Вы находитесь на странице: 1из 13

An Elementary Proof of the Existence of Isothermal Parameters on a Surface

Author(s): Shing-Shen Chern


Reviewed work(s):
Source: Proceedings of the American Mathematical Society, Vol. 6, No. 5 (Oct., 1955), pp. 771-
782
Published by: American Mathematical Society
Stable URL: http://www.jstor.org/stable/2032933 .
Accessed: 03/06/2012 10:55

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

American Mathematical Society is collaborating with JSTOR to digitize, preserve and extend access to
Proceedings of the American Mathematical Society.

http://www.jstor.org
AN ELEMENTARY PROOF OF THE EXISTENCE OF
ISOTHERMAL PARAMETERS ON A SURFACE
SHIING-SHEN CHERN

1. Introduction. Let
(1) ds2 = E(x, y)dx2 + 2F(x, y)dxdy + G(x, y)dy2,
EG - F2 > 0, E > 0,
be a positive definite Riemann metric of two dimensions defined in
a neighborhood of a surface with the local coordinates x, y. By iso-
thermal parameters we mean local coordinates u, v relative to which
the metric takes the form
(2) ds2 = X(u, v)(du2 + dv2), X(u, v) > 0.
In order that isothermal parameters exist it is necessary to impose
on the metric some regularity assumptions. In fact, it was shown
recently by Hartman and Wintnerl that it is not sufficient to assume
the functions E, F, G to be continuous. So far the weakest conditions
under which the isothermal parameters are known to exist were
found by Korn and Lichtenstein.2 To formulate their theorem we re-
call that a function f(x, y) in a domain D of the (x, y)-plane is said to
satisfy a Holder condition of order X, 0 <X _ 1, if the inequality
(3) f(x,y) -f(x', y')I <CrX
holds for any two points (x, y), (x', y') of D, where C is a constant
and r is the Euclidean distance between these two points. With this
definition the theorem of Korn-Lichtenstein can be stated as follows:
Received by the editors September 29, 1954 and, in revised form, November 15,
1954.
1P. Hartman and A. Wintner, On the existence of Riemannian manifolds which
cannot carry non-constant analytic or harmonic functions in the small, Amer. J. Math.
vol. 75 (1953) pp. 260-276. Also: S. Chern, P. Hartman, and A. Wintner, On iso-
thermic coordinates, Comment-Math. Helv. vol. 28 (1954) pp. 301-309.
2 A. Korn, Zwei Anwendungen der Methode der sukzessiven Annaherungen, Schwarz

Abhandlungen pp. 215-229; L. Lichtenstein, Zur Theorie der konformen Abbildung.


Konforme Abbildung nichtanalytischer, singularitatenfreier Flachenstiicke auf ebene
Gebiete, Bull. Int. de l'Acad. Sci. Cracovie, ser. A (1916) pp. 192-217. Cf. also the
paper of C. B. Morrey, On the solutions of quasi-linear elliptic partial differential
equations, Trans. Amer. Math. Soc. vol. 43 (1938) pp. 126-166, in which the iso-
thermal parameters are shown to exist in a generalized sense under weaker hypotheses.
Added in proof. Weaker conditions were recently found by Hartman and Wintner.
Cf. their paper, On uniform Dini conditions in the theory of linear partial differential
equations of elliptic type, Amer. J. Math. vol. 77 (1955) pp. 329-354.

771
772 S. S. CHERN [October

Suppose, in a domain D of the (x, y)-plane, the functions E, F, G


satisfy a Hdlder condition of order X, 0 <X < 1. Then every point of D
has a neighborhood whose local coordinates are isothermal parameters.
This theorem is rather useful in the global theory of surfaces, when
analyticity assumptions are not desirable. We shall present here what
seems to be an elementary and rather straightforward proof. Need-
less to say, the essential ideas of this proof are contained in the works
of Korn and Lichtenstein. We believe, however, that some simplifica-
tions are achieved by the consistent use of the complex notation.3
We first observe that the isothermal parameters are invariant un-
der conformal transformations, that is, multiplications of the Rie-
mann metric (1) by a positive factor. Under a conformal transforma-
tion the angle between two vectors remains invariant. If we further
assume that the coordinates x, y form a positive system, that is, if we
allow only those transformations of local coordinates for which the
Jacobian is positive, the angle can be defined, together with its orien-
tation. To express these relations analytically, it will be convenient
to introduce complex-valued differential forms, that is, forms
wo=a+i13, where a, ,B are real differential forms. We shall write
co=oa-if.
Since the quadratic differential form in the right-hand side of (1) is
positive definite, we can write
2 2 2
(4) ds = 01 + 02,

where 01, 02 are real linear differential forms:


01 = aldx + bidy,
02 = a2dx + b2dy.

Assuming that alb2-a2b, >0, the forms 01, 02 are determined up to a


proper orthogonal transformation, that is, one with determinant + 1.
We put
(6) 01 + i02,

so that
(7) ds2 =i.
The form q5is then determined up to a complex factor of absolute
value 1. A conformal transformation of the Riemann metric is given,
3After this paper was submitted for publication,it has come to my attention that
a similarproofwas given by L. Bers in his mimeographednotes on Riemann surfaces,
New York University, 1951-1952.
x9551 ISOTHERMALPARAMETERS ON A SURFACE 773

in terms of the form g5, by the multiplication of g5 by an arbitrary


nonzero complex factor. In the recent terminology we say that the
complex linear differential form g5,determined up to a nonzero com-
plex factor, defines an almost complex structure. As discussed above,
such a structure allows the introduction of the notion of oriented
angle.
The determination of isothermal paratneters u, v is equivalent to
that of a complex-valued function w=u+iv, such that
(8) dw = (l/p)4.
For we have then
(9) ds2 = O = I p 2dwdw = Ip 2(dU2 + dv2).

Conversely, the isothermal parameters u, v determine a function w


satisfying (8).
2. Preliminaries. We shall first make estimates of certain integrals,
which will be needed in the proof.
Let (t, q) be any point in the (x, y)-plane, and let
(10) r = + ((x - t)2 + (y - )2)1/2

If g(r) is a function of class 1, we have, by exterior differentiation,


-(y - q)dx + (x - )dy 9'(r)
d{g(r) (= g dxAdy.

It follows by Stokes Theorem that, if D is a domain bounded by a


curve C and if (t, q){D, we have

r1r g'(r) r_-(Y_-_ (x_-_)dy


)dx +
IfJ D r dxdy = fg(r) -(

This formula remains true, even if (t, 7) ED (but not on C), provided
that the integral in the left-hand side is convergent. The integral

(12)
(12) V ~~~1
A-(y- x + (x - t)dy
71)d (,r){
, 1 E
is an integer and is usually called the order of the point (t, 7) relative
to the curve C.
In particular, if g(r) =rX, X#O, formula (11) becomes

(13) ffrxdxdy fx -(y-71)dx + (x- )dy


D

and is true either when (t, 7) ED or when X> 0. It follows that, if the
774 S. S. CHERN (October

vector joining (t, 7) to a point (x, y) CC turns monotonely in the same


sense, we have
rrrXdxdy r
(14)
fJ r2 -sx ,
(15)
(15)
xxdI<< _27r sx
I1ffJDJ rXdxdy X < ot

where A and a denote respectively an upper bound and a lower bound


of the distance from (t, 7X)to a point (x, y) CC. We emphasize that
formula (15) is valid only under the assumption (Q, q) {D.
3. Main lemma. To simplify our formulas we shall use the complex
coordinate z =x+iy in the (x, y)-plane. If f(x, y) is a complex-valued
function of class 1, we define fx, f, by the equation
(16) df = f2dz + ffd2 = f2(dx + idy) + f1(dx - idy).
They are therefore related to the ordinary partial derivatives f., fv
by the equations
(17) fv = (fx - ifv)/2, fi = (f + if)/2.
We shall write the function as f(z, z), thus emphasizing the fact that
it is in general not analytic in z.
On the other hand, we can define these operators fz, fz directly,
without using the partial derivativesfX,f. For instance, we can adopt
the definitions:
fz = lim {f(z + h/2, z + h/2) -f(z, z)
hO0

( 18) + if(z - hi/2, z + hi/2) - if(z, ) }/h,


f8 = lim {f(z + h/2, z + h/2) -f(z, z)
h-O

+ if(z + hi/2, z - hi/2) - if(z, ) }/h,


which clearly give (17), when f, f,, exist.
LEMMA.Let D be a disc of radius R about the origin in the (x, y-) plane.
Let f(z, z) be a complex-valued continuous function in D, which satisfies
the condition

(19) |f(?, D)-f(r2, P) 1 Brx2, r12 = | - 2j

for any two points P1,?2CD, where X and B are constants, 0 <X < 1. Let
the function F(D, T) be defined by
1955] PARAMETERS
ISOTHERMAL ON A SURFACE 775

orf(z,z2)dzdz rrf(z,z2)dxdy
(20) -27riF(D, ) = f f = 2i f,J )d E D.
D D

Then: (1) Fr and F4 exist, and F =f(, T); (2) If If(z, 2) A, z D,


the following inequalities are valid:
(21) |F(D,) -<4RA,
(22) | F(D, ) J< (2x+1/X)RxB,
(23) IF(r1, 11) - F(r2, T2) | < 2(A + (2X+1/X)RXB)r12,

(24) | F(r1, Tj) - Fr(2, T2)| < (X)Br12,


where ,u(X) > 0 is independent of P1, ?2.
To prove the lemma we write, according to the second equation of
(18),

- 27riF- = lrim I((, 2))-(~~)


A' T))
h-,o 2 (Jz'
J(z f - (z - r - h/2)
D
1 dajr r dzdz
hj z
(z - )(z - -hi/2)l dzdz+f( -

Computation gives
rr ddz
(25) = - 21riT.
-
D Z

From these it follows that F1=f(?, T). Similarly, we prove that Fr


exists and is given by

(26) - 27 iFr f(f B) f("T) d-dz.

From (20) we have, by using (14) with X = 1,

2r | F ?2A dxdy < 87rAAR


D

which gives (21). Similarly, from (26), we have


r x 2X+2
2irIFr| < 2BffJ 2dxdy 2 7rBRX,
D

which gives (22).


Inequality (23) is trivial. To prove (24) let Do be the intersection
776 S. S. CHERN [October

of D with a circular disc of radius 2r12=21J -t21 about ?2. We can


write
- 27,i{Fr(?,, ) - Fr(?2, T2) }
= J {S ) - f(tip TO
Z)-f(zf, d2dz
Do (z -

-3 {f(z, ) -f(?2, ?2) } dz


1
(27) (Do 2)

+ {f(Z,f ) f(~i)} (T dzdz


Z
D-Do

- JJ {f(Z, 2) - f(.1, T2)


I
d2dz
(Z )2
D- Do -

a rr dzdz
+ {f(r2, T2) - f(t, T1)I -
J J

We observe that in this sum only the first two integrals are improper
integrals, while the last three are proper integrals. To estimate their
absolute values we first have

Jj {f(Z,f ) - f(i,i T } dzdz


Do
(Z
dxdy 47rB
< 2B J<J 1 (3rl2)
Do

Do ~~~~(Z
-2 2)2
dxdy 47rB
_<2B _P12-- < (2rl2) X

The sum of the third and fourth integrals in (27) is equal to

ff(z2) - f(?.,
t') d(
2ddz
D- Do

= 4i r r i 1
=4i J J f(z' 2)- f(~, I)+ f(~, I)- fa-, i) }o dtdxdy,
D-D0 (z -
I9551 ISOTHERMALPARAMETERS ON A SURFACE 777
where r is a point on the segment ?1?2. It follows that its absolute
value is
? 4r12(Fj+ F2),

where F1 and F2 are respectively upper bounds of the integrals


fZ, 2) _ r dxdy
3

f dxdy
{S(r$-f(?. -

z
D- Do

along the segment P1r2. Using (15), we have


f -,
fz, ) dxdy?B f d

4rB 1
1 -X rl2 X

Cff(,)' - f(~', T x 47r


I I ( r r)-f'rlr19
dxdy ? Br12--
D-Do

The right-hand members of these inequalities can therefore be taken


as F1 and F2 respectively.
Finally, we have, by (25),
d2dz dzdz d2dz - -

D- Do
~2
D
~ 2
Do

the integral over Do being equal to zero. It follows that the last term
of (27) is zero.
Summing up these estimates, we get
2ir I Fr(r1, )- Fr(r2, T2) I
< 47rB
; (3r12)"+ 47rB (2rl2)'+ 4 47rB + X)r2
r12
4XrB

This gives (24), if we set


2 2-X
(28) ____(3 + 2X) + 8

Thus the proof of the lemma is complete.


778 S. S. CHERN [October

4. An existence and uniqueness theorem on an integro-differential


equation; applicationto isothermal parameters.
THEOREM. In the domain D of the complex z-plane defined by Iz ? R
(= positive constant) let
(29) Zw = a(z, 2)wi + b(z, 2)w,

be a differential operator, whose coefficients a(z, s), b(z, z) satisfy a


Hdlder condition of order X, 0<X<1, and vanish at z=O. Let a(z, z)
be a function in D, satisfying a Hdlder condition of the same order X.
Then the equation
(30 (Zw + aw)(z, B)
(30) T) + JJd2dz
27riw(~, = ( CD,
D

where o() is a complex analytic function, with o(0) = 0, has exactly one
solution w(z, z), provided that R is sufficiently small.
Before proceeding to the proof of this theorem, we shall make some
further estimates of integrals, based on the lemma of the last section.
The hypothesis of the theorem implies the existence of a number M
large enough to fulfill the following inequalities:
(31) 1 < M, I a (D, T)I :_!M, I a-(r)I < M,

(32) | h(b1, O - h(r2, T2)I _ Mr'12,

(33) (Z + a)o(1) -(Z + a)of(r2) | <- rl2, r-2 = -


-2

where?, P1,P are any three pointsof D and whereh(z,z) standsfor


each of the functions a(z, z), b(z, z), a(z, z), o(z). Since a, b, (Z+a)o
all vanishat 0, it followsfromthe corresponding Holderinequalities
that
Ia(D,T)I < MIDI ?MR<x, Ib(q, I)I MRX,
( (Z + a)o(t) I_ M2R.
Considernow the functionF(D,T)definedin (20). Usingthe nota-
tion of the lemmaof ?3,we have
I(Z +a)F(q, T)I
= Ia(D,T)FF+ b(q,T)Fr+ a(q, T)F I
(35) < MR)I(A+ (2x+1/X)RXB)+ 4MRA
= MRX{ (1 + 4R1-x)A+ (2x+1/X)RXB
},
'9551 ISOTHERMALPARAMETERS ON A SURFACE 779

| (Z + a)F(v,, ) - (Z + a)F(v2, T2) I


-|a(t,, T&)(v,,Ti)+ b(v,,Tj)Fr(P,,Tj)+ at(P,,Tj)F(P,,Tj)
-3a(v2, 2)f(v2, T2) - b(r2, N2)Fr(r2,T2) - a(r2, T2)F(?2, T2)
(36) ? Mr 2{A + (2 /X)R B + 4RA} + MR (B + /(X)B)r12
+ M(A + (2 +1/X)RXB)2r12
? Mri2{A + gi(R)A + g2(R)B}
where
gi(R) = 4R + 22-xR1-',
g2(R) = { (2x+1/X) + 1i(X)+ I }RX + (8/X)R

are functions of R, which tend to zero with R.


Having the above inequalities, we shall prove the existence of a
solution of (30) by successive approximations. For reasons which will
be clear later we put the following restrictions on R:
+ 2
(38) 4RX < 1 22- 2-1 Rt--' < 1

and we choose a constant c, depending only on X, such that


X+ 2 X
X +2 X +2
(39)
2
1+ /.t(X)+3 2x<c.

We now define
27riwo(q,) =

(40) - f (ZWn
+ aw) (z Z)
n =0, 1,*
27riwn__ ________ d2dz, *
z- *
D

For these functions we shall prove the inequalities:

(41) I wn(, T) ? M(cMRX)n


(42) | (Z + aC)wn(p,T) | < M(cMRx)n+l,

(43) | wn(1, 1)- Wn(;2, T2) | ? M(cMR') r12,

(44) | (Z + a)Wn(li, T) - (Z + a)Wn(;2, T2) | X) rl2.


(cM 2/2X)(cMR

In particular, the last one implies that the function under the integral
780 S. S. CHERN [October

sign in (40) satisfies a Holder condition, thus allowing the definition


of the next integral.
The inequalities (41)-(44) will be proved by induction on n. For
n =0, (41) is a consequence of the third inequality of (31), (42) fol-
lows from the third inequality of (34), since c> 1, (43) follows from
(32), and (44) from (33).
We now suppose that (41)-(44) are true and proceed to prove the
correspondinginequalities for n + 1. By (21), (23), and the induction
hypothesis, we have
Wn+l(P, !) ? 4R. M(cMRx) n+1 < M(cMRx)n+1I
Wx+?i(?l1 I1) - Wn+1(72y T2) I
{ M(cMRX)n+l + (2x+1/X)RX(cM2/2X)(cMRX)?n}
=<2r12
X tn+1 X -
- M(cMR) r12(1+ 2/X) 2r12X.
The last relation gives (43) (for the index n+1), on account of the
second inequality of (38).
Similarly, we get from (35) and the induction hypothesis,
(Z + a)wn+i(?, 9)| + 4Rl'-)cM2RX
MRXI{(1
}(CMRII)n
+ (21+1/X)RX1(CM2/2X1)
- M2RX(cMR)?)+l(1+ 4R'- + 2/X)
S M(cMR) +2

on using the first inequality of (39). By (36), we have


(Z + at)Wn+i(r1, 11) - (Z + Ct)Wn+1(2, 12) I

< Mr 2 M(cMRX) {(1 + gl(R))cMR" + g2(R)cM/2X}


= M rl2(cMR ) I+l{1 + gi(R) + (1/(2R)X)g2(R)}
This gives (44) for n+1, if
1 + gi(R) + (1/(2R)X)g2(R)? c/2\.
But the latter follows from the second inequality of (39). Thus our
induction is complete.
It follows that the series
00

(45) Wn
w(Z, 2)

converges absolutely and uniformly in D if CMR)< 1, and defines a


function w(z, z), which satisfies (30).
'955] ISOTHERMALPARAMETERS ON A SURFACE 781

To prove the uniqueness of the solution when R is sufficiently


small, let w'(z, z) be another solution of (30) such that (Z+a)w'(z, z)
satisfies a Holder condition of order X. Then the function
w(z, z) = w(z, z) - w'(z, z)

satisfies the equation

-27i(, ) f (Z + a) (z ) d2dz.
z
D

Let A and B be respectively the least upper bounds of

|(Z + at)fv(D,), rED


and
|(Z + a) (?', ')-(Z + a) t(2, T2) | /r12, P1, P2 E D1 5 P2-

From (35) and (36) we get


A MR"{ (1 + 4R'-N)A + (2x+'/X)RXB},
B ? M{A + gi(R)A + g2(R)B}.
From these we easily conclude that, if R is sufficiently small, A =0.
The latter implies that w(z, ) =-0. Thus the proof of our theorem is
complete.
In order to derive from the above existence theorem the theorem of
Korn-Lichtenstein we follow the notation of ?1. In a neighborhood
of the point z = x +iy =0 we suppose the almost complex structure to
be given by the complex-valued linear differential form
(46) = (1 - a(z, z))dz + b(z, z)dz,
which is determined up to a nonzero factor. By a linear transforma-
tion on x, y with constant coefficients and by multiplication of 4 by
a constant factor when necessary, we can suppose that a(O, 0)
=b(O, 0) =0. Equation (8) is equivalent to the equations
wa = b/p, wz= (1-a)/p.
Elimination of p gives
(1 - a)wi - bwz= 0,
or, if we make use of the operator Z in (29),
(47) Ws= Zw.
By the above existence theorem, the equation
782 S. S. CHERN

(48) 2riw(r, ) + ZW(z Z d-dz =( ) C D, (O) =O,

has a solution w(z, z). If o() is not a constant, say o() =D, this solu-
tion has the property that wr(O, 0) z0O. Since Zw(z, z) satisfies a
Hoilder condition, it follows from our lemma in ?3 that w(z, z) satis-
fies (47). Thus we have proved the theorem of Korn-Lichtenstein.
REMARKS.1. As L. Bers observed to me, the same method can be
used to establish the existence of a local solution of the equation
(49) w=aw- + bwz + aw + #w + y,
where a, b, a, ,B,'y satisfy a Holder condition and where Ia| + Ib < 1.
2. It follows from our proof that the first partial derivatives of the
isothermal parameters also satisfy a Holder condition of order X.
INSTITUTEFORADVANCED STUDYAND
UNIVERSITYOF CHICAGO

Вам также может понравиться