ∈ T 1 M satisfies
H P (v) = c − 2g(σ(v, v), σ(v, v)). (16.17)
Hence M has constant para-holomorphic sectional curvature c if and only
if M is a null-isotropic invariant submanifold.
Similarly, we also have the following.
Proposition 16.4. The scalar curvature of an invariant submanifold Mn2n

of a para-Kähler space form M̃m 2m
(c) is equal to 21 n(n + 1)c if and only if
Sσ = 0, where Sσ is defined by
Xn
Sσ = ǫi ǫj g(σ(ei , ej ), σ(ei , ej )), ǫi = g(ei , ei ). (16.18)
i,j=1
Remark 16.2. Some simple examples of non-totally geodesic null-isotropic

invariant submanifolds of para-Kähler manifolds are given by the isometric
2n+2
immersions of (E2nn , g0 , P) into (En+1 , g0 , P) defined by
n n
!
X X
2 2
ai (xi + yi ) , x1 , . . . , xn , ai (xi + yi ) , y1 , . . . , yn ,
j=1 j=1
where a1 , . . . , an are real numbers, not all zero. Such null-isotropic invariant
submanifolds satisfy Sσ = 0.
16.4 Lagrangian submanifolds of para-Kähler manifolds
Analogous to totally real submanifolds of an almost Hermitian manifold, a

pseudo-Riemannian submanifold M of an almost para-Hermitian manifold
is called anti-invariant if P(Tx M ) ⊂ Tx⊥ M for each x ∈ M . In particular,
M is called Lagrangian if P(Tx M ) = Tx⊥ M for each x ∈ M .
Lemma 16.3. Let M be a Lagrangian submanifold of a para-Kähler man-

ifold M̃n2n . Then
(a) P(∇X Y ) = DX (PY );
(b) APX Y = −P(σ(X, Y ));

(c) g(σ(X, Y ), PZ) = g(σ(Y, Z), PX) = g(σ(Z, X), PY );
(d) P(R(X, Y )Z) = RD (X, Y )PZ
for X, Y, Z tangent to M , where ∇ and R are the Levi-Civita connection
and curvature tensor of M .
Proof. Let X, Y, Z be vector fields tangent to M . Then it follows from

formulas of Gauss and Weingarten that
P(∇X Y ) + P(σ(X, Y )) = P(∇˜ XY )
˜ X (PY ) = DX (PY ) − APY X,
=∇
which implies statement (a) and statement (b).
From (2.14) and statement (b), we find
g(σ(Z, X), PY ) = −g(P(σ(X, Z)), Y ) = g(APX Z, Y ) = g(σ(Y, Z), PX).
Similarly, we also have g(σ(X, Y ), PZ) = g(σ(Y, Z), PX). Thus we get
statement (c). Statement (d) follows from statement (a) and (2.24).
The equations of Gauss and Codazzi of a Lagrangian submanifold M of
a para-Kähler space form Mn2n (4c) are given respectively by
R̃(X, Y ; Z, W ) = g(Aσ(Y,Z) X, W ) − g(Aσ(X,Z) Y, W ) (16.19)
+ c (g(X, W )g(Y, Z) − g(X, Z)g(Y, W )),
¯ ¯ Y σ)(X, Z)
(∇X σ)(Y, Z) = (∇ (16.20)
for X, Y, Z, W ∈ T M .
Lemma 16.3(b) implies that Ricci’s equation is a consequence of Gauss’
equation for such submanifolds.
If we put σ = P ◦ σ̂, then (16.1) and Lemma 16.3(c) yield
g(Aσ(Y,Z) X, W ) = g(σ(X, W ), σ(Y, Z))
= g(σ(X, W ), Pσ(Y, Z))
= g(σ(σ(Y, Z), X), PW )
= −g(σ(σ(Y, Z), X), W ).
Therefore, equation (16.14) can be rephrased as
R(X, Y )Z = σ(σ(X, Z), Y ) − σ(σ(Y, Z), X)
+ cg(Y, Z)X − cg(X, Z)Y.
Consequently, after applying a result of [Eschenburg and Tribuzy (1993)],
we have the following existence and uniqueness theorems for Lagrangian
submanifolds in para-Kähler space forms.
Theorem 16.1. Let M be a simply-connected Riemannian n-manifold. If

σ̂ is a T M -valued symmetric bilinear form on M such that
(a) g(σ̂(X, Y ), Z) is totally symmetric;
¯
(b) (∇σ̂)(X, Y, Z) is totally symmetric;
(c) R(X, Y )Z = cg(Y, Z)X − cg(X, Z)Y + σ̂(σ(X, Z), Y )
− σ̂(σ̂(Y, Z), X),
then there exists a Lagrangian immersion L of M into a complete simply-
connected para-Kähler space form Mn2n (4c) whose second fundamental form
is given by σ = P ◦ σ̂.
Theorem 16.2. Let L1 , L2 : M → Mn2n (4c) be two Lagrangian immersions

of a Riemannian n-manifold M into a complete simply-connected para-
Kähler space form Mn2n (4c) with second fundamental forms σ 1 and σ 2 ,
respectively. If
g(σ 1 (X, Y ), PL1⋆ Z) = g(σ 2 (X, Y ), PL2⋆ Z)
holds for vectors X, Y, Z tangent to M , then there exists a P-isometry Φ of
Mn2n (4c) such that L1 = Φ ◦ L2 .
Proposition 16.5. The only totally umbilical Lagrangian submanifold M

of a para-Kähler space form Mn2n (4c) with n ≥ 2 is the totally geodesic one.
Proof. Let M be a totally umbilical Lagrangian submanifold of Mn2n (4c)

with n ≥ 2. Then its second fundamental form satisfies
σ(X, Y ) = g(X, Y )H, X, Y ∈ T M. (16.21)
If H = 0, then M is totally geodesic. Thus we may assume that H 6= 0.
It follows from (16.21) that
¯ X σ)(Y, Z) = g(Y, Z)DX H.
(∇
After applying Codazzi’s equation, we find
g(Y, Z)DX H = g(X, Z)DY H, X, Y, Z ∈ X(M ).
For any X ∈ T M , by choosing 0 6= Y = Z ⊥ X we find DH = 0. Hence
g(H, H) is constant. Therefore Gauss’ equation and p (16.21) imply that M
is of constant curvature c − ||H||2 , where ||H|| = −g(H, H).
Let us put Z = PH. Then Lemma 16.6(i) yields ∇Z = 0. Thus Z is a
nonzero parallel vector field on M , which implies that N is flat. Therefore
we get c = ||H||2 > 0.
From totally umbilicity of M , we also have [AH , Aξ ] = 0 for any normal

vector ξ. Hence, after applying DH = 0 we derive from Ricci’s equation
that
g(R(Z, Y )H, PY ) = 0, Y, Z ∈ T M.
On the other hand, by (16.1) and Proposition 16.1 we find
g(R(Z, Y )H, PY ) = c{g(Z, PH)g(Y, Y ) − g(Y, PH)g(Y, Z)}.
Hence, after choosing Y, Z in such way that Z = PH and g(Y, Z) = 0, we
find g(H, H) = 0, which is a contradiction.
16.5 P R-submanifolds in para-Kähler manifolds
Definition 16.6. A pseudo-Riemannian submanifold M of a para-Kähler

manifold M̃ is called a P R-submanifold if the tangent bundle T M of M is
the direct sum of an invariant distribution D and an anti-invariant distri-
bution D⊥ , i.e.,
T M = D ⊕ D⊥ , PD = D, PD⊥ ⊆ Tp⊥ M.
For a P R-submanifold M , let ν denote the orthogonal complement of

PD⊥ in T ⊥ M . Then we have
T ⊥ M = PD⊥ ⊕ ν, Pν = ν. (16.22)
The following proposition characterizes P R-submanifold of para-Kähler
manifolds.
Proposition 16.6. Let M → M̃ be an isometric immersion of a pseudo-

Riemannian manifold M into a para-Kähler manifold M̃ . Then a necessary
and sufficient condition for M to be a P R-submanifold is that F ◦ P = 0,
where F and P are defined by (16.11).
Proof. For any vector U tangent to M , we have the following decompo-

sition
U = P 2 U = P 2 U + F P U + tF U + f F U.
So, by identifying the tangent and the normal components of this equation
respectively, we find
P 2 + tF = I and; F P + f F = 0.
Suppose that M is a P R-submanifold. Then after choosing U = X ∈ D,

we have PX = P X and F X = 0. Hence P 2 = I and F P = 0 on D.
On the other hand, if we choose U = Z ∈ D⊥ , then we find P Z = 0.
Hence we derive F P = 0 on D⊥ .
Conversely, let us assume F P = 0. If we put
D = {X ∈ T M : PX ∈ T M },
D⊥ = {Z ∈ T M : PZ ∈ T ⊥ (M )},
then by direct computations we conclude that D and D⊥ are orthogonal
such that T M = D ⊕ D⊥ . This proves the proposition.
Lemma 16.4. If M is a P R-submanifold of a para-Kähler manifold M̃ ,

then we have:
(1) g(APZ U, PX) = g(∇U Z, X),
(2) APZ W = APW Z and APξ X = −Aξ PX,
˜ U P)V = tσ(U, V ) + AF V U ,
(3) (∇
for all X, Y ∈ D, Z, W ∈ D⊥ , U, V ∈ T M and ξ ∈ ν.
Proof. Consider the following identity

˜ U PZ
P∇U Z + Pσ(U, Z) = ∇
(16.23)
= −APZ U + DU PZ.
Hence, after taking scalar product of (16.23) with PX ∈ D, we obtain
statement (1). Similarly, by taking the scalar product of (16.23) with W ∈
D⊥ , we obtain the first equation of statement (2).
The second equation of statement (2) follows from the following fact:
g(Aξ PX, U ) = g(σ(PX, U ), ξ) = g(∇˜ U PX, ξ)
˜ U X, ξ) = −g(∇
= g(P ∇ ˜ U X, Pξ)
= −g(σ(X, U ), Pξ) = −g(APξ X, U ).
Finally, statement (3) follows easily from (2.5), (2.13) and (16.11).
For P R-submanifolds of a para-Kähler manifold, we have the following
result from [Chen (2011b)].
Theorem 16.3. Let M be a P R-submanifold of a para-Kähler manifold.

Then we have:
(1) The anti-invariant distribution D⊥ is a non-degenerate integrable dis-
tribution;
(2) The invariant distribution D is a non-degenerate minimal distribution;

(3) The distribution D is integrable if and only if the second fundamental
form σ̊ of D satisfies σ̊(P X, Y ) = σ̊(X, P Y );
(4) The distribution D is integrable if and only if we have
g(σ(P X, Y ), F Z) = f (σ(X, P Y ), F Z); (16.24)
(5) The distribution D is integrable and its leaves are totally geodesic in M
if and only if g(σ(X, Y ), F Z) = 0;
(6) the leaves of the anti-invariant distribution D⊥ are totally geodesic in
M if and only if g(σ(X, Z), PW ) = 0;
for all X, Y ∈ D and Z, W ∈ D⊥ .
Proof. Let M be a P R-submanifold of a para-Kähler manifold M̃ . If a

vector v ∈ Dx⊥ , x ∈ M, is orthogonal to every vector in Dx⊥ , then it must be
orthogonal to every vector in Tx M . Therefore we get v = 0, hence D⊥ is
non-degenerate. Consequently, its orthogonal distribution D must be also
non-degenerate.
Now, we want to prove that D⊥ is always integrable. From Lemma 16.4,
we have
P[Z, W ] = P(∇Z W − ∇W Z) = DZ PW − DW PZ (16.25)
for Z, W ∈ D⊥ .
On the other hand, for any ξ ∈ ν and Z, W ∈ D⊥ , we have
˜ Z Pξ, W ) = g(DZ ξ, PW ) = −g(ξ, DZ PW ).
g(APξ Z, W ) = −g(∇
Thus
g(ξ, DW PZ − DZ PW ) = g(APξ Z, W ) − g(APξ W, Z) = 0.
Since this holds for any ξ ∈ ν, we obtain DW PZ − DZ PW ∈ PD⊥ for
Z, W ∈ D⊥ . By combining this with (16.25), we get P[D⊥ , D⊥ ] ⊂ PD⊥ ,
which implies [D⊥ , D⊥ ] ⊂ D⊥ . Therefore D⊥ is a integrable distribution.
Thus we have statement (1).
For each vector field X ∈ D and Z ∈ D⊥ we have
g(Z, ∇X X) = −g(PZ, ∇˜ X PX)
˜ X PZ, PX)
= g(∇
= −g(APZ X, PX).
Replacing X by PX yields
g(Z, ∇PX PX) = −g(APZ PX, X) = −g(APZ X, PX).
Combining both equations gives g(Z, ∇X X + ∇PX PX) = 0. Consequently,

the mean curvature vector H̊ of D⊥ vanishes. Therefore the anti-invariant
distribution D⊥ is a minimal distribution. This gives statement (2).
For vector fields X, Y ∈ D and Z ∈ D⊥ , we get g([X, Y ], PZ) = 0. Thus
g(σ̊(X, PY ), Z) = g(∇X PY, Z) = g(∇˜ X PY, Z)
= −g(∇ ˜ X Y, PZ) = −g(∇
˜ Y X, PZ) = g(∇Y PX, Z).
Combining this with g(h̊(PX, Y ), Z) = g(∇PX Y, Z) yields
g(σ̊(JX, Y ) − σ̊(X, PY ), Z) = g( [X, PY ], Z),
which implies statement (3).
Next, let us assume that D is an integrable distribution. Let N be an
integrable submanifold, σ̊ the second fundamental form of N in M̃ and
σ̃ the second fundamental form of N in M̃ . Since D is invariant, N is a
invariant submanifold of M̃ and hence
σ̃(X, PY ) = σ̃(PX, Y ) = P σ̃(X, Y )
according to Lemma 16.2. Now, we have σ̃ = σ + σ̊ and hence
σ(X, PY ) − σ(PX, Y ) = σ̊(PX, Y ) − σ̊(X, PY ). (16.26)
Therefore, after taking the inner product of (16.26) with F Z, we obtain
(16.24).
Conversely, let us assume that (16.24) holds for all X, Y ∈ D and Z ∈
D⊥ . Since P is parallel with respect to ∇,˜ we have
0 = g(σ(X, PY ), F Z) − g(σ(PX, Y ), F Z)
= g(P ∇˜ X Y − ∇X P Y − P ∇ ˜ Y X + ∇Y P X, F Z)
= g(P[X, Y ], F Z).
Therefore P applied to the tangent vector field [X, Y ] is tangent to M and
hence [X, Y ] belongs to the distribution D. This proves statement (4).
Suppose that D is an integrable distribution and its leaves are totally
geodesic in M . Then for vector fields X, Y ∈ D and Z ∈ D⊥ , we have
∇X Y ∈ D and hence
g(σ(X, Y ), F Z) = g(σ(X, Y ), PZ)
= −g(P ∇ ˜ X Y, Z) + g(P∇X Y, Z) (16.27)
= −g(∇˜ X PY, Z) = −g(∇X PY, Z) = 0.
Conversely, the integrability of the holomorphic distribution D follows from
g(σ(X, Y ), F Z) = 0 and statement (4). By a similar computation as in
(16.27), we obtain ∇X Y ∈ D for X, Y ∈ D. Therefore, each leaf of D is
totally geodesic in M . Consequently, we have statement (5).
Statement (6) follows easily from statement (1) and Lemma 16.4(1).
16.6 P R-warped products and P -products in para-Kähler

manifolds
Definition 16.7. A P R-submanifold of a para-Kähler manifold M̃ is called

a P R-warped product if it is a warped product N T ×f N ⊥ of an invariant
submanifold N T and an anti-invariant submanifold N ⊥ of M̃ . And a P R-
submanifold of a para-Kähler manifold is called a P R-product if it is locally
a direct product N T × N ⊥ of an invariant submanifold N T and an anti-
invariant submanifold N ⊥ .
Proposition 16.7. Let M be a P R-product of a para-Kähler manifold M̃ .
Then we have
2g(σ(X, Z), σ(X, Z)) = ǫX ǫZ (K̃(X, Z) + K̃(X, PZ)), (16.28)
⊥
for any unit vector fields X ∈ D and Z ∈ D , where K̃ is the sectional
curvature of M̃ , ǫX = g(X, X) and ǫZ = g(Z, Z).
Proof. Let M be a P R-product of a para-Kähler manifold M̃ . We have
∇X Z = ∇Z X = 0
for any unit vector fields X ∈ D and Z ∈ D⊥ . Thus we find from Lemma
16.4(1) that
σ(D, D⊥ ) ⊥ PD⊥ . (16.29)
Therefore, by applying Codazzi’s equation, we get
R̃(X, PX; Z, PZ) = g(DX σ(PX, Z), PZ) − g(DPX σ(X, Z), PZ). (16.30)
Since M is a P R-product, Lemma 16.4(1) gives APZ X = 0. Hence we may
derive from (16.29) and (16.30) that
R̃(X, PX; Z, PZ) = g(σ(X, Z), DPX PZ) − g(σ(PX, Z), DX PZ)
= g(σ(X, Z), P∇PX Z) − g(σ(PX, Z), P∇X Z)
= −2g(σ(PX, Z), Pσ(X, Z))
(16.31)
= −2g(APσ(X,Z) PX, Z)
= 2g(Aσ(X,Z) X, Z)
= 2g(σ(X, Z), σ(X, Z)),
where we have applied Lemma 16.4(2).
On the other hand, the first Bianchi identity and Lemma 16.1 imply
R̃(X, PX; Z, PZ) = −R̃(Z, X; PX, PZ) − R̃(PX, Z; X, PZ)
= ǫX ǫZ (K̃(X, Z) + K̃(X, PZ)).
Combining this with (16.31) gives (16.28).
Corollary 16.1. Let M be a P R-product in a para-Kähler space form

Mn2n (4c). Then we have
g(σ(X, Z), σ(X, Z)) = cg(X, X)g(X, Z) (16.32)
for any nonzero vector fields X ∈ D and Z ∈ D⊥ . In particular, if c 6= 0
and X, Z are non-null, then σ(X, Z) 6= 0.
Proof. It follows from Proposition 16.1 that

K̃(X, Z) = K̃(X, PZ) = c
for unit vectors X, Z satisfying g(X, Z) = g(X, PZ) = 0. Therefore, by
applying Proposition 16.7 we obtain the corollary.
Recall that for any submanifold M of a para-Kähler manifold M̃ , there is
a canonical endomorphism P of the tangent bundle T M defined by (16.11).
The next result characterizes P R-products.
Proposition 16.8. [Chen and Munteanu (2012)] A P R-submanifold M of

a para-Kähler manifold is a P R-product if and only if P is parallel, i.e.,
∇P = 0.
Proof. Let M be a P R-submanifold of a para-Kähler manifold M̃ . If P

is parallel, then Lemma 16.4(3) gives
tσ(U, V ) = −AF V U (16.33)
for U, V ∈ T M . In particular, if U = X ∈ D, then F X = 0. Hence (16.33)
implies tσ(U, X) = 0, thus
AF Z X = 0 (16.34)
⊥
for any Z ∈ D and X ∈ D. Thus, by Theorem 16.3, we see that D is
integrable and also the leaves of D⊥ are totally geodesic in M . Let N T be
a leaf of the invariant distribution D and N ⊥ be a leaf of the anti-invariant
distribution D⊥ . Then for any X, Y ∈ D and Z ∈ D⊥ , (16.34) and Lemma
16.4 gives
0 = g(AF Z Y, X) = g(PAF Z Y, PX)
(16.35)
= g(∇Y Z, PX) = −g(Z, ∇Y PX).
From this we conclude that N T is totally geodesic in M and M is a P R-
product in the para-Kähler manifold.
Conversely, if M is a P R-product, then we have ∇U Y ∈ D for any
Y ∈ D and U ∈ T M . Thus, by (2.5) and (2.13), we find
Pσ(U, Y ) = σ(U, PY ).
From this, together with (2.5) and

(∇˜ U P)V = ∇˜ U (PV ) − P ∇
˜ U V, (16.36)
we derive (∇U P )Y = 0 for any Y ∈ D and U ∈ T M .
Similarly, from ∇U Z ∈ D⊥ for Z ∈ D⊥ and U ∈ T M , we may also
prove that (∇U P )Z = 0. Consequently, we obtain ∇P = 0. This proves
the theorem.
Theorem 16.4. [Chen (2011b)] Let N T ×N ⊥ be a P R-product of the para-

Kähler (h + p)-plane P h+p with h = 12 dim N T and p = dim N ⊥ . If N ⊥ is
either spacelike or timelike, then the P R-product is an open part of a direct
product of a para-Kähler h-plane P h and a Lagrangian submanifold L of
P p , i.e.,
N T × N ⊥ ⊂ P h × L ⊂ P h × P p = P h+p . (16.37)
Proof. Let N T × N ⊥ be a PR-product in P h+p . Then it follows from

Corollary 16.1 that
g(σ(X, Z), σ(X, Z)) = 0
holds for X ∈ D and Z ∈ D⊥ . Thus if m = dim N ⊥ + 12 dim N T and if
N ⊥ is either spacelike or timelike, then we obtain σ(X, Z) = 0. Hence
the immersion is mixed totally geodesic. Therefore, according to Moore’s
Lemma, the immersion is a direct product immersion. Consequently, after
applying the assumption on the dimension, we conclude that N T is an open
portion of a para-Kähler r-plane with 2h = dim N T and the immersion is
given by (16.37).
16.7 P R-products in non-flat para-Kähler space forms
Example 16.2. Some important examples of PR-products can be obtained

via the following embedding introduced in Example 16.1:
ϕr,p : Pcr (B) × Pcp (B) −→ Pcr+p+rp (B).
If N T is the para-Kähler space form Pcr (B) and if N ⊥ is a Lagrangian
submanifold of Pcp (B), then
ϕr,p
N T × N ⊥ ⊂ Pcr (B) × Pcp (B) −→ Pcr+p+rp (B)
is a PR-product of Pcr+p+rp (B) such that N T is immersed as a totally
geodesic invariant submanifold in Pcr+p+rp (B). Clearly, the codimension of
such PR-products is p + 2rp.
The next result determines the smallest codimension of a PR-product

in a non-flat para-Kähler space form.
Theorem 16.5. [Chen (2011b)] Let M = N T × N ⊥ be a PR-product in a

2m
para-Kähler space form Mm (4c) with dimR N T = 2r and dimR N ⊥ = p.
If c 6= 0, then m ≥ r + p + rp.
Moreover, if m = r + p + rp, then N T is immersed in Mm 2m
(4c) as a
totally geodesic invariant submanifold.
Proof. Let M = N T × N ⊥ be a PR-product in a para-Kähler space

2m
form Mm (4c), c 6= 0. Let u1 , . . . , u2r be an orthonormal basis of D and
v1 , . . . , vp an orthonormal basis of D⊥ . Then (16.31) and the linearity of σ
imply
g(σ(ui , vα ), σ(uj , vα )) = 0 (16.38)
for 1 ≤ i 6= j ≤ 2r and α = 1, . . . , p.
If p = 1, then (16.29), (16.31) and (16.38) imply that the codimension
2m
of the PR-product in Mm (4c) is at least p + 2rp. Thus m ≥ r + p + rp.
If p ≥ 2, then (16.38) and the linearity of σ imply that
g(σ(ui , vα ), σ(uj , vβ )) + g(σ(ui , vβ ), σ(uj , vα )) = 0 (16.39)
for 1 ≤ i 6= j ≤ 2r and 1 ≤ α 6= β ≤ p.
On the other hand, Gauss’ equation and Proposition 16.1 imply that
the curvature tensor R and the second fundamental form σ of M satisfy
0 = R(ui , uj ; vβ , vα ) = g(σ(ui , vα ), σ(uj , vβ )) − g(σ(ui , vβ ), σ(uj , vα )).
g(σ(ui , vα ), σ(uj , vβ )) = 0 (16.40)
for 1 ≤ i 6= j ≤ 2r; 1 ≤ α 6= β ≤ p. Thus, by (16.29), (16.31) and (16.40),
we find m ≥ r + p + rp.
Now, assume that m = r + p + rp. Then σ(D, D⊥ ) is the orthogonal
complementary subbundle of PD in T ⊥ M . Thus Lemma 16.4(1) gives
σ(D, D) ⊥ D⊥ . (16.41)
Also, from Gauss’ equation and Proposition 16.1, we find
g(σ(X, Y ), σ(P X, W )) = g(σ(X, W ), σ(P X, Y )) (16.42)
⊥
for X, Y ∈ D and W ∈ D . Further, (16.29) and Lemma 16.4(1) give
g(σ(X, W ), σ(P X, Y )) = g(Aσ(X,W ) P X, Y )
= −g(APσ(X,W ) X, Y ) (16.43)
= −g(σ(X, Y ), Pσ(X, W )).
From (16.6) and (16.41) we also have

g(σ(X, Y ), Pσ(X, W )) = −g(Pσ(X, Y ), σ(X, W ))
= −g(∇W X, Pσ(X, Y ))
(16.44)
= g(∇W (P X), σ(X, Y ))
= g(σ(X, Y ), σ(P X, W )).
Therefore, by combining (16.42), (16.43) and (16.44), we obtain
g(σ(X, Y ), σ(P X, W )) = g(σ(X, W ), σ(P X, Y )) = 0. (16.45)
So, by applying linearity of σ, we obtain σ(D, D) ⊥ σ(D, D⊥ ).
Since σ(D, D⊥ ) is the orthogonal complementary normal subbundle of
PD. Thus, by applying (16.41), we find σ(X, Y ) = 0, X, Y ∈ D. After
combining this with the total geodesy of N T in N T × N ⊥ , we conclude
that N T is immersed as a totally geodesic submanifold of Mm2m
(4c).
An immediate application of Theorem 16.5 is the following.
Corollary 16.2. If the direct product Rr2r × Pp2p of two para-Kähler man-
ifolds can be isometrically immersed in a non-flat para-Kähler space form
2(r+p+rp)
Mr+p+rp (4c) as an invariant submanifold, then both Rr2r and Pp2p are of
constant para-holomorphic sectional curvature 4c.
Moreover, both Rr2r and Pp2p are immersed as totally geodesic invariant
2(r+p+rp)
submanifolds of Mr+p+rp (4c).
Proof. Under the hypothesis of the corollary, let us consider the PR-
2(r+p+rp)
submanifold Rr2r ×N ⊥ of Mr+p+rp (4c) such that N ⊥ is a Lagrangian sub-
manifold of Pp2p . Then, by Theorem 16.5, Rr2r is totally geodesic invariant
2(r+p+rp)
submanifold of Mr+p+rp (4c). Thus Rr2r is of constant para-holomorphic
sectional curvature 4c.
2(r+p+rp)
Similarly, Pp2p is a totally geodesic submanifold of Mr+p+rp (4c).
2p
Hence Pp is also of constant para-holomorphic sectional curvature 4c.
16.8 Warped product P R-submanifolds
First we give the following result from [Chen and Munteanu (2012)].
Proposition 16.9. If N ⊥ ×f N T is a warped product submanifold in a

para-Kähler manifold M̃ with warping function f such that N ⊥ is an anti-
invariant submanifold and N T an invariant submanifold of M̃ , then M is
a P R-product N ⊥ × NfT . Thus f is a constant function.
Proof. Consider X, Y ∈ D and Z ∈ D⊥ . Compute

˜ X Y, PZ)
g̃(σ(X, Y ), F Z) = g̃(∇
= −g̃(Y, P ∇˜ X Z)
= g(P Y, ∇X Z) (16.46)
= g(P Y, Z(ln f )X)
= Z(ln f )g(X, P Y ).
Since σ( , ) is symmetric and g( , P( )) is skew-symmetric, it follows that
Z(ln f ) vanishes for all Z ∈ T N ⊥ . Consequently, f is a constant and thus
the warped product is nothing but a P R-product.
Proposition 16.9 shows that there do not exist proper warped product
P R-submanifolds in para-Kähler manifolds of the form N ⊥ ×f N T .
Definition 16.8. A P R-submanifold of a para-Kähler manifold M̃ is called

a P R-warped product if it is a warped product of the form: N T ×f N ⊥ ,
where N T in an invariant submanifold, N ⊥ is an anti-invariant submanifold
of M̃ and f is a non-constant function f : N T → R+ .
Since the metric on NT of a P R-warped product N T ×f N ⊥ is neutral,
we simply called the P R-warped product N T ×f N ⊥ spacelike or timelike
depending on N ⊥ is space-like or time-like, respectively.
Analogous to Theorem 11.2, the next result characterizes P R-warped

products in para-Kähler manifolds.
Proposition 16.10. Let M be a proper P R-submanifold of a para-Kähler

manifold. Then M is a P R-warped product if and only if
AF Z X = (P X(µ)) Z, ∀ X ∈ D, Z ∈ D⊥ , (16.47)
for some smooth function µ on M satisfying W µ = 0 for W ∈ D⊥ .
Proof. Let M be a P R-warped product N T ×f N ⊥ in a para-Kähler

manifold M̃ . Then the invariant distribution D is integrable and its leaves
are totally geodesic in M . Thus it follows from Lemma 16.4, Theorem
16.3(5) and Proposition 3.1(2) that
AF Z X = −((P X) ln f )Z, X ∈ D, Z ∈ D⊥ .
Since f is a function on N T , we also have W (ln f ) = 0 for W ∈ D⊥ .
Conversely, assume that M is a proper P R-submanifold of a para-Kähler
manifold M̃ satisfying (16.47) for some function µ with W µ = 0, W ∈ D⊥ .
Then we have
g(σ(D, D), F D⊥ ) = 0, (16.48)
g(σ(P X, Z), F W ) = −(Xµ)g(Z, W ), (16.49)
⊥
for vectors X, Y ∈ D and Z, W ∈ D . Thus Theorem 16.3(5) and (16.48)
imply that D is integrable and its leaves are totally geodesic in M .
On the other hand, from Lemma 16.4 and (16.49) we obtain
g(∇Z X, W ) = −g(∇Z W, X)
= −g(P APW Z, X)
(16.50)
= g(σ(P X, Z), PW )
= −(Xµ)g(Z, W )
⊥
for X in D and Z, W in D .
Since the distribution D⊥ of M is always integrable, (16.50) together
with W µ = 0, W ∈ D⊥ , imply that leaves of D⊥ are extrinsic spheres in M ,
i.e., they are totally umbilical submanifolds with parallel mean curvature
vector. Hence, after applying a result of [Hiepko (1979)], we conclude that
M is locally the warped product N T ×f N ⊥ of a invariant submanifold
and an anti-invariant submanifold N ⊥ of M , where NT is a leaf of D and
N ⊥ is a leaf of D⊥ and f is a certain warping function. Therefore M is a
P R-warped product.
The next result was proved in [Chen and Munteanu (2012)].
Theorem 16.6. Let N T ×f N ⊥ be a P R-warped product in a para-Kähler

manifold M̃ . Suppose that N ⊥ is space-like and ∇⊥ (PN ⊥ ) ⊆ PN ⊥ . Then
the second fundamental form of M satisfies
Sσ ≤ 2p||∇ ln f ||2 + ||σνD ||2 , (16.51)
where p = dim N ⊥ , Sσ = g̃(σ, σ), ∇ ln f is the gradient of ln f with respect
to the metric g and
2
||σνD || = g̃ (σν (D, D), σν (D, D)) .
Here the index σν represents the ν-component of σ.
Proof. Let g T and g ⊥ denote the metrics on N T and N ⊥ respectively,

then the warped metric on M is given by g = gT + f 2 g⊥ .
Let us consider
• on N T : an orthonormal basis {Xi , Xi∗ = P Xi }, i = 1, . . . , h, where
h = dim N T ; moreover, one can suppose that ǫi := g(Xi , Xi ) = 1 and
hence ǫi∗ := g(Xi∗ , Xi∗ ) = −1, for all i.
• on N ⊥ : an orthonormal basis {Z̃a }, a = 1, . . . , p; we put ǫa :=

g⊥ (Z̃a , Z̃a ) = 1, for all a;
1
• in each point (x, y) ∈ M : Za (x, y) = f (x) Z̃a (y);
• in ν: an orthonormal basis {ξα , ξα∗ = f ξα∗ }, α = 1, . . . , q; moreover,
one can assume ǫα := g̃(ξα , ξα ) = 1 and thus ǫα∗ := g̃(ξα∗ , ξα∗ ) = −1.
Now, we want to compute

g̃(σ, σ) = g̃ (σ(D, D), σ(D, D)) + 2g̃ σ(D, D⊥ ), σ(D, D⊥ )

+ g̃ σ(D⊥ , D⊥ ), σ(D⊥ , D⊥ ) ,
where
Xh n
g̃ (σ(D, D), σ(D, D)) = ǫi ǫj g̃ (σ(Xi , Xj ), σ(Xi , Xj ))
i,j=1
+ ǫi∗ ǫj g̃ (σ(Xi∗ , Xj ), σ(Xi∗ , Xj )) (16.52)

+ ǫi ǫj∗ g̃ (σ(Xi , Xj∗ ), σ(Xi , Xj∗ ))
o
+ ǫi∗ ǫj∗ g̃ (σ(Xi∗ , Xj∗ ), σ(Xi∗ , Xj∗ )) ,

g̃ σ(D, D⊥ ), σ(D, D⊥ )
p n
h X
X
= ǫi ǫa g̃ (σ(Xi , Za ), σ(Xi , Za )) (16.53)
i=1 a=1
o
+ ǫi∗ ǫa g̃ (σ(Xi∗ , Za ), σ(Xi∗ , Za ))
and
p
X

g̃ σ(D⊥ , D⊥ ), σ(D⊥ , D⊥ ) = ǫa ǫb g̃ (σ(Za , Zb ), σ(Za , Zb )) . (16.54)
a,b=1
To do so, first we analyze σ(D, D). Since D is totally geodesic, we have
σ(D, D) ∈ ν. Hence one can write the following
α α∗
σ(Xi , Xj ) = σij ξα + σij ξα∗ ,
α α∗
σ(Xi∗ , Xj ) = σi∗j ξα + σi∗j ξα∗ ,
α α∗
σ(Xi∗ , Xj∗ ) = σi∗j∗ ξα + σi∗j∗ ξα∗ ,
α α∗
σ(Xi , Xj∗ ) = σij∗ ξα + σij∗ ξα∗ .
It follows that
q n
h X
X α 2 α∗ 2

g̃ (σ(D, D), σ(D, D)) = (σij ) − (σij )
i,j=1 α=1
α 2 α∗ 2
α 2 α∗ 2
(16.55)
− (σi∗j− (σi∗j
) ) − (σij∗ ) − (σij∗ )
α 2 α∗ 2
o
+ (σi∗j∗ ) − (σi∗j∗ ) .
Due to the integrability of D we deduce that

α α α∗ α∗ α α α∗ α∗
σi∗j = σij∗ , σi∗j = σij∗ , σi∗j∗ = σij , σi∗j∗ = σij .
Furthermore, using Lemma 16.4, we may write
g̃ (σ(X, Y ), ξ) = −g̃ (σ(X, P Y ), f ξ) , ∀ X, Y ∈ D, ξ ∈ ν
and consequently we have
α α∗
σij = g̃ (σ(Xi , Xj ), ξα ) = −g̃ (σ(Xi , Xj∗ ), ξα∗ ) = σij∗ ,
α∗ α
σij = −g̃ (σ(Xi , Xj ), ξα∗ ) = g̃ (σ(Xi , Xj∗ ), ξα ) = σij∗ .
By replacing all these in (16.55), we obtain
q
h X
X
g̃ (σ(D, D), σ(D, D)) = ||σνD ||2 = 4 α 2
(σij α∗ 2
) − (σij ) . (16.56)
i,j=1 α=1

Let us focus now on g̃ σ(D, D⊥ ), σ(D, D⊥ ) . As before, we write
b α α∗
σ(Xi , Za ) = σia F Zb + σia ξα + σia ξα∗ ,
b α α∗
σ(Xi∗ , Za ) = σi∗a F Zb + σi∗a ξα + σi∗a ξα∗ .
It follows that
p
X q
X
b 2
α 2 α∗ 2

g̃ (σ(Xi , Za ), σ(Xi , Za )) = − (σia ) + (σia ) − (σia ) ,
b=1 α=1
Xp q
X
b
α 2
g̃ (σ(Xi∗ , Za ), σ(Xi∗ , Za )) = − (σi∗a )2 + α∗ 2
(σi∗a ) − (σi∗a ) .
b=1 α=1
We obtain
P
h p
P
g̃ σ(D, D⊥ ), σ(D, D⊥ ) = − b 2
(σia b
) − (σi∗a )2
i=1 a,b=1
q
(16.57)
P p P
h P
α 2 α∗ 2 α 2 α∗ 2

+ (σia ) − (σia ) − (σi∗a ) + (σi∗ ) .
i=1 a=1 α=1
From Lemma 16.4 we have

g̃ (σ(P X, Z), f ξ) = −g̃ (σ(X, Z), ξ)
and consequently
α α∗
σi∗a = g̃(σ(Xi∗ , Za ), ξα ) = −g̃(σ(Xi , Za ), ξα∗ ) = σia ,
(16.58)
α∗ α
σi∗a = −g̃(σ(Xi∗ , Za ), ξα∗ ) = g̃(σ(Xi , Za ), ξα ) = σia .
Moreover we know that
g(σ(P X, Z), F W ) = −X(ln f )g(Z, W ).
This yields
b
σia = P Xi (ln f )δab ,
b
(16.59)
σi∗a = Xi (ln f )δab .
By combining (16.57), (16.58) and (16.59) we get
h
X 2 2
g(σ(D, D⊥ ), σ(D, D⊥ )) = p Xi (ln f ) − P Xi (ln f )
i=1
q
p X
(16.60)
h X
X α 2 α∗ 2

+2 (σia ) − (σia ) .
i=1 a=1 α=1
As
g(σ(X, Z), f ξ) = −g(∇⊥
X F Z, ξ)
and using the hypothesis ∇⊥ ⊥ ⊥

D PD ⊂ PD , we find
σ(D, D⊥ ) ⊂ PD⊥ .
α α∗
Hence σia and σia vanish. Thus

g̃ σ(D, D⊥ ), σ(D, D⊥ ) = p g (∇ ln f, ∇ ln f ) . (16.61)
Finally, we study g(σ(D⊥ , D⊥ ), σ(D⊥ , D⊥ )). We write
c α α∗
σ(Za , Zb ) = σab F Zc + σab ξα + σab ξα∗
and hence
p X
X q p
X
α 2
g(σ(D⊥ , D⊥ , σ(D⊥ , D⊥ )) = α∗ 2
(σab ) − (σab ) − c 2
(σab ) .
a,b=1 α=1 a,b,c=1
As g(σ(Z, W ), f ξ) = −g(∇⊥ ⊥
Z F W, ξ) and using ∇D ⊥ PD
⊥
⊂ PD⊥ , we
⊥ ⊥ ⊥ α α∗
derive σ(D , D ) ⊂ PD . Thus we have σab = σab = 0. Consequently, we
obtain
X p
g(σ(D⊥ , D⊥ ), σ(σ(D⊥ , σ(D⊥ )) = − c 2
(σab ) . (16.62)
a,b,c=1
From these we obtain the theorem.
Remark 16.3. If the manifold N ⊥ in Theorem 16.6 were time-like, then

2 2
(16.51) shall be replaced by Sσ ≥ 2p||∇ ln f || + ||σνD || .
Remark 16.4. For every P R-warped product N T × N ⊥ in M̃ , we have

dim M̃ ≥ dim N T + 2 dim N ⊥ . Thus the smallest possible codimension for
such warped products in M̃ is equal to dim N ⊥ .
Theorem 16.7. [Chen and Munteanu (2012)] Let N T ×f N ⊥ be a P R-

warped product in a para-Kähler manifold M̃ . If N ⊥ is space-like (resp.,
time-like) and dim M̃ = dim N T + 2 dim N ⊥ , then the second fundamental
form of M satisfies
2 2
Sσ ≤ 2p||∇ ln f || (respectively, Sσ ≥ 2p||∇ ln f || ). (16.63)
If the equality sign of (16.63) holds identically, we have
σ(D, D) = σ(D⊥ , D⊥ ) = {0}. (16.64)
Proof. Inequality (16.63) follows from (16.51). If the equality sign holds,
then (16.64) follows from the proof of Theorem 16.6.
16.9 P R-warped products satisfying the basic equality
In this section, we will use letters i, j, k for indices running from 1 to h;

a, b, c for indices from 1 to p; and A, B for indices between 1 and m with
m = h + p.
2(h+p)
On Eh+p we consider the global coordinates (xi , xh+a , yi , yh+a ) and
2(h+p)
use the canonical flat para-Kähler structure on Eh+p defined as above.
The next result from [Chen and Munteanu (2012)] determines space-like
P R-warped products in P h+p satisfying the equality case of (16.63).
Proposition 16.11. Let M = N T ×f N ⊥ be a space-like P R-warped
product in the para-Kähler (h + p)-plane P h+p with h = 12 dim N T and
p = dim N ⊥ . If M satisfies the equality case of (16.63) identically, then
(a) N T is a totally geodesic submanifold in P h+p , and hence it is congruent
to an open part of P h ;
(b) N ⊥ is a totally umbilical submanifold in P h+p .
Moreover, if N ⊥ is a real space form of constant curvature k, then the
2
warping function f satisfies |∇f | = k.
Proof. Under the hypothesis, we know from the proof of Theorem 16.6
that the second fundamental form satisfies σ(D, D) = σ(D⊥ , D⊥ ) = {0}.
On the other hand, since M = N T ×f N ⊥ is a warped product, N T is
totally geodesic and N ⊥ is totally umbilical in M . Thus we have the first
two statements.
The last statement of the proposition can be proved as follows. If R⊥
denotes the Riemann curvature tensor of N ⊥ , then we have

RZV W = RZV⊥
W − |∇ ln f |2 g(V, W )Z − g(Z, W )V
for any Z, V, W tangent to N ⊥ .

If N ⊥ is a space form of constant curvature k, then R takes the form

k 2
RZV W = 2
− |∇ ln f | g(V, W )Z − g(Z, W )V . (16.65)
f
The equation of Gauss may be written, for vectors tangent to N ⊥ , as
g (RZV W, U ) = g(R̃ZV W, U ) + g(σ(V, W ), σ(Z, U ))
− g(σ(Z, W ), σ(V, U )).
Since the ambient space is flat and σ(D⊥ , D⊥ ) = 0 due to the equality
in (16.63), it follows that g(RZV W, U ) = 0. Combining this with (16.65)
gives ||∇ ln f ||2 = k/f 2 . This gives the statement.
Para-complex numbers have the expression v = x + jy, where x, y are
real numbers and j satisfies j2 = 1, j 6= 1. The conjugate of v is v̄ = x − jy.
The multiplication of two para-complex numbers is defined by
(a + jb)(s + jt) = (as + bt) + j(at + bs).
For each natural number m, we put
Dm = {(x1 + jy1 , . . . , xm + jym ) : xi , yi ∈ R}.
With respect to the multiplication of para-complex numbers and the canon-
ical flat metric, Dm is a flat para-Kähler manifold of dimension 2m. Once we
identify (x1 + jy1 , . . . , xm + jym ) ∈ Dm with (x1 , . . . , xm , y1 , . . . , ym ) ∈ E2m
m ,
we may identify Dm with the para-Kähler m-plane P m in a natural way.
We denote by S p , Ep and H p the unit p-sphere, the Euclidean p-space
and the unit hyperbolic p-space, respectively.
Theorem 16.8. [Chen and Munteanu (2012)] Let N T ×f N ⊥ be a space-
like P R-warped product in the para-Kähler (h + p)-plane P h+p with h =
1 T
2 dim N and p = dim N ⊥ . Then we have
Sσ ≤ 2p|∇ ln f |2 . (16.66)
T
The equality sign of (16.66) holds identically if and only if N is an open
part of a para-Kähler h-plane, N ⊥ is an open part of S p , Ep or H p , and
the immersion is given by one of the following.
(1) Φ : D1 ×f S p → P h+p ;

h
X h
X
Φ(z, w) = z1 + v̄1 (w0 − 1) vj zj , . . . , zh + v̄h (w0 − 1) vj zj ,
j=1 j=1

h
X h
X
w1 jvj zj , . . . , wp jvj zj  , h ≥ 2,
j=1 j=1
p
with warping function f = hv̄, zi2 − hjv̄, zi2 , where v = (v1 , . . . , vh ) ∈
S 2h−1 ⊂ Dh , w = (w0 , w1 , . . . , w

p
p ) ∈ S , z = (z1 , . . . , zh ) ∈ D1 and
D1 = z ∈ Dh : hv̄, zi2 > hjv̄, zi2 .
(2) Φ : D1 ×f H p → P h+p ;

X h h
X
Φ(z, w) = z1 + v̄1 (w0 − 1) vj zj , . . . , zh + v̄h (w0 − 1) vj zj ,
j=1 j=1

h
X h
X
w1 jvj zj , . . . , wp jvj zj  , h ≥ 1,
j=1 j=1
p
with the warping function given by f = hv̄, zi2 − hjv̄, zi2 , where v =
(v1 , . . . , vh ) ∈ H 2h−1 ⊂ Dh , w = (w0 , w1 , . . . , wp ) ∈ H p and z =
(z1 , . . . , zh ) ∈ D1 .
(3) Φ(z, u) : D1 ×f Ep −→ P h+p ;

v̄1 X 2 X v̄h X 2 X
p h p h
Φ(z, u) = z1 + ua vj zj , . . . , zh + ua vj zj ,
2 a=1 j=1
2 a=1 j=1

Xh X h
u1 jvj zj , . . . , up jvj zj  , h ≥ 2,
j=1 j=1
p
with warping function f = g(v̄, z)2 −g(jv̄, z)2 ,
where v = (v1 , . . . , vh )
is a light-like vector in Dh , z = (z1 , . . . , zh ) ∈ D1 and u = (u1 , . . . , up )
∈ Ep . Moreover, in this case, each leaf Ep is quasi-minimal in P h+p .
(4) Φ(z, u) : D2 ×f Ep −→ P h+p ;
p p
!
v1 X 2 vh X 2 v0 v0
Φ(z, u) = z1 + u , . . . , zh + u , u1 , . . . , up ,
2 a=1 a 2 a=1 a 2 2
h ≥ 1,
p √ √
with the warping function f = −hv, zi, where v0 = b1 + ǫj b1 with
b1 > 0, ǫ = ±1, z = (z1 , . . . , zh ) ∈ D2 and u = (u1 , . . . , up ) ∈ Ep ,
D2 = {z ∈ Dh : hv, zi < 0}, and
v = (v1 , . . . , vh ) = (b1 + ǫjb1 , . . . , bh + ǫjbh ).
2(h+p)
In each of the four cases the warped product is minimal in Eh+p .
Remark 16.5. For a time-like P R-warped product N T ×f N ⊥ in the para-
Kähler (h + p)-plane P h+p , a similar classification result as Theorem 16.8
holds as well (cf. Remark 6.4 of [Chen and Munteanu (2012)]).
Chapter 17
Warped Products in Sasakian

Manifolds
17.1 Sasakian manifolds and submanifolds
A (2m + 1)-dimensional Riemannian manifold M̃ is a Sasakian manifold if

it admits a (1, 1)-tensor field φ, a vector field ξ and a 1-form η satisfying
the following structure equations:
φ2 X = −X + η(X)ξ, (17.1)
g(φX, φY ) = g(X, Y ) − η(X)η(Y ), (17.2)
η(ξ) = 1, η(X) = g(X, ξ), (17.3)
˜ X φ)Y = g(X, Y )ξ − η(Y )X,
(∇ (17.4)
˜ X ξ = −φX,
∇ (17.5)
for vector fields X, Y on M̃ , where g is metric on M̃ . The vector field ξ is
called the structure vector field or Reeb vector field. The 1-form η defines
the contact distribution D = {X ∈ T M̃ : η(X) = 0} of rank 2m.
It follows immediately from (17.1) and (17.2) that
g(X, φY ) = −g(φX, Y ) (17.6)
for X, Y ⊥ ξ. In particular, we have g(X, φX) = 0 for all X ⊥ ξ.
Definition 17.1. A submanifold M of a Sasakian manifold M̃ is called a

C-totally real submanifold if it is normal to the structure vector field ξ.
The following result can be found in page 43 of [Yano and Kon (1983)].
Proposition 17.1. If M is a C-totally real submanifold of a Sasakian

manifold M̃ 2m+1 , then M is anti-invariant, i.e.,
φ(Tx M ) ⊂ Tx⊥ M, ∀x ∈ M. (17.7)
Moreover, we have dim M ≤ m.
409
Proof. Since ξ is normal to M , Weingarten’s formula and (17.5) imply

˜ X ξ, Y ) = g(Aξ X, Y )
g(φX, Y ) = −g(∇
for vector fields X, Y ∈ X(M ). Since Aξ is symmetric and φ is skew-

symmetric, we get Aξ = 0 and φX ∈ T ⊥ M . Hence dim M ≤ m.
For submanifolds of M̃ tangent to ξ, there are several special families
of submanifolds defined based on the action of φ on the submanifolds.
Definition 17.2. Let M be a submanifold of a Sasakian manifold M̃ which

is tangent to ξ. Then
(a) M is called an invariant submanifold if each tangent space of M is

invariant under φ, i.e., φ(Tx M ) ⊂ Tx M , ∀x ∈ M .
(b) M is called an anti-invariant submanifold if φ(Tx M ) ⊂ Tx⊥ M , ∀x ∈ M .
(c) M is called a contact CR-submanifold if its tangent bundle admits an
orthogonal decomposition T M = DT ⊕ D⊥ , where DT is a φ-invariant
distribution and D⊥ an anti-invariant distribution on M .
(d) A contact CR-submanifold M is called a contact CR-product if it is
locally the Riemannian product N T × N ⊥ of an invariant submanifold
N T and an anti-invariant submanifold N ⊥ of M̃ .
Let M be a submanifold tangent to ξ in a Sasakian manifold M̃ . For a

vector X ∈ T M , we put
φX = P X + F X, (17.8)
where P X and F X denote the tangential and normal components of φX,

respectively. Then P is an endomorphism of T M of and F is a normal
bundle valued 1-form on M .
Since ξ is tangent to M , we derive from (17.1) and (17.5) that
P ξ = F ξ = 0, (17.9)
∇X ξ = −P X, σ(X, ξ) = −F X. (17.10)
Similarly, for a vector ζ ∈ T ⊥ M , we put
φζ = tζ + f ζ, (17.11)
where tζ and f ζ are the tangential and the normal components of φζ,
respectively.
By applying Proposition 5.6 we know that the Riemann curvature tensor
of a submanifold M in a Sasakian space form M̃ (c) of constant φ-sectional
Warped Products in Sasakian Manifolds 411
curvature c satisfies
c + 3
R(X, Y )Z = g(Y, Z)X − g(X, Z)Y
4
c − 1
− η(Y )η(Z)X − η(X)η(Z)Y
4 (17.12)
+ g(Y, Z)η(X)ξ − g(X, Z)η(Y )ξ − g(P Y, Z)P X

+ g(P X, Z)P Y + 2g(P X, Y )P Z
+ Aσ(Z,Y ) X − Aσ(Z,X) Y
for vectors X, Y, Z ∈ T M . Moreover, we know that the equations of Gauss
and Codazzi of M in M̃ (c) are given respectively by
R(X, Y ; Z, W ) = R̃(X, Y ; Z, W ) + hσ(X, W ), σ(Y, Z)i
(17.13)
− hσ(Y, W ), σ(X, Z)i
and
¯ X σ)(Y, Z) − (∇
(∇ ¯ Y σ)(X, Z)
c−1 (17.14)
= {g(P Y, Z)F X − g(P X, Z)F Y − 2g(P X, Y )F Z} .
4
Definition 17.3. The second fundamental form σ of M in M̃ (c) is said to
satisfy the classical Codazzi equation if
¯ X σ)(Y, Z) = (∇
(∇ ¯ Y σ)(X, Z) (17.15)
holds for all X, Y, Z ∈ T M.
Lemma 17.1. Let M be a submanifold tangent to ξ in a Sasakian space

form M̃ (c) with c 6= 1. If the second fundamental form σ of M satisfies
the classical Codazzi equation, then M is either an invariant submanifold
or an anti-invariant submanifold.
Proof. By applying (17.13) and (17.15), we get

g(P Y, Z)F X − g(P X, Z)F Y − 2g(P X, Y )F Z = 0 (17.16)
for all X, Y, Z ∈ T M . Suppose that there exist a vector U ∈ Tx M, x ∈
M , such that P U 6= 0 and F U 6= 0. Then we deduce from (17.16) that
g(P U, P U )F U = 0 is false. Hence we must have either P U = 0 or F U = 0.
Moreover, we can also prove that there do not exist U, V ∈ Tx M such that
F U = P V = 0 and P U, F V 6= 0. Consequently, we obtain either P = 0 or
F = 0. Consequently, M is either an invariant submanifold (if F = 0) or
an anti-invariant submanifold (if P = 0).
17.2 Warped products in Sasakian manifolds
For C-totally real warped products in a Sasakian space form, we have.
Theorem 17.1. [Matsumoto and Mihai (2002)] Let ψ : N1 ×f N2 → M̃ (c)

be a C-totally real immersion of a warped product N1 ×f N2 into a Sasakian
space form M̃ (c). Then we have
∆f (n1 + n2 )2 2 n1 (c + 3)
≤ H + , (17.17)
f 4n2 4
where ni = dim Ni , i = 1, 2.
The equality sign of (17.17) holds if and only if ψ is mixed totally
geodesic and n1 H1 = n2 H2 , where Hi , i = 1, 2 denote the partial mean
curvature vectors.
Remark 17.1. Theorem 17.1 was proved by modifying the proof of The-
orem 10.1, using (17.12) instead of (2.21). For the details, see [Matsumoto
and Mihai (2002)].
Some immediate applications of Theorem 17.1 are the following.
Corollary 17.1. Let N1 ×f N2 be a warped product whose warping function

f is harmonic. Then
(i) N1 ×f N2 admits no minimal C-totally real immersion into a Sasakian

space form M̃ (c) with c < −3.
(ii) Every minimal C-totally real immersion of N1 ×f N2 in the standard
Sasakian space form R2m+1 (−3) is a warped product immersion.
Proof. (a) Assume f is a harmonic function on N1 and N1 ×f N2 admits

a minimal C-totally real immersion in a Sasakian space form M̃ (c). Then
the inequality (17.17) becomes c ≥ −3.
(b) If c = −3, the equality case of (17.17) holds identically. Hence, by
Theorem 17.1, it follows that N1 ×f N2 is mixed totally geodesic in M̃ (c)
and H1 = H2 = 0 holds. Consequently, a well-known result of [Nölker
(1996)] implies that the immersion is a warped product immersion.
Corollary 17.2. If the warping function f of a warped product N1 ×f N2

is an eigenfunction of the Laplacian on N1 with corresponding eigenvalue
λ > 0, then N1 ×f N2 does not admit a minimal C-totally real immersion
in a Sasakian space form M̃ (c) with c ≤ −3.
The following is an example of a C-totally real submanifold satisfying

the equality case of (17.17) identically.
Example 17.1. Consider a totally geodesic S 5 (1) ⊂ S 7 (1) and let ζ be a

unit normal vector field orthogonal to the linear subspace containing S 5 (1).
Let N be a minimal C-totally real surface of S 5 (1) and define the warped
product manifold M = (− π2 , π2 ) ×cos t N . Then
ψ : (− π2 , π2 ) ×cos t N → S 7 (1); ψ(t, p) = (sin t)ζ + (cos t)p
is an isometrical immersion of M in S 7 (1). It is direct to verify that ψ is
C-totally real and it satisfies the equality case of (17.17) identically.
For a warped product N1 ×f N2 in a Sasakian manifold M̃ with ξ tangent

to N2 , we have the following.
Proposition 17.2. [Matsumoto and Mihai (2002)] Let N1 ×f N2 be a

warped product in a Sasakian manifold M̃ . If ξ is tangent to N2 , then
f is a constant and N1 is an anti-invariant submanifold of M̃ .
Proof. Assume N1 ×f N2 is a warped product submanifold of a Sasakian

manifold M̃ such that ξ is tangent to N2 . For vector fields X tangent to
N1 and Z tangent to N2 , Proposition 3.1(2) yields
∇X Z = (X ln f )Z. (17.18)
If we put Z = ξ in (17.18), then it follows from (17.10) and (17.18) that
−P X = ∇X ξ = (X ln f )ξ.
Since X lies in the contact distribution, P X and ξ are linearly independent.
Hence we get X ln f = 0 and P X = 0 for X ∈ T N1 . Therefore f is constant
and N1 is an anti-invariant submanifold.
Similarly, we also have the following result from [Uddin et al. (2010)].
Proposition 17.3. Let N1 ×f N2 be a warped product in a Sasakian man-

ifold M̃ . If ξ is tangent to N1 , then N2 is an anti-invariant submanifold of
M̃ . Moreover, we have ξ(ln f ) = 0.
Proof. Assume that N1 ×f N2 is a warped product submanifold of a

Sasakian manifold M̃ such that ξ is tangent to N1 . Hence we obtain from
Proposition 3.1(2) and (17.5) that
−P Z = ∇Z ξ = (ξ ln f )Z. (17.19)
Since P Z ⊥ Z, (17.19) implies P Z = 0 and ξ(ln f ) = 0. Consequently, N2
is an anti-invariant submanifold of M̃ .
By using Lemma 10.2, we also have the following result.
Proposition 17.4. [Uddin et al. (2010)] Let f1 N1 ×f2 N2 be a doubly warped

product in a Sasakian manifold M̃ .
(1) If ξ is tangent to N1 , then f1 is constant and N2 is anti-invariant.

(2) If ξ is tangent to N2 , then f2 is constant and N1 is anti-invariant.
Consequently, the warped product is a singly warped product in both cases.
Proof. Let f1 N1 ×f2 N2 be a doubly warped product in a Sasakian man-

ifold M̃ . Then Lemma 10.2 gives
∇X Z = ∇Z X = (X ln f2 )Z + (Z ln f1 )X (17.20)
for X tangent to N1 and Z tangent to N2 . If ξ is tangent to N1 , we get
∇Z ξ = (ξ ln f2 )Z + (Z ln f1 )ξ.
Combining this with (17.5) and Gauss’ formula gives
−P Z − F Z = ∇ ˜ Z ξ = (ξ ln f2 )Z + (Z ln f1 )ξ + σ(ξ, Z). (17.21)
Since Z, P Z and ξ are linearly independent, (17.21) implies P Z = 0 and
Z ln f1 = 0 for any Z ∈ T N2 . Hence we obtain statement (1).
Similarly, we also have statement (2).
For a warped product submanifold N1 ×f N2 in a Sasakian space form
with ξ ∈ T N1 , we have the following.
Theorem 17.2. [Matsumoto and Mihai (2002)] If N1 ×f N2 is a warped

product submanifold of a Sasakian space form M̃ (c) such that ξ is tangent
to N1 , then we have
∆f (n1 + n2 )2 2 n1 (c + 3) c − 1
≤ H + − , (17.22)
f 4n2 4 4
where ni = dim Ni , i = 1, 2. Moreover, the equality sign of (17.22) holds if
and only if ψ is mixed totally geodesic and n1 H1 = n2 H2 , where Hi , i = 1, 2
are the partial mean curvature vectors.
17.3 Contact CR-submanifolds
Given a contact CR-submanifold M of a Sasakian manifold M̃ with ξ ∈ D,

or ξ ∈ D⊥ , the tangent space Tx M at x ∈ M decomposes orthogonally as
Tx M = Hx (M ) ⊕ Span {ξx } ⊕ Hx⊥ (M ),
such that
φ(Hx (M )) = Hx (M ), φ(Hx⊥ (M )) ⊂ Tx⊥ M.
Definition 17.4. Given a contact CR-submanifold M of a Sasakian man-

ifold M̃ , if the dimension Hx (M ) is constant along M , then H(M ) is the
maximally φ-invariant distribution of M in M̃ . In this case,
H(M ) = {Hx (M ), x ∈ M }
is called the Levi distribution of M .
Consider a warped products N1 ×f N2 in M̃ which is a contact CR-

submanifold of M̃ . Such a submanifold is always tangent to the structure
vector field ξ.
We distinguish 2 cases:
(a) ξ is tangent to N1 ;
(b) ξ is tangent to N2 .
In Case (a), there are 2 subcases:

(a.1) N1 is anti-invariant and N2 is invariant;
(a.2) N1 is invariant and N2 is anti-invariant in M̃ .
Theorem 17.3. Let M = N1 ×f N2 be a warped product submanifold of

a Sasakian manifold M̃ . If N1 is an anti-invariant submanifold and N2 is
an invariant submanifold of M̃ , then the warping function f is a constant.
Hence M is a contact CR-product.
Proof. Under the hypothesis of the theorem, let X be a vector field

tangent to N2 and Z a vector field tangent to N1 . Then it follows from
Proposition 3.1(2) that
∇X Z = (Z ln f )X. (17.23)
Now we distinguish two cases:
Case (1): ξ is tangent to N1 . We put Z = ξ. Since we have
˜ X ξ = −φX ∈ T N2 ,
∇
we obtain from (17.23) that φX = (ξ ln f )X. But this is impossible when-
ever dim N2 6= 0.
Case (2): ξ is tangent to N2 . In this case, we put X = ξ. Since we have
∇Z ξ = −P Z = 0 as well as ∇Z ξ = ∇ξ Z for any vector field Z tangent to
N1 , we obtain Z(ln f )ξ = 0 and hence we have Z(ln f ) = 0 for all Z tangent

to N1 . Consequently, f is constant and hence the warped product above is
nothing but a Riemannian product N1 × N2f where N2f is the manifold N2
with metric f 2 gN2 which is homothetic with the original metric.
Remark 17.2. Theorem 17.3 was proved in [Munteanu (2005)]. This result
is due to [Hasegawa and Mihai (2003)] under an additional condition that
ξ is tangent to N1⊥ .
Now, we consider the subcase (a.2).

In views of Proposition 17.4, Theorem 17.3 and Definition 11.1, we call
a submanifold M in a Sasakian manifold a contact CR-warped product if it
is the warped product N T ×f N ⊥ such that N T is an invariant submanifold
tangent to ξ and N ⊥ is a C-totally real submanifold.
The next result is due to [Hasegawa and Mihai (2003); Munteanu
(2005)].
Theorem 17.4. Let N T ×f N ⊥ be a warped product in a Sasakian manifold

M̃ such that N T is an invariant submanifold tangent to ξ and N ⊥ is a β-
dimensional C-totally real submanifold of M̃ . Then we have:
(i) The second fundamental form σ of M satisfies
kσk2 ≥ 2β{|∇(ln f )|2 + 1}. (17.24)
(ii) If the equality sign of (17.24) holds identically, then N T is a totally

Moreover, M is a minimal submanifold of M̃
Proof. Let M = N T ×f N ⊥ be a warped product in a Sasakian manifold

M̃ such that N T is an invariant submanifold tangent to ξ and N ⊥ is a
C-totally real submanifold of M̃ .
For a unit vector field X tangent to N T and vector fields Z, W tangent
to N ⊥ , we have
g(σ(φX, Z), φZ) = g(∇˜ Z φX, φZ) = g(φ∇
˜ Z X, φZ)
(17.25)
˜ Z X, Z) = g(∇Z X, Z) = X ln f.
= g(∇
On the other hand, since M̃ is Sasakian, it is easily seen that
σ(ξ, Z) = −φZ. (17.26)
Now, inequality (17.24) follows easily from (17.25) and (17.26).

Let σ ′′ denote the second fundamental form of N ⊥ in M . Then

g(σ ′′ (Z, W ), X) = g(∇Z W, X) = −(X ln f )g(Z, W ),
or, equivalently,
σ ′′ (Z, W ) = −g(Z, W )∇(ln f ). (17.27)
If the equality sign of (17.24) holds identically, then we have
σ(D, D) = 0, σ(D⊥ , D⊥ ) = 0, σ(D, D⊥ ) ⊂ φD⊥ . (17.28)
T
Since N is totally geodesic in M , the first condition in (17.28) shows that
N T is also totally geodesic in M̃ .
The second condition in (17.28) and (17.27) imply that N2 is totally
umbilical in M̃ . Furthermore, it follows from (17.28) that M is a minimal
submanifold of M̃ . This completes the proof of the theorem.
The following result is a consequence of Theorem 17.4.
Corollary 17.3. [Hasegawa and Mihai (2003)] Let M = N T ×f N ⊥ be
a warped product in a Sasakian space form M̃ (c) of constant φ-sectional
curvature c such that N T is an invariant submanifold tangent to ξ and
N ⊥ is a β-dimensional C-totally real submanifold of M̃ . If the second
fundamental form σ of M satisfies
kσk2 = 2β{|∇(ln f )|2 + 1}. (17.29)
Then
(a) N T is a totally geodesic invariant submanifold of M̃ (c). Hence N T is
a Sasakian space form of constant φ-sectional curvature c.
(b) N ⊥ is a totally umbilical C-totally real submanifold of M̃ (c). Hence
N ⊥ is a real space form of sectional curvature ε > c+3
4 .
(c) If β > 1, then the warping function f satisfies

2 c+3
|∇f | = ε − f 2.
4
Proof. Statement (a) of this corollary follows from Theorem 17.4. Also,
it follows that N ⊥ is totally umbilical in the Sasakian space form M̃ (c).
Gauss’ equation implies that N ⊥ is of constant curvature ε ≥ (c + 3)/4.
Moreover, (17.23) implies that ε = (c + 3)/4 holds if and only if the warping
function f is constant.
Let R′′ denote the Riemann curvature tensor of N ⊥ . Then we find from
Proposition 3.2(5) that the Riemann curvature tensor R′′ of N ⊥ satisfies
R′′ (Z, W )V = R(Z, W )V + k∇(ln f )k2 {g(W, V )Z − g(Z, V )W },
for Z, W, V tangent to N2 . Consequently, after applying the equation of
Gauss, we obtain statement (c).
17.4 CR-warped products with smallest codimension
Analogous to Theorem 12.4, the next result from [Hasegawa and Mihai
(2003)] determines the smallest codimension for contact CR-warped prod-
ucts in a standard Sasakian sphere.
Theorem 17.5. Let N T ×f N ⊥ be a contact CR-warped product in the

Sasakian (2m + 1)-dimensional sphere S 2m+1 . If N T is compact, then
m ≥ α + β + αβ, (17.30)
where dim N T = 2α + 1 and dim N ⊥ = β.
Proof. Let N T ×f N ⊥ be a contact CR-warped product in the Sasakian

sphere S 2m+1 such that N T is an invariant submanifold tangent to ξ and
N ⊥ is a C-totally real submanifold of S 2m+1 . We denote by ν the normal
subbundle orthogonal to φD⊥ . Obviously, we have
T ⊥ M = φD⊥ ⊕ ν, φν = ν.
T
For any vector fields X tangent to N and orthogonal to ξ and for vector
field Z tangent to N ⊥ , equation (17.23) gives R̃(X, φX; Z, φZ) = 0.
On the other hand, by Codazzi equation we find
R̃(X, φX; Z, φZ) = −g(DX σ(φX, Z), φZ)
+ g(σ(∇X φX, Z), φZ) + g(σ(φX, ∇X Z), φZ) (17.31)
+ g(DφX σ(X, Z) − σ(∇φX X, Z) − σ(X, ∇φX Z), φZ).
By using (17.23) and structure equations of the Sasakian manifold, we get
g(DX σ(φX, Z), φZ) = Xg(σ(φX, Z), φZ) − g(σ(φX, Z), DX φZ)
= Xg(∇Z X, Z) − g(σ(φX, Z), φ∇˜ X Z)
= X((X ln f )g(Z, Z)) − (X ln f )g(σ(φX, Z), φZ)
− g(σ(φX, Z), φσν (X, Z))
= (X 2 ln f )g(Z, Z) + (X ln f )2 g(Z, Z) − |σν (X, Z)|2 ,
where σν (X, Z) denotes the ν-component of σ(X, Z).
Also, we have
g(σ(∇X φX, Z), φZ) = g(∇˜ Z ∇X φX, φZ)
˜ Z∇
= g(∇ ˜ X φX, φZ) − g(∇
˜ Z σ(X, φX), φZ)
= −g(X, X)g(Z, Z) + ((∇X X) ln f )g(Z, Z),
g(σ(φX, ∇X Z), φZ) = (X ln f )g(σ(φX, Z), φZ) = (X ln f )2 g(Z, Z).
Substituting the above relations into (17.31), we obtain
R̃(X, φX; Z, φZ) = 2|σν (X, Z)|2 − (X 2 ln f )g(Z, Z)

+ ((∇X X) ln f )g(Z, Z) − ((φX)2 ln f )g(Z, Z) (17.32)
+ ((∇φX φX) ln f )g(Z, Z) − 2g(X, X)g(Z, Z),
which implies
n 1
kσν (X, Z)k2 = g(X, X) + (H ln f (X, X)
2 o (17.33)
+ H ln f (φX, φX)) g(Z, Z).
Let {X0 = ξ, X1 , ..., X2α , Z1 , ..., Zβ } be a local orthonormal frame on M

such that X0 , ..., X2α are tangent to N1 and Z1 , ..., Zβ are tangent to N ⊥ .
Since the Hessian and the second fundamental form are bilinear, we obtain
from (17.33) via polarization that
g(σν (Xi , Zs ), σν (Xj , Zt ))

1 (17.34)
= 1 + (H ln f (Xi , Xj ) + H ln f (φXi , φXj )) δij δst .
2
Therefore {σν (Xi , Zt ) : i = 1, ..., 2α; t = 1, ..., β} are mutually orthogonal
vector fields.
Suppose that N T is compact. Then ln f has an absolute minimum at
some point u ∈ N T . At this critical point, the Hessian H ln f is non-negative
definite. Thus each σν (Xi , Zt ) 6= 0 by (17.33). Therefore the rank of ν is
at least 2αβ. Consequently, we obtain m ≥ α + β + αβ.
Definition 17.5. A contact CR submanifold M in a Sasakian manifold

with ξ ∈ DT is called strictly proper if dim H(M ) > 0 and dim D⊥ > 0,
where H(M ) is the Levi distribution of M (cf. Definitions 17.2 and 17.3).
For strictly proper contact CR-products we have the following two re-
sults from [Munteanu (2005)].
Proposition 17.5. Let N T × N ⊥ be a strictly proper contact CR-product

in a Sasakian space form M̃ 2m+1 (c) with c 6= −3. Then m ≥ αβ + α + β,
where α and β are given as in Theorem 17.5.
Proposition 17.6. Let N T × N ⊥ be a contact CR-product in a Sasakian

space form M̃ 2m+1 (c) with c 6= −3. If m = αβ + α + β, then N T is a totally
geodesic invariant submanifold of M̃ .
The following example shows that Proposition 17.5 is false if c = −3.
Example 17.2. [Munteanu (2005)] Let R2α+1 (−3) denote the standard
Sasakian space form of constant φ-sectional curvature −3. Denote by S β (1)
the unit hypersphere in the Euclidean space Eβ+1 .
For a positive integer a ≤ α, let us consider the map
ϕ : R2α+1 (−3) × S β (1) → R2aα+2β+1 (−3) (17.35)
defined by
ϕ(x1 , y1 , . . . , xα , yα , z, w0 , w1 , . . . , wβ )
= w0 x1 , w0 y1 , . . . , wβ x1 , wβ y1 , . . . , w0 xa ,

w0 ya , . . . , wβ xa , wβ ya , xa+1 , ya+1 , . . . , xα , yα , z ,
where w02 + w12 + . . . + wp2 = 1.
It is direct to verify that ϕ is an isometric immersion of the warped
product R2α+1 (−3) ×f S β (1) intoqthe Sasakian space form R2aα+2β+1 (−3)
Ph
with the warped function f = 12 2 2
i=1 (xi + yi ) such that R
2α+1
(−3) is
β
immersed as an invariant submanifold and S as an anti-invariant subman-
ifold in R2aα+2β+1 (−3).
It is direct to verify that ϕ : R2α+1 (−3) × S β (1) → R2aα+2β+1 (−3)
satisfies the equality: ||σ||2 = 2β |∇ ln f |2 − ∆ ln f + 1 .
17.5 Another inequality for contact CR-warped products

in Sasakian manifolds
Analogous to Theorem 13.1 for CR-warped products in complex space

forms, there is another inequality for all contact CR-warped product sub-
manifolds in a Sasakian space form.
Theorem 17.6. [Mihai (2004); Munteanu (2005)] Let M = N T ×f N ⊥ be

a contact CR-warped product in a Sasakian space form M̃ (c) such that N T
is a (2α + 1)-dimensional invariant submanifold tangent to ξ and N ⊥ is a
β-dimensional C-totally real submanifold of M̃ . Then
(a) The second fundamental form σ of M satisfies
||σ||2 ≥ 2β{|∇(ln f )|2 − ∆(ln f ) + 1} + αβ(c + 3). (17.36)
(b) The equality sign of (17.36) holds identically if and only if
(b.1) N T is a totally geodesic invariant submanifold of M̃ (c). Hence N T
is a real space form of constant curvature ǫ ≥ (c + 3)/4.
(b.2) N ⊥ is a totally umbilical C-totally real submanifold of M̃ (c). Hence

N ⊥ is a Sasakian space form of constant φ-sectional curvature c.
Proof. Let M = N T ×f N ⊥ be a CR-warped product submanifold in
a Sasakian space form such that N T is a (2α + 1)-dimensional invariant
submanifold tangent to ξ and N ⊥ is a β-dimensional C-totally real sub-
manifold. We choose a local orthonormal frame on M :
{X0 = ξ, X1 , . . . , Xα , Xα+1 = φX1 , . . . , X2α = φXα , Z1 , . . . , Zβ }
so that X0 , . . . , X2α are tangent to N T and Z1 , . . . , Zβ are tangent to N ⊥ .
For unit vector fields X tangent to N T and Z, W tangent to N ⊥ , we
have
g(σ(φX, Z), φZ) = g(∇ ˜ Z φX, φZ) = g(φ∇˜ Z X, φZ)
(17.37)
= g(∇˜ Z X, Z) = g(∇Z X, Z) = X(ln f ).
On the other hand, since the ambient manifold is Sasakian, we have
σ(ξ, Z) = φZ. (17.38)
⊥
Let σφD⊥ (X, Z) denote the φ(D )-component of σ(X, Z). It follows
from (17.37) and (17.38) that
β
2α X
X
|σφD⊥ (Xi , Zt )|2 = β{|∇(ln f )|2 + 1}. (17.39)
i=0 t=1
Let ν denote the normal subbundle of M orthogonal to φ(D⊥ ), i.e.,
T ⊥ M = φ(D⊥ ) ⊕ ν, φ(ν) = ν.
For any unit vector field X tangent to N T and orthogonal to ξ and for Z
tangent to N ⊥ , (17.12) gives
c−1
R̃(X, φX; Z, φZ) = .
2
On the other hand, by Codazzi’s equation we have
R̃(X, φX; Z, φZ)
= g(DφX σ(X, Z) − σ(∇φX X, Z) − σ(X, ∇φX Z), φZ) (17.40)
− g(DX σ(φX, Z) − σ(∇X φX, Z) − σ(φX, ∇X Z), φZ).
By applying Proposition 3.1(2) and the structure equations of M̃ , we get
g(DX σ(φX, Z), φZ) = Xg(σ(φX, Z), φZ) − g(σ(φX, Z), DX φZ)
˜ X Z)
= Xg(∇Z X, Z) − g(σ(φX, Z), φ∇
= X((X ln f )g(Z, Z)) − (X ln f )g(σ(φX, Z)φZ)
− g(σ(φX, Z), φσν (X, Z))
= (X 2 ln f )g(Z, Z) + (X ln f )2 g(Z, Z) − |σν (X, Z)|2 ,
where σν (X, Z) denotes the ν-component of σ(X, Z). Also, we have

˜ Z ∇X φX, φZ)
g(σ(∇X φX, Z), φZ) = g(∇
= g(∇ ˜ X φX, φZ) − g(∇
˜ Z∇ ˜ Z σ(X, φX), φZ)
= −g(X, X)g(Z, Z) + ((∇X X) ln f )g(Z, Z),
g(σ(φX, ∇X Z), φZ) = (X ln f )g(σ(φX, Z), φZ)
= (X ln f )2 g(Z, Z).
Thus, after substituting the above relations into (17.40), we obtain
R̃(X, φX; Z, φZ) = 2|σν (X, Z)|2 − (X 2 ln f )g(Z, Z)
+ ((∇X X) ln f )g(Z, Z) − 2g(X, X)g(Z, Z) (17.41)
2
− ((φX) ln f )g(Z, Z) + ((∇φX φX) ln f )g(Z, Z).
By summing (17.41), we find
β
2α X
X αβ(c + 3)
|σν (Xi , Zt )|2 = − β∆(ln f ). (17.42)
i=0 t=1
2
After combining (17.39) and (17.42), we obtain inequality (17.36).
Let σ ′′ denote the second fundamental form of N ⊥ in M . Then we get
σ ′′ (Z, W ) = −g(Z, W )∇(ln f ). (17.43)
If the equality sign of (17.36) holds identically, then we have
σ(D, D) = {0}, σ(D⊥ , D⊥ ) = {0}, (17.44)
which shows that M is minimal in M̃ (c).
Since N T is totally geodesic in M , the first condition in (17.44) implies
that N T is totally geodesic in M̃ (c). So, N T is a Sasakian space form of
constant φ-sectional curvature c.
The second condition in (17.44) and (17.43) imply that N ⊥ is totally
umbilical in M̃ (c). Hence, by applying Gauss’ equation, we conclude that
N ⊥ is a real space form of constant curvature ǫ ≥ (c + 3)/4. Finally, we
see from Proposition 3.2 that ǫ = (c + 3)/4 holds identically if and only if
the warping function f is constant.
The next two corollaries from [Mihai (2004)] are easy consequences of
Theorem 17.6.
Corollary 17.4. For every contact CR-warped product N T ×f N ⊥ in the

Sasakian S 2m+1 (1) with compact N T and every q ∈ N ⊥ , we have
Z
||σ||2 dVT ≥ 2β(2α + 1)vol(N T ), (17.45)
N T ×{q}
where dVT and vol(N T ) are the volume element and the volume of N T ,
respectively.
The equality sign of (17.45) holds identically if and only if
(a) The warping function f is constant;
(b) (N T , gT ) is isometric to S 2α+1 (1) and it is isometrically immersed in
S 2m+1 as a totally geodesic invariant submanifold, and
(c) (N ⊥ , f 2 g⊥ ) is isometric to an open portion of the sphere S β (1) of
constant sectional curvature one and it is isometrically immersed in
S 2m+1 (1) as a totally geodesic C-totally real submanifold.
Proof. Since N T is compact, (17.36) and Hopf’s lemma imply
Z Z
||σ||2 dVT ≥ 2β |∇(ln f )|2 dVT
N T ×{q} NT (17.46)
+ 2β(2α + 1)vol(N T )
for each q ∈ N ⊥ . Obviously, inequality (17.46) implies (17.45). Clearly,
the equality sign of (17.45) holds if and only if we have
(1) f is constant and

(2) the equality sign of (17.36) holds identically.
Consequently, we have statements (a), (b) and (c).

The converse is clear.
Corollary 17.5. Let N T ×f N ⊥ be contact CR-warped product in the
Sasakian S 2m+1 (1) with compact N T . If f is non-constant, then for each
q ∈ N ⊥ we have
Z Z
||σ||2 dVT ≥ 2βλ1 (ln f )2 dVT + 2β(2α + 1)vol(N T ), (17.47)
N T ×{q} NT
where λ1 is the first positive eigenvalue of the Laplacian on N T .
Moreover, the equality sign of (17.47) holds identically if and only if we
have
(a) ∆ ln f = λ1 ln f and
(b) the contact CR-warped product is N T -totally geodesic and N ⊥ -totally
geodesic.
Proof. Assume that the warping function f is non-constant. Then the
minimum principle on λ1 yields
Z Z
2
||∇ ln f || dVT ≥ λ1 (ln f )2 dVT , (17.48)
NT NT
with the equality holding if and only if ∆ ln f = λ1 ln f . Thus, by combining

(17.46) and (17.48), we obtain (17.47).
Clearly, the equality sign of (17.47) holds identically if and only if we
have statements (a) and (b).
17.6 Pointwise bi-slant and hemi-slant warped products in

Sasakian manifolds
Let M be a submanifold of a Sasakian manifold M̃ . For a given nonzero

vector X ∈ Tx M linearly independent from ξ, we denote by θ(X) the angle
between φX and P X, where P X is defined by (17.8).
Pointwise slant and slant submanifolds of a Sasakian manifold were
defined in [Lotta (1996)] as follows.
Definition 17.6. A submanifold M of a Sasakian manifold M̃ is called

pointwise slant, if, for each point x ∈ M , the angle θ(X) is independent of
the choice of X ∈ Tx M − Span{ξ} at x.
The function θ on a pointwise slant submanifold M is called the slant
function. If the slant function is a constant function, then M is called a
slant submanifold with slant angle θ.
Obviously, if θ ≡ 0 on a pointwise slant submanifold M , then M is an

invariant submanifold. Similarly, if θ ≡ π2 on M , then M is an anti-invariant
submanifold of M̃ .
Lemma 17.2. Let M be a submanifold of a Sasakian manifold M̃ tangent

to ξ. Then M is pointwise slant if and only if there exists a function
λ : M → [0, 1] such that
P 2 = λ(−I + η ⊗ ξ). (17.49)
Proof. Suppose that there exists a function λ : M → [0, 1] such that

(17.49) holds. Then, for any X ∈ T M − Span{ξ}, we have
g(φX, P X) g(X, P 2 X) g(X, φ2 X) |φX|
cos θ(X) = =− = −λ =λ .
|φX||P X| |φX||P X| |φX||P X| |P X|
On the other hand, by using cos θ(X) = |P X|/|φX| and the above
equality, we find cos2 θ(X) = λ. Thus θ(X) is independent of the choice of
X for each given point p ∈ M . Thus M is pointwise slant.
The converse is clear.
For a pointwise slant submanifold M , Lemma 17.2 implies

g(P X, P Y ) = cos2 θ{g(X, Y ) − η(X)η(Y )}, (17.50)
2
g(F X, F Y ) = sin θ{g(X, Y ) − η(X)η(Y )}, (17.51)
for X, Y ∈ T M , where θ is the slant function.
Definition 17.7. A submanifold M of a Sasakian manifold M̃ is called

pointwise bi-slant if there exist two pointwise slant distributions Dθ1 and
Dθ2 with slant functions θ1 and θ2 , respectively, such that T M admits the
following orthogonal direct decomposition:
T M = Dθ1 ⊕ Dθ2 ⊕ Span{ξ}. (17.52)
A pointwise bi-slant submanifold M is called pointwise semi-slant if
θ1 ≡ 0 on M , i.e., Dθ1 is a φ-invariant distribution. Similarly, a pointwise
bi-slant submanifold is called pointwise hemi-slant if θ1 ≡ π2 on M , i.e.,
Dθ1 is an anti-invariant distribution. A pointwise bi-slant submanifold M
is called semi-invariant if θ1 ≡ 0 and θ2 ≡ π2 on M , i.e., Dθ1 is φ-invariant
and Dθ2 is anti-invariant [Bejancu and Papaghiuc (1981)].
Remark 17.3. In Definition 17.7, if we assume that both slant functions θ1

and θ2 are constant, then we obtain the corresponding notions of bi-slant,
semi-slant and hemi-slant submanifolds.
Example 17.3. For any θ1 , θ2 ∈ [0, π2 ], the map

φ(u, v, w, s, t) = 2 u, 0, w, 0, v cos θ1 , v sin θ1 , s cos θ2 , s sin θ2 , t
defines a bi-slant submanifold with bi-slant angles θ1 , θ2 in R9 (−3).
For semi-slant submanifolds of a Sasakian manifold, we have the follow-

ing two propositions from [Cabrerizo et al. (1999)].
Proposition 17.7. If M is a semi-slant submanifold of a Sasakian mani-

fold M̃ such that dim Dθ1 6= 0, θ1 = 0, then the invariant distribution Dθ1
is always non-integrable.
Proof. Follows from the fact that g([X, φX], ξ) 6= 0 for X ∈ Dθ1 .
Given a semi-slant submanifold M in a Sasakian manifold M̃ , we denote
by Pi the projection on the distribution Dθi for i = 1, 2. Let us put
Ti = Pi ◦ P,
where P is defined by (17.8). Then we have
φX = φP1 X + P P2 X + F P2 X, X ∈ T M.
Proposition 17.8. Let M be a semi-slant submanifold of a Sasakian man-

ifold M̃ . Then the slant distribution Dθ2 is integrable if and only if θ2 ≡ π2 ,
i.e., M is a semi-invariant submanifold.
Proof. For any X, Y ∈ Dθ2 , it is easy to see that

g([X, Y ], ξ) = 2g(Y, T2 X).
Thus, if D is integrable, then T2 ≡ 0 and so, θ2 ≡ π2 . Therefore M is a
θ2
semi-invariant submanifold. The converse is easy to verify.

Pointwise bi-slant and hemi-slant warped product submanifolds of a
Sasakian manifold are defined as follows.
Definition 17.8. A submanifold M of a Sasakian manifold M̃ is called a

pointwise bi-slant warped product if M is the warped product N1θ1 ×f N2θ2
of two pointwise slant submanifolds of M̃ with slant functions θ1 and θ2 ,
respectively.
In particular, if either θ1 ≡ π2 or θ2 ≡ π2 , then M is called a pointwise
hemi-slant warped product.
Remark 17.4. In Definition 17.8, if both slant functions θ1 and θ2 are

constant, then we obtain the corresponding notions of bi-slant, semi-slant
and hemi-slant warped products.
By using Propositions 17.1, 17.2 and 17.3 we have the following.
Proposition 17.9. Let M = N1θ1 ×f N2θ2 be a pointwise bi-slant warped

product in a Sasakian manifold M̃ .
(1) If the structure vector field ξ of M̃ is normal to M , then θ1 ≡ θ2 ≡ π2 .

(2) If ξ is tangent to N1 , then θ2 ≡ π2 . Thus N1θ is pointwise slant and N2⊥
is anti-invariant.
(3) If ξ is tangent to N2 , then f is constant and θ1 ≡ π2 . Therefore, M
is a Riemannian product of an anti-invariant submanifold N1⊥ and a
pointwise slant submanifold N2θ with the metric gθ = f 2 gN θ2 .
2
In cases (2) and (3), M is a pointwise hemi-slant warped product.
Remark 17.5. The bibliography of this book contains many articles on

warped product submanifolds in various contact metric manifolds other
than Sasakian manifolds. However these topics will not be treated in the
present book.
Chapter 18
Warped Products in Affine Spaces
18.1 Affine spaces and hypersurfaces
Let Rm be the standard real vector m-space. For an ordered pair (p, q)
of points in Rm , let −→
pq denote the vector q − p ∈ Rm . The vector space
Rm together with the mapping Rm × Rm → Rm : (p, q) 7→ − →
pq is called an
affine m-space.
Let x(t), a ≤ t ≤ b, be a smooth curve in Rm and let Y be a vector field
on Rm . If we take an affine coordinate system {x1 , . . . , xm } and write
m
X ∂
Y = Yi and x(t) = (x1 (t), . . . , xm (t)),
i=1
∂xi
then the covariant derivative Dt Y of Y along the curve x(t) is defined by

Xm Xm
dY i (xi ) ∂ ∂Y i dxj ∂
Dt Y = = .
i=1
dt ∂xi i,j=1 ∂xj dt ∂xi
For X ∈ Txo Rm , ∇X Y is defined by ∇X Y = (Dt Y )t , where x(t) is a

curve with initial point xo and initial tangent vector X. It is direct to verify
that, for f ∈ F (Rm ) and X, Y ∈ X(Rm ), we have
(1) ∇f X Y = f ∇X Y ,
(2) ∇X (f Y ) = X(f )Y + f ∇X Y .
Thus ∇ is an affine connection.

Consider an affine m-space Rm with the affine connection ∇ as above.
Then the curvature tensor vanishes. Thus ∇ is a flat affine connection.
Definition 18.1. An immersion φ : N → Rn+k of an n-manifold N into

Rn+k is called an affine immersion if there exists a rank-k distribution E
427
on N with Ex ⊂ Tφ(x) Rn+k for each x ∈ N such that
Tφ(x) Rn+k = φ∗ (Tx (M )) ⊕ Ex (direct sum), (18.1)

(∇X (φ∗ (Y )))x = φ∗x (∇′X Y ) + αx (X, Y ), (18.2)
for X, Y ∈ X(N ), where αx (X, Y ) is the Ex -component of (∇X φ∗ (Y ))x

according to the direct sum (18.1). And (18.2) decomposes ∇X φ∗ (Y ) into
the tangential component φ∗x (∇′X Y ) and the component αx (X, Y ) in Ex
according to (18.2).
By an affine hypersurface N in Rn+1 we mean a codimension-one affine

immersion φ : N → Rn+1 with a transversal vector field ξ, i.e., ξ is a vector
field which is not tangent to N at each point on N .
The formulas of Gauss and Weingarten for an affine hypersurface φ :
N → Rn+1 are given respectively by
∇X φ∗ (Y ) = φ∗ (∇′X Y ) + h(X, Y )ξ, (18.3)

∇X ξ = −φ∗ (SX) + τ (X)ξ, (18.4)
for X, Y ∈ X(N ), where φ∗ (∇′X Y ) and −φ∗ (SX) are the tangential com-
ponents of ∇X φ∗ (Y ) and ∇X ξ, respectively.
The induced affine connection ∇′ , the (1, 1)-tensor S, the 1-form τ , and
the symmetric (0, 2)-tensor h are called the induced affine connection, the
affine shape operator, the torsion form, and the affine fundamental form
(relative to the transversal vector field ξ), respectively.
The transversal vector field ξ of φ : N → Rn+1 is called a relative
normalization (or equiaffine) if τ = 0 holds identically, i.e., ∇X ξ is tangent
to N for each X ∈ T N .
Notice that ∇′ and h depend upon the choice of transverse vector field
ξ. We consider only those hypersurfaces for which h is non-degenerate.
Interestingly, this is a property of the hypersurface N and not depend upon
the choice of transverse vector field ξ. When h is non-degenerate, it defines
a pseudo-Riemannian metric on N , called the relative metric. Throughout
this chapter, we assume that the affine fundamental form h is definite.
For tangent vectors X1 , . . . , Xn of N let (hij ) be the n × n matrix given
by hij = h(Xi , Xj ). Define an affine volume element v on N by
v(X1 , . . . , Xn ) = det(hij )1/2 .
Since h depends on the choice of transverse vector field ξ, it follows that v

does too.
Warped Products in Affine Spaces 429
ˆ K̂, R̂ and κ̂ denote the Levi-Civita connection, the sectional

Let ∇,
curvature, the curvature tensor and the normalized scalar curvature of h,
respectively.
The difference tensor K is the symmetric (1, 2)-tensor field defined by
ˆ X Y.
KX Y = K(X, Y ) = ∇X Y − ∇ (18.5)
The Tchebychev form T , the Tchebychev vector field T # , the Pick invariant
J, and the relative theorema egregium are given by
1
T (X) = Tr { Y → K(X, Y )}, (18.6)
n
h(T # , X) = T (X), (18.7)
h(C, C) = 4h(K, K) = 4n(n − 1)J, (18.8)
n
κ̂ = J + H − h(T # , T # ), (18.9)
n−1
1
where H = n Tr S is the relative mean curvature.
18.2 Centroaffine hypersurfaces
Let φ : N → Rn+1 be a non-degenerate hypersurface whose position vector

field is nowhere tangent to N . Then φ can be regarded as a transversal
field along itself. We call ξ = −φ the centroaffine normal. The φ together
with this normalization is called a centroaffine hypersurface.
The centroaffine structure equations are given by
∇X φ∗ (Y ) = φ∗ (∇′X Y ) + h(X, Y )ξ, (18.10)
∇X ξ = −φ∗ (X), (18.11)
where ∇ is the canonical flat connection of Rn+1 , the induced centroaffine
connection ∇′ is a torsion-free connection, and h is a non-degenerate sym-
metric (0, 2)-tensor field, called the centroaffine metric.
The corresponding equations of Gauss and Codazzi are given by
R(X, Y )Z = h(Y, Z)X − h(X, Z)Y, (18.12)
(∇′X h)(Y, Z) = (∇′Y h)(X, Z). (18.13)
Moreover, we have
(∇′X h)(Y, Z) = −2h(KX Y, Z). (18.14)
From now on we assume that the centroaffine hypersurface is definite, i.e.,
h is definite, so h defines a pseudo-Riemannian metric on N . In order to
consider only a positive-definite metric, we now make the following changes:

if h is negative definite, we introduce a transversal vector field ξ = −φ and
a (0, 2)-tensor h is replaced by −h. In both cases, (18.10) and (18.13) hold,
whereas (18.11) and (18.12) change sign. In case ξ = −φ, we call N positive
definite; in case ξ = φ, we call N negative definite.
It is known that the centroaffine metric h is definite if and only if the
hypersurface is locally strongly convex.
We use the following terminology:
(i) The centroaffine hypersurface N is said to be of elliptic type if, for any
point φ(x) ∈ Rn+1 with x ∈ N , the origin of Rn+1 and the hypersurface
are on the same side of the tangent hyperplane φ∗ (Tx N ); in this case
the centroaffine normal vector field is given by ξ = −φ.
(ii) The centroaffine hypersurface N is said to be of hyperbolic type if, for
any point φ(x) ∈ Rn+1 , the origin of Rn+1 and the hypersurface are
on the different side of the tangent hyperplane φ∗ (Tx N ); in this case
the centroaffine normal vector field is given by ξ = φ.
Consider a hypersurface N of the equiaffine space. If the Tchebychev
form T vanishes on N , then N is called a proper affine hypersphere centered
at the origin. If the difference tensor K vanishes, then N is a quadric,
centered at the origin. In particular, an ellipsoid if N is positive definite;
and a two-sheeted hyperboloid if N is negative definite. For centroaffine
geometry one has
h(KX Y, Z) = h(Y, KX Z), (18.15)
R̂(X, Y )Z = KY KX Z − KX KY Z (18.16)
+ ε(h(Y, Z)X − h(X, Z)Y ),
ˆ
(∇K)(X, ˆ
Y, Z) = (∇K)(Y, ˆ
Z, X) = (∇K)(Z, X, Y ), (18.17)
n
κ̂ = J + ε − h(T #, T # ), (18.18)
n−1
where ε = 1 or −1 according to N is of elliptic or of hyperbolic type.
Equation (18.15) implies that KX is self-adjoint with respect to h.
18.3 Graph hypersurfaces
An affine hypersurface φ : N → Rn+1 is called a graph hypersurface if the

transversal vector field ξ is a constant vector field. A result of [Nomizu
and Pinkall (1987)] states that N locally is affine equivalent to the graph
immersion of some function F . In case that h is non-degenerate, it defines

a pseudo-Riemannian metric, called the Calabi metric.
For graph hypersurfaces, the following equations hold as well:
∇X φ∗ (Y ) = φ∗ (∇′X Y ) + h(X, Y )ξ, (18.19)
(∇′X h)(Y, Z) = (∇′Y h)(X, Z), (18.20)
h(KX Y, Z) = h(Y, KX Z), (18.21)
R̂(X, Y )Z = KY KX Z − KX KY Z, (18.22)
(∇′X h)(Y, Z) = −2h(KX Y, Z). (18.23)
However, equations (18.11), (18.12), (18.16) and (18.18) shall be replaced
respectively by
∇X ξ = 0, (18.24)
R(X, Y )Z = 0, (18.25)
R̂(X, Y )Z = KY KX Z − KX KY Z, (18.26)
n
κ̂ = J − h(T # , T #). (18.27)
n−1
If the Tchebychev form T = 0, then the graph hypersurface is called an
improper affine hypersphere.
If we choose ξ = (0, . . . , 0, 1), then we can assume that locally N is
given by xn+1 = F (x1 , . . . , xn ). It turns out that (x1 , . . . , xn ) are ∇′ -flat
coordinates on N and that the Calabi metric is given by

∂ ∂ ∂2F
h , = . (18.28)
∂xi ∂xj ∂xi ∂xj
Furthermore, N is an improper
affine sphere if and only if the Hessian
∂2 F
determinant det ∂xi ∂xj is constant.
Let N1 and N2 be two improper affine hyperspheres in Rp+1 and Rq+1
defined respectively by the equations
xp+1 = F1 (x1 , . . . , xp ) and yq+1 = F2 (y1 , . . . , yq ).
Then one can define a new improper affine hypersphere N in Rp+q+1 given
by
z = F1 (x1 , . . . , xp ) + F2 (y1 , . . . , yq ), (18.29)
p+q+1
where (x1 , . . . , xp , y1 , . . . , yq , z) are the coordinates on R . The affine
normal of N is (0, . . . , 0, 1). Clearly, the Calabi metric on N is the product
metric. Following [Dillen and Vrancken (1994)] we call this composition the
Calabi composition of N1 and N2 .
18.4 A realization problem for affine hypersurfaces
For a Riemannian n-manifold (N, g) equipped with Levi-Civita connec-

tion ∇′ , Cartan and Norden studied non-degenerate affine immersions
ψ : N → Rn+1 with a transversal vector field ξ such that the induced
affine connection is exactly the Levi-Civita connection ∇′ . The Cartan-
Norden theorem states that if ψ is such an affine immersion, then either
∇′ is flat and ψ is a graph immersion or ∇′ is not flat and Rn+1 admits a
parallel Riemannian metric relative to which ψ is an isometric immersion
and ξ is perpendicular to ψ(N ) (cf. [Nomizu and Sasaki (1994)]).
We present another realization problem from a different view point,
namely, we consider the following [Chen (2005c)].
Realization Problem: Which Riemannian manifolds (N, g) can be im-

mersed as affine hypersurfaces in an affine space such that the induced affine
metric h on N is exactly the given Riemannian metric g?
Definition 18.2. We say that a warped product N1 ×f N2 can be realized
as an affine hypersurface if there exists a codimension one affine immersion
from N1 ×f N2 into some affine space in such a way that the induced
fundamental form h is exactly the warped product metric of N1 ×f N2 .
As before, let S n (a2 ), H n (−a2 ), and En denote the n-sphere of constant
curvature a2 , the hyperbolic n-space of constant curvature −a2 , and the
Euclidean n-space, respectively.
The following results show that there exist many warped product which
can be realized either as graph or as centroaffine hypersurfaces.
Proposition 18.1. [Chen (2005c)] Let f be a positive function defined on
an open interval I ⊂ R. Then we have:
(a) Every warped product I ×f R can be realized as a graph surface in an
affine 3-space R3 .
(b) For each integer n > 2, the warped product I ×f H n−1 (−a2 ) can be
realized as a graph hypersurface in Rn+1 .
(c) If f ′ (s) 6= 0 on I, then the warped product I ×f En−1 , n > 2, can be
realized as a graph hypersurface in Rn+1 .
(d) If f ′ (s)2 > a2 on I for some positive number a, then the warped product
I ×f S n−1 (a2 ), n > 2, can be realized as a graph hypersurface in Rn+1 .
Proof. Without loss of generality we assume that 0 ∈ I. Consider the
warped product surface I ×f R with warped product metric ds2 + f 2 (s)dt2 .
If we put
(Z
p )
s
f ′2 + 1
ψ(s, t) = f (s) exp ds sinh t,
0 f
(Z p ) Z s ! (18.30)
s
f ′2 + 1 f ds
f (s) exp ds cosh t, p ,
0 f 0 f′ − f ′2 + 1
then ψ satisfies
p
f f ′′ + f ′2 + 1 f ′ + f ′2 + 1
ψss = p ψs + ξ, ψst = ψt ,
f f ′2 + 1 f
p
ψtt = f f ′2 + 1 − f ′ ψs + f 2 ξ, ξ = (0, 0, 1).
Thus ψ is a realization of I ×f R as a graph surface in R3 with the metric
ds2 + f 2 (s)dt2 as the Calabi metric h and ξ as the Calabi normal.
−
Next, assume that a is a positive real number. Let Na,f be the warped
n−1 2
product manifold I ×f H (−a ) with the warped product metric
 
n−1
Y
2 2
ds + f (s) du2 + cosh (au2 )du3 + · · · +
2 2 2 2
cosh (auj )dun  . (18.31)
2
j=2
−
Consider the immersion φ of Na,f into Rn+1 given by
(Z p )
s
f ′2 + a2
φ(s, u1 , . . . , un ) = f (s) exp ds sinh(au2 ),
0 f
n−1 n
!
Y Y
sinh(au3 ) cosh(au2 ), . . . , sinh(aun ) cosh(auj ), cosh(auj ), 0
j=2 j=2
Z s p
1 ′

− 2 0, . . . , 0, f ′2 2
f + a + f ds .
a 0
Then a direct computation yields
f f ′′ + f ′2 + a2
φss = p φs + ξ, ξ = (0, . . . , 0, 1),
f f ′2 + a2
p
f ′ + f ′2 + a2
φsuj = φuj ,
f
φui uj = a tanh(aui )φuj , 2 ≤ i < j ≤ n,
(18.32)
p j−1
Y
φuj uj ′2 2 ′ 2
= f f + a φs − f f φs + f ξ cosh2 (aui )
i=2
j−1 j−1
!
X sinh(2auk ) Y 2
−a cosh (aui ) φuk ,
2
k=2 i=k+1
for j = 2, . . . , n. It follows from (18.32) that φ is a graph hypersurface whose

−
Calabi metric is given by (18.31). Therefore φ is a realization of Na,f as a
graph hypersurface with the metric (18.31) as the Calabi metric h and ξ as
the Calabi normal.
If f ′ (s) 6= 0, let Nfo denote I ×f En−1 with the warped product metric:

ds2 + f 2 (s) du22 + du23 + · · · + du2n . (18.33)
Consider the immersion φo of Nfo into Rn+1 defined by
 
Z s Xn Z s
ds f (s) 
φo (s, u2 , . . . , un ) = u2 , . . . , un , ′ (s)
− u2j , ′ (s)
ds .
0 f (s)f j=2 0 f
Then a straightforward computation yields
f ′2 + f f ′′ o
φoss = − φs + ξ, φosuj = φoui uj = 0,
ff′ (18.34)
φouj uj = −2f f ′φos 2
+ f ξ, 2 ≤ i 6= j ≤ n,
where ξ = (0, . . . , 0, 2) is a transversal vector field. It follows from (18.34)
that φo is a graph hypersurface whose Calabi metric is (18.33). Hence φo
is a realization of I ×f En−1 as a graph hypersurface with (18.33) as the
Calabi metric h and ξ as the Calabi normal.
+
Finally, assume that f satisfies f ′ (s)2 > a2 with a > 0. Let Na,f denote
n−1 2
I ×f S (a ) equipped with the warped product metric
n−1
!
Y
2 2 2 2 2 2 2
ds + f (s) du2 + cos (au2 )du3 + · · · + cos (auj )dun . (18.35)
j=2
+
Consider the immersion φ+ of Na,f into Rn+1 defined by
(Z p )
s ′2 − a2
f
φ+ (s, u2 , . . . , un ) = f (s) exp ds ×
0 f
n−1 n
!
Y Y
sin(au2 ), . . . , sin(aun ) cos(auj ), cos(auj ), 0 (18.36)
j=2 j=2
Z !
s
f ds
− p
0, . . . , 0, .
f ′2 − a2 − f ′
0

f f ′′ + f ′2 − a2 +
φ+ss = p φs + ξ,
f f ′2 − a2
p
+ f ′ + f ′2 − a2 +
φsuj = φuj , j = 2, . . . , n, (18.37)
f
φ+ +
ui uj = −a tan(aui )φuj , 2 ≤ i < j ≤ n,
p j−1
Y
φ+
uj uj
′2 2 + ′ + 2
= f f − a φs − f f φs + f ξ cos2 (aui )
i=2
j−1
X j−1
sin(2auk ) Y
+a cos2 (aui )φ+
uk , j = 2, . . . , n,
2
k=2 i=k+1
where ξ = (0, . . . , 0, 1) is a transversal vector field. It follows from (18.37)

that φ+ is a graph hypersurface whose Calabi metric is (18.35). Hence, φ+
+
is a realization of Na,f as a graph hypersurface with the warped product
metric (18.35) as the Calabi metric h and ξ as the Calabi normal.
Remark 18.1. If the warping function f of I ×f (s) En−1 satisfies f ′ (s) = 0

on I, then f is a positive real number, say c. In this case, I ×f En−1 can
be realized as a graph hypersurface in Rn+1 by the following immersion
n
!
s2 c2 X 2
φ(s, u2 , . . . , un ) = s, u2 , . . . , un , + u .
2 2 j=2 j
Proposition 18.2. [Chen (2005c)] Let n > 2 and f = f (s) be a positive

function defined on an open interval I and a a positive number. Then
(a) if f ′ (s)2 > f 2 (s) − a2 , then I ×f H n−1 (−a2 ) can be realized as a cen-
troaffine hypersurface in Rn+1 ;
(b) if f ′ (s)2 > f (s)2 , then I ×f En−1 can be realized as a centroaffine
hypersurface in Rn+1 ;
(c) if f ′ (s)2 > f (s)2 + a2 , then I ×f S n−1 (a2 ) can be realized as a graph
hypersurface in Rn+1 .
Proof. Without loss of generality, we may assume that the open interval
− +
I contains 0. Let Na,f , Nfo and Na,f be the warped products defined as
before. Suppose that f satisfies f (s)2 > f 2 (s)−a2 . Consider the immersion
′
−
η − of Na,f into Rn+1 given by
(Z p )
s
− f ′2 − f 2 + a2
η (s, u2 , . . . , un ) = f (s) exp ds
0 f
n−1 n
!
Y Y
× sinh(au2 ), . . . , sinh(aun ) cosh(auj ), cosh(auj ), 0
j=2 j=2
Z !!
s
f ds
+ 0, . . . , 0, exp p .
0 f′ − f ′2 − f 2 + a2
Then by a straightforward computation we find
− f ′2 + f f ′′ − 2f 2 + a2 −
ηss = p ηs + η − ,
f f ′2 − f 2 + a2
p
− f ′ + f ′2 − f 2 + a2 −
ηsuj = ηuj , j = 2, . . . , n,
f
ηu−i uj = a tanh(aui )ηu−j , 2 ≤ i < j ≤ n,
(18.38)
p X
j−1
− ′2 2 2 − ′ − 2 −
ηuj uj = f f − f + a ηs − f f ηs + f η cosh2 (aui )
i=2
j−1 j−1
!
X sinh(2auk ) Y
−a cosh2 (aui ) φ−
uk , j = 2, . . . , n,
2
k=2 i=k+1
−
which implies that η is a centroaffine hypersurface with (18.31) as the
−
centroaffine metric. Hence η − is a realization of Na,f as a centroaffine
hypersurface with (18.31) as the centroaffine metric.
If f satisfies f ′ (s)2 > f 2 (s), we consider the immersion
Z s p ′2 !
f − f2
η(s, u2 , . . . , un ) = f (s) exp ds 1, u2 , . . . , un , 0
0 f
Z s !!
f ds
+ 0, . . . , 0, exp p .
0 f ′ − f ′2 − f 2
A direct computation shows that η satisfies (18.38) with a = 0. Hence η
defines a centroaffine hypersurface with (18.33) as its centroaffine metric.
Thus η is a realization of Nfo as a centroaffine hypersurface with the warped
product metric as its centroaffine metric.
If f ′ (s)2 > f 2 (s) + a2 , we consider
(Z p )
s
+ f ′2 − f 2 − a2
η (s, u2 , . . . , un ) = f (s) exp ds
0 f
 
n−1
Y Y n
× sin(au2 ), . . . , sin(aun ) cos(auj ), cos(auj ), 0
j=2 j=2
Z !!
s
f ds
+ 0, . . . , 0, exp p .
0 f′ − f ′2 − f 2 − a2
+
Then η + is a realization of Na,f as a centroaffine hypersurface with the
warped product metric as its centroaffine metric.
18.5 Warped products as centroaffine hypersurfaces
For warped products realized as centroaffine hypersurfaces we have the

following result from [Chen (2005c)].
Theorem 18.1. Let N1 ×f N2 be the warped product of two Riemannian

manifolds N1 and N2 . If N1 ×f N2 is realized as a centroaffine hypersurface
in Rn+1 , then the warping function satisfies
∆f (n1 + n2 )2
≥ n1 ε − h(T # , T # ), (18.39)
f 4n2
where n = n1 + n2 , ni = dim Ni , i = 1, 2, ∆ is the Laplacian of N1 ,
and ε = 1 or −1 according to the centroaffine hypersurface is elliptic or
hyperbolic.
Proof. Let N = N1 ×f N2 be a warped product manifold equipped with

the warped product metric h = h1 + f 2 h2 , where h1 , h2 are the metrics on
N1 , N2 , respectively.
Assume that (N, h) is realized via φ : N1 ×f N2 → Rn+1 as a locally
strongly convex centroaffine hypersurface in Rn+1 with h as the centroaffine
metric. Denote by ∇, ˆ K̂, etc. the Levi-Civita connection, sectional curva-
ture, etc. of (N, h) as in section 17.1.
Since N1 ×f N2 is a warped product, we have
∇ˆ XZ = ∇ ˆ Z X = (X ln f )Z (18.40)
for h-unit vector fields X, Z tangent to N1 , N2 , respectively. Hence

K̂(X ∧ Z) = h(∇ ˆ Z∇ˆ XX − ∇ˆ X∇ˆ Z X, Z) = 1 (∇ ˆ X X)f − X 2 f . (18.41)
f
For simplicity, we use the following ranges of indices:
1 ≤ i, j, k ≤ n1 ; n1 + 1 ≤ s, t ≤ n; 1 ≤ α, β, γ ≤ n,
unless mentioned otherwise. If we choose a local h-orthonormal frame
{e1 , . . . , en } such that e1 , . . . , en1 and en1 +1 , . . . , en are tangent to N1 and
N2 , respectively, then for each s ∈ {n1 + 1, . . . , n} we have
X n1
∆f
= K̂(ej ∧ es ), (18.42)
f j=1
where K̂(ej ∧ es ) denotes the sectional curvature of the plane section

spanned by ej and es . Let us put
n2
δ= h(T # , T #) + n(n − 1)κ̂ − 2n1 n2 ε. (18.43)
2
Then (18.18) and (18.43) imply that

n2
h(T # , T #) = h(K, K) − δ + n(n − 1) − 2n1 n2 ε. (18.44)
2
At a point p ∈ N , we choose a h-orthonormal basis e∗1 , . . . , e∗n of Tx N such
that e∗1 is in the direction of the Tchebychev vector field T # at p (when
T # = 0 at x, we may choose e∗1 , . . . , e∗n to be an arbitrary h-orthonormal
basis).
γ∗
If we put Kαβ = h(K(eα , eβ ), e∗γ ), then (18.44) becomes
Xn X Xn X
n2
1∗ 2

1∗ 2 r∗ 2
h(T # , T #) − Kαα = Kαβ + (Kαβ )
2 α=1 r=2 α,β (18.45)
α6=β

+ n(n − 1) − 2n1 n2 ε − δ.
Hence we can apply Corollary 14.1 to the left-hand side of (18.45) to obtain
X ∗ ∗ X ∗ ∗ X n n

1∗ 2 1 X X r∗ 2
Kii1 Kjj
1
+ 1
Kss 1
Ktt ≥ Kαβ + (Kαβ )
s<t
2 r=2
i<j α<β α,β (18.46)
n(n − 1) − 2n1 n2 δ
+ ε− ,
2 2
with equality holding if and only if
∗ ∗ ∗ ∗
1
K11 + · · · + Kn11 n1 = Kn11 +1,n1 +1 + · · · + Knn
1
. (18.47)
K̂(eα ∧ eβ ) = ε − h(K(eα , eα ), K(eβ , eβ )) + h(K(eα , eβ ), K(eα , eβ )).
Combining with (18.42) gives
n2 ∆f n(n − 1) X X
= κ̂ − K̂(ej ∧ ek ) − K̂(es ∧ et )
f 2 s<t
j<k
n(n − 1) n1 (n1 − 1) XX
α∗ α∗ α∗ 2
= κ̂ − ε+ Kjj Kkk − (Kjk )
2 2 α j<k
n2 (n2 − 1) XX
α∗ α∗ α∗ 2
− ε+ Kss Ktt − (Kst )
2 α s<t
n X (18.48)
n(n − 1) X
r∗ r∗ r∗ 2
= (κ̂ − ε) + n1 n2 ε + Kjj Kkk − (Kjk )
2 r=2 j<k
n X
X
r∗ r∗ r∗ 2
X 1∗ 1∗
+ Kss Ktt − (Kst ) + Kjj Kkk
r=2 s<t j<k
X X X
1∗ 1∗ 1∗ 2 ∗
1 2
+ Kss Ktt − (Kjk ) − (Kst ) .
s<t j<k s<t
Therefore (18.46) and (18.48) yield

Xn X
n2 ∆f n(n − 1) r∗ r∗ r∗ 2
X 1∗ 2
≥ κ̂ + Kjj Kkk − (Kjk ) + (Kjt )
f 2 r=2 j,t
j<k
n n X
1 XX r 2∗ X
r∗ r∗ r∗ 2
δ
+ (Kαβ ) + Kss Ktt − (Kst ) −
2 r=2 α,β r=2 s<t
2
n X
X n
n(n − 1) r∗ 2 1 X X r∗ 2 (18.49)
= κ̂ + (Kjt ) + Kjj
2 r=1 j,t
2 r=2 j
n X
X 2
1 r ∗ δ
+ Ktt −
2 r=2 t
2
n(n − 1) δ
≥ κ̂ − .
2 2
Thus, by combining (18.43) and (18.49), we find
n2 ∆f n2
≥ n1 n2 ε − h(T # , T #),
f 4
Two immediate consequences of Theorem 18.1 are the following results
from [Chen (2005c)].

satisfies ∆f ≤ 0 at some point on N1 , then N1 ×f N2 cannot be realized as
an elliptic proper affine hypersphere in Rn+1 .

satisfies (∆f )/f < − dim N1 at some point on N1 , then N1 ×f N2 cannot
be realized as a hyperbolic proper affine hypersphere in Rn+1 .
If the warping function is harmonic, Corollary 18.5 yields
Corollary 18.3. Every warped product manifold N1 ×f N2 with harmonic

warping function cannot be realized as an elliptic proper affine hypersphere
in Rn+1 .
Another interesting application of Theorem 18.1 is the following.
Corollary 18.4. If N1 is a compact Riemannian manifold, then every

warped product manifold N1 ×f N2 with arbitrary warping function can-
not be realized as an elliptic proper affine hypersphere in Rn+1 .
Proof. Assume N1 is compact and that the warped product N1 ×f N2

can be realized as an elliptic proper affine hypersphere in Rn+1 . Then we
find from Theorem 18.1 that (∆f )/f ≥ n1 = dim N1 .
Since the warping function is positive, ∆f > 0 on N1 . But this is
impossible since N1 is a compact Riemannian manifold according to Hopf’s
lemma.
Theorem 18.2. [Chen (2005c)] Let φ : N1 ×f N2 → Rn+1 be a realization
of a warped product N1 ×f N2 as a centroaffine hypersurface. If the warping
function satisfies the equality case of (18.39) identically, then we have
(a) The Tchebychev vector field T # vanishes identically.

(b) The warping function f is an eigenfunction of the Laplacian ∆ with
eigenvalue n1 ε.
(c) N1 ×f N2 is realized as a proper affine hypersphere centered at the
origin.
Proof. Assume that φ : N1 ×f N2 → Rn+1 is a realization of the warped

product N1 ×f N2 as a centroaffine hypersurface which satisfies the equality
case of (18.39) identically.
Let us choose local h-orthonormal frames e1 , . . . , en and e∗1 , . . . , e∗n in
the same way as in the proof of Theorem 18.1. If we put
L1 = Span{e1 , . . . , en1 }, L2 = Span{en1 +1 , . . . , en },
then it follows from the proof of Theorem 18.1 that all of the inequalities
in (18.46), (18.48) and (18.49) become equalities. Hence we obtain
Xn1 X n
1∗ 1∗
Kjj = Kss , (18.50)
j=1 s=n1
n1
X n
X
∗ ∗ ∗ ∗
1 r r r
Kjs = Kjs = Kjj = Kss =0 (18.51)
j=1 s=n1 +1
for 1 ≤ j ≤ n1 ; n1 + 1 ≤ s ≤ n; 2 ≤ r ≤ n. In particular, (18.50) implies
that K(L1 , L2 ) = {0}. Thus we get
K(L1 , L1 ) ⊂ L1 , K(L2 , L2 ) ⊂ L2
according to (18.16). Hence, by using (18.50), we find
Xn1 Xn
1∗ 1∗
Kjj = Kss = 0.
j=1 s=n1
Consequently, we obtain T # = 0 which implies statements (a) and (c).
Statement (b) follows from easily statement (a) and the equality case
of (18.39).
The following examples illustrate that all of the results obtained in this
section are best possible.
Example 18.1. Let M = N1 ×cos s N2 be the warped product of the open
interval N1 = (−π, π) and an open portion N2 of the unit (n − 1)-sphere
S n−1 (1) equipped with the warped product metric:
n−1
!
Y
2 2 2 2 2 2 2
h = ds + cos s du2 + cos u2 du3 + · · · + cos uj dun . (18.52)
j=2
Consider the immersion of N into the affine (n + 1)-space defined by
n−1 n
!
Y Y
2
sin s, sin u2 cos s, . . . , sin un cos s cos uj , cos s cos uj .
j=2 j=2
Then N is a centroaffine elliptic hypersurface whose centroaffine metric is
(18.52) and it satisfies T # = 0. Moreover, the warping function f = cos s
satisfies (∆f )/f = 1 = εn1 . Hence, this centroaffine hypersurface satisfies
the equality case of (18.39) identically. Consequently, the estimate given
in Theorem 18.1 is optimal for centroaffine elliptic hypersurfaces.
Example 18.2. Let N = R×cosh s H n−1 (−1) be the warped product of the
real line and the unit hyperbolic space H n−1 (−1) equipped with warped
product metric:
n−1
!
Y
2 2 2 2 2 2 2
h = ds + cosh s du2 + cosh u2 du3 + · · · + cosh uj dun . (18.53)
j=2
n−1 n
!
Y Y
sinh s, sinh u2 cosh s, . . . , sinh un cosh s cosh2 uj , cosh s cosh uj .
j=2 j=2
Then N is a centroaffine hyperbolic hypersurface whose centroaffine metric
is (18.53) and it satisfies T # = 0. Moreover, f = cosh s satisfies
∆f
= −1 = εn1 .
f
Thus, this centroaffine hypersurface satisfies the equality case of (18.39)
identically. Consequently, the estimate given in Theorem 18.1 is optimal
for centroaffine hyperbolic hypersurfaces as well.
Remark 18.2. Example 18.1 shows that conditions “∆f ≤ 0” in Corollary
18.1 and “harmonicity” in Corollary 18.3 are both necessary. Example 18.1
implies that condition “N1 is a compact Riemannian manifold” in Corollary
18.4 is necessary. Example 18.2 shows that condition “(∆f )/f < − dim N1 ”
in Corollary 18.2 is sharp.
18.6 Warped products as graph hypersurfaces
For warped product graph hypersurfaces we have the following results from
[Chen (2005c)].
Theorem 18.3. If a warped product N1 ×f N2 can be realized as a graph

hypersurface in Rn+1 , then the warping function satisfies
∆f (n1 + n2 )2
≥− h(T # , T # ), (18.54)
f 4n2
where n = n1 + n2 , n1 = dim N1 and n2 = dim N2 .
Proof. For graph hypersurfaces in Rn+1 we have (18.22) instead of

(18.16). Therefore, by an argument similar to the proof of Theorem 18.1,
we obtain the theorem.
Theorem 18.2 implies the following.

satisfies ∆f < 0 at some point on N1 , then N1 ×f N2 cannot be realized as
an improper affine hypersphere in Rn+1 .
Another application of Theorem 18.3 is the following.
Corollary 18.6. If N1 is a compact Riemannian manifold, then every

warped product N1 ×f N2 cannot be realized as an improper affine hyper-
sphere in Rn+1 .
Proof. Assume N1 is a compact Riemannian manifold and that the

warped product N1 ×f N2 can be realized as an improper affine hyper-
sphere in Rn+1 . Then Theorem 18.2 yields ∆f ≥ 0.
Because N1 is assumed to be a compact Riemannian manifold, the Hopf
lemma implies that the warping function f is a positive constant, say c.
Consequently, the warped product N1 ×f N2 is the Riemannian product
(N1 , h1 ) × (N2 , c2 h2 ).
Since T M1 ⊕ ξ is parallel along N1 , N1 is contained in an (n1 + 1)-
dimensional affine subspace as a graph hypersurface with ξ as the Calabi
normal. But this is impossible since N1 is compact and ξ is a nonzero
constant transversal vector field.
Theorem 18.4. Let φ : N1 ×f N2 → Rn+1 be a realization of a warped

product as a graph hypersurface. If the warping function satisfies the equal-
ity case of (18.54) identically, then
(a) the Tchebychev vector field T # vanishes identically;

(b) the warping function f is a harmonic function;
(c) N1 ×f N2 is realized as an improper affine hypersphere.
Proof. This can be proved in the same way as Theorem 18.2.

The following examples show that the results given in this section are
sharp.
Example 18.3. Let N = R ×s N2 be the warped product of the real line

and an open part N2 of the unit sphere S n−1 (1) equipped with the warped
product metric:
n−1
!
Y
2 2 2 2 2 2 2
h = ds + s du2 + cos u2 du3 + · · · + cos uj dun . (18.55)
j=2

n−1
!
Y Yn s
2
s sin u2 , sin u3 cos u2 , . . . , sin un cos uj , cos uj , .
j=2
j=2 2
Then N is a graph hypersurface whose Calabi normal is
ξ = (0, . . . , 0, 1)
#
and it satisfies T = 0. Moreover, the Calabi metric is given by the warped
product metric (18.55).
Clearly, the warping function is a harmonic function. Therefore, this
hypersurface satisfies the equality case of (18.54) identically. Consequently,
the estimate given in Theorem 18.2 is optimal.
Remark 18.3. Example 18.3 implies that the condition “∆f < 0” given
in Corollary 18.5 is optimal.
18.7 Realization of Robertson-Walker spaces as affine

hypersurfaces
In this final section, we provide explicit realizations of Robertson-Walker

spacetimes Ln1 (k, f ) in affine spaces obtained in [Chen (2007)].
With respect to a Euclidean coordinate system {u2 , . . . , un } on En−1 ,
the metric tensor g0 on En−1 is given by
Xn
g0 = du2j .
j=2
With respect to a spherical coordinate system {u2 , . . . , un } on the unit

(n − 1)-sphere S n−1 the metric tensor g1 on S n−1 is given by
n−1
Y
g1 = du22 + cos2 u2 du23 + · · · + cos2 uj du2n . (18.56)
j=2
Similarly, for the unit hyperbolic (n − 1)-space H n−1 the corresponding
metric tensor g−1 on H n−1 is given by
n−1
Y
g−1 = du22 + cosh2 u2 du23 + · · · + cosh2 uj du2n .
j=2
We have the following realization theorem from [Chen (2007)] for
Robertson-Walker spacetimes as centroaffine hypersurfaces in Rn+1 .
Theorem 18.5. Let f be a positive function defined on an open interval
I ∋ 0. We have
(a) Every 2D Robertson-Walker spacetime M 2 (0, f ) can always be realized
as centroaffine hypersurfaces in R3 .
(b) For any integer n ≥ 2, the Robertson-Walker spacetime M n (1, f ) can
always be realized as centroaffine hypersurfaces in Rn+1 .
(c) If f > 1 holds on I, then the Robertson-Walker spacetime M n (−1, f )
can be realized as a centroaffine hypersurface in Rn+1 .
Proof. (i) It follows from (4.1) that the metric tensor g of the Robertson-
Walker spacetime M 2 (0, f ) = I ×f E1 is given by the warped product
Lorentzian metric:
g = −dt2 + f 2 (t)dx2 . (18.57)
Let us consider the embedding φ : M 2 (0, f ) → R3 defined by
√ R t f ′ +√f ′2 +f 2 +1
Rt f′+ f ′ 2 +f 2 +1
φ= ex+ 0 f dt
, xex+ 0 f dt
,
√ ! (18.58)
Rt f (f ′ + f ′2 +f 2 +1)
dt
e 0 f 2 +1 .
It is direct to verify that the embedding φ satisfies

1 + 2f 2 + f ′2 + f f ′′
φtt = p φt − φ,
f 1 + f 2 + f ′2
p
f ′ + 1 + f 2 + f ′2
φtx = φx , (18.59)
f
f (1 + f 2 )
φxx = 2φx − p φt + f 2 (t)φ,
f ′ + 1 + f 2 + f ′2
∂ ∂
where φt = φ∗ ( ∂t ), φx = φ∗ ( ∂x ) are tangent vector fields of M .
If we choose the transversal vector field ξ to be φ, then it follows
from (18.3) and (18.59) that the induced affine fundamental form h on
the Robertson-Walker spacetime M 2 (0, f ) is given by
h = −dt2 + f 2 (t)dx2 , (18.60)
2
which is exactly the Lorentzian metric (18.57) on M (0, f ). This is done by
comparing the component of ξ from (18.3) and of φ from (18.59). Thus the
embedding defined by (18.58) is a realization of M 2 (0, f ) as a centroaffine
hypersurface. This proves statement (a).
(ii) Consider the Robertson-Walker spacetime M n (1, f ) = I ×f S n−1 with
the warped product Lorentzian metric:
( n−1
)
Y
2 2 2 2 2 2 2
g = −dt + f (t) du2 + cos u2 du3 + · · · + cos uj dun . (18.61)
j=2
n n+1
Define η : M (1, f ) → R by
√
Rt f (f ′ + f ′ 2 +f 2 +1)
dt
η(t, u2 , . . . , un ) = 0, . . . , 0, e 0 f 2 +1
!
R t √f ′2 +f 2 +1 n−1
Y n
Y
dt
+ f (t)e 0 f sin u2 , . . . , sin un cos uj , cos uj , 0 .
j=2 j=2
By a long direct computation we obtain
f ′2 + f f ′′ + 2f 2 + 1
ηtt = p ηt − η,
f f ′2 + f 2 + 1
p
f ′ + f ′2 + f 2 + 1
ηtuj = ηuj , j = 2, . . . , n,
f
ηui uj = −(tan ui )ηuj , 2 ≤ i < j ≤ n,
j−1
! j−1
!
Y f + f3 Y
2 2 2
ηuj uj = f cos ui η − p cos ui ηt
i=2 f ′ + f ′2 + f 2 + 1 i=2
j−1 j−1
!
X sin(2uk ) Y 2
+ cos ui ηuk , 2 ≤ j ≤ n.
2
k=2 i=k+1
Comparing this with (18.3) in the same way as above shows that with ξ = η
the induced affine fundamental form on the Robertson-Walker spacetime
M n (1, f ) is given by
( n−1
)
Y
2 2 2 2 2 2 2
h = −dt + f (t) du2 + cos u2 du3 + · · · + cos uj dun ,
j=2
which is exactly the Lorentzian metric (18.61) on M n (1, f ). Consequently,

η : M n (1, f ) → Rn+1 gives rise to a realization of M n (1, f ) as a centroaffine
hypersurface. This proves statement (b).
(iii) Assume that f > 1 holds on I. Let us consider the Robertson-Walker
spacetime M n (−1, f ) = I ×f H n−1 with the warped product Lorentzian
metric:
( n−1
)
Y
2 2 2 2 2 2 2
g = −dt + f (t) du2 + cosh u2 du3 + · · · + cosh uj dun .
j=2
Define ψ : M n (−1, f ) → Rn+1 by

√ !
Rt f (f ′ + f ′ 2 +f 2 −1 )
dt
ψ= 0, . . . , 0, e 0 f 2 −1
!
R t √f ′ 2 +f 2 −1 n−1
Y n
Y
dt
+ f (t)e 0 f sinh u2 , . . . , sinh un cosh uj , cosh uj , 0 .
j=2 j=2
A long straightforward computation yields

f ′2 + f f ′′ + 2f 2 − 1
ψtt = p ψt − ψ,
f f ′2 + f 2 − 1
p
f ′ + f ′2 + f 2 − 1
ψtuj = ψuj , j = 2, . . . , n,
f
ψui uj = (tanh ui )ψuj , 2 ≤ i < j ≤ n,
! j−1 (18.62)
f − f3 Y
ψuj uj = f ψ +2
p ψt cosh2 ui
′ ′2
f + f +f −1 2
i=2
j−1 j−1
!
X sinh(2uk ) Y
− cosh2 ui ψuk , j = 2, . . . , n.
2
k=2 i=k+1
After comparing (18.3) with (18.62) in a same way as before shows that
with ξ = ψ the induced affine fundamental form on the manifold M n (−1, f )
is given by
 
 n−1
Y 
h = −dt2 + f 2 (t) du22 + cosh2 u2 du23 + · · · + cosh2 uj du2n ,
 
j=2
n
which is exactly the Lorentzian metric on M (−1, f ). Therefore we obtain a
realization of the Robertson-Walker spacetime M n (−1, f ) as a centroaffine
hypersurface. This proves statement (c).
We may also realize Robertson-Walker spacetimes M n (k, f ) as graph

hypersurfaces as follows.
First, we observe that the Minkowski spacetime En1 , n ≥ 2, with the
canonical Lorentzian metric:
n
X
g = −dt2 + du2j (18.63)
j=2
can be realized as a graph hypersurface in an affine (n + 1)-space Rn+1 .

This can be easily done as follows:
Consider the embedding φ : En1 → Rn+1 defined by
n
!
1 X 2 t2
φ(t, u2 , . . . , un ) = t, u2 , . . . , un , u − . (18.64)
2 j=2 j 2
It is easy to verify that the embedding φ satisfies
φtt = −ξ, φuj uk = δjk ξ, φtuj = 0, (18.65)
where ξ is the constant transversal vector field given by ξ = (0, . . . , 0, 1).

It follows from (18.3) and (18.65) that the induced Calabi metric h on
the Minkowski spacetime En1 via φ is exactly the Minkowski metric (18.63)
on E41 . Therefore (18.64) gives rise to a realization of En1 in Rn+1 as a graph
hypersurface.
For the realization of Robertson-Walker spacetimes as graph hypersur-
faces, we have the following result from [Chen (2007)].
Theorem 18.6. Let f be a positive function defined on an open interval

I. We have
(a) Every 2-dimensional Robertson-Walker spacetime M 2 (0, f ) = I ×f R

can be realized as a graph surface in an affine 3-space R3 .
(b) For n ≥ 3, we have the following:
(b.1) The Robertson-Walker spacetime M n (1, f ) = I ×f S n−1 can always
be realized as a graph hypersurface in Rn+1 .
(b.2) If f has no critical points in I, then the Robertson-Walker spacetime
M n (0, f ) = I ×f En−1 can be realized as a graph hypersurface in
Rn+1 .
(b.3) If f ′2 (t) > 1 holds for each t ∈ I, then the Robertson-Walker space-
time M n (−1, f ) = I ×f H n−1 can be realized as a graph hypersurface
in Rn+1 .
Proof. Without loss of generality, we may assume that the interval I

contains 0.
(A) Consider the 2-dimensional Robertson-Walker spacetime equipped with
the warped product Lorentzian metric: g = −dt2 + f 2 (t)du2 .
Let us put
(Z p )
t
f ′2 + 1
ψ(t, u) = f (t) exp dt sin u,
0 f
(Z p ) Z t !
t
f ′2 + 1 p
′
f (t) exp dt cos u, f (f + f ′2 + 1)dt .
0 f 0
Then a direct computation shows that ψ satisfies
f f ′′ + f ′2 + 1
ψtt = p ψt − ξ,
f f ′2 + 1
p
f ′ + f ′2 + 1 (18.66)
ψtu = ψu ,
f
p
ψuu = f f ′ − f ′2 + 1 ψt + f 2 ξ
with ξ = (0, 0, 1). It follows from (18.3) and (18.66) that the induced Calabi
metric h is exactly the Lorentzian metric g = −dt2 + f 2 (t)du2 . This shows
that every 2-dimensional Robertson-Walker spacetime can be realized as a
graph surface in R3 . This proves statement (a).
(B) Consider the Robertson-Walker spacetime M n (1, f ) = I ×f S n−1
equipped with Lorentzian metric:
( n−1
)
Y
2 2 2 2 2 2 2
g = −dt + f (t) du2 + cos u2 du3 + · · · + cos uj dun . (18.67)
j=2
Let us define η : M n (1, f ) → Rn+1 by

Z !
t p
η(t, u2 , . . . , un ) = 0, . . . , 0, f f ′ − f ′2 + 1 dt
0
Rt p
{ f ′ 2 + 1/f }dt
+ f e− 0 sin u2 , sin u3 cos u2 , . . . ,
n−1 n
!
Y Y
sin un cos uj , cos uj , 0
j=2 j=2
A straightforward long computation yields

1 + f f ′′ + f ′2
ηtt = − p ηt − ξ,
f f ′2 + 1
p
f ′ − f ′2 + 1
ηtuj = ηuj , j = 2, . . . , n,
f
ηui uj = −(tan ui )ηuj , 2 ≤ i < j ≤ n,
( ) j−1 (18.68)
f Y
2
ηuj uj = p ηt + f ξ cos2 ui
′2
f +1−f ′
i=2
j−1
( j−1
)
X sin 2uk Y
2
+ cos ui ηuk , j = 2, . . . , n,
2
k=2 i=k+1
with ξ = (0, . . . , 0, 1).

It follows from (18.3) and (18.68) that the induced Calabi metric on
the Robertson-Walker spacetime M n (1, f ) via η is exactly the Lorentzian
metric (18.67) on M n (1, f ). Thus η is a realization of M n (1, f ). This
proves statement (b.1).
Now, let us consider the Robertson-Walker spacetime
M (0, f ) = I ×f En−1
equipped with the Lorentzian metric:

g = −dt2 + f 2 (t) du22 + du23 + · · · + du2n . (18.69)
Suppose that the warping function f does not have critical points. Let us
define an immersion φ : M (0, f ) → Rn+1 by
ζ(t, u2 , . . . , un ) = u2 , . . . , un ,
 (18.70)
n
X Z t Z t
dt f (t) 
u2j + , dt .
j=2 0 f (t)f ′ (t) 0 2f ′ (t)

f ′2 + f f ′′
ζtt = − ζt − ξ,
ff′
(18.71)
ζtuj = ζui uj = 0,
ζuj uj = 2f f ′ ζt + f 2 ξ,
for 2 ≤ i 6= j ≤ n, where ξ = (0, . . . , 0, 1) is a constant transversal vector
field.
It follows from (18.3) and (18.70) that the induced Calabi metric h via
ζ is exactly the Lorentzian metric (18.69) on M n (0, f ). Consequently, ζ
is a realization of M n (0, f ) as a graph hypersurface in Rn+1 . This proves
statement (b.2).
(C) Finally, assume that f ′2 (t) > 1 holds for each t ∈ I. Let us consider
the Robertson-Walker spacetime
M n (−1, f ) = I ×f H n−1
with the Lorentzian(metric: )
n−1
Y
g = −dt2 + f 2 (t) du22 + cosh2 u2 du23 + · · · + cosh2 uj du2n . (18.72)
j=2
Define φ : M n (−1, f ) → Rn+1 by !
Z t p
′

φ(t, u2 , . . . , un ) = 0, . . . , 0, f ′2
f − 1 − f dt
0
f
+ Rt p sinh u2 , sinh u3 cosh u2 , . . . ,
{ f ′2 − 1/f }dt
e 0
n−1 n
!
Y Y
sinh un cosh uj , cosh uj , 0 .
j=2 j=2
Then a straightforward long computation yields
1 − f f ′′ − f ′2
φtt = p φt − ξ,
f f ′2 − 1
p
f ′ − f ′2 − 1
φtuj = φuj ,
f
φui uj = (tanh ui )φuj ,
( ) j−1 (18.73)
f Y
φuj uj = p φt + f 2 ξ cosh2 ui
f ′ − f ′2 − 1 i=2
j−1
( j−1
)
X sinh 2uk Y
2
− cosh ui φuk ,
2
k=2 i=k+1
for 2 ≤ i < j ≤ n with ξ = (0, . . . , 0, 1).
It follows from (18.3) and (18.73) that the induced Calabi metric h on
the Robertson-Walker spacetime M n (−1, f ) via the map φ is exactly the
Lorentzian metric given in (18.72) on M n (−1, f ). Therefore φ gives rise to
a realization of M n (−1, f ) as a graph hypersurface in Rn+1 . This proves
statement (b.3) of the theorem. Consequently, we complete the proof of
the theorem.
Bibliography
Abedi, E., Haghighatdoost, G., Ilmakchi, M. and Nazari, Z. (2013). Contact 3-

structure QR-warped product submanifold in Sasakian space form, Turkish
J. Math. 37, 195–201.
Ahn, S. S., Kim D. S, and Kim, Y.-H. (1996). Totally umbilical Lorentzian sub-
manifolds, J. Korean Math. Soc. 33, 507–512.
Aledo, J. A., Romero, A. and Rubio, R. M. (2014). Upper and lower bounds for
the volume of a compact spacelike hypersurface in a generalized Robertson-
Walker spacetime, Int. J. Geom. Methods Mod. Phys. 11, no. 1, 1450006.
Alegre, P., Chen, B.-Y., Munteanu, M. I. (2014). Riemannian submersions, δ-
invariants, and optimal inequality, Ann. Global Anal. Geom. 42, 317–331.
Al-Ghefari, R., Al-Solamy, F. R. and Shahid, M. H. (2006). Contact CR-warped
product submanifolds in generalized Sasakian space forms, Balkan J. Geom.
Appl. 11, no. 2, 1–10.
Ali, A., Othman, W. A. M. and Ozel, C. (2016a). Characterizations of contact
CR-warped product submanifolds of nearly Sasakian manifolds, Balkan J.
Geom. Appl. 21, no. 2, 9–20.
Alı́as, L. J., Romero, A. and Sánchez, M. (1995). Uniqueness of complete spacelike
hypersurfaces of constant mean curvature in generalized Robertson-Walker
spacetimes, Gen. Relativity Gravitation 27, no. 1, 71–84.
Alı́as, L. J., Colares, A. G. and de Lima, H. F. (2013). On the rigidity of com-
plete spacelike hypersurfaces immersed in a generalized Robertson-Walker
spacetime, Bull. Braz. Math. Soc. (N.S.) 44, no. 2, 195–217.
Al-Luhaibi, N. S., Al-Solamy, F. R. and Khan, V. A. (2009). CR-warped product
submanifolds of nearly Kaehler manifolds, J. Korean Math. Soc. 46, 979–
995.
Al-Solamy, F. R., Chen, B.-Y., Deshmukh, S. (2014). Two optimal inequalities for
anti-holomorphic submanifolds and their applications, Taiwanese J. Math.
18, 199–217.
Al-Solamy F. R. and Khan, M. A. (2008b). Non-existence of non-trivial generic
warped product in Kaehler manifolds, Note Mat. 28, no. 2, 63–68.
Al-Solamy F. R. and Khan, M. A. (2012a). Semi-invariant warped product sub-
manifolds of almost contact manifolds, J. Inequal. Appl. 2012, 2012:127.
451
Al-Solamy F. R. and Khan, M. A. (2014a). Hemi-slant warped product subman-

ifolds of nearly Kaehler manifolds, Abstract Appl. Anal. 2014, no.404851.
Al-Solamy, F. R., Khan, V. A. and Uddin, S. (2016). Geometry of warped product
semi-slant submanifolds of nearly Kaehler manifolds, Result. Math. DOI
10.1007/s00025-016-0581-4.
Antić, M., Djorić, M. and Vrancken, L. (2006). Characterization of totally
geodesic totally real 3-dimensional submanifolds in the 6-sphere, Acta Math.
Sin. (Engl. Ser.) 22, 1557–1564.
Arroyo, J., Barros, M. and Garay, O. J. (1999) Willmore-Chen tubes on ho-
mogeneous spaces in warped product spaces. Pacific J. Math. 188 (1999),
201–207.
Arslan, K., Ezentas, R. Mihai, I. and Murathan, C. (2005). Contact CR-warped
product submanifolds in Kenmotsu space forms, J. Korean Math. Soc. 42,
1101–1110.
Arslan, K., Carriazo, A., Chen, B.-Y. and Murathan, C. (2010). On slant sub-
manifolds of neutral Kaehler manifolds, Taiwanese J. Math. 14, 561–584.
Atceken, M. (2008). Warped product semi-slant submanifolds in locally Rieman-
nian product manifolds, Bull. Aust. Math. Soc. 77, no. 2, 177–186.
Atceken, M. (2009a). Geometry of warped product semi-invariant submanifolds
of a locally Riemannian product manifold, Serdica Math. J. 35, 273–286.
Atceken, M. (2009b). Warped product semi-invariant submanifolds in almost
paracontact Riemannian manifolds, Math. Probl. Eng. no. 621625, 16 pages.
Atceken, M. (2010a). Warped product semi-invariant submanifolds in almost
paracontact metric manifolds, Math. Morav. 14, 15–21.
Atceken, M. (2010b). Warped product semi-slant submanifolds in Kenmotsu man-
ifolds, Turkish J. Math. 34, 425–432.
Atceken, M. (2011a). Contact CR-warped product submanifolds in cosymplectic
space forms, Collect. Math. 62, no. 1, 17–26.
Atceken, M. (2011b). Warped product semi-invariant submanifolds in locally de-
composable Riemannian manifolds, Hacet. J. Math. Stat. 40, 401–407.
Atceken, M. (2013). Contact CR-warped product submanifolds in Kenmotsu
space forms, Bull. Iranian Math. Soc. 39, 415–429.
Atceken, M. (2015). Contact CR-warped product submanifolds in Sasakian space
forms, Hacet. J. Math. Stat. 44, 2332.
Atceken, M., Sahin, B. and Kilic, E. (2003). On invariant submanifolds of Rie-
mannian warped product manifold, Turkish J. Math. 27, 407–423.
Anciaux, H. (2011). Minimal submanifolds in pseudo-Riemannian geometry,
World Scientific (Hackensack, NJ).
Barros, M. and Romero, A. (1982). Indefinite Kähler manifolds, Math. Ann. 261,
55–62.
Beem, J. K., Ehrlich, P. E. and Easley, K. L. (1996). Global Lorentzian geometry,
2nd Ed., (Marcel Dekker, New York).
Bejancu, A. (1978). CR-submanifolds of a Kaehler manifold I, Proc. Amer. Math.
Soc. 69, 134–142.
Bejancu, A. (1979). On the geometry of leaves on a CR-submanifold, Ann. St.
Univ. Al. I. Cuza Iasi 25, 393–398.
Bibliography 453
Bejancu, A. (1980). Umbilical CR-submanifolds of a Kähler manifold. Rend. Mat.

13, 431-446
Bejancu, A. (1986). Geometry of CR-Submanifolds (D. Reidel, Dordrecht).
Bejancu, A. (2007). Oblique warped products, J. Geom. Phys. 57, 1055–1073.
Bejancu, A. and Duggal, K. L. (1993). Real hypersurfaces of indefinite Kaehler
manifolds, Int. J. Math. Math. Sci. 16, 545–556.
Bejancu, A., Kon, M. and Yano, K. (1981). CR-submanifolds of a complex space
form, J. Differential Geom. 16, 137145.
Bejancu, A. and Papaghiuc, N. (1981). Semi-invariant submanifolds of a Sasakian
manifold, An. Stiint. Al. I. Cuza. Univ. Iasi. 27, 163–170.
Beltrami, E. (1864). Ricerche di analisi applicata alla geometria, Giornale di
Math. II, 150–162.
Berger, M., Gauduchon, P. and Mazet, E. (1971). Le Spectre d’une Variété Rie-
mannienne, Lecture Notes in Mathematics, Vol. 194.
Berndt, J. (1989). Real hypersurfaces with constant principal curvatures in com-
plex hyperbolic space, J. Reine Angew. Math. 395, 132–141.
Bertola, M. and Gouthier, D. (2001). Warped products with special Riemannian
curvature, Bol. Soc. Brasil. Mat. (N.S.) 32, 45–62.
Besse, A. (1987). Einstein Manifolds, (Springer-Verlag, New York-Berlin).
Binh, T. Q. and De, A. On contact CR-warped product submanifolds of a quasi-
Sasakian manifold, Publ. Math. Debrecen 84, 123–137.
Birkhoff, G. D. (1923). Relativity and Modern Physics, Harvard University Press.
Bishop, R. L. and O’Neill, B. (1969). Manifolds of negative curvature, Trans.
Amer. Math. Soc. 145, 1–49.
Blair, D. E. (2010). Riemannian Geometry of Contact and Symplectic Manifolds,
2nd edition (Birkhäuser Boston, Inc., MA).
Blair, D. E. and Chen, B. Y. (1979). On CR-submanifolds of Hermitian manifold,
Israel J. Math. 34, 353–363.
Blair, D. E. and Dragomir, S. (2002). CR products in locally conformal Kähler
manifolds, Kyushu J. Math. 56, 337–362.
Bolton, J., Vrancken, L. and Woodward, L. M. (2007). On almost complex curves
in the nearly Kähler 6-sphere, Quart. J. Math. Oxford Ser. 45, 407–427.
Bolton, J., Rodriguez Montealegre, C. and Vrancken, L. (2009). Characterizing
warped-product Lagrangian immersions in complex projective space, Proc.
Edinb. Math. Soc. (2) 52, no. 2, 273–286.
Bonanzinga, V. and Matsumoto, K. (2004). Warped product CR-submanifolds in
locally conformal Kaehler manifolds, Period. Math. Hungar. 48, 207–221.
Bonanzinga, V. and Matsumoto, K. (2008). On doubly warped product CR-
submanifolds in a locally conformal Kaehler manifold, Tensor 69, 76–82.
Bonnet, O. (1867). Mémoire sur la theéorie des surfaces applicables sur une surface
donnée, École Polytech. 42, 31–151.
Borrelli, V., Chen, B. Y. and Morvan, J. M. (1995). Une caractérisation géomé-
trique de la sphère de Whitney, C.R. Acad. Sc. Paris, 321, 1485–1490.
Brozos-Vázquez, M., Garcı́a-Rı́o, E. and Vázquez-Lorenzo, R. (2005). Some re-
marks on locally conformally flat static space-times, J. Math. Phys. 46
(2005), no. 2, 022501, 11 pages.
Bryant, R. L. (1982). Submanifolds and special structure on the octonians, J.

Diff. Geom., 17, 185–232.
Burstin, C. (1931). Ein Betrag zum Problem der Einbettung der Riemannschen
Räume in eukildische Räume, Math Sb. 38, 74–85.
Butruille, J.-B. (2005). Classification of homogeneous nearly Kähler manifolds,
Ann. Global Anal. Geom. 27, 201–225.
Caballero, M., Romero, A. and Rubio, R. M. (2011). Constant mean curvature
spacelike hypersurfaces in Lorentzian manifolds with a timelike gradient
conformal vector field, Class. Quantum Gravity 28(14), 145009.
Cabrerizo, J. L., Carriazo, A., Fernández, L. M. and Fernández, M. (1999). Semi-
slant submanifolds of a Sasakian manifold, Geom. Dedicata 78, 183–199.
Cabrerizo, J. L., Carriazo, A., Fernández, L. M. and Fernández, M. (2000). Slant
submanifolds in Sasakian manifolds, Glasgow Math. J. 42, 125–138.
Cabrerizo, J. L., Carriazo, A., Fernández, L. M. and Fernández, M. (2002). Rie-
mannian submersions and slant submanifolds, Publ. Math. Debrecen 61,
523–532.
Calabi, E. (1953). Isometric imbedding of complex manifolds, Ann. Math. 58,
1–23.
Calabi, E. (1958). Construction and properties of some 6-dimensional almost
complex manifolds, Trans. Amer. Math. Soc. 87, 407–438.
Calabi, E. (1967). Minimal immersions of surfaces in Euclidean spheres, J. Dif-
ferential Geom. 1, 111–125.
Calabi, E. and Gluck, H. (1993). What are the best almost-complex structures
on the 6-sphere?, Proc. Sympos. Pure Math., 54, Part 2, 99–106zw.
Calvaruso, G. and Perrone, D. (2010). Contact pseudo-metric manifolds, Differ-
ential Geom. Appl. 28, 615–634.
Carriazo, A. (2000). Bi-slant immersions, in: Proceedings ICRAMS, (J. C. Misra,
S. B. Sinha, Eds.), Narosa Publishing House, 88–97.
Cartan, É. (1927). Sur la possibilité de plonger un espace riemannien donné dans
un espace euclidien, Ann. Soc. Math. Polon. 6, 1–17.
Cartan, É. (1939). Sur les familles remarquables d’hypersurfaces isoparamétriques
dans les espaces sphériques, Math. Z. 45, 335–367.
Castro, I. and Chen, B. Y. (2006). Lagrangian surfaces in complex Euclidean
plane via spherical and hyperbolic curves, Tohoku Math. J. 58, 565–579.
Castro, I. and Urbano, F. (2006). A characterization of the Lagrangian pseudo-
sphere, Proc. Amer. Math. Soc. 132, 1797–1804.
Chen, B. Y. (1973b). Geometry of Submanifolds, (M. Dekker, New York, NY).
Chen, B. Y. (1974). Some conformal invariants of submanifolds and their appli-
cations, Boll. Un. Mat. Ital. 10, 380–385.
Chen, B. Y. (1979). Extrinsic spheres in Riemannian manifolds, Houston J. Math.
5, no. 3, 319–324.
Chen, B. Y. (1981a). Geometry of Submanifolds and Its Applications (Science
University of Tokyo, Tokyo, Japan).
Chen, B. Y. (1981b). CR-submanifolds of a Kähler manifold I, J. Differential
Geom. 16, 305–322.
Bibliography 455
Chen, B. Y. (1981c). CR-submanifolds of a Kähler manifold II, J. Differential

Geom. 16, 493–509.
Chen, B. Y. (1981d). Cohomology of CR-submanifolds, Ann. Fac. Sc. Toulouse
Math. 3, 167–172.
Chen, B. Y. (1981e). Differential geometry of real submanifolds in a Kähler man-
ifold, Monatsh. Math. 91, 257–274.
Chen, B. Y. (1981f). Totally umbilical submanifolds of Kaehler manifolds, Arch.
Math. 36, 83–91.
Chen, B. Y. (1984). Total Mean Curvature and Submanifolds of Finite Type
(World Scientific Publ., New Jersey).
Chen, B. Y. (1990a). Slant immersions, Bull. Austral. Math. Soc. 41, 135147.
Chen, B. Y. (1990b). Geometry of Slant Submanifolds, (KULeuven, Belgium).
Chen, B. Y. (1993a). Differential geometry of semiring of immersions, I: General
theory, Bull. Inst. Math. Acad. Sinica 21, 1–34.
Chen, B. Y. (1993b). Some pinching and classification theorems for minimal sub-
manifolds, Arch. Math. 60, 568–578.
Chen, B. Y. (1996a). Mean curvature and shape operator of isometric immersions
in real-space-form, Glasgow Math. J. 38, 87–97.
Chen, B. Y. (1996b). A general inequality for submanifolds in complex-space-
forms and its applications, Arch. Math. 67, 519–528.
Chen, B. Y. (1997a). Interaction of Legendre curves and Lagrangian submanifolds,
Israel J. Math. 99, 69–108.
Chen, B. Y. (1997b). Complex extensors and Lagrangian submanifolds in complex
Euclidean spaces, Tohoku Math. J. 49, 277–297.
Chen, B. Y. (1998a). Strings of Riemannian invariants, inequalities, ideal immer-
sions and their applications, The Third Pacific Rim Geometry Conference
(Seoul, 1996), 7–60, Monogr. Geom. Topology, 25 (Intern. Press).
Chen, B. Y. (1998b). Special slant surfaces and a basic inequality, Results Math.
33, 65–78.
Chen, B. Y. (1999a). Relations between Ricci curvature and shape operator for
submanifolds with arbitrary codimension, Glasgow Math. J. 41, 33–41.
Chen, B. Y. (1999b). On slant surfaces, Taiwanese J. Math. 3, 163–179.
Chen, B. Y. (1999c). Representation of flat Lagrangian H-umbilical submanifolds
in complex Euclidean space, Tohoku Math. J. 51, 13–20.
Chen, B. Y. (2000a). Some new obstructions to minimal and Lagrangian isometric
immersions, Japan. J. Math. 26, 105–127.
Chen, B. Y. (2000b). Riemannian Submanifolds, Handbook of Differential Geom-
etry, vol. I, 187–418 (eds. F. Dillen and L. Verstraelen).
Chen, B. Y. (2000c). Twisted product CR-submanifolds in Kaehler manifolds,
Tamsui Oxford J. Math. Sci. 16 (2000), no. 2, 105–121.
Chen, B. Y. (2000d). Ideal Lagrangian immersions in complex space forms, Math.
Proc. Cambridge Philos. Soc. 128, 511–533.
Chen, B. Y. (2001a). Geometry of warped product CR-submanifolds in Kähler
manifolds, I, Monatsh. Math. 133, 177–195.
Chen, B. Y. (2001b). Geometry of warped product CR-submanifolds in Kähler
manifolds, II, Monatsh. Math. 134, 103–119.
Chen, B. Y. (2001c). Riemannian geometry of Lagrangian submanifolds, Tai-

wanese J. Math. 5, 681–723.
Chen, B. Y. (2001d). A general inequality for Kählerian slant submanifolds and
related results, Geom. Dedicata 85, 253–271.
Chen, B. Y. (2001e). Constant-ratio hypersurfaces, Soochow J. Math. 27, 353–
362.
Chen, B. Y. (2002a). On isometric minimal immersions from warped products
into real space forms, Proc. Edinburgh Math. Soc. 45, 579–587.
Chen, B. Y. (2002b). Geometry of warped products as Riemannian submanifolds
and related problems, Soochow J. Math. 28, 125–156.
Chen, B. Y. (2002c). Real hypersurfaces in complex space forms which are warped
products, Hokkaido Math. J. 31, 363–383.
Chen, B. Y. (2002d). Non-immersion theorems for warped products in complex
hyperbolic spaces, Proc. Japan Acad. Ser. A Math. Sci. 78, 96–100.
Chen, B. Y. (2002e). Classification of flat slant surfaces in complex Euclidean
plane, J. Math. Soc. Japan 54, 719–746.
Chen, B. Y. (2002f). Classification of slumbilical submanifolds in complex space
forms, Osaka J. Math. 39, 23–47.
Chen, B. Y. (2002g). Convolution of Riemannian manifolds and its applications,
Bull. Austral. Math. Soc. 66, 177–191.
Chen, B. Y. (2002h). Segre embedding and related maps and immersions in dif-
ferential geometry, Arab J. Math. Sci. 8, no. 1, 1–39.
Chen, B. Y. (2002i). Geometry of position functions of Riemannian submanifolds
in pseudo-Euclidean space, J. Geom. 74, 61–77.
Chen, B. Y. (2002j). Ricci curvature of real hypersurfaces in complex hyperbolic
space, Arch. Math. (Brno) 38, no. 1, 73–80.
Chen, B. Y. (2003a). Another general inequality for CR-warped products in com-
plex space forms, Hokkaido Math. J. 32, 415–444.
Chen, B. Y. (2003b). What can we do with Nash’s embedding theorem?, Soochow
J. Math. 30, 303–338.
Chen, B. Y. (2003c). A general optimal inequality for warped products in complex
projective spaces and its applications, Proc. Japan Acad. Ser. A Math. Sci.
79, 89–94.
Chen, B. Y. (2003d). Flat slant surfaces in complex projective and complex hy-
perbolic planes, Results Math. 44, 54–73.
Chen, B. Y. (2003e). More on convolution of Riemannian manifolds, Beiträge
Algebra Geom. 44, 9–24.
Chen, B. Y. (2003f). Constant-ratio space-like submanifolds in pseudo-Euclidean
space, Houston J. Math. 29 , 281–294
Chen, B. Y. (2004a). CR-warped products in complex projective spaces with
compact holomorphic factor, Monatsh. Math. 141, 177–186.
Chen, B. Y. (2004b). Warped products in real space forms, Rocky Mountain J.
Math. 34, 551–563.
Chen, B. Y. (2004c). CR-warped products in complex projective spaces with
compact holomorphic factor, Monatsh. Math. 141, 177–186.
Chen, B. Y. (2005a). Riemannian submersions, minimal immersions and coho-
mology class, Proc. Japan Acad. Ser. A Math. Sci. 81, 162–167.
Bibliography 457
Chen, B. Y. (2005b). A general optimal inequality for arbitrary Riemannian sub-

manifolds, J. Inequal. Pure Appl. Math. 6, no. 3, Article 77, 10 pages.
Chen, B. Y. (2005c). On warped product immersions, J. Geom. 82, no. 1-2, 36–49.
Chen, B. Y. (2005d). Geometry of affine warped product hypersurfaces, Results
Math. 48 (2005), no. 1-2, 9–28.
Chen, B. Y. (2005e). First normal bundle of ideal Lagrangian immersions in
complex space forms, Math. Proc. Camb. Phil. Soc. 138, 461–464.
Chen, B. Y. (2007). Realization of Robertson-Walker space-times as affine hyper-
surfaces, J. Phys, A 40, no. 15, 4241–4250.
Chen, B. Y. (2008). Classification of marginally trapped Lorentzian flat surfaces
in E42 and its application to biharmonic surfaces, J. Math. Anal. Appl. 340,
861–875.
Chen, B. Y. (2009a). Dependence of the Gauss-Codazzi equations and the Ricci
equation of Lorentz surfaces Publ. Math. Debrecen 74, 341–349.
Chen, B. Y. (2009b). Classification of spatial surfaces with parallel mean cur-
vature vector in pseudo-Euclidean spaces with arbitrary codimension, J.
Math. Phys. 50, 043503, 14 pages.
Chen, B. Y. (2009c). Complete classification of spatial surfaces with paral-
lel mean curvature vector in arbitrary non-flat pseudo-Riemannian space
forms, Cent. Eur. J. Math. 7, 400–428.
Chen, B. Y. (2010a). Submanifolds with parallel mean curvature vector in Rie-
mannian and indefinite space forms, Arab J. Math. Sci. 16, 1–45.
Chen, B. Y. (2010b). Lagraignan submanifolds in para-Kähler manifolds, Non-
linear Anal. 73, 3561–3571.
Chen, B. Y. (2010c). Explicit classification of parallel Lorentz surfaces in 4D
indefinite space forms with index 3, Bull. Inst. Math. Acad. Sinica (N.S.)
5, 311–345.
Chen, B. Y. (2011a). Classification of minimal Lorentz surfaces in indefinite space
forms with arbitrary codimension and arbitrary index, Publ. Math. Debre-
cen, 78, 485–503.
Chen, B. Y. (2011b). Pseudo-Riemannian Geometry, δ-invariants and Applica-
tions, (World Scientific, Hackensack, NJ).
Chen, B. Y. (2011c). Classification of flat Lagrangian H-umbilical submanifolds
in para-Kähler n-plane, Int. Electron. J. Geom. 4, no. 1, 1–14.
Chen, B. Y. (2012). An optimal inequality for CR-warped products in complex
space forms involving CR δ-invariant, Internat. J. Math. 23(3), 1250045.
Chen, B. Y. (2013a). A tour through δ-invariants: From Nash embedding theorem
to ideal immersions, best ways of living and beyond, Publ. Inst. Math.
(Beograd) (N.S.), 94(108) (2013), 67–80.
Chen, B. Y. (2013b). On ideal hypersurfaces of Euclidean 4-space, Arab J. Math.
Sci. 19, 129–144.
Chen, B. Y. (2013c). Geometry of warped product submanifolds: a survey, J.
Adv. Math. Stud. 6, no. 2, 1–43.
Chen, B. Y. (2014). A simple characterization of generalized Robertson-Walker
spacetimes, Gen. Relativ. Gravit. 46, no.1833.
Chen, B. Y. (2015). Some results on concircular vector fields and their applications
to Ricci solitons, Bull. Korean Math. Soc. 52, 1535-1547.
Chen, B. Y. (2016a). Concircular vector fields and pseudo-Kaehler manifolds,

Kragujevac J. Math. 40, 7–14.
Chen, B. Y. (2016b). CR-warped product submanifolds in Kaehler manifolds, in:
Geometry of Cauchy-Riemann Submanifolds, 1–25, Springer-Verlag.
Chen, B. Y. (2016c). CR-submanifolds and δ-invariants, in: Geometry of Cauchy-
Riemann Submanifolds, 27–55, Springer-Verlag.
Chen, B.Y. (2017a). Rectifying submanifolds of Riemannian manifolds and
torqued vector fields, Kragujevac J. Math. 41, no. 1, 93–103.
Chen, B.Y. (2017b). Classification of torqued vector fields and its applications to
Ricci solitons, Kragujevac J. Math. 41, no. 2, 239–250.
Chen, B.Y. (2017c). Topics in differential geometry associated with position vec-
tor fields on Euclidean submanifolds, Arab J. Math. Sci. 23, no. 1, 1–17.
Chen, B. Y. and Dillen, F. (2008a). Optimal inequalities for multiply warped
product, Int. Electron. J. Geom. 1, 1–11. Erratum, ibid, 4 (2011), 138.
Chen, B. Y. and Dillen, F. (2008b). Warped product decompositions of real
space forms and Hamiltonian-stationary Lagrangian submanifolds, Non-
linear Anal. 69, no. 10, 3462–3494.
Chen, B. Y. and Dillen, F. (2011). Optimal general inequalities for Lagrangian
submanifolds in complex space forms, J. Math. Anal. Appl. 379, 229–239.
Chen, B. Y., Dillen, F., Van der Veken, J. and L. Vrancken (2013). Curvature
inequalities for Lagrangian submanifolds: the final solution, Differential
Geom. Appl. 31, 808–819.
Chen, B. Y., Dillen, D., Verstraelen, L. and Vrancken, L. (1994). Totally real
submanifolds of CP n satisfying a basic equality, Arch. Math. 63, 553–564.
Chen, B. Y., Dillen, D., Verstraelen, L. and Vrancken, L. (1995a). Characterizing
a class of totally real submanifolds of S 6 by their sectional curvatures,
Tohoku Math. J. 47, 185–198.
Chen, B. Y., Dillen, D., Verstraelen, L. and Vrancken, L. (1995b).Two equiv-
ariant totally real immersions into the nearly Kähler 6-sphere and their
characterization, Japanese J. Math. 21, 207–222.
Chen, B. Y., Dillen, D., Verstraelen, L. and Vrancken, L. (1996). An exotic totally
real minimal immersion of S 3 in CP 3 and its characterization, Proc. Roy.
Soc. Edinburgh Sec. A, 126, 153–165.
Chen, B. Y. and Garay, O. J. (2012a). δ(2)-ideal null 2-type hypersurfaces of
Euclidean space are spherical cylinders, Kodai Math. J. 35, 382–391.
Chen, B. Y. and Garay, O. J. (2012b). Pointwise slant submanifolds in almost
Hermitian manifolds, Turkish J. Math. 36, no. 4, 630–640.
Chen, B. Y. and Kuan, W. E. (1985). The Segre imbedding and its converse,
Ann. Fac. Sc. Toulouse Math. 7, 1–28.
Chen, B. Y. and Maeda, S. (1996). Extrinsic characterizations of circles in a
complex projective space imbedded in a Euclidean space, Tokyo J. Math.
19, 169–185.
Chen, B. Y. and Maeda, S. (2003). Real hypersurfaces in nonflat complex space
forms are irreducible, Osaka J. Math. 40, 121–138.
Chen, B. Y. and Munteanu, M. I. (2012). Geometry of PR-warped products in
para-Kähler manifolds, Taiwanese J. Math. 16, 1293–1327.
Bibliography 459
Chen, B. Y. and Ogiue, K. (1974a). On totally real submanifolds, Trans. Amer.

Math. Soc. 193, 257–266.
Chen, B. Y. and Ogiue, K. (1974b). Two theorems on Kähler manifolds, Michigan
Math. J. 21, 257–266.
Chen, B. Y. and Tazawa, Y. (1990). Slant surfaces of codimension two. Ann. Fac.
Sci. Toulouse Math. 11, no. 3, 29–43.
Chen, B. Y. and Tazawa, Y. (1991). Slant submanifolds in complex Euclidean
spaces. Tokyo J. Math. 14), no. 1, 101–120.
Chen, B. Y. and Tazawa, Y. (2000). Slant submanifolds of complex projective
and complex hyperbolic spaces, Glasgow Math. J. 42, 439–454.
Chen, B. Y. and Uddin, S. (2016). Warped product pointwise bi-slant submani-
folds of Kaehler manifolds, Publ. Math. Debrecen 92, 16 pages.
Chen, B. Y. and Van der Veken, J. (2007). Spatial and Lorentzian surfaces in
Robertson-Walker space times, J. Math. Phys. 48, 073509, 12 pages.
Chen, B. Y. and Vanhecke, L. (1981). Differential geometry of geodesic spheres,
J. Reine Angew. Math. 325, 28–67.
Chen, B. Y. and Verheyen, P. (1983). Totally umbilical submanifolds of Kaehler
manifolds II, Bull. Soc. Math. Belg. Ser. B 35, no. 1, 27-44.
Chen, B. Y. and Vrancken, L. (1997). Existence and uniqueness theorem for slant
immersions and its applications, Results Math. 31, 28–39.
Chen, B. Y., Vrancken, L. (1999). CR-submanifolds of complex hyperbolic spaces
satisfying a basic equality, Israel J. Math. 110, 341–358.
Chen, B. Y. and Vrancken, L. (2001). Addendum to: Existence and uniqueness
theorem for slant immersions and its applications, Results Math. 39, 18–22.
Chen, B. Y. and Vrancken, L. (2002a). Slant surfaces with prescribed Gaussian
curvature, Balkan J. Geom. Appl. 7, no. 1, 29–36.
Chen, B. Y., Vrancken, L. (2002b). Lagrangian submanifolds of the complex
hyperbolic space, Tsukuba J. Math. 26, 95–118.
Chen, B. Y. and Wei, S. W. (2008a). Submanifolds of warped product manifolds
I ×f S m−1 (k) from a p-harmonic viewpoint, Bull. Transilv. Univ. Brasov
Ser. III 1(50) (2008), 59–77.
Chen, B. Y. and Wei, S. W. (2008b). Differential geometry of submanifolds of
warped product manifolds I ×f S m−1 (k), J. Geom. 91, 21–42.
Chen, B. Y. and Wei, S. W. (2009). Growth estimates for warping functions and
their geometric applications, Glasgow Math. J. 51, 579–592.
Chen, B. Y. and Wu, B. Q. (1988). Mixed foliate CR-submanifolds in a complex
hyperbolic space are nonproper, Internat. J. Math. Math. Sci. 11, 507–515.
Chen, B. Y. and Yano, K. (1971). Minimal submanifolds of a higher dimensional
sphere, Tensor (N.S.) 22, 369–393.
Chen, C.-H. (1999). Warped products of metric spaces of curvature bounded from
above, Trans. Amer. Math. Soc. 351, 4727–4740.
Chern, S. S. (1968). Minimal Submanifolds in a Riemannian Manifold, (Univer-
sity of Kansas).
Cherubini, C., Bini, D., Capozziello, S. and Ruffini, R. (2002). Second order scalar
invariants of the Riemann tensor: applications to black hole spacetimes, Int.
J. Modern Phys. D. 11, no. 6, 827–841.
Chojnacka-Dulas, J., Deszcz, R., Glogowska, M. and Prvanović, M. (2013). On

warped product manifolds satisfying some curvature conditions, J. Geom.
Phys. 74, 328–341.
Christoffel, E. B. (1869), ÜUber die Transformation der homogenen Differen-
tialausdrücke zweiten Grades, Jour. Reine Angew. Math. 70, 46–70.
Clarke, C. J. S. (1970). On the global isometric embedding of pseudo-Riemannian
manifolds, Proc. Roy. Soc. London, A, 314, 417–428.
Dajczer, M. (1980). Reducao da Codimensao de Imersoes Isometricas Regulares,
Doctoral thesis, IMPA.
Dajczer, M. and Florit, L. A. (1998). On Chen’s basic equality, Illinois J. Math.
42, 97–106.
Dajczer, M. and Nomizu, K. (1981). On flat surfaces in S13 and H13 , Manifolds
and Lie Groups, Birkhäuser, 71–108.
Dajczer, M. and Tojeiro, R. (2004). Isometric immersions in codimension two of
warped products into space forms, Illinois J. Math. 48, 711–746.
Dajczer, M. and Vlachos, T. (2012). Isometric immersions of warped products,
arXiv:1108.3905.
De, U. C., Murathan, C. and Ozgur, C. (2010). Pseudo symmetric and pseudo
Ricci symmetric warped product manifolds, Commun. Korean Math. Soc.
25, 615–621.
Deahna, F. (1840). Über die Bedingungen der Integrabilitat, J. Reine Angew.
Math. 20, 340–350.
Decruyenaere, F., Dillen, F., Verstraelen, L. and Vrancken, L. (1994). The semir-
ing of immersions of manifolds, Beiträge Algebra Geom. 34, 209–215.
Deng, S. (2009). An improved Chen-Ricci inequality, Int. Electron. J. Geom. 2,
no. 2, 39–45.
Deprez, J. (1986). Semiparallel hypersurfaces, Rend. Sem. Mat. Univ. Politec.
Torino 44, 303–316
de Rham, G. (1952). Sur la reductibilit d’un espace de Riemann, Comment. Math.
Helv. 26, 328–344.
Defever, F., Deszcz, R., Glogowska, M., Goldberg, V. V. and Verstraelen, L.
(2002). A class of four-dimensional warped products, Demonstratio Math.
35, 853–864.
Deszcz, R., Glogowska, M., Jelowicki, J. and Zafindratafa, G. (2016). Curva-
ture properties of some class of warped product manifolds, Int. J. Geom.
Methods Mod. Phys. 13, no. 1, 1550135, 36 pages.
Deszcz, R., Hotloś, M. and Sentürk, Z. (2001). On curvature properties of quasi-
Einstein hypersurfaces in semi-Euclidean spaces, Soochow J. Math. 27, 375–
389.
Deszcz, R. and Kucharski, M. (1999). On curvature properties of certain gener-
alized Robertson-Walker spacetimes, Tsukuba J. Math. 23, 113–130.
Deszcz, R., Verstraelen, L. and Vrancken, L. (1991). The symmetry of warped
product space-times, Gen. Relativity Gravitation 23, 671–681.
Dillen, F. and Nölker, S. (1993). Semi-parallelity, multi-rotation surfaces and the
helix-property, J. Reine Angew. Math. 435, 33–63.
Bibliography 461
Dillen, F., Opozda, B., Verstraelen, L. and Vrancken, L. (1987). On totally real
3-dimensional submanifolds of the nearly Kaehler 6-sphere, Proc. Amer.
Math. Soc. 99, 741–749.
Dillen, F., Opozda, B., Verstraelen, L. and Vrancken, L. (1988). On totally real
surfaces of the nearly Kaehler 6-sphere, Geom. Dedicata 27, 325–334.
Dillen, F., Verstraelen, L. and Vrancken, L. (1990). Classification of totally real
3-dimensional submanifolds of S 6 (1) with K ≥ 1/16, J. Math. Soc. Japan
42, 565–584.
Dillen, F. and Vrancken, L. (1994). Calabi-type composition of affine spheres,
Differential Geom. Appl. 4, 303–328.
Dillen, F. and Vrancken, L. (1996). Totally real submanifolds in S 6 (1) satisfying
Chen’s equality, Trans. Amer. Math. Soc. 348, 1633–1646.
Dobarro, F. and Lami Dozo, E. (1987). Scalar curvature and warped products of
Riemann manifolds, Trans. Amer. Math. Soc. 303, 161–168.
Dobarro, F. and Ünal, B. (2005). Curvature of multiply warped products, J.
Geom. Phys. 55, 75–106.
Dobarro, F. and Ünal, B. (2007). About curvature, conformal metrics and warped
products, J. Phys. A 40, no. 46, 13907–13930.
do Carmo, M. and Dajczer, M. (1983) Rotation hypersurfaces in spaces of con-
stant curvature, Trans. Amer. Math. Soc. 277, 685–709.
Dragomir, S. and Ornea, L. (1998). Locally conformal Kähler geometry, (Progress
in Mathematics, 155. Birkhäuser Boston, Inc., Boston, MA.)
Duggal, K. L. and Sahin, B. (2010). Differential geometry of lightlike submani-
folds, (Birkhäuser Verlag, Basel).
Duggal, K. L. and Sharma, R. (1999). Symmetries of spacetimes and Riemannian
manifolds, Kluwer Academic Publishers, Dordrecht, 1999.
Dwivedi, M. K. and Kim, J.-S. (2012). Chen-Tripathi inequality for warped prod-
uct submanifolds of S-space forms, An. Stiint. Univ. Al. I. Cuza Iasi. Mat.
(N.S.) 58, 195–208.
Ejiri, N. (1981). Totally real submanifolds in a 6-sphere, Proc. Amer. Math. Soc.
83, 759–763.
Erbacher, J. (1971). Reduction of the codimension of an isometric immersion, J.
Differential Geom. 5, 333–340.
Eschenburg, J.-H. and Tribuzy, R. (1993). Existence and uniqueness of maps into
affine homogeneous spaces, Rend. Sem. Mat. Univ. Padova, 89, 11–18.
Escobales, R. (1975). Riemannian submersions with totally geodesic fibers, J.
Differential Geom. 10, 253–276.
Escobales, R. (1978). Riemannian submersions from complex projective spaces,
J. Differential Geom. 13, 93–107.
Etayo, F. (1998). On quasi-slant submanifolds of an almost Hermitian manifold,
Publ. Math. Debrecen 53, 217–223.
Faghfouri, M. and Ghaffarzadeh, N. (2015). On doubly warped product subman-
ifolds of generalized (κ, µ)-space forms, Afr. Mat. 26, 1443–1455.
Faghfouri, M. and Majidi, A. (2015). On doubly warped product immersions. J.
Geom. 106, 243–254.
Fernández-Lápez, M., Garcia-Rio, E., Kupeli, D. N. and Unal, B. (2001). A cur-

vature condition for a twisted product to be a warped product, Manuscripta
Math. 106, 213–217.
Friedrich, T. and Grunewald, R. (1985). On the first eigenvalue of the Dirac
operator on 6-dimensional manifolds, Ann. Global Anal. Geom. 3, 265–273.
Friedrich, T. and Ivanov, S. (2002). Parallel spinors and connections with skew-
symmetric torsion in string theory, Asian J. Math. 6, 303–335.
Gadea, P. M. and Montesinos Amilibia, A. (1989). Spaces of constant para-
holomorphic sectional curvature, Pacific J. Math. 136, 85–101.
Gauss, C. F. (1827). Disquisitiones generales circa superficies curvas, Comment.
Soc. Sci. Gotting. Recent. Classis Math. 6.
Germain, S. (1831). Mémoire sur la coubure des surfaces, J. Reine Angrew. Math.
7, 1–29.
Germain, S. (1921). Recherches sur la theories des surfaces élastiques, Paris.
Gheysens, L., Verheyen, P. and Verstraelen, L. (1983). Characterization and ex-
amples of Chen submanifolds, J. Geom. 20, 47–62.
Graves, J. T. (1845). On a connection between the general theory of normal
couples and the theory of complete quadratic functions of two variables,
Phil. Magaz. 26, 315–320.
Graves, L. K. (1979a). On codimension one isometric immersions between indef-
inite space forms, Tsukuba J. Math. 3, 17–29.
Graves, L. K. (1979b). Codimension one isometric immersions between Lorentz
spaces, Trans. Amer. Math. Soc. 252, 367–392.
Gray, A. (1969). Almost complex submanifolds of the six sphere, Proc. Amer.
Math. Soc. 20, 277–279.
Gray, A. (1970). Nearly Kähler manifolds, J. Differential Geometry, 4, 283–309.
Greene, R. E. (1970). Isometric embeddings of Riemannian and pseudo-Rieman-
nian manifolds, Mem. Amer. Math. Soc. 97, 1–63.
Greenfield, S. (1968). Cauchy-Riemann equations in several variable, Ann. Sc.
Norm. Sup. Pisa, 22, 275–314.
Gromoll, D. and Grove, K. (1988). The low-dimensional metric foliations of Eu-
clidean spheres, J. Differential Geom. 28, 143–156.
Gromov, M. L. (1971). A topological technique for the construction of solutions
of differential equations and inequalities, Intern. Congr. Math., 2, 221–225.
Gromov, M. L. (1985). Isometric immersions of Riemannian manifolds, The math-
ematical heritage of Élie Cartan (Lyon, 1984), Astérisque 1985, 129–133.
Gromov, M. L. and Rokhlin, V. A. (1970). Embeddings and immersions in Rie-
mannian geometry, Russian Math. Surveys, 25, 1–57.
Gülbahar, M., Kilic, E. and Celik, S. S. (2015). Special proper pointwise slant
surfaces of a locally product Riemannian manifold, Turkish J. Math. 39,
884–899.
Haesen, S., Nistor, A. I. and Verstraelen, L. On growth and form and geometry
I, Kragujevac J. Math. 36 (2012), 5–25.
Haider, S. M. K., Thakur, M. and Advin (2012). Warped product skew CR-
submanifolds of a cosymplectic manifold, Lobach. J. Math. 33, 262–273.
Harvey, R. and Lawson, H. B. (1982). Calibrated foliation, Amer. J. Math. 104,
607–633.
Bibliography 463
Hasegawa, I. and Mihai, I. (2003). Contact CR-warped product submanifolds in

Sasakian manifolds, Geom. Dedicata 102, 143–150.
Hashimoto, H. and Mashimo, K. (1999). On some 3-dimensional CR submanifolds
in S 6 , Nagoya Math. J. 156, 171–185.
Hawking, S. W. and Ellis, G. F. R. (1973). The large scale structure of space-time,
(Cambridge University Press).
He, C., Petersen, P. and Wylie, W. (2014). Warped product Einstein metrics over
spaces with constant scalar curvature, Asian J. Math. 18, 159–189.
He, C., Petersen, P. and Wylie, W. (2015). Warped product rigidity, Asian J.
Math. 19, 135–170.
Helfrich, W. (1973). Elastic properties of lipid bilayers: Theory and possible
experiments, Z. Naturforsch. 28, 693–703.
Helgason, S. (1978). Differential Geometry, Lie groups and Symmetric Spaces,
(Academic Press, NY).
Henry, R. C. (2000). Kretschmann scalar for a Kerr-Newman black hole, Astro-
phys. J. 535, 350–353.
Hiepko, S. (1979). Eine innere Kennzeichnung der verzerrten Produkte, Math.
Ann. 241, 209–215.
Hilbert, D. (1915). Die Grundlagen der Physik, Konigl. Gesell. d. Wiss. Gttingen,
Nachr. Math.-Phys. Kl. 395–407.
Hilbert, D. (1924). Die grundlagen der physik, Math. Ann. 92, 1–32.
Hotloś, M. (2004). On conformally symmetric warped products, Ann. Acad. Ped-
agog. Crac. Stud. Math. 4, 75–85.
Hou, Z., Deng, S. and Kaneyuki, S. (1997). Dipolarizations in compact Lie alge-
bras and homogeneous para-Kähler manifold, Tokyo J. Math. 20, 381–388.
Hui, S. K., Uddin, S., Ozel, C. and Mustafa, A. (2012). Warped product subman-
ifolds of LP-Sasakian manifolds, Discrete Dyn. Nat. Soc. Art. ID 868549.
Jamal, N., Khan, K. A. and Khan, V. A. (2010). Generic warped product sub-
manifolds of locally conformal Kähler manifolds, Acta Math. Sci. Ser. B
Engl. Ed. 30, no. 5, 1457–1468.
Jamali, M. and Shahid, M. H. (2014). Multiply warped product submanifolds of
a generalized Sasakian space form, Int. Electron. J. Geom. 7, no. 2, 72–83.
Janet, M. (1926). Sur la possibilité de plonger un espace riemannien donné dans
un espace euclidien, Ann. Soc. Math. Polon. 5, 38–43.
Jebsen, J. T. (1921). On the general spherically symmetric solutions of Einsteins
gravitational equations in vacuo, Ark. Mat. Ast. Fys,. 15, 1–9.
Kazan, S. and Sahin, B. (2013). Characterizations of twisted product manifolds
to be warped product manifolds, Acta Math. Univ. Comenian. 82, 253–263.
Khan, K. A., Ali, S. and Jamal, N. (2008). Generic warped product submanifolds
in a Kaehler manifold, Filomat 22, 139–144.
Khan, K. A. and Chahal, K. S. (2010). Warped product pseudo-slant submanifold
of trans-Sasakian manifolds, Thai J. Math. 8, 263–273.
Khan, K. A. and Chahal, K. S. (2014). Semi-slant warped product submanifolds
of a trans-Sasakian manifold, An. Univ. Craiova Ser. Mat. Inf. 41, 38–46.
Khan, K. A., Khan, V. A. and Uddin, S. (2007). Warped product contact CR-
submanifolds of trans-Sasakian manifolds, Filomat 21, 55–62.
Khan, K. A., Khan, V. A. and Uddin, S. (2008). Warped product submanifolds

of cosymplectic manifolds, Balkan J. Geom. Appl. 13, no. 1, 55–65.
Khan, K. A., Uddin, S. and Sachdeva, R. (2012). Semi-invariant warped product
submanifolds of cosymplectic manifolds, J. Inequal. Appl. 2012, 2012:19.
Khan, M. A., Uddin, S. and Sachdeva, R. (2012). Semi-invariant warped product
submanifolds of cosymplectic manifolds, J. Inequal. Appl. 2012:19, 12 pages.
Khan, V. A. and Jamal, N. (2010). CR-warped product submanifolds of locally
conformal Kaehler manifolds, Toyama Math. J. 33, 1–19.
Khan, V. A., Jamal, N. and Al-Kathiri, A. (2009). A note on warped products
submanifolds of nearly Kaehler manifolds, Toyama Math. J. 32, 59–73.
Khan, V. A. and Khan, K. A.(2014). Semi-slant warped product submanifolds of
a nearly Kaehler manifold, Differ. Geom. Dyn. Syst. 16, 168–182.
Khan, V. A. and Khan, K. A. (2009). Generic warped product submanifolds in
nearly Kaehler manifolds, Beiträge Algebra Geom. 50, no. 2, 337–352.
Khan, V. A., Khan, K. A. and Uddin, S. (2007). Warped product CR-
submanifolds in nearly Kaehler manifolds, SUT J. Math. 43, 201–213.
Khan, V. A., Khan, K. A. and Uddin, S. (2008). Contact CR-warped product
submanifolds of Kenmotsu manifolds, Thai J. Math. 6, 139–145.
Khan, V. A., Khan, K. A. and Uddin, S. (2009a). CR-warped product submani-
folds in a Kaehler manifold, Southeast Asian Bull. Math. 33, 865–874.
Khan, V. A., Khan, K. A. and Uddin, S. (2009b). Contact CR-warped product
submanifolds of Kenmotsu manifolds, Thai J. Math. 6, 307–314.
Khan, V. A., Khan, K. A. and Uddin, S. (2011). A note on warped product
submanifolds of Kenmotsu manifolds, Math. Slovaca 61, 79–92.
Khan, V. A. and Khan, M. A. (2007). Semi-slant submanifolds of a nearly Kaehler
manifold, Turkish J. Math. 31, 341–353.
Khan, V. A. and Shuaib, M. (2007). Some warped product submanifolds of a
Kenmotsu manifold, Bull. Korean Math. Soc. 51, 863–881.
Kim, D.-S. (2000). Einstein warped product spaces, Honam Math. J. 22, 107–111.
Kim, D.-S. and Kim, Y.-H. (2003). Compact Einstein warped product spaces with
nonpositive scalar curvature, Proc. Amer. Math. Soc. 131, 2573–2576.
Kim, Y.-H. and Yoon, D. W. (2004). Inequality for totally real warped products
in locally conformal Kaehler space forms, Kyungpook Math. J. 44, 585–592.
Kobayashi, S. and Nomizu, K. (1963). Foundations of Differential Geometry, I,
(J. Wiley, New York).
Kobayashi, S. and Nomizu, K. (1968). Foundations of Differential Geometry, II,
(J. Wiley, New York).
Kruchkovich, G. I. (1957). On semi-reducible Riemannian spaces (in Russian),
Dokl. Akad. Nauk SSSR 115, 862–865.
Kühnel, W. (1988). Conformal transformations between Einstein spaces, in: Con-
formal geometry (Bonn, 1985), 105–146, Vieweg, Braunschweig.
Kulkarni, R. S. (1979). The values of sectional curvature in indefinite metrics,
Comm. Math. Helv. 54, 173–176.
Kumar, R. (2013). Characterization of GCR-lightlike warped product of indefinite
Kenmotsu manifolds, Mat. Vesnik 65, 533–546.
Bibliography 465
Kumar, R., Rani, R., Gogna, M. and Nagaich, R. K. (2014). Characterization

of GCR-lightlike warped product of indefinite Sasakian manifolds, Arab J.
Math. Sci. 20, 191–204.
Lawn, M.-A. and Ortega, M. (2015). A fundamental theorem for hypersurfaces
in semi-Riemannian warped products, J. Geom. Phys. 90, 55–70.
Lawson, H. B. (1969). Local rigidity theorems for minimal hypersurfaces, Ann.
Math. 89, 187–197.
Lee, J. W. and Sahin, B. (2014). Pointwise slant submersions, Bull. Korean Math.
Soc. 51, 1115–1126.
Li, H., Ma, H. and Su, L. (2008). Lagrangian spheres in the 2-dimensional complex
space forms, Israel J. Math. 166, 113–124.
Lone, M. A. and Shahid, M. H. (2016). Hemi-slant submanifolds of nearly Kaehler
manifolds, arXiv:1603.014031v1 [math.DG].
Lotta, A. (1996). Slant submanifolds in contact geometry, Bull. Math. Soc.
Roumanie 39, 183–198.
Magid, M. A. (1984). Isometric immersions of Lorentz space with parallel second
fundamental forms. Tsukuba J. Math. 8, 31–54.
Malek, F. and Nejadakbary, V. (2011). Warped product submanifold in general-
ized Sasakian space form, Acta Math. Acad. Paedagog. Nyházi. 27, 325–338.
Mantica, C. A., Luca, G. M. and De, U. C. (2016a). A condition for a perfect-
fluid space-time to be a generalized Robertson-Walker space-time, J. Math.
Phys. 57, no. 022508, 6 pages; Erratum, ibid, 57, no. 049901.
Mantica, C. A. and Molinari, L. G. (2016a). On the Weyl and Ricci tensors of
generalized Robertson-Walker space-times, J. Math. Phys. 57, no. 102502.
Mantica, C. A., Suh, Y.-J. and De, U. C. (2016b). A note on generalized
Robertson-Walker space-times, Int. J. Geom. Methods Mod. Phys. 13(6),
no. 1650079, 1–9.
Mantica, C. A. and Molinari, L. G. (2017). Generalized Robertson-Walker space-
times - A survey, Intern. J. Geom. Meth. Modern Phys. 14(3), no.1730001.
Matsumoto, K. and Bonanzinga, V. (2007). Doubly warped product CR-subma-
nifolds in a locally conformal Kaehler space form II, An. Stiint. Univ. Al.
I. Cuza Iasi. Mat. (N.S.) 53, suppl. 1, 235–248.
Matsumoto, K. and Bonanzinga, V. (2008). Doubly warped product CR-subman-
ifolds in a locally conformal Kaehler space form, Acta Math. Acad. Paeda-
gog. Nyhzi. (N.S.) 24, no. 1, 93–102.
Matsumoto, K. and Mihai, I. (2002). Warped product submanifolds in Sasakian
space forms, SUT J. Math. 38, 135–144.
Meumertzheim, M., Reckziegel, H. and Schaaf, M. (1999). Decomposition of
twisted and warped product nets, Results Math. 36, 297–312.
Mihai, A. (2004). Warped product submanifolds in complex space forms, Acta
Sci. Math. (Szeged) 70, 419–427.
Mihai, A. (2005a). Warped product submanifolds in quaternion space forms, Rev.
Roumaine Math. Pures Appl. 50, 283–291.
Mihai, A. (2005b). Warped product submanifolds in quaternion space forms, Acta
Univ. Apulensis Math. Inform. 10, 31–38.
Mihai, A. (2005c). Warped product submanifolds in generalized complex space

forms, Acta Math. Acad. Paedagog. Nyhazi. (N.S.) 21, 79–87.
Mihai, A. and Radulescu, I. N. (2012). An improved Chen-Ricci inequality for
Kaehlerian slant submanifolds in complex space forms, Taiwan. J. Math.
16, 761-770.
Mihai, I. (2004). Contact CR-warped product submanifolds in Sasakian space
forms, Geom. Dedicata 109, 165–173.
Milousheva, V. (2013). Marginally trapped surfaces with pointwise 1-type Gauss
in Minkowski 4-space, Intern. J. Geom. 2, 34–43.
Mirandola, H. and Vitório, F. (2012). Multiple warped product metrics into
quadrics, arXiv:1210.1812v1.
Mirandola, H. and Vitório, F. (2015). Global isometric embeddings of multiple
warped product metrics into quadrics, Kodai Math. J. 38, 119–134.
Moore, J. D. (1971). Isometric immersions of riemannian products, J. Differential
Geom. 5, 159–168.
Morimoto, A. (1964). On normal almost contact structure with regularity, Tohoku
Math. J. 16, 90–104.
Munteanu, M. I. (2005). Warped product contact CR-submanifolds of Sasakian
space forms, Publ. Math. Debrecen 66 , 75–120.
Munteanu, M. I. (2007a). A note on doubly warped product contact CR-
submanifolds in trans-Sasakian manifolds, Acta Math. Hung. 116, 121–126.
Munteanu, M. I. (2007b). Doubly warped product CR-submanifolds in locally
conformal Kähler manifolds, Monatsh. Math. 150, 333–342.
Murathan, C., Arslan, K., Ezentas, R. and Mihai, I. (2006), Warped product
submanifolds in Kenmotsu space forms, Taiwanese J. Math. 10, 1431–1441.
Mustafa, A., De, A. and Uddin, S. (2015). Characterization of warped product
submanifolds in Kenmotsu manifolds, Balkan J. Geom. Appl. 20, 86–97.
Mustafa, A., Uddin, S., Khan, V. A. and Wong, B. R. (2013). Contact CR-warped
product submanifolds of nearly trans-Sasakian manifolds, Taiwan. J. Math.
17, 1473–1486.
Mustafa, A., Uddin, S. and Wong, B. R. (2014). Generalized inequalities on
warped product submanifolds in nearly trans-Sasakian manifolds, J. In-
equal. Appl. 2014:346, 16 pages.
Nagy, P.-A. On nearly-Kähler geometry, Ann. Global Anal. Geom. 22 , 167–178.
Nash, J. F. (1956). The imbedding problem for Riemannian manifolds, Ann.
Math. 63, 20–63.
Newlander, A. and Nirenberg, L. (1957). Complex analytic coordinates in almost
complex manifolds, Annals of Math. 65, 391–404.
Nölker, S. (1996). Isometric immersions of warped products, Differential Geom.
Appl. 6, 1–30.
Nomizu, K. and Pinkall, U. (1987). On the geometry of affine immersions, Math.
Z. 195, 165–178.
Nomizu, K. and Sasaki, T. (1994). Affine Differential Geometry: Geometry of
Affine Immersions, (Cambridge University Press).
Ogawa, Y. (1978). On conformally flat spaces with warped product Riemannian
metric, Nat. Sci. Rep. Ochanomizu Univ. 29, 117–127.
Bibliography 467
Ogiue, K. (1964). On almost contact manifolds admitting axiom of planes or

axiom of free mobility, Kodai Math. Sem. Rep. 16, 223–232.
Ogiue, K. (1974). Differential geometry of Kähler submanifolds, Adv. Math. 13,
73–114.
Olteanu, A. (2008). CR-doubly warped product submanifolds in Sasakian space
forms, Bull. Transilv. Univ. Brasov Ser. III 1(50), 269–277.
Olteanu, A. (2009). Contact CR-doubly warped product submanifolds in Ken-
motsu space forms, J. Inequal. Pure Appl. Math. 10, no. 4, Article 119.
Olteanu, A. (2010a). Multiply warped product submanifolds in Sasakian space
forms, Int. Electron. J. Geom. 3, no. 1, 1–10.
Olteanu, A. (2010b). A general inequality for doubly warped product submani-
folds, Math. J. Okayama Univ. 52, 133–142.
O’Neill, B. (1966). The fundamental equations of a submersion, Michigan Math.
J. 13, 459–469.
O’Neill, B. (1983). Semi-Riemannian Geometry with Applications to Relativity,
(Academic Press, New York).
Oprea, T. (2007). Chen’s inequality in the Lagrangian case, Colloq. Math. 108,
163–169.
Osserman, R. (1990). Curvature in the eighties, Amer. Math. Mon. 97, 731–756.
Palma, A. R. (2007). Pseudo-slant warped products in trans-Sasakian manifolds,
Bull. Transilv. Univ. Brasov Ser. B (N.S.) 14(49), suppl., 235–240.
Papaghiuc, N. (1994). Semi-slant submanifolds of a Kaehlerian manifold, An.
Stiint. Univ. Al. I. Cuza Iasi Sec. I a Mat. 40, 55–61.
Park, K.-S. (2015). Pointwise almost h-semi-slant submanifolds, Internat. J.
Math. 26, no. 12, 1550099, 26 pages.
Park, K.-S. and Sahin, B. (2014). Semi-slant Riemannian maps into almost Her-
mitian manifolds, Czechoslovak Math. J. 64, 1045–1061.
Penrose, R. (1965). Gravitational collapse and space-time singularities, Phys. Rev.
Lett. 14, 57–59.
Perktas, S. Y., Kilic, E. and Keles, S. (2012). Warped product submanifolds of
Lorentzian paracosymplectic manifolds, Arab. J. Math. 1, 377–393.
Petrović, M., Rosca, R. and Verstraelen, L. (1989). Exterior concurrent vector
fields on Riemannian manifolds I. Some general results, Soochow J. Math.
15, 179–187.
Petrović, M., Rosca, R. and Verstraelen, L. (1993). Exterior concurrent vector
fields on Riemannian manifolds II. Examples of exterior concurrent vector
fields on manifolds, Soochow J. Math. 19, 357–368.
Poincaré, H. (1895). Analysis situs, J. Ecole polytech. (2) 1, 1–121.
Ponge, R. and Reckziegel, H. (1993). Twisted products in pseudo-Riemannian
geometry, Geom. Dedicata, 48, 15–15.
Ranjan, A. (1985). Riemannian submersions of spheres with totally geodesics
fibers, Osaka J. Math. 22, 243–260.
Rashevsky, P. K. (1948). The scalar field in a stratified space, Trudy Sem. Vektor.
Tenzor. Anal. 6, 225–248.
Reckziegel, H. (1981). On the problem whether the image of a given differentiable
map into a Riemannian manifold is contained in a submanifold with parallel
second fundamental form, J. Reine. Angew. Math. 325, 87–104
Reckziegel, H. (1985). Horizontal lifts of isometric immersions into the bundle

space of a pseudo-Riemannian submersion, Lecture Notes Math. 1156, 264–
279.
Riemann, B. (1854). Über die Hypothesen welche der Geometrie zu Grunde liegen,
Habili. Abhand. Königlichen Ges. Wissen. Göttingen, 13.
Rimoldi, M. (2011). A remark on Einstein warped products, Pacific J. Math.
252, 207–218.
Rodriguez Montealegre, C. and Vrancken, L. (2009). Warped product minimal
Lagrangian immersions in complex projective space, Results Math. 56, 405–
420.
Romero, A., Rubio, R. M. and Salamanca, J. J. (2013a). Parabolicity of spacelike
hypersurfaces in generalized Robertson-Walker spacetimes. Applications to
uniqueness results, Int. J. Geom. Methods Mod. Phys. 10, no. 8, 1360014.
Romero, A., Rubio, R. M. and Salamanca, J. J. (2013b). Uniqueness of complete
maximal hypersurfaces in spatially parabolic generalized Robertson-Walker
spacetimes, Classical Quantum Gravity 30, no. 11, 115007, 13 pages.
Rosca, R. (1972). On null hypersurfaces of a Lorentzian manifold. Tensor (N.S.)
23, 66–74.
Rozenfeld, B. A. (1949). On unitary and stratified spaces, Trudy Sem. Vektor.
Tenzor. Anal. 7, 260–275.
Sachdeva, R., Kumar, R. and Bhatia, S. S. (2015). Warped product slant lightlike
submanifolds of indefinite Sasakian manifolds, Balkan J. Geom. Appl. 20,
no. 1, 98–108.
Sahin, B. (2006a). Nonexistence of warped product semi-slant submanifolds of
Kaehler manifolds. Geom. Dedicata 117, 195–202.
Sahin, B. (2006b). Slant submanifolds of an almost product Riemannian manifold,
J. Korean Math. Soc. 43, 717–732.
Sahin, B. (2006c). Warped product semi-invariant submanifolds of a locally prod-
uct Riemannian manifold, Bull. Math. Soc. Sci. Math. Roumanie (N.S.)
49(97), 383–394.
Sahin, B. (2007a). Notes on doubly warped and doubly twisted product CR-
submanifolds of Kaehler manifolds, Mat. Vesnik 59, 205–210.
Sahin, B. (2007b). Slant submanifolds of quaternion Kaehler manifolds, Commun.
Korean Math. Soc. 22, 123-135.
Sahin, B. (2009a). Warped product submanifolds of Kaehler manifolds with a
slant factor, Ann. Polon. Math. 95, 207–226.
Sahin, B. (2009b). Every totally umbilical proper slant submanifold of a Kähler
manifold is totally geodesic, Results Math. 54, 167–172.
Sahin, B. (2009c). Warped product semi-slant submanifolds of a locally product
Riemannian manifold, Studia Sci. Math. Hungar. 46, 169–184.
Sahin, B. (2010). Skew CR-warped products of Kaehler manifolds, Math. Com-
mun. 15, 189–204.
Sahin, B. (2011). Slant submersions from almost Hermitian manifolds, Bull. Math.
Soc. Sci. Math. Roumanie (N.S.) 54, 93–105.
Sahin, B. (2013a). Slant Riemannian maps to Kähler manifolds, Int. J. Geom.
Methods Mod. Phys. 10, no. 2, 1250080.
Bibliography 469
Sahin, B. (2013b). Warped product pointwise semi-slant submanifolds of Kähler

manifolds, Port. Math. 70, 251–268.
Sahin, B. (2013c). Slant Riemannian maps from almost Hermitian manifolds,
Quaest. Math. 36, 449–461.
Sahin, B. (2013d). Warped product pointwise semi-slant submanifolds of Kähler
manifolds, Port. Math. 70, 251–268.
Sahin, B. and Günes, R. (2008). CR-warped product submanifolds of nearly
Kaehler manifolds, Beiträge Algebra Geom. 49, 383–397.
Sahin, B. and Sahin, F. (2015). Homology of contact CR-warped product sub-
manifolds of an odd-dimensional unit sphere, Bull. Korean Math. Soc. 52,
215–222.
Salavessa, I. M. C. (2003). On the Kähler angles of submanifolds, Port. Math.
(N.S.) 60, 215-235.
Salavessa, I. M. C. and Valli, G. (2002). Minimal submanifolds of Kähler-Einstein
manifolds with equal Kähler angles, Pacific J. Math. 205, 197-235.
Sánchez, M. (1999). On the geometry of generalized Robertson-Walker space-
times: curvature and Killing fields, J. Geom. Phys. 31, no. 1, 1–15.
Sasahara, T. (1999). CR-submanifolds in complex hyperbolic spaces satisfying an
equality of Chen, Tsukuba J. Math. 23, no. 3, 565–583.
Sasahara, T. (2014). Ideal CR submanifolds in non-flat complex space forms,
Czechoslovak Math. J. 64(139) , no. 1, 79–90.
Sasaki, S. (1960). On differentiate manifolds with certain structures which are
closely related to almost contact structure I, Tohoku Math. J. 12, 459-476.
Sato, N. (1981). Certain CR-submanifolds of almost-Hermitian manifolds,
preprint.
Sato, N. (1982). Certain anti-holomorphic submanifolds of almost Hermitian man-
ifolds, Sci. Rep. Niigata Univ. Ser. A 18, 1–9.
Schaaf, M. (2000). Decomposition of isometric immersions between warped prod-
ucts, Beiträge Algebra Geom. 41, 427–436.
Schläfli, L. (1873). Nota alla Memoria del sig. Beltrami, Ann. Mat. Pura Appl. 5,
178–193.
Sergey, S. (2016). Liouville-type theorems for twisted and warped products man-
ifolds, arXiv:1608.03590.
Schouten, J. A. (1954). Ricci-calculus, 2nd edition, (Springer, London).
Schwarzschild, K. (1916). Über das Gravitationsfeld eines Massenpunktes nach
der Einsteinschen Theorie, Preuss. Akad. Wissen. Sitz. 7, 189–196.
Segre, C. (1891). Sulle varietà che rappresentano le coppie di punti di due piani
o spazi, Rend. Cir. Mat. Palermo 5, 192–204.
Sekigawa, K. (1983). Almost complex submanifolds of a 6-dimensional sphere,
Kodai Math. J., 6, 174–185.
Shenawy, S. and Unal, B. (2015). 2-Killing vector fields on warped product man-
ifolds, Internat. J. Math. 26, no. 9, 1550065, 17 pages.
Shukla, S. S. and Tiwari, S. K. (2010). Contact CR-warped product submanifolds
in locally conformal almost cosymplectic manifolds, Bull. Calcutta Math.
Soc. 102, no. 2, 121–128.
Sikorav, J. C. (1986). Non-existence de sous-variété lagrangienne exacte dans Cn

(d’après Gromov), Aspects dynamiques et topologiques des groupes infinis
de transformation de la mcanique (Lyon, 1986), 95–110.
Slesar, V., Sahin, B. and Vilcu, G.-E. (2014). Inequalities for the Casorati curva-
tures of slant submanifolds in quaternionic space forms, J. Inequal. Appl.
2014, 2014:123.
Spivak, M. (1979). A Comprehensive Introduction to Differential Geometry, vol.
IV, (Publish or Perish, Berkeley, CA).
Sternberg, S. (1964). Lectures on Differential Geometry, (Prentice-Hall, NJ).
Suceavă, B. (2001). The Chen invariants of warped products of hyperbolic planes
and their applications to immersibility problems, Tsukuba J. Math. 25,
311–320.
Sular, S. (2015). Doubly warped product submanifolds of a Riemannian manifold
of quasi-constant curvature, An. Stiint. Univ. Al. I. Cuza Iasi. Mat. (N.S.)
61, 235–244.
Sular, S. and Ozgur, C. (2011). Doubly warped product submanifolds of (κ, µ)-
contact metric manifolds, Ann. Polon. Math. 100, 223–236.
Sular, S. and Ozgur, C. (2012a). Contact CR-warped product submanifolds in
generalized Sasakian space forms, Turkish J. Math. 36, 485–497.
Sular, S. and Ozgur, C. (2012b). On quasi-Einstein warped products, An. Stiint.
Univ. Al. I. Cuza Iasi. Mat. (N.S.) 58, 353–362.
Sular, S. and Ozgur, C. (2012c). Contact CR-warped product submanifolds in
generalized Sasakian space forms, Turk. J. Math. 36, 485–497.
Tachibana, S. (1959). On almost-analytic vectors in certain almost-Hermitian
manifolds, Tohoku Math. J. 11, 351–363.
Takagi, R. (1975). Real hypersurfaces in a complex projective space with constant
principal curvatures I, II, J. Math. Soc. Japan, 27, 43–53, 507–516.
Takahashi, T. (1966). Minimal immersions of Riemannian manifolds, J. Math.
Soc. Japan, 18, 380–385.
Takahashi, T. (1969). Sasakian manifold with pseudo-Riemannian metric, Tohoku
Math. J. 21, 271–290.
Takeuchi, M. (1981). Parallel submanifolds of space forms, in: Manifolds and Lie
Groups, 429–447 (Birkhäuser, Boston).
Takeuchi, M. and Kobayashi, S. (1968). Minimal imbeddings of R-spaces, J. Dif-
ferential Geom. 2, 203–215.
Tastan, H. M. (2015). Warped product skew semi-invariant submanifolds of order
1 of a locally product Riemannian manifold, Turkish J. Math. 39, 453–466.
Tastan, H. M. (2016). Biwarped product submanifolds of a Kähler manifold,
(preprint).
Tazawa, Y. (1994a). Construction of slant immersions, Bull. Inst. Math. Acad.
Sinica 22, 153–166.
Tazawa, Y. (1994b). Construction of slant immersions II, Bull. Belg. Math. Soc.
Simon Stevin 1, 569–576.
Tojeiro, R. (2007). Conformal immersions of warped products, Geom. Dedicata
128, 17–31.
Tojeiro, R. (2016). A decomposition theorem for immersions of product manifolds,
Proc. Edinb. Math. Soc. 59, 247–269.
Bibliography 471
Tripathi, M. M. (2012). C-totally real warped product submanifolds, An. Stiint.

Univ. Al. I. Cuza Iasi. Mat. (N.S.) 58, 417–436.
Tsukamoto, Y. (1957). On Kählerian manifolds with positive holomorphic curva-
ture, Proc. Japan Acad. 33, 333–335.
Ucci, J. (1983). On the nonexistence of Riemannian submersions from CP 7 and
QP 3 Proc. Amer. Math. Soc. 88, 698–700.
Udagawa, S. (1986). Spectral geometry of Kaehler submanifolds of a complex
projective space, J. Math. Soc. Japan, 38, 453–471.
Uddin, S. (2010a). Warped product CR-submanifolds of LP-cosymplectic mani-
folds, Filomat 24, 87–95.
Uddin, S. (2010b). CR-warped product submanifolds of Lorentzian manifolds,
Math. Morav. 14, 129–136.
Uddin, S. (2010c). On doubly warped and doubly twisted product submanifolds,
Int. Electron. J. Geom. 3 (2010), no. 1, 35–39.
Uddin, S. and Alqahtani, L. S. (2016). Chen type inequality for warped product
immersions in cosymplectic space forms, J. Nonlin. Sci. Appl. 9, 2914–2921.
Uddin, S., Al-Solamy, F. R. and Khan, K. A. (2016). Geometry of warped product
pseudo-slant submanifolds of nearly Kaehler manifolds, An. Stiint. Univ.
Al. I. Cuza Iasi Sect. I a Mat. 62, 223–234.
Uddin, S., Chen, B. Y. and Al-Solamy, F. R. (2017). Warped product bi-slant
immersions in Kaehler manifolds, Mediterr. J. Math. 14, no. 2, 14:95.
Uddin, S. and Chi, A. Y. M. (2011a). Warped product pseudo-slant submanifolds
of nearly Kaehler manifolds, An. Stiint. Univ. “Ovidius” Constanta Ser.
Mat. 19, 195–204.
Uddin, S. and Chi, A.Y.M. (2011b). Warped product CR-submanifolds of cosym-
plectic manifolds, Ric. Mat. 60, 143–149.
Uddin, S. and Chi, A. Y. M. (2015). Warped product pseudo-slant submanifolds
of nearly Kaehler manifolds, An. Stiint. Univ. ”Ovidius” Constanta Ser.
Mat. 19, no. 3, 195–204.
Uddin, S.and Khan, K. A. (2011). Warped product CR-submanifolds of cosym-
plectic manifolds, Ric. Mat. 60, 143–149.
Uddin, S., Khan, V. A. and Khan, H. H. (2010). Some results on warped product
submanifolds of a Sasakian manifold, Int. J. Math. Math. Sci. 2010, Art.
ID 743074, 9 pages.
Uddin, S., Khan, V. A. and Khan, K. A. (2012). Warped product submanifolds
of a Kenmotsu manifold, Turkish J. Math. 36, 319–330.
Uddin, S., Mustafa, A., Wong, B. R. and Ozel, C. (2014). A geometric inequality
for warped product semi-slant submanifolds of nearly cosymplectic mani-
folds, Rev. Un. Mat. Argentina 55, 55–69.
Ünal, B. (2000). Multiply warped products, J. Geom. Phys. 34,287–301.
Ünal, B. (2001). Doubly warped products, Differential Geom. Appl. 15, 253–263.
Vaisman, I. (1976). On locally conformal almost Kähler manifolds, Israel J. Math.,
24, 338-351.
Vanhecke, L. (1975). Almost Hermitian manifolds with J-invariant Riemann cur-
vature tensor, Rend. Sem. Mat. Univ. e Politec. Torino 34, 487–498.
Vanhecke, L. (1977). On immersions with trivial normal connection in some al-

most Hermitian manifolds, Rend. Mat. (6) 10, 75–86.
Vı̂lcu, G.-E. (2012). Ruled CR-submanifolds of locally conformal Kähler mani-
folds, J. Geom. Phys., 62, 1366–1372.
Vı̂lcu, G.-E. (2013). On Chen invariants and inequalities in quaternionic geometry,
J. Inequal. Appl., 2013:66.
Wallach, N. R. (1972). Minimal immersions of symmetric spaces into spheres, in:
Symmetric spaces, 8, 1–40.
Wang, Y. (2013). Curvature of multiply warped products with an affine connec-
tion, Bull. Korean Math. Soc. 50, 1567–1586.
Weingarten, J. (1861). Ueber eine Klasse auf einander abwickelbarer Flächen, J.
Reine Angew. Math. 59, 382–393.
Wettstein, B. (1978). Congruence and existence of differentiable map, Ph.D. The-
sis (ETH Zürich, no. 6252).
Weyl, H. (1912). Das asymptotische Verteilungsgesetz der Eigenwerte linearer
partieller Differentialgleichungen, Math. Ann. 71, 411-479.
Weyl, H. (1918). Reine Infinitesimalgeometrie, Math. Z. 26, 384–411.
Wilking, B. (2001). Index parity of closed geodesics and rigidity of Hopf fibrations,
Invent. Math. 144, 281–295.
Wolf, J. and Gray, A. (1968). Homogeneous spaces defined by Lie group auto-
morphisms II, J. Differential Geometry, 2, 115–159.
Yano, K. (1940). Concircular geometry I. Concircular transformations, Proc. Imp.
Acad. Tokyo 16, 195–200.
Yano, K. (1965). Differential geometry on complex and almost complex spaces,
(Macmillan Company, New York).
Yano, K. and Chen, B. Y. (1971). On the concurrent vector fields of immersed
manifolds, Kodai Math. Sem. Rep. 23, 343–350.
Yano, K. and Kon, M. (1983). CR submanifolds of Kaehlerian and Sasakian
manifolds, (Birkhäuser, Boston, MA).
Yoon, D. W. (2004). Some inequalities for warped products in cosymplectic space
forms, Differ. Geom. Dyn. Syst. 6, 51–58.
Yoon, D. W. and Cho, K. S (2004). Inequality for warped products in generalized
Sasakian space forms, Int. Math. J. 5, 225–235.
Yoon, D. W., Cho, K. S. and Han, S. G. (2004). Some inequalities for warped
products in locally conformal almost cosymplectic manifolds, Note Mat. 23,
51–60.
Yüksel Perkas, S., Kilic, E. and Keles, S. (2012). Warped product submanifolds
of Lorentzian paracosymplectic manifolds, Arab. J. Math. 1, 377–393.
Zhang, P. (2015). Remarks on Chen’s inequalities for submanifolds of a Rieman-
nian manifold of nearly quasi-constant curvature, Vietnam J. Math. 43,
557–569.
Zhong, H. H. (2001). On totally real submanifolds in a nearly Kähler manifold,
Port. Math. (N.S.) 58, 219–231.
Zhong, Z. (2014). Warping functions of some warped products, J. Math. Anal.
Appl. 412, 1019–1024.
General Index
Cn s∗ , 79 φ-section, 100
A, 28 σ, 26
CHsn (4c), CH n (4c), 118 σ-submanifold, 364
CPsn (4c), CP n (4c), 118 r
σij , 28
D, 28 τ , 14
F , 125 | · |, 2
H, 29 h, 180
H(M ), 415 p, 180
Hsk (c), 16 Cm+1
∗ , 270
H P , 384 Cm+1
1∗ , 271
P , 125 C∗ , 233
Pcr (B), 387 Rm , 427
Q(u, v), 12 R∗ , 271
R, 10 Em ∗ , 234
Ric, 14 En s, 4
Ssk (c), 16 R2n+1 , 108
Sh , 32 A, 103
#S(n), 326 F(M ), 4
∆, 17 O, 361
Ω, 116 S(n), 326
Ωαβ , 46 X(M ), 5
¯ 31
∇σ, Åπ , 103
¯ j σ, 31
∇
Ăπ , 103 Adapted pseudo-orthonormal frame,
δ-invariant, 326 127
R2n+1 (−3), 108 Affine connection, 5
LC, 4 Affine space, 427
L, 48 Affine volume element, 428
ω i , 46 almost complex structure
ωαβ , 46 compatible, 141
k · k, 2 Almost Hermitian manifold
473
constant type, 228 Schläfli, 25

holomorphically planar, 359 Thurston’s geometrization, 1
pointwise constant type, 227 Connection
Anti-holomorphic submanifold, 195 affine, 5
doubly twisted product, 76
Base, 71 induced affine, 428, 429
Base of warped product, 48 Levi-Civita, 5
Basic equality, 257 normal, 30
Basic vector field, 102 torsion free, 5
Basis twisted product, 76
orthonormal, 2 van der Waerden-Bortolotti, 31
pseudo-orthonormal, 126 warped product, 48
Bi-slant submanifold, 252 Weyl, 255
Bilinear form Contact CR-submanifold, 410
non-degenerate symmetric, 2 Contact CR-warped product, 416
symmetric, 1 Contact distribution, 409
Black hole, 98 Contact slant submanifold, 177
Convolution, 78
C-totally real submanifold, 409 Covariant derivative, 5
Calabi composition, 431 CR δ-invariant, 338
Calabi metric, 431 CR-manifold, 196
Canonical fibration, 107 CR-product, 199
Cartan-Janet’s theorem, 25 proper, 199
Cayley algebra, 361 CR-submanifold, 195
Cayley number, 360, 361 doubly twisted product, 242
Centroaffine hypersurface, 429 doubly warped product, 238
Centroaffine normal, 429 mixed foliate, 205
Chen class, 201 twisted product, 238
Chen invariant, 327 CR-warped product, 231
Chen’s basic equality, 332 contact, 416
Chen’s basic inequality, 332 partial Segre, 235
Christoffel symbols, 6 Curvature
Codazzi equation, 31 φ-sectional, 100
classical, 411 doubly twisted product, 76
Complex extensor, 131 pseudo-Kähler manifold, 117
Complex Hessian, 277 Ricci, 14
Complex hyperbolic space, 118 scalar, 14
Complex point, 121 sectional, 13
Complex projective space, 118 submersion, 104
Complex space form, 117 twisted product, 76
generalized, 228 warped product, 50
Complex structure, 116 Curvature-like function, 13
Complex submanifold, 121, 363
Conformal mapping, 19 D-homothetic deformation, 101
Conjecture Derivative
Poincaré, 1 covariant, 5
General Index 475
induced covariant, 8 First normal space, 35

Differential of map, 7 parallel, 35
Direct product, 48 Flat manifold, 13
Distant parallelism, 9 Ricci, 14
Distribution Foliation, 59
contact, 409 autoparallel, 60
holomorphic, 180, 195 spherical, 60
Levi, 415 totally geodesic, 60
maximally φ-invariant, 415 totally umbilical, 60
minimal, 199 Form
pointwise slant, 251 connection, 46
purely real, 180 curvature, 46, 111
slant, 248 fundamental, 116, 383
totally geodesic, 199 second fundamental, 28
totally real, 195 Tchebychev, 429
Divergence, 17 torsion, 428
Doubly twisted product, 76 Formula
Gauss, 26, 428
Einstein manifold, 14 Koszul, 5
Einstein’s static model, 81 Weingarten, 28, 428
Einstein, A., 96 Function
Einstein-Hilbert action, 15 convex, 52
Ellipse of curvature, 365 curvature-like, 13
Ellipse of second curvature, 365 slant, 171, 424
Embedding, 24 strictly convex, 52
partial Segre, 235 warping, 48
Segre, 124 Fundamental 2-form, 116
Veronese, 124 Fundamental equation, 31
Equation Fundamental form, 383
fundamental, 31 affine, 428
Cartan’s structure, 46 second, 28
centroaffine structure, 429 third, 365
Codazzi, 31, 429 Fundamental theorems of submani-
Gauss, 31, 429 folds, 34
O’Neill, 104
Ricci, 31 Gauss’ equation, 31
Equiaffine, 428 Gauss’ formula, 26
Existence theorem, 34, 129, 391 Generic submanifold, 179
Extension of vector field, 26 proper, 179
Extrinsic sphere, 73 Generic warped product, 231
trivial, 231
Fiber, 48, 71 Geodesic, 9
Fibration maximal, 9
canonical, 107 Geodesic ball, 10
Hopf, 105 Geodesic complete, 9
First Bianchi identity, 12 Gradient, 16, 52
Graph hypersurface, 430 Chen, 327

Kretschmann, 98
H-submanifold, 83 Pick, 429
Hessian, 17 Invariant submanifold, 387
complex, 277 Isometry, 19
Hopf fibration, 105
hyperbolic, 271 K-contact manifold, 100
Hypersphere Killing spinor, 361
improper affine, 431 Killing vector field, 18
proper affine, 430 Koszul’s formula, 5
Hypersurface
affine, 428 Lagrangian pseudo-sphere, 133
centroaffine, 429 Lagrangian submanifold, 389
graph, 430 H-umbilical, 130
totally umbilical, 129
Ideal immersion, 332 Laplacian, 17
Identity Lee vector field, 255
first Bianchi, 12 Legendre curve, 314
Ricci, 12 Legendre submanifold, 113
second Bianchi, 12 Lemma
Immersion, 24 Moore, 398
N2 -pseudo umbilical, 68 Length of a vector, 2
δ(n1 , . . . , nk )-ideal, 332 Levi distribution, 415
affine, 427 Levi-Civita connection, 5
contact slant, 177 Lift of covariant tensor, 50
horizontal, 111 Light cone, 4
ideal, 332 Linear isometry, 3
isometric, 24 Local isometry, 19
minimal, 30 Locally strongly convex hypersur-
warped product, 62 face, 430
Whitney, 133, 143 Lorentz manifold, 3
Improper affine hypersphere, 431 Lorentzian complex space form, 117
Indefinite complex space form, 117 Lorentzian Kähler manifold, 116
Indefinite Kähler manifold, 116 Lorentzian metric, 3
Indefinite real space form, 16 Lorentzian real space form, 16
Index
complex, 116 Manifold
pseudo-Riemannian manifold, 3 CR–, 196
pseudo-Riemannian metric, 3 K-contact, 100
symmetric bilinear form, 2 RK-, 227
Induced covariant derivative, 8 3-symmetric, 360
Inequality hyperbolic Kähler, 384
universal, 328 almost Hermitian, 115
Invariant almost Kähler, 115
CR δ-, 338 almost para-Hermitian, 383
δ-, 326 almost Tachibana, 359
General Index 477
conformally flat, 58 Lorentzian, 3

constant curvature, 15 neutral, 3
contact metric, 100 pseudo-Hermitian, 116
convolution, 78 pseudo-Kähler, 116
Einstein, 14, 25, 92 pseudo-Riemannian, 3
flat, 13 relative, 428
geodesic complete, 9 Robertson-Walker, 81
globally conformal Kähler, 254 semi-Riemannian, 3
Hermitian, 116 Minimal distribution, 199
indefinite Hermitian, 116 Minimal submanifold, 30
indefinite Kähler, 116 Minkowski space, 4
Kähler, 115 Mixed totally geodesic, 186
locally conformal Kähler, 174, Moore lemma, 64
254 Moore’s lemma, 64
Lorentz, 3
Lorentzian Kähler, 116 Nash’s embedding theorem, 25
nearly Kähler, 359 Nearly Kähler manifold, 359
para-Kähler, 384 strict, 359
pseudo-Hermitian, 116 Nebulae, 81
pseudo-Kähler, 116 Nijenhuis tensor, 370
pseudo-Riemannian, 3 Non-degenerate plane section, 13
Ricci flat, 14 Non-degenerate subspace, 2
Ricci simple, 96 Norm of a vector, 2
Riemannian, 3 Normal connection, 30
Sasakian, 100 Normal coordinate system, 10
semi-Riemannian, 3 Null curve, 4
strict nearly Kähler, 359
symplectic, 170 O’Neill’s equations, 104
Map O’Neill’s integrability tensor, 103
dual, 7 Operator
Euclidean Segre, 235 affine shape, 428
exponential, 9 blacket, 5
Weingarten, 28 shape, 28
Mean curvature, 29
k-th power, 29 P-isometric, 384
squared, 29 Para-complex number, 383
Mean curvature vector, 29 Para-Kähler n-plane, 384
distribution, 199 Para-Kähler space form, 384
parallel, 30 Parallel normal field, 30
partial, 63 Parallel mean curvature vector, 30
Metric Parallel submanifold, 36
Calabi, 431 Parallel translation, 8
centroaffine, 429 Parallel vector field, 8
conformal flat, 58 Partial mean curvature vector, 63
convolution, 78, 79 Partition, 207, 327
indefinite Kähler, 116 Plane section, 12
non-degenerate, 13 Reduction theorem, 36

Point Reeb vector field, 409
complex, 121 Relative mean curvature, 429
totally real, 127 Relative normalization, 428
Pointwise bi-slant submanifold, 252, Relative nullity space, 357
425 Representation
Pointwise hemi-slant submanifold, warped product, 68
425 Rest space, 82
Pointwise semi-slant submanifold, Ricci curvature, 14
425 warped product, 52
Pointwise slant submanifold, 424 Ricci equation, 31
PR-product, 396 Ricci flat manifold, 14
PR-submanifold, 392 Ricci identity, 12
PR-warped product, 396, 401 Ricci simple manifold, 96
spacelike, 401 Ricci tensor, 14
timelike, 401 Riemann curvature tensor, 10
Problem Riemannian manifold, 3
realization, 432 RK-manifold, 227
Product Rotation hypersurface, 216
CR–, 199
P R-, 396 Sasakian manifold, 100
doubly twisted, 76, 242 Sasakian space form, 108
doubly warped, 238 Scalar curvature, 14
generic, 190 Scalar product, 2
inner, 2 Schläfli’s conjecture, 25
Riemannian, 48 Schwarzschild chart, 97
scalar, 2 Schwarzschild radius, 97
standard generic, 190 Second Bianchi identity, 12
twisted, 71 Second fundamental form, 28
warped, 47 Sectional curvature
Proper affine hypersphere, 430 φ-, 100
Pseudo-Hermitian manifold, 116 holomorphic, 117
Pseudo-hyperbolic subspace, 38 para-holomorphic, 384
Pseudo-Kähler submanifold, 122 Semi-invariant submanifold, 425
Pseudo-orthonormal frame, 126 Semi-parallel submanifold, 65
Pseudo-Riemannian manifold, 3 Slant
complete, 9 pointwise proper, 171
Pseudo-Riemannian submersion, Slant angle, 141
102 Slant frame, 165
Pseudo-sphere, 15, 38 Slant function, 171, 424
Pseudo-umbilical submanifold, 42 Slant submanifold, 424
Purely real submanifold, 126 bi-, 252
Kählerian, 143
Quasi-minimal submanifold, 30 pointwise bi-, 252
pointwise hemi-, 253
Radius of black hole, 98 pointwise semi-, 253
General Index 479
proper, 141 austere, 122

Space bi-slant, 252
complex hyperbolic, 118 complex, 121, 363
complex projective, 118 constant ratio, 80
first normal, 35 contact CR-, 410
pseudo-Euclidean, 15 contact CR-product, 410
pseudo-hyperbolic, 15 curvature-invariant, 86
relative nullity, 357 generic, 179
rest, 82 hemi-slant, 244
scalar product, 2 integral, 113
Space form invariant, 387, 410
complex, 117 isotropic, 123
generalized complex, 228 Kählerian slant, 143
indefinite complex, 117 Lagrangian, 389
indefinite real, 16 Lagrangian-umbilical, 133
Lorentzian, 16 Legendre, 113
para-Kähler, 384 marginally trapped, 30
Sasakian, 108 minimal, 30
Space slice, 82 mixed totally geodesic, 60
Spacetime null-isotropic, 123
anti de Sitter, 16 parallel, 36, 87
de Sitter, 16 pointwise bi-slant, 252, 425
generalized Robertson-Walker, pointwise hemi-slant, 425
94 pointwise semi-slant, 251
Minkowski, 4 pointwise slant, 424
perfect fluid, 96 proper CR-, 195
Reissner-Nordström, 98 pseudo-Kähler, 122
Robertson-Walker, 47, 82 pseudo-minimal, 30
Schwarzschild, 81 pseudo-Riemannian, 24
Standard space, 65 pseudo-umbilical, 42
Structure purely real, 126, 180
almost contact, 100 quasi-minimal, 30
almost contact metric, 100 semi-invariant, 425
almost product, 383 semi-parallel, 65
complex, 116 semi-slant, 248
contact metric, 100 slant, 141, 424
symplectic, 115 totally geodesic, 29, 87
Structure vector field, 409 totally real, 127
Submanifold totally umbilical, 29, 89, 90
C-totally real, 409 transverse, 82, 86, 88
CR–, 195 warped product bi-slant, 253
P R-, 392 warped product pointwise bi-
H–, 83, 85, 86, 90 slant, 252
σ-, 364, 377 Surface
anti-holomorphic, 195 marginally trapped, 30
anti-invariant, 389, 410 quasi-minimal, 30
superminimal, 366 horizontal, 48, 71

Symplectic manifold, 115 lightlike, 4
mean curvature, 29
Tchebychev form, 429 non-spacelike, 4
Tchebychev vector field, 429 null, 4
Tensor spacelike, 4
conformal curvature, 58 timelike, 4
difference, 429 unit, 2
fundamental, 103 vertical, 48, 71
metric, 3 Vector field
Nijenhuis, 370 basic, 102
O’Neill integrability, 103 characteristic, 100
Ricci, 14 concircular, 19
Riemann curvature, 10 conformal, 18, 19, 82
torsion, 5 Killing, 18
Theorem Lee, 255
Cartan-Janet, 25 parallel, 8
Cartan-Norden, 432 Reeb, 100, 409
Clarke-Greene, 26 structure, 409
Darboux, 113 Tchebychev, 429
de Rham’s decomposition, 61 torqued, 74
Erbacher-Magid, 36
Escobales-Ranjan, 106 Warped product, 47
fundamental, 34, 129, 391 CR-, 231
Hiepko’s decomposition, 62 P R-, 396, 401
Kulkarni, 13 bi-slant, 253
Nölker, 64 contact CR-, 416
Nash, 25 doubly, 212
Ponge-Reckziegel, 76 Einstein, 52
Theorema egregium, 10 generic, 231
relative, 429 hemi-slant, 245
Totally geodesic distribution, 199 multiply, 59
Totally geodesic submanifold, 29, 37 pointwise bi-slant, 252, 426
Totally real point, 127 pointwise hemi-slant, 426
Totally real submanifold, 127 pointwise semi-slant, 251
Totally umbilical submanifold, 29 proper, 229
Transverse submanifold, 82 semi-slant, 248
Twisted product, 71 trivial, 48
doubly, 76 Warped product connection, 48
Warped product curvature, 50
Umbilical section, 29 Warped product representation, 62
Uniqueness theorem, 35, 129, 391 Warping function, 48
Universal inequality, 328 Weingarten formula, 28
Weyl connection, 255
Vector Whitney sphere, 132
causal, 4 Wirtinger angle, 141
Author Index
Abedi, E., 451 Besse, A., 52, 55–57, 105, 453

Advin, 462 Bhatia, S. S., 468
Ahn, S.-S., 38, 451 Binh, T. Q., 453
Al-Ghefari, R., 451 Bini, D., 98, 459
Al-Kathiri, A., 373, 464 Birkhoff, G. D., 97, 453
Al-Luhaibi, N. S., 372, 373, 451 Bishop, R. L., xxiii, 34, 47, 52, 453
Al-Solamy, F. R., 242, 247, 254, 341, Blair, D. E., xxiv, 99, 196, 199, 202,
344–348, 372–374, 451, 452, 471 255, 256, 453
Alı́as, L. J., 94, 451 Bolton, J., 366, 453
Aledo, J. A., 94, 451 Bolyai, J., viii
Alegre, P., 349, 351, 354, 451 Bonanzinga, V., 256, 453, 465
Ali, A., 451 Bonnet, O., 453
Ali, S., 242, 463 Borrelli, V., 345, 453
Alqahtani, L. S., 471 Bronowski, J., xi, xxi
Anciaux, H., 452 Brozos-Vázquez, M., 58, 59, 74, 453
Antić, M., 369, 452 Bryant, R. L., 366, 454
Arroyo, J., 452 Burstin, C., 25, 454
Arslan, K., 452, 466 Butruille, J.-B., 360, 454
Atceken, M., 452
Caballero, M., 94, 454
Bachmann, F., xi Cabrerizo, J. L., 425, 454
Barros, M., 117, 452 Calabi, E., 194, 362, 369, 431, 433–
Beem, J. K., 452 435, 442, 443, 447–450, 454
Behnke, H., xi Calvaruso, G., 99, 454
Bejancu, A., 110, 195, 198, 203, 206, Capozziello, S., 98, 459
425, 452, 453 Carriazo, A., 244, 425, 452, 454
Beltrami, E., 43, 453 Cartan, É., xi, xiv–xvi, 25, 432, 454
Berger, M., xi, 282, 453 Castro, I., 133, 315, 454
Berndt, J., 453 Cayley, A., 360, 361
Bernoulli, J., xix Chahal, K. S., 463
Bertola, M., 453 Chen, B. Y., vii, xi, xix, xx, xxiv,
481
19, 20, 30, 40, 58, 65, 71, 72, 74, 96, 460
78–80, 89, 96, 99, 119, 124, 127, Dillen, F., vii, xx, 65, 78, 212, 213,
129–131, 133, 134, 136, 140, 141, 337, 355, 357, 358, 367, 369, 381,
143, 144, 146–148, 151, 153, 158, 382, 431, 455, 458, 460, 461
162, 165, 170, 171, 173, 174, 177– Dimitric, I., xxiv
179, 188, 190–194, 196, 199, 201– Djorić, M., 369, 452
204, 206, 207, 209, 212, 213, 215, do Carmo, M., 216, 217, 221, 461
217, 222, 225, 229, 231, 235, 238, Dobarro, F., 60, 461
247, 252–254, 258, 259, 275, 282, Dombrowski, P., xiv, xx
283, 296, 299–301, 308, 315, 316, Dragomir, S., 174, 201, 254–256,
318, 321, 324, 326, 328, 329, 331– 453, 461
333, 336–338, 340, 341, 344–351, Duggal, K. L., 110, 230, 453, 461
354, 355, 357, 358, 366, 367, 381, Dwivedi, M. K., 461
382, 393, 397–400, 402, 406–408,
432, 435, 437, 439, 440, 442–444, Easley, K. L., 452
447, 451–459, 471, 472 Ehrlich, P. E., 452
Chen, C.-H., 459 Einstein, A., ix, xi, xvi, xviii, xx,
Chern, S. S., v, vii–ix, xi, 360, 459 xxiii, 1, 14, 15, 23, 47, 52, 54, 55,
Cherubini, C., 98, 459 78, 92, 96–98, 164, 325, 359, 360
Chi, A. Y., 374, 471 Ejiri, N., 366, 367, 369, 461
Cho, K. S., 472 Ellis, G, F. R., 98, 463
Chojnacka-Dulas, J., 460 Erbacher, J., 36, 221, 461
Christoffel, E. B., 6, 460 Eschenburg, J.-H., 34, 129, 153, 390,
Clarke, C. J. S., 26, 460 461
Codazzi, D., 31, 39, 70, 87, 88, 90, Escobales, R., 106, 107, 461
127, 128, 134, 138, 139, 153, 164, Etayo, F., 171, 175, 461
184, 189, 203–205, 221, 278, 283, Ezentas, R., 452, 466
304, 305, 368, 379, 390, 391, 396,
411, 418, 421, 429 Faghfouri, M., 71, 214, 461
Colares, A. G., 94, 451 Fernández, L. M., 425, 454
Crittenden, R. J., 34 Fernández, M., 425, 454
Fernández-Lápez, M., 73, 77, 78,
Dajczer, M., 216, 217, 221, 332, 460, 462
461 Fladt, K., xi
de Lima, H. F., 94, 451 Florit, L. A., 332, 460
de Rham, G., 61, 460 Freudenthal, H., vii, x, xi
de Sitter, W., 16, 81 Friedmann, A., xx, 81
De, A., 453 Friedrich, T., 361, 462
De, U. C., 96, 460, 465
Deahna, F., 460 Günes, R., 372, 373, 469
Decruyenaere, F., 78, 460 Gödel, K., viii
Defever, F., 460 Gadea, P. M., 385, 387, 462
Deng, S., 384, 460, 463 Garay, O. J., xxiv, 173, 452, 458
Deprez, J., 65 Garcı́a-Rı́o, E., 58, 59, 73, 74, 77,
Deshmukh, S., 341, 344–348, 451 78, 453, 462
Deszcz, R., xii, xiii, xvii, xviii, xx, Gauduchon, P., 282, 453
Author Index 483
Gauss, C. F., viii, xiii–xv, xx, 10, 23, Hesse, L. O., 17

26, 30–32, 37, 39–41, 44, 45, 51, Hiepko, S., 62, 95, 138, 233, 273,
56, 57, 64, 70, 92, 97, 112, 123, 298, 323, 355, 402, 463
127–129, 134, 135, 138–140, 144, Hilbert, D., viii, 15, 463
147, 150, 153, 158, 161–163, 181, Hotloś, M., 96, 460, 463
184–186, 188, 192, 193, 197, 203, Hou, Z., 384, 463
209, 211, 218, 221, 222, 224, 226, Houh, C. S., xix
229, 230, 239, 243, 247, 249, 258, Hubble, E., 81
264, 265, 277, 301, 317, 318, 329– Hui, S. K., 463
331, 333, 339, 341, 346, 352, 358, Huntley, H. E., xxi
364, 366, 370–372, 378–380, 389–
391, 399, 407, 411, 414, 417, 422, Ilmakchi, M., 451
428, 429, 462 Ivanov, S., 361, 462
Germain, S., 462
Ghaffarzadeh, N., 461 Jahanara, B., xvii
Gheysens, L., 462 Jamal, N., 242, 373, 463, 464
Glogowska, M., 460 Jamali, M., 463
Gluck, H., 362, 454 Janet, M., 25, 463
Gogna, M., 465 Jebsen, J. T., 97, 463
Goldberg, V. V., 460 Jelowicki, J., 460
Gouthier, D., 453
Graves, J. T., 383, 462 Kähler, E., 99, 107, 109, 115–119,
Graves, L. K., 462 121–130, 143–149, 158, 163–165,
Gray, A., 359, 360, 363, 365, 462, 170, 171, 174–176, 179–188, 190,
472 191, 193, 194, 197–207, 228, 229,
Greene, R. E., 26, 462 231–235, 238, 239, 242–245, 248–
Greenfield, S., 196, 462 252, 254–257, 270, 278, 284, 338,
Gromoll, D., 107, 462 341, 346, 350, 359, 360, 383–385,
Gromov, M. L., 26, 133, 325, 462 387–394, 396–402, 406–408
Grove, K., 107, 462 Kühnel, W., xv, 56, 464
Grunewald, R., 361, 462 Kagan, V. A., xix
Gulbahar, M., 462 Kaluza, T., 23
Kaneyuki, S., 384, 463
Haesen, S., xi, xvii, 80, 462 Kazan, S., 73, 463
Haghighatdoost, G., 451 Keles, S., 467, 472
Haider, S. M. K., 462 Khan, H. H., 413, 414, 471
Han, S. G., 472 Khan, K., 464
Harvey, R., 122, 462 Khan, K. A., 242, 372, 373, 463, 464,
Hasegawa, I., 416–418, 463 471
Hashimoto, H., 376, 377, 463 Khan, M. A., 242, 372–374, 451,
Hawking, S. W., 98, 463 452, 464
He, C., 463 Khan, V. A., 372–374, 413, 414, 452,
Helfrich, W., 463 463, 464, 466, 471
Helgason, S., 463 Kilic, E., 452, 462, 467, 472
Helmholtz, H., xiv Killing, W., 18
Henry, R. C., 98, 463 Kim, D.-S., 38, 53, 55, 56, 464
Kim, J.-S., 461 Minkowski, H., 4

Kim, S. S., 38, 451 Mirandola, H., 466
Kim, Y.-H., 38, 53, 55, 451, 464 Molinari, L. G., 96, 465
Klein, F., xiv, xv Montesinos Amilibia, A., 385, 387,
Klein, O., 23 462
Kobayashi, S., 464, 470 Moore, J. D., 64, 191, 466
Kon, M., 206, 409, 453, 472 Morimoto, A., 107, 466
Koszul, J.-L., 5, 6, 19 Morvan, J.-M., 345, 453
Kretschmann, E., 98 Munteanu, M. I., 256, 349, 351, 354,
Kruchkovich, G. I., xxiii, 47, 464 397, 400, 402, 406–408, 416, 417,
Kuan, W. E., 124, 194, 458 419, 420, 451, 458, 466
Kucharski, M., 460 Murathan, C., 452, 460, 466
Kulkarni, R. S., xii, 13, 464 Mustafa, A., 463, 466, 471
Kumar, R., 464, 465, 468
Kunle H., xi Nölker, S., 60, 62–65, 211, 213, 355,
Kupeli, D. N., 73, 77, 78, 462 412, 460, 466
Nagaich, R. K., 465
Lami Dozo, E., 461 Nagano, T., v
Laplace, P.-S., 17 Nagy, P.-A., 360, 466
Lawn, M.-A., 465 Nash, J. F., xxiii, 25, 26, 207, 325,
Lawson, H. B., 122, 462, 465 466
Lee, J. W., 465 Nazari, Z., 451
Lemaı̂tre, G., viii, xx, 81 Nejadakbary, V., 465
Li, H., 133, 465 Neumann, J. von, vii, viii, xi
Lie, S., xiv, xv Newlander, A., 370, 466
Lobachevsky, N. I., viii Newton, I., vii, xi, 325
Lone, M. A., 374, 465 Nirenberg, L., 370, 466
Lorentz, H., 4, 16 Nistor, A. I., 80, 462
Lotta, A., 424, 465 Nomizu, K., xii, 430, 432, 460, 464,
Luca, G. M., 96, 465 466
Norden, A., 432
Ma, H., 133, 465 Nordström, G., 98
Maeda, S., 300, 458
Magid, M., 36, 64, 465 O’Neill, B., xi, xxiii, 47, 52, 99, 102–
Majidi, A., 71, 214, 461 104, 111, 453, 467
Malek, F., 465 Ogawa, Y., 58, 59, 466
Mantica, C. A., 96, 465 Ogiue, K., 108, 129, 301, 308, 459,
Mashimo, K., 376, 377, 463 467
Matsumoto, K., 256, 412–414, 453, Olteanu, A., 213, 467
465 Opozda, B., 369, 461
Mazet, E., 282, 453 Oprea, T., 337, 467
Meumertzheim, M., 465 Ornea, L., 174, 201, 254, 255, 461
Mihai, A., 228, 465, 466 Ortega, M., 465
Mihai, I., xxiv, 412–414, 416–418, Osserman, R., xi, 467
420, 422, 452, 463, 465, 466 Othman, W. A. M., 451
Milousheva, V., 466 Otsuki, T., v
Author Index 485
Ozel, C., 451, 463, 471 Sachdeva, R., 464, 468

Ozgur, C., 460, 470 Sahin, B., xxiv, 73, 149, 230, 238,
242, 244, 245, 248, 249, 251, 252,
Palma, A. R., 467 254, 372, 373, 452, 461, 463, 465,
Papaghiuc, N., 248, 425, 453, 467 467–470
Park, K.-S., 467 Sahin, F., 469
Penrose, R., 30, 467 Salamanca, J. J., 468
Perelman, G. Y., 1 Salavessa, I. M. C., 164, 469
Perktas, S. Y., 467 Saracoglu Celik, S., 462
Perrone, D., 99, 454 Sasahara, T., 355, 469
Petersen, P., 463 Sasaki, S., 99, 469
Petrović-Torgašev, M., xxiv, 19, 467 Sasaki, T., 432, 466
Pinkall, U., 430, 466 Sato, N., 370, 371, 469
Poincaré, H., 1, 467 Schaaf, M., 465, 469
Ponge, R., 60, 75, 76, 467 Schläfli, L., 25, 469
Prvanović, M., 460 Schouten, J. A., xvi, xvii, 19, 469
Pythagoras, viii, xiv Schur, I., xii
Schwarzschild, K., 47, 81, 96–98, 469
Rachevsky, P., xix Segre, C., 124, 469
Radulescu, I. N., 466 Sekigawa, K., 363, 366, 375, 377, 469
Rani, R., 465 Sentürk, Z., 96, 460
Ranjan, A., 106, 107, 467 Sentürk, A., xvii
Rashevsky, P. K., 383, 467 Sergey, S., 469
Reckziegel, H., 60, 75, 76, 111, 113, Shahid, M. H., 374, 451, 463, 465
157, 465, 467, 468 Shapiro, Ya. L., xix
Reeb, G., 100, 409 Sharma, R., 461
Reissner, H., 98 Shenawy, S., 469
Ricci, G., xvi–xviii, xx, 12, 14, 15, Shukla, S. S., 469
31, 32, 52, 53, 55, 56, 58, 84, 92, Sikorav, J. C., 133, 470
96, 97, 127–129, 153, 327, 331, Slesar, V., 470
346, 360, 368, 369, 384, 390, 392 Spivak, M., 470
Riemann, B., xiii, xiv, xvii, 1, 23, Steiner, H.-G., x
468 Sternberg, S., 470
Rimoldi, M., 468 Stewart, I., viii
Robertson, H., 81–83, 86, 87, 89, 93, Struik, D. J., xvi
94, 96, 443–450 Su, L., 133, 465
Rodriguez Montealegre, C., 453, 468 Suceavă, B., xxiv, 331, 470
Rokhlin, V. A., 26, 462 Suh, Y.-J., 96, 465
Romero, A., 94, 117, 451, 452, 454, Sular, S., 470
468 Szabó, Z. I., xii, xiii, xv
Rosca, R., 19, 30, 467, 468
Rozenfeld, B. A., 383, 468 Tachibana, S., 359, 370, 470
Rubio, R. M., 94, 451, 454, 468 Takagi, R., 470
Ruffini, R., 98, 459 Takahashi, T., 99, 101, 108, 110, 470
Takeuchi, M., 470
Sánchez, M., 94, 451, 469 Tastan, H. M., 470
Tazawa, Y., 144, 162, 177, 178, 459, 147, 148, 150, 163, 185, 188, 197,
470 203, 229, 230, 239, 243, 249, 250,
Thakur, M., 462 301, 302, 307, 352, 368, 371, 372,
Thurston, W. P., xv, xvi, xix, 1 379, 380, 390, 410, 472
Tits, J. H., xiv Wettstein, B., 34, 153, 472
Tiwari, S. K., 469 Weyl, H., xii, xvii–xix, 58, 255, 472
Tojeiro, R., 460, 470 Wilking, B., 107, 472
Tribuzy, R., 34, 129, 153, 390, 461 Wolf, J., 360, 472
Tripathi, M. M., 471 Wong, B. R., 466, 471
Tsukamoto, Y., 359, 471 Woodward, L. M., 366, 453
Wu, B. Q., 206, 459
Ucci, J., 106, 471 Wylie, W., 463
Udagawa, S., 471
Uddin, S., 247, 252–254, 372, 374, Yüksel Perkas, S., 472
413, 414, 452, 459, 463, 464, 466, Yano, K., v, xix, 19, 206, 359, 409,
471 453, 459, 472
Unal, B., 60, 73, 77, 78, 212, 461, Yoon, D. W., 464, 472
462, 469, 471
Urbano, F., 133, 454 Zafindratafa, G., 460
Zhang, P., 472
Vázquez-Lorenzo, R., 58, 59, 74, 453 Zhong, H. H., 369, 472
Vaisman, I., 471 Zhong, Z., 472
Valli, G., 164, 469
Van der Veken, J., xxiv, 89, 337,
458, 459
Vanhecke, L., 228, 364, 459, 471, 472
Verheyen, P., 188, 459, 462
Veronese, G., 124
Verstraelen, L., vii, xi, xvii, xx, xxi,
xxiv, 19, 78, 80, 357, 358, 367,
369, 381, 382, 455, 458, 460–462,
467
Vilcu, G.-E., 254, 470, 472
Vitório, F., 466
Vlachos, T., 460
Vrancken, L., 78, 151, 153, 337, 355,
357, 358, 366, 367, 369, 381, 382,
431, 452, 453, 458–461, 468
Walker, A. G., 81–83, 86, 87, 89, 93,

94, 96, 443–450
Wall, C. T. C., xv
Wallach, N. R., 472
Wang, Y., 472
Wei, S. W., xxiv, 89, 215, 459
Weingarten, J., 28, 30, 41, 89, 112,

Differential Geometry of Warped Product Manifolds and Submanifolds

Загружено:

Сведения о документе

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Differential Geometry of Warped Product Manifolds and Submanifolds

Загружено:

Авторское право:

Доступные форматы

b2530 International Strategic Relations and China’s National Security: World at the Crossroads

This page intentionally left blank

b2530_FM.indd 6 01-Sep-16 11:03:06 AM

Library of Congress Cataloging-in-Publication Data

British Library Cataloguing-in-Publication Data

Copyright © 2017 by World Scientific Publishing Co. Pte. Ltd.

LaiFun - 10419 - Differential Geometry.indd 1 02-05-17 12:27:33 PM

This page intentionally left blank

b2530_FM.indd 6 01-Sep-16 11:03:06 AM

viii Differential Geometry of Warped Product Manifolds and Submanifolds

x Differential Geometry of Warped Product Manifolds and Submanifolds

mentals of Mathematics. Volume II: Geometry” (eds. Behnke-Bachmann-

xii Differential Geometry of Warped Product Manifolds and Submanifolds

around all infinitesimal co-ordinate parallelograms cornered at these points.

xiv Differential Geometry of Warped Product Manifolds and Submanifolds

sion of the Theorem of Pythagoras of the ambient space E 3 to the surface

The surfaces of revolution in E 3 with constant Gauss curvatures respec-

xvi Differential Geometry of Warped Product Manifolds and Submanifolds

directions are geometrically distinguished from others. To begin with, fol-

Ricci curvatures” with Jahanara, Haesen and Sentürk –. In analogy with

xviii Differential Geometry of Warped Product Manifolds and Submanifolds

the corresponding curvature functions LS and L are the same). In résumé :

two distinct corresponding to the wildest possible anisotropy. For spaces of

xx Differential Geometry of Warped Product Manifolds and Submanifolds

Dombrowski’s “The Brachistochrone Problem: A Problem of Elementary

rewarded by enjoying their participation in the creation of the results that

This page intentionally left blank

b2530_FM.indd 6 01-Sep-16 11:03:06 AM

xxiv Differential Geometry of Warped Product Manifolds and Submanifolds

1. Riemannian and Pseudo-Riemannian Manifolds 1

xxvi Differential Geometry of Warped Product Manifolds and Submanifolds

3. Warped Product Manifolds 47

4. Robertson-Walker Spacetimes and Schwarzschild Solution 81

5. Contact Metric Manifolds and Submersions 99

6. Kähler and Pseudo-Kähler Manifolds 115

6.4 Segre and Veronese embeddings . . . . . . . . . . . . . . . 124

7. Slant Submanifolds 141

8. Generic Submanifolds of Kähler Manifolds 179

9. CR-submanifolds of Kähler Manifolds 195

10. Warped Products in Riemannian and Kähler Manifolds 207

xxviii Differential Geometry of Warped Product Manifolds and Submanifolds

10.3 Some applications of Theorems 10.1 and 10.2 . . . . . . . 213

11. Warped Product Submanifolds of Kähler Manifolds 229

12. CR-warped Products in Complex Space Forms 257

13. More on CR-warped Products in Complex Space Forms 283

14. δ-invariants, Submersions and Warped Products 325

15. Warped Products in Nearly Kähler Manifolds 359

16. Warped Products in Para-Kähler Manifolds 383

xxx Differential Geometry of Warped Product Manifolds and Submanifolds

17. Warped Products in Sasakian Manifolds 409

18. Warped Products in Affine Spaces 427

Riemannian and Pseudo-Riemannian

Riemannian geometry was first put forward in generality by B. Riemann

1.1 Symmetric bilinear forms and scalar products

A symmetric bilinear form on a finite-dimensional real vector space V is

2 Differential Geometry of Warped Product Manifolds and Submanifolds

v 6= 0. Similarly, a symmetric bilinear form B is called negative definite

p On a scalar product space V , the norm ||v|| of a vector v is defined to be

Riemannian and Pseudo-Riemannian Manifolds 3

Proof. Since g is non-degenerate, there is a unit vector e1 ∈ V . Let U1

For an orthonormal basis e1 , . . . , en of a scalar product space V , the number

1.2 Riemannian and pseudo-Riemannian manifolds

A (pseudo-Riemannian) metric tensor g on a manifold M is a symmetric

4 Differential Geometry of Warped Product Manifolds and Submanifolds