Академический Документы
Профессиональный Документы
Культура Документы
r
X
i
v
:
s
u
b
m
i
t
/
0
4
6
8
8
2
7
[
m
a
t
h
-
p
h
]
5
M
a
y
2
0
1
2
Finsler Spinors and Twistors in
Einstein Gravity and Modications
Sergiu I. Vacaru
Science Department, University Al. I. Cuza Iasi,
54, Lascar Catargi street, Iasi, Romania, 700107
May 5, 2012
Abstract
We present a generalization of the spinor and twistor geometry
for FinslerCartan spaces modelled on tangent Lorentz bundles, or on
(pseudo) Riemanian manifolds. Nonholonomic (Finsler) twistors are
dened as solutions of generalized twistor equations determined by spin
connections and frames adapted to nonlinear connection structures.
We show that the constructions for local twistors can be globalized
using nonholonomic deformations with auxiliary metric compatible
connections completely determined by the metric structure and/or the
Finsler fundamental function. We explain how to perform such an
approach in the Einstein gravity theory formulated in Finsler like vari-
ables with conventional nonholonomic 2+2 splitting.
Keywords: spinors and twistors, Finsler geometry, nonlinear con-
nections, nonholonomic manifolds, Einstein spaces.
PACS: 04.50.Kd, 02.40.Tt, 04.90.+e, 04.20.Gz
MSC: 53B40, 32L25, 53C28, 83C60
Contents
1 Introduction 2
2
F
y
a
y
b
in order to dene a class of twistor structures related to Finslerian
geodesics, which are integral curves of certain systems of ordinary dierential
equations (ODEs) and some projective classes of isotropic sprays. Such con-
structions were performed for Finsler generating functions with scalar ag
curvature, in special, for the Randers metric. It was possible to formulate a
variational principle for twistor curves arising from such examples of Finsler
geometries with scalar ag structure. The (nonlinear) geodesic/ (semi) spray
4
Historical Remarks: The author of this paper began his research on twistor
gauge models of gravity when he was a postgraduate at physics department of M.
A. Lomonosov State University at Moscow, during 1984-1987. At that time, there were
translated in Russian some fundamental papers on Twistor Theory and published a se-
ries of important works by soviet authors, for instance, the Russian variant of [2].
Various subjective issues related to the crash of former Soviet Union resulted in de-
fending authors PhD thesis [37] in 1994, at University Alexandru Ioan Cuza, UAIC, at
Iasi (Yassy), Romania. There were also certain important scientic arguments to transfer
such a research on geometry and physics from Russia to Romania, where two schools on
nonholonomic geometry and generalized Finsler spaces (G. Vranceanu and R. Miron, A.
Bejancu and others) published a number of works beginning 20ths of previous Century
but worked in isolation during dictatorial socialist period.
Twistor equations are generically nonintegrable for arbitrary curved spacetime. Such
equations became integrable, for instance, if the Weyl spinor vanishes, see details in [1].
Our main idea is to use for denition of twistors another class of equations with an
auxiliary metric compatible connection, completely determined by the same metric but
with nontrivial torsion. For certain conditions on noholonomic structure, various non
integrable twistor congurations can be globalized in a selfconsistent form. Imposing
additional anholonomi conditions, we can extract, for instance, certain real vacuum Ein-
stein manifolds. Such methods were formalized by G. Vranceanu in his geometry of non-
holonomic manifods [38, 39, 40], see further developments in [41] and, with applications
in modern classical and quantum gravity and generalized Finsler geometries and (non)
commutative geomeric ows, [5, 17, 23, 31].
In some sense, Finsler spaces are modelled by geometries with nonholonomic distribu-
tions on tangent bundles or on manifolds with bered structure. The directions summa-
rized in monograph by R. Miron and M. Anastasiei [13] were developed following some
fundamental geometric ideas and results due to E. Cartan [28], M. Matsumoto [12] and
others (on almost Kahler Finsler structures, Einstein equations for the CartanFinsler
connection, J. Kerns geometrization of mechanics via Finsler methods [42] etc). Using
nonholonomic distributions, it was possible to formulate Cliord and spinor analogs of
metric compatible Finsler geometries and generalizations and dene DiracFinsler oper-
ators [25, 26], see also reviews of results and complete lists of references in [27, 17, 31].
Approaches based on Berwald and Chern connections, semisprays and sectional curvature
[16, 43] have been developed in modern literature but because of nonmetricity and spectic
nonminimal relations between the Finsler metric and curvature seem to be less related
to standard theories in physics [34, 24, 23].
5
congurations reect only partially the geometric properties, and possible
relations to systems of dierential equations (with partial derivatives ones,
PDEs, and/or ODEs) of spaces endowed with fundamental Finsler functions.
As we emphasized above, complete geometric and possible physically viable
models of Finsler spacetimes can be formulated after additional assumptions
on linear connection structures and conditions on their (non) compatibility
with the metric and Nconnection structure. Only in such cases, we can con-
clude if there are, or not, Finsler analogs of spinors and anisotropic models
of bosonic and fermionic elds and interactions for certain classes of Finsler
connections.
This paper is organized as follows: In section 2, we present a brief in-
troduction into the geometry of metric compatible nonholonomic manifolds
and bundles enabled with nonlinear connection (Nconnection) structure. It
is elaborated an unied Nadapted formalism both for the FinslerCartan
spaces and the Einstein gravity theory reformulated in Finsler like variables.
We also provide new results on the geometry of conformal transforms and
Nconnection structures. Section 3 is devoted to the dierential geome-
try of spinors on FinslerCartan and EinsteinFinsler spaces. We dene
nonholonomic (Finsler) twistors in section 4 considering auxiliar metric
compatible Finsler like connections. There are studied the conditions when
nonholonomic twistors can describe global structures in Einstein gravity and
modications.
Acknowledgements: Authors research is partially supported by the
Program IDEI, PN-II-ID-PCE-2011-3-0256.
2 Nonholonomic (Finsler) Geometry and Gravity
Finsler type geometries can be modelled on (pseudo) Riemannian man-
ifolds and/or tangent bundles enabled with necessary types nonholonomic
distributions and supersymmteric and/or noncommutative generalizations.
In this section, we x notations and provide necessary results. For details
and proofs, we refer to [9, 24, 23, 17] where the socalled geometry of Finsler
Einstein gravity and modications is formulated in a language familiar to
researchers in mathematical relativity.
6
2.1 The geometry of nonholonomic bundles and manifolds
2.1.1 Nonlinear and distinguished connections
Let us consider a (n + m)dimensional real or complex manifold V of
necessary smooth/analytic/holomorphic class, where 1) n = m = 2 (this
will be used for constructing exact solutions in general relativity applying
Finsler and twistor methods), or 2) n = m = 4 (for Finsler twistor models
on tangent bundles to Lorentz manifolds).
Denition 2.1 A nonholonomic manifold is a pair (V, N) dened by a
nonintegrable distribution N on V.
The geometry of nonholonomic real manifolds is studied in Refs. [38, 39,
40, 23, 17] for various nite dimensions n 2 and m > 1. For simplicity,
we shall consider a subclass of nonholonomic distributions N stating bered
structures : V V with constant rank , where V is a two dimensional, 2-
d, or 4-d, (for instance, pseudoRiemanian manifold, or any its complexied
version). In general, we can consider any such map with dierential map
T (X, Y) :=
X
Y
Y
X[X, Y] = 0.
We note that is not a dconnection because it does not preserve un-
der parallelism the Nconnection splitting. For a N (1) with nonzero N
curvature (2), there is a preferred dconnection which can be considered
as the analog of the LeviCivita connection for nonholonomic manifolds:
Theorem 2.2 Any metric structure g denes a unique canonical dconnecti-
on D which is metric compatible, Dg = 0, and with zero pure hori-
zontal and vertical dtorsions, i.e., respectively, hT (hX, hY) = 0 and
vT (vX, vY), for T (X, Y) := D
X
YD
Y
X[X, Y] .
For a xed N, both linear connections and Dare completely dened by
the same metric g.
6
There is a substantial dierence between the canonical
dconnection D and the connections used in RiemannCartan geometry
(see, for instance, [1] and detailed discussions with respect Einstein and
Finsler geometries and metricane generalizations in [9, 23, 17]). The
torsion T is completely dened by data (g, N), i.e. by the metric structure
if the value N is prescribed, but in the EinsteinCartan gravity we need
additional algebraic equations for torsion.
Corollary 2.1 There is a canonical distortion relation
D = +Q, (5)
where both linear connections D and and the distortion tensor Q are
completely dened by the metric tensor g for a prescribed Nconnection
splitting.
Using the torsion tensor R (4) of D, we can introduce in standard form
the Ricci dtensor Ric, which is nonsymmetric because of nontrivial non-
holonomic/ torsion structure, the curvature scalar
s
R and the Einstein
6
In a series of our works, we wrote
D for the canonical dconnection and used hats
for the coecients and values computed for this linear connection. In this article we shall
use only the symbol D; in next sections,
D will be used for a conformal transforation of
D.
9
dtensor E. The (canonical nonholonomic) Einstein equations for D are
written geometrically
E = Ric
1
2
g
s
R = , (6)
where the source can be computed as in general relativity (GR) with
spacetime metric but D. A Nadapted variational calculus with pos-
sible matter elds (uids, bosons, fermions etc, all with respect to non-
holonomic frames) can be formulated but this will result in nonsymmetric
dtensors , which is not surprising because of nonholonmic character of
such constructions with induced torsion. On (pseudo) Riemannian mani-
folds with singnature (+ ++) and a prescribed nonintegrable decomposi-
tion 2 +2, we can perform local constructions with {T
}, where T
= (x
i
, y
a
), where the hindices run values
i, j, ... = 1, 2, ...n (for nonholonomic deformations in GR, i, j, ... = 1, 2 or,
on tangent to Lorentz bundles, i, j, ... = 1, 2, 3, 4) and the vindices take
10
values a, b, c, ... = n + 1, n + 2, n + m (for nonholonomic deformations in
GR, a, b, ... = 3, 4 and, on tangent to Lorentz bundles, or a, b, ... = 5, 6, 7, 8).
For bundle spaces, y
a
are typical ber coordinates and x
i
are coordinates
on base manifolds. We can introduce on V certain local coordinate bases
= /u
= (
i
= /x
i
,
a
= /y
a
) and their duals du
= (dx
j
, dy
b
)
[we shall emphasize some indices if it is necessary that they are coordinate
ones but omit underlining when that will not result in ambiguities].
Transforms to arbitrary local frames, e
, are given
by nondegenerate vierbein matrices, e
(u), re-
spectively, e
= e
and e
= e
du
=
_
e
i
=
x
i
N
b
i
y
b
, e
a
=
y
a
_
, (8)
and its dual
e
=
_
e
j
= dx
j
, e
b
= dy
b
+N
b
i
dx
i
_
. (9)
Proof. This follows from the possiblity to construct Nadapted bases
of type
e
= e
and e
= e
du
, (10)
where
e
(u) =
_
e
i
i
(u) N
b
i
(u)e
a
b
(u)
0 e
a
a
(u)
_
, e
(u) =
_
e
i
i
(u) N
b
k
(u)e
k
i
(u)
0 e
a
a
(u)
_
.
(11)
, e
] = e
= W
, (12)
where the (antisymmetric) nontrivial anholonomy coecients are computed
W
b
ia
=
a
N
b
i
and W
a
ji
=
a
ij
.
Proposition 2.2 Any metric structure g on V can be written in Nadapted
form as a distinguished metric (dmetric)
g =
h
g +
v
h = g
ij
(u) e
i
e
j
+ h
ab
(u) e
a
e
b
. (13)
Proof. Via frame/coordinate transforms, g
= e
, any metric
g = g
(u) du
du
(14)
can written in the form
g
=
_
g
ij
+N
a
i
N
b
j
h
ab
N
e
j
h
ae
N
e
i
h
be
h
ab
_
. (15)
Introducing formulas (9) and (11) into (13) we obtain the coordinate form
(14) and (15). Inverse transforms are similar.
Using the last two propositions, we can compute the Nadapted coe-
cients
=
_
L
i
jk
, L
a
bk
, C
i
jc
, C
a
bc
_
, with respect to frames (8) and (9), of
the canonical dconnection D,
L
i
jk
=
1
2
g
ir
(e
k
g
jr
+e
j
g
kr
e
r
g
jk
) , (16)
L
a
bk
= e
b
(N
a
k
) +
1
2
h
ac
_
e
k
h
bc
h
dc
e
b
N
d
k
h
db
e
c
N
d
k
_
,
C
i
jc
=
1
2
g
ik
e
c
g
jk
, C
a
bc
=
1
2
h
ad
(e
c
h
bd
+e
c
h
cd
e
d
h
bc
) .
The Nadapted coecients of dtorsion
T = {T
= (T
i
jk
, T
i
ja
, T
a
ji
, T
a
bi
, T
a
bc
)} (2.1.1) of D are computed
T
i
jk
= L
i
jk
L
i
kj
= 0, T
i
ja
= T
i
aj
= C
i
ja
, T
a
ji
=
a
ji
,
T
a
bi
=
b
N
a
i
L
a
bi
, T
a
bc
= C
a
bc
C
a
cb
= 0. (17)
12
We provide also the Nadapted coecients of dcurvature
R = {R
= (R
i
hjk
, R
a
bjk
, R
i
jka
, R
c
bka
, R
i
jbc
, R
a
bcd
)} (4) of D,
R
i
hjk
= e
k
L
i
hj
e
j
L
i
hk
+L
m
hj
L
i
mk
L
m
hk
L
i
mj
C
i
ha
a
kj
,
R
a
bjk
= e
k
L
a
bj
e
j
L
a
bk
+L
c
bj
L
a
ck
L
c
bk
L
a
cj
C
a
bc
c
kj
,
R
i
jka
= e
a
L
i
jk
D
k
C
i
ja
+C
i
jb
T
b
ka
, (18)
R
c
bka
= e
a
L
c
bk
D
k
C
c
ba
+C
c
bd
T
c
ka
,
R
i
jbc
= e
c
C
i
jb
e
b
C
i
jc
+C
h
jb
C
i
hc
C
h
jc
C
i
hb
,
R
a
bcd
= e
d
C
a
bc
e
c
C
a
bd
+C
e
bc
C
a
ed
C
e
bd
C
a
ec
.
Contracting indices, we can compute the h- vcomponents R
of
the Ricci tensor Ric,
R
ij
R
k
ijk
, R
ia
R
k
ika
, R
ai
R
b
aib
, R
ab
R
c
abc
. (19)
The scalar curvature is
s
R g
= g
ij
R
ij
+h
ab
R
ab
. (20)
In component form, the analog of Theorem 2.3 is
Theorem 2.4 The Einstein equations in GR are equivalent to
R
1
2
g
s
R =
, (21)
L
c
aj
= e
a
(N
c
j
), C
i
jb
= 0,
a
ji
= 0, (22)
written for the canonical dconnection coecients (16) if
The main reason to work with equations of type (6) and (21) is that
such equations for D (we say in nonholonomic variables) decouple with
respect to Nadapted frames (for spaces with splitting of dimension 2, or
3, +2 + 2 + 2 + ...) for generic odiagonal ansatz for metric g and cer-
tain parametrizations of N depending on all coordinates. This allows us
to integrate such nonlinear PDE in very general forms. We construct inte-
gral varieties determined by corresponding classes of generating and integra-
tion functions and integration constant which may be dened from certain
13
boundary/Cauchy conditions and additional physical arguments. Imposing
additional LeviCivita (LC) conditions (7), which constrain nonholonomi-
cally the integral varieties of solutions of E = , we can extract solutions
in GR. We note that we can not decouple and integrate in such odiagonal
forms the Einstein equations if we work from the very begining and only
with . The main trick is that we relax the constraints of zero torsion
in the standard Einstein equations by considering an auxiliary connection
D (in next section, we shall see that this is a Finsler type dconnection);
such constructions are provided in Refs. [6, 7, 5, 17, 23].
2.2 Metric compatible FinslerCartan geometries
We outline some results from the Finsler geometry on tangent bundles
[13, 14, 28, 11, 12, 16] and show how the constructions can be redened for
nonholonomic (pseudo) Riemannian manifolds [23, 17, 34, 5].
2.2.1 The Finsler fundamental/generating function
Let us consider a tangent bundle TM =
xM
T
x
M, where T
x
M are
the tangent spaces at points x M, for the base space M being a real C
on
TM := TM\{0}, for {0} denoting the set of zero
sections of TM on M;
14
2. F(x, y) = F(x, y), for any > 0, i.e. it is a positive 1homogeneous
function on the bers of TM;
3. for any y
T
x
M, the Hessian
v
g
ij
(x, y) =
1
2
2
F
2
y
i
y
j
(25)
is considered as s a vertical (v) metric on typical ber, i.e. it is
nondegenerate and positive denite, det |
v
g
ij
| = 0.
If the base M is taken to be a Lorentz manifold in GR, we can constr-
cut generalizations on TM with a good physical axiomatic system which
is very similar to that of Einstein gravity when the LeviCivita connection
is substituted by a metric compatible Finsler variant of the canonical
dconnection D, see discussions in [47, 48] and next subsections.
Remark 2.1 The condition 3 above should be relaxed to not positive de-
nite for models of Finsler gravity with nite, in general, locally anisotropic
speed of light.
Considering a background (pseudo) Riemannian metric g
ij
(x) with sin-
gature (+, +, +, ) on M, we can elaborate various geometric and physical
models on TM with locally anisotropics metrics g
ij
(x, y) depending on ve-
locity type coordinates y
a
. The main dierence between (pseudo) Rieman-
nian and Finsler geometries is that the rst type ones are completely dened
by a metric structure g
ij
(x) (from which a unique LeviCivita connection
can be constructed) but the second type ones can not be completely derived
from a Finsler metric F(x, y) and/or its Hessian
v
g
ij
(x, y).
Remark 2.2 A complete Finsler geometry model (F :
F
N,
F
g,
F
D) can
be dened by additional assumptions on how three fundamental geometric
objects (the Nconnection
F
N, the total metric
F
g, the dconnection
F
D) can be determined uniquely by a fundamental Finsler function F.
Finsler like geometries can be elaborated on a generic nonholonomic
bundle/manifold V following selfconsistent geometric and physically im-
portant principles (for instance, V = TM, V is a (pseudo) Riemannian
manifold with nonholonomic 2 + 2 splitting [6, 7, 5]; there were performed
similar generalizations for supermanifolds/superbundles and/or noncommu-
tative generalizations, aneFinsler spaces etc, see [30, 31, 17]).
15
2.2.2 The canonical Finsler connections and lifts of metrics
Let us consider L = F
2
is considered as an eective regular Lagrangian
on TM and action integral S() =
1
_
0
L(x(), y())d, for y
k
() = dx
k
()/d,
where x() parametrizes smooth curves on a manifold M with [0, 1].
Lemma 2.1 The EulerLagrange equations
d
d
L
y
i
L
x
i
= 0 are equiva-
lent to the nonlinear geodesic (equivalently, semispray) equations
d
2
x
k
d
2
+
2
G
k
(x, y) = 0, where
G
k
=
1
4
g
kj
_
y
i
2
L
y
j
x
i
L
x
j
_
, (26)
for g
kj
being inverse to
v
g
ij
g
ij
(25).
Certain geometric properties of fundamental Finsler functions can be
studied via semispray congurations not concerning the problem of de-
nition of connections and metrics for such spaces. For instance, J. Kern
[42] suggested to consider nonhomomgeneous regular Lagrangians instead of
those considered in Finsler geometry. That resulted in socalled Lagrange
Finsler geometry studied in details in [13], on applications in modern physics
see [23, 34].
Denition 2.7 -Corollary: There is a canonical Nconnection
N = {
N
a
j
},
N
a
j
:=
G
a
(x, y)
y
j
, (27)
completely dened by the fundamental Finsler function F.
Proof. Using the above Lemma and local computations we can verify
that the conditions DenitionTherem (2.2) for Nconnections are satised.
See also details of such a proof in [13, 17].
= (e
i
, e
a
) and e
= (e
i
, e
a
) : we have to substi-
tute N
a
j
N
a
j
into, respectively, (8) and (9).
Denition 2.8 -Corollary: A total metric structure on TM can be dened
by a Sasaky type lift of g
ij
,
g = g
ij
(x, y) e
i
e
j
+ g
ij
(x, y) e
i
e
j
. (28)
16
It is possible to use other geometric principles for lifts and projections
when, for instance, from a given F it is constructed a complete homo-
geneous metric on total/horizontal spaces of TM. For models of locally
anisotropic/Finsler gravity on TM, or on V, a generalized covariance princi-
ple has to be considered following geometric and physical considerations [47,
48]. Such constructions are performed up to certain frame/coordinate trans-
forms e
= e
and N = {N
a
i
= e
a
a
e
i
i
N
a
}. We
can dene a subclass of frame/coordinate transforms preserving a prescribed
splitting (1).
2.2.3 Models of FinslerCartan spaces
Using last two DenitionCorollaries, we prove
Theorem 2.5 A fundamental Finsler function F(x, y) denes naturally a
nonholonomic RiemannCartan model on
TM determined by geometric data
(F :
F
N =
N,
F
g = g,
F
D = D), where D is determined by Nadapted
coecients
= ( L
i
jk
, L
a
bk
, C
i
jc
, C
a
bc
) computed using formulas (16) for
g g (28) and N
N (27).
Introducing coecients
L
i
jk
,
C
i
jc
) is dened by
introducing g = g with
h
ij
= g
ij
and N =
N in (29).
Using formulas (17) and (18) for Nadapted coecients of
D, we prove
17
Theorem 2.6 The nontrival components of torsion
T
= {
T
i
jc
,
T
a
ij
,
T
a
ib
}
and curvature
R
= {
R
i
hjk
,
P
i
jka
,
S
a
bcd
} of
D are respectively
T
i
jk
= 0,
T
i
jc
=
C
i
jc
,
T
a
ij
=
a
ij
,
T
a
ib
= e
b
_
N
a
i
_
L
a
bi
,
T
a
bc
= 0, (30)
and
R
i
hjk
= e
k
L
i
hj
e
j
L
i
hk
+
L
m
hj
L
i
mk
L
m
hk
L
i
mj
C
i
ha
a
kj
, (31)
P
i
jka
= e
a
L
i
jk
D
k
C
i
ja
,
S
a
bcd
= e
d
C
a
bc
e
c
C
a
bd
+
C
e
bc
C
a
ed
C
e
bd
C
a
ec
.
A very important property of
_
g;
h
ij
= g
ij
,
D
_
is that such geometric
data can be encoded equivalently into an almost K ahler structure [12] (see
also section 2.2.4 in [24]). This allows us to perform deformation quantiza-
tion and or Abrane quantizations of Finsler geometry and generalizations,
see [9, 8]. Such constructions are important for denition of almost K ahler
spinors and Dirac operators in Finsler geometry (we do not present details
in this work but emphasize that almost symplectic Finsler structures can be
encoded into corresponding spinor and twistor structures).
2.2.4 On metric noncompatible Finsler geometries
Mathematicians elaborated dierent models of Finsler geometry gener-
ated by a fundamental Finsler function F(x, y). Most known are construc-
tions due to L. Berwald [49] and S. Chern [50] (see details in [16]) and
nonstandard denition for the Ricci curvature by H. AkbarZadeh [43].
For instance,
the Berwald dconnection is
B
D : = (
B
L
i
jk
=
N
i
j
/y
k
,
B
C
i
jc
= 0);
the Chern dconnection is
Ch
D : = (
Ch
L
i
jk
=
L
i
jk
,
Ch
C
i
jc
= 0).
The Cherns dconnection is very similar to the LeviCivita connection,
for geometric constructions on the hsubspace. The Finsler geometries de-
termined by such dconnections are not metric compatible on total space
of TM and characterized by nontrivial nonmetricity elds, Q := Dg,
B
Q =
0 and
Ch
Q = 0. We studied various generalizations aneFinsler and
aneLagrange spaces in Part I of [17]. Nontrivial nonmetricity elds (and
nonstandard denitions of scalar and Ricci curvatures of Finsler spaces)
present, in general, diculties for denition of spinors and Dirac type oper-
ators, formulating conservation laws etc, see critical remarks in [34, 23, 24].
So, thre are substantial geometric and physical reasons to work with Finsler
Cartan type spaces and similar metric compatible congurations for appli-
cations in modern gravity and cosmology.
18
2.2.5 Finsler variables in general relativity
In this section, we show how the Einstein gravity can rewritten equiv-
alently in Finsler like variables.
Let us consider a (pseudo) Riemannian space V with nonholonomic 2+2
splitting N = {N
a
i
} and dmetric g = {g
} = {g
}, which can be
written in the form (13) and/or (14) and (15). We can always introduce on
a welldened cart for an atlas covering V a homogeneous function F(x, y)
satisfying the conditions of Denition 2.6 and Remark 2.1. Using such a
formal (pseudo) Finsler generating function, we can construct a Sasaki d
metric of type (28), for
f
ij
:=
1
2
2
F
2
y
i
y
j
and
N
a
j
obtained for F F following
formulas (27) and (26). With respect to dual local basis du
= (dx
i
, dy
a
),
such a total metric can be written in the form
f
=
_
f
ij
+
N
a
i
N
b
j
f
ab
N
e
j
f
ae
N
e
i
f
be
f
ab
_
.
Solving a quadratic algebraic equation for e
and
f
(u),
g
(u) = e
(u)e
(u)f
(u), (32)
we can rewrite connections and tensors on V, up to frame/coordinate trans-
forms, in terms of variables
_
F :
f
_
or (g, N) . We may change the carts and
coordinates and F in order to get real welldened solutions for vierbeins
e
.
The above constructions depend on arbitrary generating function F,
which states a 2+2 splitting via formulas (27) and (26) and respective frames
(8) and (9), in their turn admitting tranforms to Nelongated values deeter-
mined by N
a
i
and/or
N
a
j
. This reects the principle of general covariance
when some additional nonholonomic constraints are imposed on frame struc-
ture. If a relation (32) is established on V, we can compute the LeviCivita
connection using the values f
and/or, equivalently, g
. We can also
compute the coecients of D (16) and
D (29) with distortion relations of
type (5). All such values are completely determined by g
(equivalently
by f
f = T
f and
_
_
V
= R
,
for
:= D
= 2D
[
D
]
(33)
and arbitrary scalar function f(x, y) and dvector V
= g
+ 8GT
, where,
for 2 + 2 splitting, and G are respectively the cosmological and Newton
constants (such values can be dened via Sasaki lifts, for 4 + 4 models on
tangent bundles). The Einstein equations for D
1
2
g
s
R +g
= 8GT
, (34)
20
where R
:= R
and
s
R := g
, (35)
:= 3g
1
2
R
= 8G(
1
4
T
)
and the conformal dtensor
C
:= R
+ 2R
[
[
]
]
+
1
3
s
R
]
= R
+ 4P
[
[
]
]
(36)
where
=
1
6
s
Rg
. (37)
Such dtensor formulas are related to similar ones for the LeviCivita con-
nection via distortions D = + Q (5),where all values are determined
by a corresponding dmetric (13) or (28). This results in distortions of
dtensors,
R
= R
+Q
,
s
R = R +
s
Q, (38)
R
= R
+Q
, C
= C
+
W
Q
,
were the left label
W
Q
:=
2
(u)g
, (39)
preserving the Nconnection structure N = {N
a
i
}, the conformal dtensor
(36) satises the conditions
=
2
C
and
C
= C
.
Proof. Such transforms can be veried by a Nadapted calculus with
respect to xed Nelongated (8) and (9). We note here that with respect
to a coordinate frame, for a metric (14) with coecients (15), a transform
(39) dene a nonlinear transform of metric. The property of rescalling holds
only for the dmetric coecients with respect to xed data N = {N
a
i
}.
= 2D
[
P
]
, (40)
are standard ones with possible h- and vprojections [13, 17].
21
Theorem 2.7 For any xed data (g
, N
a
i
) , there is a nonholonomic de-
formation to some
_
g
, N
a
_
for which C
= 0 with respect to a
re-dened e
= (e
j
, e
b
) (9).
Proof. Let us x a dmetric (13) with coecients g
:=
2
(u)
with
= (e
j
= dx
j
, e
b
= dy
b
+ N
b
dx
i
). For such
a dmetric and Nadapted cobases, we can verify that C
= 0, as
a consequence of Proposition 2.3. We can redene data ( (13),g
) in a
coordinate form (14) with coecients (15) (with primed indices, g
).
Then considering arbitrary frame transforms e
we can compute g
=
e
, N
a
i
) , for which, in general, C
= 0, and
the corresponding to , C
= 0, for . Nev-
ertheless, we can always associate a nonholonomic manifold V enabled with
the same metric structure but with such a Nconnection N when the corre-
sponding canonical dconnection D is with zero Weyl dtensor C
= 0
(we omit priming of indices).
The above values C
and C
j
+
j
i
= 2 g
ij
h
I, in any point x
T hV.
The Cliord dmodule of a vector bundle E (in general, we can consider
a complex vector bundle
E
: E V) is dened by the C(V)module
Sec(E) of continuous sections in E, c : Sec(
N
Cl(V)) End(Sec(E)).
Prescribing a Nconnection structure, a Cliord Nanholonomic bundle on
V is by denition
N
Cl(V)
.
=
N
Cl(T
V), where T
vV).
7
in certain cases, we shall consider only horizontal geometric constructions if they
are similar to vertical ones
23
Let V
n
be a vector space provided with Cliord structure. We write
h
V
n
if its tangent space is provided with a quadratic form
h
g and consider
h
Cl(V
n
) Cl(
h
V
n
) using the subgroup SO(
h
V
n
) O(
h
V
n
). A standard
denition of spinors is possible using sections of a vector bundle S on a
manifold M being considered an irreducible representation of the group
Spin(M)
.
= Spin(T
x
M) dened on the typical ber. The set of sections
Sec(S) denes an irreducible Cliord module.
The space of complex hspins is dened by the subgroup
h
Spin
c
(n) Spin
c
(
h
V
n
)
h
Spin
c
(V
n
) Cl(
h
V
n
),
determined by the products of pairs of vectors w
h
V
C
when w
.
= pu where
p is a complex number of module 1 and u is of unity length in
h
V
n
. Similar
constructions can be performed for the vsubspace
v
V
m
, which allows us
to dene similarly the group of real vspins. A hspinor bundle
h
S on a h
space hV is a complex vector bundle with both dened action of the hspin
group
h
Spin(V
n
) on the typical ber and an irreducible representation
of the group
h
Spin(V) Spin(hV)
.
= Spin(T
x
hV). The set of sections
Sec(
h
S) denes an irreducible Cliord hmodule.
Denition 3.3 A distinguished spinor (dspinor) bundle S
.
= (
h
S,
v
S)
for V, dimV = n + m, is a complex vector bundle with an action of
the spin distinguished (dgroup) Spin V
.
= Spin(V
n
) Spin(V
m
) with
an irreducible representation Spin(V)
.
= Spin(T
.
= e
i
(x, y) e
i
and e
i
.
= e
i
i
(x, y) e
i
, when g
ij
e
i
e
j
=
and
g
ij
e
i
i
e
j
j
= g
ij
. This allows us to dene the algebra of Diracs gamma
hmatrices with selfadjoint matrices M
k
(C), where k = 2
n/2
is the dimen-
sion of the irreducible representation of Cl(hV) derived from the relation
24
= 2
h
I. The action of dx
i
Cl(hV) on a spinor
h
h
S
can be parametrized by formulas
h
c(dx
)
.
=
and
h
c(dx
i
)
h
.
=
i h
e
i
h
. The algebra of Diracs gamma vmatrices is dened by self
adjoint matrices M
k
(C), where k
= 2
m/2
is the dimension of the irreducible
representation of Cl(F), for a typical ber F), when
a
b
+
a
= 2
a
b v
I.
The action of dy
a
Cl(F) on a spinor
v
v
S is
v
c(dy
a
)
.
=
a
and
v
c(dy
a
)
v
.
=
a v
e
a
a
a v
.
In general, a matrix calculus with gamma dmatrices can be elaborated
for a total dmetric structure g =
h
g
v
h. We consider dspinors
.
=
(
h
,
v
) S
.
= (
h
S,
v
S) and dgamma matrix relations
= 2
I. The action of du
Cl(V) on a dspinor
S resulting
in distinguished irreducible representations c(du
)
.
=
and
c = (du
.
=
.
We obtain dmetric dgamma matrix relations
(u)
(u) +
(u)
(u) = 2g
(u) I,
which can rewritten for boldface coecients of metric. In irreducible
form
.
=
h
v
and
.
=
h
v
, or,
.
= (
h
i
,
v
a
) and
.
= (
h
,
v
).
The spin connection
S
for (pseudo) Riemannian manifolds is stan-
dardly determined by the LeviCivita connection,
S
.
= d
1
4
i
jk
j
dx
k
.
Similar constructions are possible for nonholonomic manifolds enabled with
metric compatible dconnections (for instance, in FinslerCartan geometry).
The spin dconnection operators
S
can be similarly constructed from any
metric compatible dconnection
f) u
= (e
i
f) dx
i
+ (e
a
f) y
a
,
for u
= e
(9).
Denition 3.4 The canonical (FinslerCartan) spin dconnection is de-
ned by D = {
} (
D = {
}) following formula
S
D
.
=
1
4
(
S
D
.
=
1
4
).
For the purposes of this work, we shall consider abstract index formula-
tions of dspinor calculus for nonholonomic manifolds/bundles with splitting
2 + 2, or 4 + 4.
25
3.1.2 Abstract dtensor and dspinor indices
Indices of dtensors are considered as a set of labels which can changed
into respective sets of dspinor indices, primed and unprimed (with dots
and without dots) following, for instance, such rules:
=
A
A
B
B
M
M
,
where dot spinor capital indices correspond small Greek tensor indices. For
h- and/or vdecompositions, when
= (
i
,
a
), we shall write
A
A
=
(
II
,
AA
), where
II
is for ver-
tical spinorvector. In similar forms, we can consider h- and v- and spinor
decompositions for forms and tensors with mixed indices. So, we shall fol-
low the formalism from [1] but redened in a form to be able to encode
spinorially dtensors with possible Nadapted splitting. Primed spinor in-
dices are complex conjugated with corresponding unprimed, for instance,
A
A
A
A
,
II
=
I
I
etc both for up and low indices.
We can consider antisymmetric spinors on total spaces and h- and
v-subspaces with the properties,
A
B
: =
A
B
=
A
B
,
A
B
:=
A
B
=
A
B
;
: =
I
=
IJ
,
I
J
:=
I
J
=
IJ
for hspinor indices;
: =
A
=
AB
,
A
B
:=
A
B
=
AB
for vspinor indices.
This is related to the rules of transforming low indices into up ones, and
inversely, using metrics and/or dmetrics, for instance, g
= [g
ij
, g
ab
] and
g
= [g
ij
, g
ab
]. In brief, the spinor decompostions of metrics are written in
the form,
g
=
A
B
A
B
, g
=
A
B
A
B
; (41)
g
ij
=
IJ
J
, g
ij
=
IJ
, for hmetrics;
g
ab
=
AB
B
, g
ab
=
AB
, for hmetrics.
In our works [25, 26, 27, 31, 17], we used also Nadapted gamma matri-
ces generating corresponding Cliord algebras for spinors (some authors
call them symbols or transition indices from Minkowski tetrads to spin
systems of reference, on corresponding tangent bundles). In brief, such
a formalism is related to orthonormalized (co) bases, e
= (e
i
, e
a
) and
e
= (e
j
, e
b
= (
A
A
= 2
, (42)
26
where the Minkowski metric
= e
and their
inverse, we can write
= 2g
and
= 2g
where dmetric (13) and, respectively, (14) and (15), including Ncoecients,
are considered for e
= e
and e
= e
du
=
A
A
B
B
A
B
A
B
(43)
for
A
A
:= e
A
A
= D
A
A
= D
A
A
, D
i
= D
II
= D
I
I
, D
a
= D
AA
= D
A
A
etc. Fixing spin diads
A
A
= (
0
A
,
1
A
),
I
I
= (
0
I
,
1
I
),
A
A
= (
0
A
,
1
A
), and
theirs respective duals,
A
A
,
I
I
,
A
A
,
9
we can introduce Nadapted dspin
coecients,
A
A
C
:=
B
A
D
A
A
A
C
=
A
C
D
A
A
B
A
,
J
II
K
:=
J
I
D
II
I
K
=
I
I
D
II
J
I
,
B
AA
C
:=
B
A
D
AA
A
C
=
A
A
D
AA
B
A
,
which are equivalent to the spin dconnection from Denition 3.4. This way,
we can introduce a canonical and/or Cartan type nulltetradic type calculus
etc. For simplicity, we shall omit in the future spinor h- and vindex formulas
if that will not result in ambiguities or lost of some important properties.
8
on tangent bundles to Lorentz manifolds of dimension 8=4+4, we can use one such a
gamma relation for the hsubspace and another one for the vsubspace
9
in our approach the underlined indices are equivalent to boldface indices in [1]
27
3.2.2 Spinor dcurvature and Bianchi identities
Following a Nadapted dspinor calculus with abstract indices for non-
honomic splitting 2+2 (for simplicity, we shall consider dot spinor indices;
only some examples for FinslerCartan congurations will be considered),
we prove:
Theorem 3.2 In dspinor variables,
the canonical dcommutator (33) (dtorsion (17)) is
=
A
B
A
B
+
A
B
A
B
,
ij
=
IJ
J
+
I
J
IJ
, ,
ia
=
IA
A
+
I
A
IA
, ...
where
A
B
:= D
A
(
A
D
A
B)
and
A
B
:= D
A(
A
D
A
)
etc;
the spinor dcommutators acting on a dspinor
C
result in
A
B
C
= [
C
T
A
B
+ (
A
C
B
T
+
A
T
B
C
)]
T
,
A
B
C
=
C
T
A
B
T
;
the Riemann dtensor (18) is
R
=
T
C
A
B
A
B
T
C
+
T
C
A
B
A
B
T
C
+
T
C
A
B
A
B
T
C
+
T
C
A
B
A
B
T
C
+ 2(
A
C
B
T
A
C
B
T
A
T
B
C
A
T
B
C
),
for = =
1
24
s
R;
the Weyl conformal dtensor (36) splits into antiselfdual,
C
:=
A
B
C
T
A
B
C
T
,
and selfdual parts,
+
C
:=
A
B
C
T
A
B
C
T
,
when C
=
C
+
+
C
.
Similar Nadapted 2 + 2 decompositions can be computed for
D
i
=
D
II
=
D
I
I
,
D
a
=
D
AA
=
D
A
A
. For instance, the components of the
28
FinslerCartan curvature (31) can be written
R
ijkh
=
IJKH
J
K
H
+
H
IJ
KH
+
IJK
H
I
J
KH
+ (44)
KH
IJ
H
+ 2(
IK
JH
K
J
H
IH
JK
IH
JK
),
P
ijka
=
IJKA
J
K
A
+
A
IJ
KA
+
IJK
A
I
J
KA
+
KA
IJ
A
+ 2(
IK
JA
K
J
A
IA
JK
IA
JK
),
...
We note that we can apply the formalism from [1] for any h- and vvalues
(with not dot indices) and, in general form for small Greek dtensor in-
dices, with dot indices), if we work in Nadapted frames and keep in mind
that Finsler like dconnections D
, or
D
A
B
C
T
= D
A
(
B
C
T)
A
B
, D
C
A
C
T
A
B
= 3D
T
B
;
the Einstein dequations (6), (21) and/or (34) transform the rst iden-
tity into D
A
A
B
C
T
= 8GD
A
(
B
T
C
T)
A
B
C
T
A
B
= 0, D
A
B
A
B
C
T
= 0;
in CartanFinsler gravity models with 2+2 splitting, similar formulas
hold for D
D;
all equations from this theorem transform into similar ones for the
LeviCivita connection if and only if
= 0, or
= 0, see
(33), which is equivalent to (7).
3.2.3 Nadapted conformal transforms
Let us introduce the value
:=
1
D
= D
ln , (45)
29
where the nonzero positive function (u) is taken for conformal tranforms
g
:=
2
g
A
B
A
B
=
A
B
and
A
B
A
B
=
A
B
.
A corresponding abstract dspinor calculus for last formula in Theorem 3.2
results in proof of
Proposition 3.1 In Nadapted form,
A
B
C
T
=
A
B
C
T
.
Applying statements of Theorem 3.3, we obtain formulas
D
A
A
A
B
C
T
=
A
A
A
B
C
T
and D
A
B
C
T
= D
B
(
B
P
A)
C
A
B
,
and
D
B
(
B
C
T)
A
B
=
A
A
B
C
T
.
Remark 3.1 Vacuum solutions of Einstein equations in general relativity
and models of FinslerCartan gravity with D, or
D, are not conformally
invariant. This follows from the fact that even
C
T
A
B
(
B
C
T)
A
B
C
T
A
B
. Nevertheless, it
should be emphasized here that such results are for a xed Nconnection
structure N, or
N. We can transform some data (g, N) with (non) zero
C
T
A
B
into certain (
g,
C
T
A
B
= 0, and/or
D
B
(
B
C
T)
A
B
A
B
C
T
= 0, for instance, following the anholo-
nomic deformation method [6, 7, 5, 17, 23].
Via Nadapted dspinor calculus, we can prove
Theorem 3.4 Under Nadapted conformal transforms, the values deter-
mining the vacuum Einstein equations for D transform following rules
C
T
A
B
=
C
T
A
B
= D
C(
B
)
T
+
C(
B
)
T
= D
C(
B
D
A
)
T
1
=
1
D
C(
B
D
A
)
T
,
4
2
= 4 + (D
) = 4 +
1
,
for := D
.
30
Using the operators D
C
B
and/or
D
C
B
B)
= 0. (46)
Let us formulate the conditions when such equations are conformally
invariant in Nadapted form. Choosing
B
=
B
, we can compute
D
A
A
B
=
D
A
A
B
+
B
C
A
and
D
(
A
B)
=
1
D
(
A
B)
(47)
where
C
A
T
A
B
C
T
= 0. (48)
Proof. It follows from D
A
(
C
D
A
B)
=
(
C
A
B)
=
A
B
C
T
T
.
B
=
B
iu
B
B
B
and
B
=
B
, (49)
D
B
B
C
= i
C
B
,
10
For simplicity, we shall consider that the spinor
B
does not posses an electromagnetic
charge.
32
where the point u
B
B
V,i
2
= 1, and
B
and
B
are constant values with
respect to Nadapted frames (8) and (9) for which
A
B
C
T
= 0.
We can generalize the concept of twistors for at spaces to nonholonomic
manifolds which are conformally at in Nadapted form:
Denition 4.2 An nonholonomic (equivalently, anholonomic) twisor space
T
is a four dimensional complex vector space (with real eight dimensions)
determined by elements of type Z
= (
A
,
A
=
A
=
A
), where
A
=
A
and
A
=
A
+iu
A
A
A
, (50)
D
A
A
B
= i
B
A
,
are solutions of the dual nonholonomic twistor equations D
(
A
A
B)
= 0.
The complex conjugation of nonholonomic (dual) twistors follows the
rules
Z
= Z
:= (
A
,
A
) and W
= W
:= (
A
,
A
).
We can conser higher valence twistors, for instance, X
where Nadapted
twistor indices transform respectively following rules (49) and (50) taken -
or + before complex unity i.
4.1.2 Geometric/physical meaning of anholonomic twistors
A nonholonomic frame structure prescribes a corresponding spiral con-
guration for twistors and their conformal transforms.
33
Denition 4.3 -Corollary: The class of curved spaces generated by an-
holonomy relations (12) subjected to the compatibility conditions (48) is
characterized by anholonomic spirality
s :=
1
2
Z
Z
, (51)
which is invariant under Nadapted conformal transforms.
Proof. It follows from verication that Z
Z
=
Z
Z
(using formulas
(47) and (49)).
/i
+
B
B
, R,
describes a light ray propagating in N-adapted form in an eective locally
anisotropic media and/or a curved spacetime with geometric objects induced
by nontrivial anholonomy coecients . If
B
= 0, such a light ray is moved
to innity.
We can also characterize masless particles with momentum, rotation and
spirality propagating in eective curved spaces derived for certain anholon-
omy relations of moving frames. Taking Z
= (
A
,
A
) with
A
= 0, we
construct
p
A
A
:=
A
, M
A
A
B
B
:= i
(
A
B)
A
B
i
(
A
A
B
and spin dvector S
=
1
2
e
= sp
, where e
is the absolute
antisymmetric dtensor and s is computed as in (51). In local Nadapted
form, such a physical interpretation of nonolonomic twistors is possible with
respect to bases of type (8) and (9) for which
A
B
C
T
= 0. This describes
a massless particle moving in a subclass of curved spaces with nontrivial
curvature for when certain anholonomic constraints are imposed.
4.2 Finsler twistors on tangent bundles
Originally, the FinslerCartan geometry was constructed on tangent
bundles with D =
D and 4 + 4 splitting as we explained in section 2.2.3.
The corresponding Weyl dtensor
C
J)
= 0, D
(A
A
B)
= 0. (52)
Choosing
J
=
J
,
B
=
B
, we can compute
D
II
J
=
D
II
J
+
J
I
KJ
and
D
(I
I
J)
=
1
D
(I
I
J)
,
D
AA
B
=
D
AA
B
+
B
A
CA
and
D
(A
A
B)
=
1
D
(A
A
B)
,
where
II
:=
D
II
ln
C
A
AA
:=
D
AA
ln are constructed similarly
to the conformal dvector (45).
If D =
D, we can obtain from the Theorem 3.2 the
Corollary 4.1 The antisedual Weyl dspinors corresponding to the Car-
tan dcurvature
R
= {
R
i
hjk
,
P
i
jka
,
S
a
bcd
} (31) are characterized by h
and vcomponents {
LIJK
,
DIJK
,
DABC
}.
This results in a set of three conditions of compatibility:
Lemma 4.2 The FinslerCartan twistor equations (46) are compatible if
and only if
LIJK
L
= 0,
DIJK
D
= 0,
DABC
D
= 0. (53)
All results on FinslerCartan twistors can be proved using formal Sasaki
lifts g g (28) and N
N (27) with spinor coecients (41) for the con-
structions with nonholnomic twistors and canonical dconnections.
Theorem 4.2 If the compatibility conditions (53) are satised, we can solve
the nonholonomic twistor equations (52) in general form for hcomponents,
J
=
J
iu
JJ
J
and
J
=
J
, (54)
D
JJ
K
= i
K
J
J
,
and for vcompnoents
B
=
B
iu
BB
B
and
B
=
B
, (55)
D
BB
C
= i
C
B
B
,
where the point u
= (u
II
, u
AA
) TM and constant values are considered
with respect to Nadapted frames (8) and (9) when the conditions (53) are
satised.
35
We can generalize the concept of twistors for at spaces to nonholonomic
manifolds which are conformally at in Nadapted form:
Denition 4.5 A horizontal twisor space hT
is a four dimensional
complex vector space (with real eight dimensions) determined by ele-
ments of type hZ
i
= (
I
,
I
),
i = 1, 2, 3, 4 with the two spinor com-
ponents hZ
I
=
I
and Z
I
=
I
dened by solutions of type (54).
A vertical twisor space vT
is a four dimensional complex vector space
(with real eight dimensions) determined by elements of type hZ
a
=
(
A
,
A
), a = 5, 6, 7, 8 with the two spinor components vZ
A
=
A
and
Z
A
=
A
dened by solutions of type (55).
We conclude that nonholonomic twistor constructions for the Finsler
Cartan spaces dub as h- and vcomponents the values introduced via canon-
ical dconnections on V. Redening the abstract index formalism for d
tensors and dspinors, all formulas can be proved by similarity in Nadapted
frames.
4.3 Nonholonomic local and global twistors
On Nadapted conformally at nonholonomic manifolds, the solutions
(49) of generalized twistor equations (46) dene certain global anholonomic
twistor structures. If the conditions (53) are not satised, we can only
dene a nonholonomic twistor bundle on a V when the geometric object
depend on base manifold points. This does not dene an alternative de-
scription of nonholonomic manifolds (and FinslerCartan geometries) in
terms of certain generalized nonholonomic twistor spaces. For a prescribed
Nconnection structure N, we can construct Nadapted local twistors with
properties similar to those of holonomic twistors considered in Chapeter 6,
paragraph 9, in [1].
Nevertheless, nonholonomic/ Finsler spaces are characterized by more
rich geometric structures which provide us new possibilities and methods
for constructing new classes of generalized twistor Finsler spaces and ap-
plications in general relativity and modications. We study two models of
Nadapted twistor spaces in local and global forms.
4.3.1 Nadapted local twistors and torsionless conditions
Let us consider a point u V for nonholonomic data (g, N, D).
36
Denition 4.6 A local Nadapted twistor
u
Z
(in brief, local dtwistor)
in a point u is given by a couple of Nadapted twospinors (
u
A
,
u
) in
this point, which in a chosen anholonomic frame (8) and (9) satised the
rules: if g
:=
2
g
then
u
Z
= (
u
A
,
A
)
u
Z
= (
u
A
=
u
A
,
u
A
=
u
+i
A
A
u
A
).
The local dvectors
u
Z
and
u
Z
depend functionally, respectively, on
(g, N, D,). The set Z
=
uV
(
u
Z
) of all local twistors
u
Z
taken in all
points u of V denes a vector bundle, when the ber in u is a complex four
dimensional vector space (i. e. the spaces of local Nadapted twistors in
u). Such a vector bundle is nonholonomic being endowed with Nconnection
structure. For simplicity, we shall omit the left low lable u and write a
local twistor as Z
if that will not result in ambiguities.
In Nadapted (and/or general local) form, the connection can be
constructed to possess zero coecients in a point and/or along a curve
though such a point (the socalled normal coordinates). This allows us
to dene transports of usual local twistors along curves with tangent vec-
tor elds t
A
A
, when
t
A
A
D
A
A
B
+it
B
A
= 0, (56)
t
A
A
D
A
A
+it
A
A
P
A
A
B
B
B
= 0,
where P
A
A
B
B
() =
0
u
A
A
+ t
A
A
B)
= 0 and
A
=
1
2
iD
A
A
A
. (57)
Corollary 4.2 The solutions for local dtwistors (57) can be globalized on
V if and only if the conditions (48) are satised, for instance, if the Weyl
dspinor vanishes.
Denition 4.7 The Nadapted covariant derivative operator (dconnection)
along t
D
A
A
(58)
37
The local dtwistor dconnection (58) allows us to compute the variation
of Z
along u(), following formulas (56) with nonzero right sides,
t
DZ
=
_
t
A
A
D
A
A
B
+it
B
A
, t
A
A
D
A
A
+it
A
A
P
A
A
B
B
B
_
.
Here we note that in similar form we can dene dual local dtwisotrs of type
W
= (
A
,
A
A
+
A
= Z
K
(t, v) = t
,
for two dvectors t
= t
M
M
, v
= v
N
N
=
_
_
i
M
N
A
M
N
M
N
D
A
A
M
N
+
M
N
D
B
A
M
N
0 i
M
N
A
M
N
_
_
.
Above constructions are determined by data (g, N, D =+Q), see (5).
In general, they can be redened for data (g,) using nonolonomic defor-
mations.
Theorem 4.3 We can globalize in nonholonomic form the local twistor con-
structions for if there is a Nconnection structure N and associtated D
for which K
= 0 and
= 0.
Proof. It is a consequence of conditions of Theorem 3.2 when
= 0
are equivalent to (7), i.e. the nonholonomically induced torsion (by (g, N))
became zero. This is compatible with Conclusion 2.2 when the conformal
Weyl dtensor/dspionor for D can be zero but similar values for are not
trivial. Such linear connections are dierent even in some Nadapted frames
they can be characterized by the same set of coecients (transformation laws
under frame/coordinate changing are dierent).
:=
2
(u)
with
= (e
j
= dx
j
, e
b
= dy
b
+ N
b
dx
i
). For such a d
metric and Nadapted cobases, we can verify that C
= 0 as a conse-
quence of Proposition 2.3. We can redene data ( (13), g
) in a coordinate
form (14) with coecients (15) (with primed indices, g
). Then consid-
ering arbitrary frame transforms e
we can compute g
= e
.
Finally, we can redene for a nonholonomic 2+2, or 4+4, splitting certain
data (g
, N
a
i
) , for which, in general, C