0% found this document useful (0 votes)

80 views73 pages

Comm Pure Appl Math - May 1978 - Yau - On The Ricci Curvature of A Compact K Hler Manifold and The Complex Monge Amp Re

Uploaded by

operatorjohn433

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

0% found this document useful (0 votes)

80 views73 pages

Comm Pure Appl Math - May 1978 - Yau - On The Ricci Curvature of A Compact K Hler Manifold and The Complex Monge Amp Re

Uploaded by

operatorjohn433

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

On The Ricci Curvature of a Compact Kahler

Manifold and the Complex Monge-Ampere

Equation, I*
SHING-TUNG YAU
Stanford University

0. Introduction

Let M be a compact Kahler manifold with Kahler metric C,,,g,rdz'@ddzI.

Then it is well known that the Ricci curvature of M can be computed in the
following way. If R,; dz' C3 dZJ is the Ricci tensor, then

In particular, the (1,l)form ( & i / 2 ~ ) I ~ R, ~, i d z ' ~ d f is' a closed (1, 1) form

which is equal to ( - f i / 2 7 r ) & l o g det (gS7)]. A well-known theorem of S. S.
Chern [6] shows that the cohomology class of this (1,l) form depends only on
the complex structure of M and is equal to the first Chern class of M.
I,,,
Therefore a necessary condition for a ( 1 , l ) form ( G I a ' r r ) Rlr dz' A d? to
be the Ricci form of some Kahler metric is that it must be closed and its
cohomology class must represent the first Chern class of M.
More than twenty years ago, E. Calabi [3] conjectured that the above
necessary condition is in fact sufficient. This conjecture of Calabi can be
reduced to a problem in non-linear partial differential equation.
In fact, if (d?/27r) Rlr dz' A dZJ represents the first Chern class of M ,
then we can find a smooth function F defined on M such that

If XI,, R l j d z ' @ d 2 ' is the Ricci tensor of some Kahler metric Ci,jgijdz'@ddz1,
then

"The research f o r this paper was done while the author was a Visiting Member at the
Courant Institute of Mathematical Sciences. Reproduction in whole or in part is permitted for
any purpose of the United States Government.

Comniunicitticins on Pure and Applied Mathematics, Vol. XXXI, 339-41 1 (1978)

@ 197X John Wiley & Sons. Inc. 00 10-36401781003 1-0339/$01 .OO
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
340 SHING-TUNG YAW

and the above equation shows that

Since M is compact and det (g,r)/det (g,;) is a globally defined function,

the maximal principle shows that log det (g,r)/det (g,T)- F is a constant.
Hence, for some constant C>O,

(0.3) det (is:)

= C exp {F)det (g,r) .

If ;& giidzi A d.2' is cohomologous to 4 J - 1 giTdzi~ d 9 then

, for
some smooth function Q we have gii = gij +aZcplaziaZi. In this case, the above
equation is equivalent to the following equation:

= C exp {F)det ( g s r ) .

Conversely, if for some constant C>O, we can solve (0.4) so that the
solution cp is smooth and C,,J(gii +a*cp/az'dZ') dz'@&f defines a Kahler
metric, then we can solve the conjecture of Calabi. It turns out, by integrating
(0.4), that there is only one choice of C:

(0.5) C
Iexp { F } = Vol ( M ) .

In this paper, we shall solve equation (0.4) with the (necessary) compati-
bility condition (0.5). In view of its application in Kahler geometry, we
generalize (0.4) to the following more general case.
Let L , and L, be holomorphic Hermitian line bundles over M. Let
sl,. * , s, be holomorphic sections of L , and t , , - - , t, be holomorphic
-
sections of L,. Then for hl 2 0, * , h, 2 0, * * * , k, 2 0, we study the equation

where F is a smooth function defined on M x R such that aF/acp 2 0.

We assume that we can find a smooth function JI defined on M such that

(0.7) (1~~)~'1+$-. - -+)tq)2kq)-1exp{F(x,

*
~+)spJZh~)()fllZkl+* J/)}=Vol ( M ) .

It is also necessary to assume that the right-hand side of (0.6) cannot tend
to infinity too fast. Hence we assume that, for m =dim M,
c

and
for some q >0, the function ( l t l ) z k l + - * + Itq(2kq)-q-1/m
+

(0.9) )A log + * * * + 1 tq)2kq))(m-1)'m is integrable over A4 and over

(JtlJ2k1

every analytic disc of M.

We shall show that, under the conditions (0.7)-(0.9),equation (0.6) has a
bounded solution cp which is smooth and defines a Kahler metric

outside the variety defined by the si and the variety defined by the ti. Near
the variety defined by the si, the Kahler metric is uniformly bounded from
above and near the variety defined by the ti the Kahler metric grows upward
at most like a "pole". Furthermore, the function 50 is unique up to a constant.
Equation (0.6) is considerably more complicated than equation (0.4) and
will be proved only at the end of the paper. However, when the right-hand
side of (0.6) specializes to the form exp{q+F(x)}, then equation (0.6) is
much easier than (0.4) and can be used to deduce the existence and
uniqueness of a canonical Kahler-Einstein metric on a compact Kahler
manifold with ample canonical line bundle. This was also conjectured by E.
Calabi.
It should be mentioned that Calabi proved the uniqueness of the solution
of (0.4) a long time ago. He also proved that when the right-hand side of
(0.4) is close enough to a constant, then (0.4) has a smooth solution. (This
part of Calabi's work was also made precise by T. Ochiai.) His later paper [ 5 ]
on affine hyperspheres is very important for the study of the Monge-Ampere
equation. The computation of the Laplacian of the third derivatives originated
from there and was shown to us by L. Nirenberg. Furthermore, in [l], T.
Aubin published a proof of Calabi's conjecture assuming the original Kahler
manifold has non-negative holomorphic bisectional curvature. This class of
manifolds is rather restrictive. Moreover, he used the variational method and
his procedure is rather difficult to comprehend.
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
342 SHING-TUNG YAU

My proof of Calabi's conjecture was completed in the middle of 1976. A

large amount of work was done while I was working on the real Monge-
Ampere equation (cf. [8], [9]). For example, Pogorelov's second order
estimate in the real Monge-Ampere equation had direct bearing on my work
here. Indeed, besides the uniform estimate, the higher order estimates were
discovered at that time. A special case of equation (0.6) when the right side is
the function exp { i +F ( x ) } had been studied independently by Aubin [2].
The plan of this paper consists in starting from special cases of the
theorem to make it easier for the reader to comprehend what is going on
before treating the more involved equation (0.6). We use the continuity
method and the required estimates are worked out for equation (0.4) in
Sections 2 and 3. We shall give applications in the second part of the paper.
I am very grateful to all my teachers and my friends for their constant
encouragement during this work. Among them I should mention E. Calabi, S.
S. Chern, S. Y. Cheng, S. Kobayashi, J. J. Kohn, B. Lawson, L. Nirenberg
and Y. T. Siu. Professor Chern kindly lectured on this paper and made
significant improvements on its presentation. His help and his encouragement
can hardly be exaggerated. My joint papers with Cheng have definite influence
on the present paper. I am also very grateful to L. Nirenberg for many
helpful conversations on the complex Monge-Amp6re equation, especially on
the complex analogue of Calabi's computation.
I dedicate this paper to my mother Leung Yeuk Lam and to the memory
of my father Chiou Chen Ying. My father worked in both philosophy,
economics and Chinese literature. The depth of his insight had a tremendous
influence on me. H e died before he was well-known. My mother took over
the most difficult job of teaching and feeding seven children in a poor family.
Without my parents' persistent teaching, I would never have been able to
become a mathematician.
Most of the results of this paper and their applications were announced in
1151.
1. Notations and Local Formulae
Let M be a rn-dimensional Kahler manifold with Kahler metric ds2=
xi,,g,r dz'C3ddz'. The Hermitian condition of the Kahler metric shows that, for
all i, j ,

be the Kahler form, then the Kahler condition is that

(1.3) dH=O,
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ON THE RICCI CURVATURE OF A COMPACT KAHLER MANIFOLD 343

or, for all i, j , k ,

(1.4)

The connection form w ! of the KBhler metric is characterized by the fact

that it is of type ( 1 , O ) and it satisfies the following equation (of preserving
scalar product):

where

w : = w -:
(1.7) 11 11 .
The matrix ( g i r ) satisfies the equation

Hence

The curvature form is given by

Differentiating (1.8), we have

(1.11)

Substituting (1.11) into (l.lO), we obtain

Therefore the tensor R i j k T is symmetric in the first and third indices, as well
SHING-TUNG YAU

is of type (1.1) and we can set

as in the second and fourth indices.

The Ricci tensor is
By Bianchi's identity

On the other hand,

we have

so that
Hence

(1.17)

(1.20)
(1.13)

(1.19)
(1.15)

(1.16)

(1.18)
(1.14)
where

and
344

or
ON THE RICCI CURVATURE OF A COMPACT KAHLER MANIFOLD 345

By direct computation, we get

-
(1.22) a log det (gij) = giT agil
i.i
Hence,

Therefore, the Ricci tensor is given by

(1.23) R.7= - a* log det (gij)

IJ a 2 azj
The curvature tensor will be an obstruction for the commuting of
covariant differentiations. Thus let us set up the terminology for covariant
differentiation.
Let f be any real-valued function defined on M. Then we define

(1.24) df=Cfioi+Cf+ii,
I 1

(1.26)

By exterior differentiating (1.24), one verifies that

(1.27)

For convenience, we shall write Af as c,fi;.

Then, as is well-known, 2 6 is
the Laplace-Beltrami operator of the Kahler manifold M. In local coordi-
nates, it is given by C,,, gLTazf/az1 d?.
Exterior differentiating ( 1 . 2 9 , we have

c( - c
dfij
k
fiko: - & fkjw:) A 0;

(1.28)
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
346 SHING-TUNG YAU

Hence we define

(1.29)

(1.30)

By (1.28), we have

(1.31) fijk =fikj 1

(1.32) fiyE =fin; 9

(1.33) f i k-l - f q k- = - .
- z f , R ' qk
I

Continuing in this way, we can find the commutation formulae of the higher
covariant derivatives of f. For example,

(1.21)
n n

The main fact that we shall use is that, when we apply covariant
differentiation successively with respect to two indices without bar, we can
interchange these two indices. The same principle applies when the two
indices have a bar at the same time. If one of the indices has a bar and the
other does not, then after commuting them, we still have terms which are
contraction of a curvature tensor with the covariant derivatives of f whose
order is the order of the original one minus two.

2. Estimates up to Second Order

From now on, we assume that M is a compact Kahler manifold. We shall

study the equation

(2.1) (
det giT+-
azi azT
) (det (gi;))F1 = exp { F } ,

where F is assumed to be in C 3 ( M ) .
We are going to look for solutions (p of (2.1) such that ( g i r + a 2 ( p / d z iaz;) is
a positive definite Hermitian matrix. The tensor Ci,j( g i r +a2(p/azia Z i ) dz'@dZ'
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
O N THE RICCI CURVATURE OF A COMPACT KAHLER MANIFOLD 347

will define another Kahler metric on M. We shall denote it by xi,,g:rdzl@&'

so that

g!:=g:+-. a2 cp
'I 82, azj

Before proving the existence of cp, we need a priori estimates of cp.

Therefore we assume that c p C~ 5 ( M ) .Our first aim will be to obtain the
estimates of cp up to second derivatives. This will be achieved under the
normalization

(2.3)
I,
q=O.

Differentiating (2.1), we have

(2.4)

where (gfii) is the inverse matrix of (gii). Now we differentiate (2.4) again and
obtain

Let A' be the normalized Laplacian associated with the metric CijgT; dz' 8
dz'. Then
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
348 SHING-TUNG YAU

Since the left-hand side of (2.6) is independent of the choice of coordinate

system, we can compute the right-hand side by picking a coordinate system at
one point so that gir = Sii, agi,-lazk= 6’gir/6’2’= 0. Then inserting (2.5) and (1.9)
into (2.6), we have

Note that each term on the right-hand side is a scalar function and is
independent of the choice of the coordinate system if we contract the terms
suitably with the tensor {gti}.
If we choose another coordinate system so that g,, = S,, and qI;= Sl,q1;,
then we have

(2.8) g’l’ = Sll(l + qI;)-*

and

Combining (2.7) and (2.9), we see that

On the other hand,

(2.1 1)
ON THE RICCI CURVATURE OF A COMPACT KAHLER MANIFOLD 349

Let C be a positive constant. Then we have

-C exp {-Cq} c
i,i
gfiicpi(Aq)7
(2.12)
- C exp {-Cq} ci.i
gfii(Aq)icpy

-Cexp{-Cq}A'q(m + A q ) + exp{-Cq}A'(Aq) .
Applying Schwarz' inequality to the first three terms of (2.13), we have

A'(exp{-Cq}(m+Aq))z -(m+Aq)-l exp{-Cq} gfir(Aq)i(Aq)T

k,i
(2.13)
- c exP - Cp>(A'co>(m+ Aq) + expi - Cq}A'(Aq).
In computing the right-hand side of (2.13), we assume that gii=Sii and
'pi;= qir6,, at a point. Then, using (2.10), we have

-(m + A q ) - ' C g'ii(Aq)i(Atp)T+A'(Aq)

(2.14)
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
350 SHING-TUNG YAU

(Here we use the fact that q k r i= (P&i = (PEtk = q , and

~ ~qrkk;= q k K ; = q k ; ~see
,
Section 1). Hence (2.14) and (2.15) imply

(2.16)

Inserting (2.16) into (2.13), we obtain

A’(exp{-Cq}(m +Aq))Zexp{-Cq}
(2.17)

’ Inserting (2.11) into (2.17), we get

+ exp {-cq}inf ( R ~ F ~ ~ )
i# I

(2.18) 1
-C exp {- Cq}m(m + Aq) + C exp { - C q } ( m + A q ) c -
i l+qi:

-Cexp{-Cq}m(m+Aq)

exp{-Cq)(m+~q)

Now let us notice the following inequality:

(2.19)

which can be verified by taking the (rn - 1)-th power of both sides. Therefore,
by (2.1),

(2.20)
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ON THE RICCI CURVATURE OF A COMPACT KAHLER MANIFOLD 35 1

Choose C so that

(2.21) +
C inf
iff
>1 .

Then it follows from (2.18) and (2.20) that

A'(exp {-Ccp}(m +Acp))

(2.22)
- C exp {-Ccp}m(m +Acp)

Using (2.22), we can now estimate exp{-Ccp}(m+Acp). In fact, it must achieve

its maximum at some point p so that the right-hand side of (2.22) is non-
positive. At this point,

0 2 A F - m 2 inf (Rinr)- Cm(m + Acp)

i+f
(2.23)
(Rini))exp [x}
+
m-1
( m Acp)l+l/(m-l)

Hence ( m + Acp)(p) has an upper bound C1 depending only on sup, (-AF),

-
sup, linfizl(Ri:fi)I,C m and sup, F. (It is clear that the inequality Y'"'(~-')<
-
ay + b implies either y1+'/(m-1)S2ay or y1+"("-')52b. In either case, we have
an upper estimate of y in terms of a and b.)
Since exp {-Ccp}(m + Acp) achieves its maximum at p, we have the follow-
ing inequality

(2.24)

We are going to make use of (2.24) to estimate sup, IcpI. First of all, since
xi
Acp + m = (1 + pi;) = gi7g;7 > 0, we can estimate sup, cp as follows.
Let G(p,q ) be the Green's function (cf. [lo]) of the operator A on M. Let
K be a constant (depending only on M) such that

(2.25) G(p,q)+KZO.
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
352 SHING-TUNG YAU

Then (2.3) shows that

(2.26)

Therefore,

(2.27) supcpSmsup
M pEM I
M
(G(p,q)+K)dq.

The inequalities (2.3) and (2.27) also imply

(2.28) s
(
sup $0 Vol ( M ) -
M I cp + ( sup
M
cp Vol (M)

= 2m Vol (MI sup

paM M I (G(p,4 ) + K ) dq

To estimate infMcp, we offer two different proofs. One works for rn = 2 and
the other works in general.
For m = 2 , we renormalize cp so that ~ ~ p ~ q ( P -(This
1 . is possible
because we have estimated SUP,^ already so that we still have an estimate

Let p be any positive number greater than or equal to 1. Then

When m = 2. we find

(2.30)

Multiplying (2.30) by exp {F)and integrating with respect to the volume form
of the original metric Ci,igi;dzi@ddZj, we have

(2.31)

Since cp is negative, (2.31) implies that, for p % 2 ,

where C, depends only on sup F.

Now applying the Sobolev inequality (this is true on a compact manifold
because we can cut the manifold into a finite number of pieces which are
diffeomorphic to a domain in Euclidean space and we can apply the standard
Sobolev inequality in Euclidean space), we can find a constant C, depending
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
354 SHING-TUNG YAU

only on M such that

sc31,IcpIP+PClC,IM IcpIp-l.

Since cpS-1, we derive from (2.33) another constant C4 depending only on

M such that

(2.34) I, lcpI’PSC4P2(I, lcpl’)1

In order to use (2.34), we need to find an estimate of

follows from the estimate of
I IqI2. But this

IcpI (see (2.28)), the Poincarh inequality and

(2.32). l M

We now claim that we can find a constant C5 depending only on M such

that

(2.35)

for all integers p 2 1. In fact, let po be the first integer such that, for p 2 po,

(2.36)

Then we choose C5 so that (2.34) is valid for 1 S p 5 2p,. (This is possible by

(2.34), the Lz-estimate of cp and the Holder inequality.) We are going to
prove (2.34) for all p 2 1 by induction.
There are two cases. If p + 1 is divisible by two, then from (2.34) we have

(2.37)

where the last inequality follows by the induction hypothesis. If we apply

(2.36) to (2.37), then (2.35) is valid with p replaced by p + 1.
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ON THE RICCI CURVATURE OF A COMPACT KAHLER MANIFOLD 355

If p + 1 is not divisible by two, then, by (2.34), the Schwarz inequality and

the induction hypothesis,

(2.38)

Applying (2.36) to (2.38), we verify (2.35) with p replaced by p + 1.

In conclusion, (2.35) is valid for all p. Therefore,

(2.39)

Since the Stirling’s formula tells us that

(2.40)

for all positive integers p, we find

(2.41) exp { k q 2 ) 5 (2a)-1’2(kCge)Pp-1’2.

p=o

When

(2.42) k < Ci2e-’ ,

the right-hand side of (2.41) is finite and we have obtained an estimate of

I, exp {kq’} with any k satisfying (2.42).

We are now in a position to use (2.24) to estimate SUP, Iql. Rewrite
(2.24) as
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
356 SHING-TUNG YAU

where

(2.44) - m S f S C , exp{Csup cp}exp{-inf cp}.

Then by the Schauder estimate (see [13] p. 156 inequality (5.5.23)), there is a
6 depending only on M and CI exp { C sup cp} such that
constant c

(2.45)

Since (2.27) and (2.28) already provide estimates for sup cp and 1, (cpl, we
have a constant C, depending only on M and C such that

(2.46) sup [ V q l C,(exp

~ {-C inf q}+ 1 ) .
M

Let q be a point in M where cp(q)=inf cp. Then in the geodesic ball, with
center q and radius -4(inf cp) C;’(exp {-C inf cp}+ l)-*,cp is not greater than
iinfcp. Since we may assume -inf cp to be large, we may assume that the
radius of this geodesic ball is smaller than the injectivity radius of M.
Therefore, the integral of exp{kq2} in this ball is not less than
C8exp {$k(inf cp)’}(-$ inf(p)2mC;2m(exp{ - C inf cp}+ l)-2m,
where C, is a posi-
tive constant depending only on M. Since (2.40) gives an estimate of the
integral of exp{kcp2} over M, the last quantity is estimated. This of course
implies an estimate of linf,cpl.
Let us now give an estimate of linfMcpl without assuming rn = 2. Let N be
any positive number. Then (2.18) shows that

(2.47) - N exp {-Ncp}rn(m + Acp) +

Choose N so that

(2.48)
ON THE RICCI CURVATURE OF A COMPACT K.&HLER MANIFOLD
357
Then, by (2.20),

There is a constant C, depending only on s u p F and m such that

-F
(2.50) $N exp ( , 1 } ( m +Aq)m/(”’-l)Z2Nm(m + A @ ) - NC, .

Inserting (2.48), (2.49) and (2.50) into (2.47), we find

A’(exp{-Nq}(m+Aq))Zexp{-Nq}
(2.5 1)
+Nexp{-Nq}rn(m+Aq).
Therefore.

2 exp {-Nq}exp {F}

+ N exp { - N q } exp {inf F}m(m+ A q )

(2.52)
+ m 2 N exp {inf F)
11
- mN exp {inf F)exp { - N q } A q

+ m Z Nexp {inf F )
I1
+ m exp {inf F)(-A exp { - N q } + N 2 exp {-&} IVq12)
2-Cloexp{-Nq}+m exp{inf F)(-Aexp{-Nq}+N2exp{-Nq}IVq12),

where Clo depends only on N, F and M. Integrating (2.52), we obtain

(2.53)
JV exp {-+Nq}12 = +NZ L exp {-Nql I V d 2

d $Clom-’ exp {-inf F }

We claim that for each N satisfying (2.48), the inequalities (2.53) and
(2.28) furnish an estimate of
i, exp{-Nq} (depending on N, F and M ) . We
are going to prove this statement by contradiction.
Suppose there exists a sequence { q i }satisfying (2.28) and (2.53) such that
1imi+,- J
M exp {-Nq,} = CQ. Then we define

It follows from (2.53) that the sequence

I, IV exp {-iN+}lz is uniformly
bounded from above by a constant depending only on N, F and M . Since

I exp {-N+i}= 1 for all i, this last fact implies (see [13]) that a subsequence
of {exp(-$NGi}} converges in L'(A.4) to some function f in L 2 ( M ) . We
assume this subsequence is {exp {-$N&}} itself.
On the other hand, we know that, for any A > 0,

(2.55) Vol {X 1 A S Iq 1 (x)} 5 -

Furthermore,

(2.56) Vol{x 1 A S e x p { - ~ N ~ i } } = V o l N exp{-Nqi}.-pi].

Since 1imi-= exp {-fNqi}= m, we conclude that, for i large enough,

By (2.28) [I, lqi1} is uniformly bounded and the inequality (2.57) then
implies that

(2.58) limVol{x
t-m 1~ ~ e x p { - $ ~ ~ ~ ) = o
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ON THE RICCI CURVATURE OF A COMPACT KAHLER MANIFOLD 359

for all A > 0. Clearly,

Vol{x I A Sf}SVol{x I $A SIf-exp{-$N@i}l}+Vol{x [ $ A Sexp{-$N@i}}

(2.59)

Since exp{-$N@i} converges to f in L'(M), (2.58) and (2.59) shows that, for
all A > 0,

(2.60) Vol {x I A Sf)=0 I

Since f is the L2-limit of exp{-$N@i}, (2.59) implies that f is zero almost

everywhere. This is a contradiction because
IM f'= 1 which is a consequence
of (2.54) and the L2 convergence of exp(-$N&} to f. Therefore we have
arrived at the conclusion that whenever N verifies (2.53), (2.28) and (2.48)

I, exp{-Ncp} has an estimate from above depending only on N, F and M.

We can now repeat the previous argument to find an estimate of linf,cpJ.
Inequality (2.46) is valid for arbitrary dimension. As before, we find a
geodesic ball with radius -; inf cp C;'(exp {-C inf cp}+ l)-' (which is not
greater than the injectivity radius) such that cp is not greater than $inf cp in
this ball. Then we choose N so large that (2.48) is satisfied and that N is
greater than 4mC. Since the integral of exp {-lVcp} in the above geodesic ball
is not less than

C12 exp {-$N inf cp}(-$ inf qIZmC;'(exp {-cinf cp)+ ~ ) - ~ m ,

where Clz is a positive constant depending only on M, we have an estimate of

-inf, cp.
Since we have already found an estimate of sup cp, this gives an estimate of
sup, Iq(. The inequalities (2.46) and (2.24) then give estimates of sup, IVcpl
and sup, (fn+Acp). On the other hand, since (S,,+cp,,) is a positive definite
Hermitian matrix, we can find upper estimates of 1t cpir for each i. The
equation n L l (1+ cp,;) = exp {F} then gives a positive lower estimate of
(1+ cpi;) for each i. In particular, the metric tensor z,,,
(g,i + a2cp/az' a5') dz' Q? df'
is uniformly equivalent to zI,l g,T dz'8ddz".
In conclusion, we have proved the following
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
360 SHING-TUNG YAU

PROPOSITION
2.1. Let M be a compact Kahler manifold with metric tensor
2 Ci,j gi:dzi@ddiJ.Let cp be a real-valued function in c"(M) such that
and xi,,
I,
(gi;+a2rp/azia?') dzi@ddZi defines another metric tensor on M. Suppose
cp = 0

det (gi;+a2cp/azi85') det (gi;)-l = exp {F). Then there are positive constants
C , , C,, C, and C,, depending on inf, F, sup,F, inf AF and M such
that sup, JrpI 5 C1, sup, IVql S C,, 0 < C, 5 1+ c p i 5
~ C, for all i.

3. Third-Order Estimates
In this section, we estimate the third derivatives of ViTk assuming cp solves
the equation (2.1) and F is C 3 ( M ) .
Following E. Calabi [ 5 ] , we consider the function

(3.1)

We are going to compute the Laplacian of S. For convenience, we shall

introduce the following convention. We say that A - B if ( A - B ( d
C,&+ C,,where C, and C2 are constants that can be estimated. We also say
that A r B if \ A- B l S C,S + C,&+ C , , where C , , C, and C, are constants
that can be estimated.
As in the last section, we shall diagonalize our metric tensor and the
Hessian (cp,~) at a point that is under consideration. Then by the computation
which will be carried out in the Appendix A, we have

(3.2)

On the other hand, by (2.7)

where c6 is a constant that can be estimated.

Therefore by letting C, be a large positive constant, we find

where C,, C,, C, are positive constants that can be estimated.

At the point where S+C,Acp achieves its maximum, (3.4) shows that
(3.5) C8(S + C7A cp) S C, + C, C7Acp .
Since we have estimated Acp already, (3.5) gives an estimate of the quantity
sup, (S + C,Ap) and hence of sup, S. This in turn gives estimates of ( P i F for
all i, j , k.
3.1. Let M be a compact Kuhler manifold with metric tensor
PROPOSITION
gi;dzi@ddZi. Let cp be a real-valued function in C 5 ( M )such that
and

defines another metric tensor on M . Suppose

det (g1;)-' = exp {F}.

4. Solution of the Equation

With the estimates of Section 1 and Section 3, we can now solve the
equation det [gii+(d2(p/aziaij)] det (gij)-' = exp {F), where F E C 3 ( M ) .
If R is the Kahler form of M, then the above equation is equivalent to

(4.1) (R + = (exp {F))Rm .

Hence integrating (4.1), we see immediately that

Conversely, we shall now prove that if F E C k ( M ) with k Z 3 and F

satisfies (4.2), then we can find a solution cp of (4.1) where p E Ck+'+(M)for
any 0 5 a < 1. (Ck-',a(M) are the functions whose ( k - 1)-derivatives are Holder
continuous with exponent a.)
We are going to use the continuity method. We consider the set

the equation det

(4.3)
= Vol (M) [ exp { t f l ]
-1
exp { t f l has a solution in Ckt'."(M)].
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
362 SHING-TUNG YAU

Obviously O E S. Hence we need only to show that S is both closed and open
in [0, 13. This will imply that 1 E S and that our original equation has a
solution in Ck+lzu(M).
To see that the set S is open, we use the standard inverse function
theorem (cf. [14]). Let 8 = {QE Ck+l*"(A4)1 1 + qir > 0 for all i and IM Q = 0)
{
and B = f~ Ck-',"(M)
space Ck+'@(A4)
IL I
f =Vol A4 . Then 8 is an open set in the Banach
and B is a hyperplane in the Banach space Ck-',"(A4).
We have a map G mapping 8 into B:

(4.4) G(rp)=det

The differential of G at a point Qo is given by det ( g i r +

aZQo/aziazj) det(gir)-'Aao, where A*,, is the (normalized) Laplacian of the
metric 1 (g,r+d2Lpn/8ziazi) d z ' @ d f ' . The tangent space of B is the space of
Ck-1,"
-functions such that
I,
It is well-kncwn that the condition for
f = 0.
Aqp0(p= g to have a weak solution
g dV, = 0. Hence the condition for

det (g,;)-lA,,cp =f

to have a weak solution is that

sure that Q E C k + l * u ( Mwhen

)
I, f = 0. The Schauder theory (cf. [13]) makes
f~ Ck-l,s(M).The solution is clearly unique if
we require that
L Q = 0. Hence the differential d G of G at cp0 is invertible
and G maps an open neighborhood of 'Po to an open neighborhood of G(Q,,)
in B. this proves that the set S is open.
It remains to prove that the set S is closed, If {t,} is a sequence in S, then
we have a sequence Q ~ E C ~ + ' * ~such
( M )that

) det (gi7)-' = Vol ( M ) (I, exp {t,F})-' exp {t,F}.

Normalizing, we can assume qq = 0.

Differentiating the above equation, we have

(4.5)
[exp {tqF}det kf)l ,

where (g6') is the inverse matrix of (gl;+d2cp,/az' d F J ) for each q.

Proposition 2.1 shows that the operator on the left-hand side of (4.5) is
uniformly elliptic. Proposition 3.1 shows that the coefficients are Holder
continuous with exponent a for any 0 5 a 5 1. The Schauder estimate then
gives an estimate on the C2,"-norm of acp,/azp. Similarly we can find the
CZ70r-norm of dcpq,'aiP. From this information we kqow that the coefficients of
the elliptic operator on the left-hand side of (4.5) have better differentiability.
The Schauder estimate again provides better differentiability of dcp,/dzp and
acp,/d?'. Iterating, one finds Ck++ ' ' -estimates of cp, . Therefore the sequence
{cp,} converges in the Ck+13a-normto a solution of the equation

det (gi;)-' = Vol ( M ) (I exp { t0F})-' exp { t,F) ,

where to = 1imq+- t,. Hence S is closed.

THEOREM
1. Assume that M is a compact Kahler manifold with metric
r

C g i ; d z i @ d i ' . Ler F be C k ( M ) with k 2 3 and exp{F)=Vol(M). Then

Jh4
there is a function cp in Ck+'."(M) for any OS'aCl such that
2 C (giy+d2cp/dzi 132') d z i @ d f i defines a Kahler metric and

As a consequence of Theorem 1, we can prove the conjecture of Calabi

[3]. Let 1 gg dx" €3d f a be a Kahler metric on a complex manifold M . Let
1 R a d~r o r @ d f Pbe the corresponding Ricci tensor. Then Chern [6] proved
that the ( 1 , l ) form (i/27r)1 Rap dz" A d I P represents the first Chern class
of M. Calabi asked the following question: Given a (1,l) form
(i/27r)CRapdzar\dZP which represents the first Chern class of M, can one
find a Kahler metric on M so that 1 kpdz" a d z P is the Ricci tensor of this
metric. Calabi himself observed the uniqueness of this problem. Namely,
there is at most one Kahler metric in each cohomology class which does this
job. He also observed the local version of this problem. He proved that, if the
tensor C Rap dz"@dZP can be realized, then all tensors which are close to
this one and which represent the first Chern class can also be realized.
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
364 SHING-TUNG YAU

To see how Calabi's conjecture follows from Theorem 1, we notice the

following well-known fact (see (1.10)):

(4.6)

Since we assume that ( i / 2 7 ~1) R,B dz, A dZp represents the first Chern class,
we conclude immediately from Chern's theorem that

(4.7)

for some smooth real-valued function f .

According to Theorem 1, we can find a smooth function p so that
1(gap +a2p/az" aZB) d z " @ d P defines a Kahler metric and that

(4.8) a2p ) det (gap)-' = C exp { f } ,

where C is a constant chosen to satisfy the equation

(4.9)

From (4.6), (4.7) and (4.8), it is easy to see that C R , p dz" @ d Z P is the Ricci
tensor of C (gap +a2cplaza a i p ) dz"@dZP.

THEOREM 2. Let M be a compact Kahler manifold with Kahler metric

C gap dz" @ dZP. Let C R a dz"@
~ d Z P be a tensor whose associated ( 1 , 1 ) form
(i/27r) R e p dz" A d i p represents the first Chern class of M. Then we can find
a Kahter metric gapdzU@dZ@whose Ricci tensor is given by C R a p dz" @ d P .
Furthermore, we can require that this Kahler metric has the same Kahler class
as the original one. In this case, the required Kahler metric is unique.
Remark. The uniqueness was proved by Calabi [4] and will also be
indicated and proved in Theorem 3.

5. Complex Monge-Ampere Equation with Degenerate Right-Hand Side

Let L be a line bundle over a compact Kahler manifold M. Let s be a
(non-trivial) holomorphic section of L. Suppose L is equipped with a
Hermitian metric so that we have a globally defined function (sI2 on M. Then,
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ON THE RICCI CURVATURE OF A COMPACT KAHLER MANIFOLD 365

for k 2 0 , we shall study an equation of the form

(5.1)

where giT d z i @ dZi is a Kahler metric and F is a smooth function such that
,.
(5.2) exp {F)= V O (~M ) .
JM

In order to solve (5.1), we approximate the equation by the following

non-degenerate elliptic equation:

(5.3) ( -
det gi7+ a:rrzj)det (gi7)-l = C,(lslz+E ) exp
~ {F},

where E > 0 is a small constant and

(5.4)

It was proved in Section 4 that (5.3) has a smooth solution (P, such that
(gii+a2cp,/dzi 82') is positive definite and

(5.5)
I,cp,=o.
We are going to prove that when E approaches zero, cp, tends to a solution of
(5.1). This depends on estimates of cps which are independent of E .
Since the right-hand side of (2.28) depends only on the Green's function
of the fixed Kahler metric, we can find a constant C, independent of E such
that

In order to estimate inf, cpE and AcpE we notice that, when s # 0,

(5.7) A I O ~ ( J S ~ ~ + E"I 2 ) Z -(A log 1~1') .

(s(Z+ E
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
366 SHING-TUNG YAW

By direct computation, one knows that, when sZ 0, A log Is(' depends only on
the fixed Hermitian metric of L. In fact, it is the trace of the first Chern form
of L with respect to the tensor XI, gl;dzi@ddzJ. Hence Alog ( I s ~ * + E ) is
uniformly bounded from below by a constant independent of E .
Let A be the normalized Laplacian of the metric I,,,
(gi;+
(82cp,/8z1 d 2 ' ) d z ' @ d Z J .Then according to (2.22),we have

By (5.7), A log E ) is bounded from below. Hence one can conclude as in

Section 2 that the following inequality holds:

where C, is a constant independent of E.

If m = 2, we multiply (5.8) by (lsI2+ E ) exp
~ {F)and integrate the resulting
inequality. We find

Here C, is the trace of the first Chern form of L with respect to the tensor
Ci,i gir dz'@ dz'.
For any 6 > 0 , we can find a constant C, such that

(5.11) m+Aq,Si(rn+Acp,)2+Cg.
ON THE RICCI CURVATURE OF A COMPACT KAHLER MANIFOLD 367

Hence by choosing C +infi+[(Rirlr)L 1, we can obtain an estimate of the form

where C3 is a positive constant independent of E.

One can now proceed as in Section 2 to find an estimate of infMcp,
independent of E and hence also an estimate of supMAcp, independent of E.
For m > 2 , we proceed as follows. Let p be any non-negative number.
Then by Schwarz’ inequality, we have

-C(lsl*+ E ) exp
~ {-Ccp,}A’,cp,.

For s f 0, A: log IsI2 is dominated from below by the trace of the first Chern
form of L with respect to (gi~+82cpe/8zi
82’) dzi@d2j. Hence there is a
positive constant C, independent of E such that

(5.14)

As in (5.7), we have

(5.15)

Hence
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
368 SHING-TUNG YAU

Inserting (5.16) into (5.13), we obtain

-
(5.17)
= -rnC(ls12+ E)P exp {-Cq,}

Using the arithmetic-geometric mean inequality and (5.3), we deduce that

(5.18)

x exp }(:- exp 1- ~ c p , ) .

Multiplying (5.18) by (lsI2+ E ) exp

~ {F) and integrating, we find a positive
constant C, independent of E such that

In (5.19), C and p are arbitrary non-negative constants. By iterating (5.19),

we see that, for all fixed p 2 0 , we can find (large) constants C and C, such
that

On the other hand, by (5.8), we can find positive constants C, and C,

independent of E such that

Multiplying (5.21) by (Is/'+ E ) exp

~ {F)and integrating, we obtain

where C9 is independent of E . Since rn + A q , > O , it follows from (5.22) that

we can find a positive constant Clo independent of E such that

(5.23) JM (Is/'+ E)k+l(m+Acp) exp{-Ccp,}S CloJM E ) exp

~ {-Ccp,}.

Integrating by part in (5.23), we derive

(5.24)

(5.26)

where Cll is independent of E. (We use the fact that IV lsI2l2 is dominated by
Isl")
By making use of (5.20),
I, ( l s 1 2 + ~ ) k exp{-Cq,} is dominated by

I, (lsI2+ E)k+l exp {-CCp,).

Therefore we can find a constant C,,independent of E such that

C,, independent of E such that

(5.27) JM l ~ [ ( l s 1 2 + &)(k+1)/2 exp { - ~ C ~ & } c'./

] I Z ~ (\s12+ &)'+I exp {-cp,).
M

Since (5.5) and (5.6) provide an estimate of

see applying the arguments of Section 2 that (5.27) gives an estimate of
I, lqEI independent of E, we

I, (Is['+ E ) ~ + 'exp {-Ccp,} independent of E. (Suppose there is no estimate.

Then one can find a sequence si +0 such that

tends to infinity. Then one defines
JM (Is\'+ E~)~+'exp {-Ccp,}

exp {-c+,}= exp{-cq,,~( I, (1sl2+ Ej)k+1 exp {-eve,})-' .

By (5.27), (Is\'+ E ~ ) ( ~ + exp
~ ) ' {-&jjj}
~ converges in L 2 ( M ) to a non-negative
function f. Using the estimate of 1p.l and restricting to sets of the form
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ON THE RICCI CURVATURE OF A COMPACT KAHLER MANIFOLD 37 1

Is/ 2 l/n, one shows that f is zero almost everywhere and hence obtains a
contradiction.)
As in Section 2, inequality (5.9) and an upper estimate of cp, give an
estimate of (Vp,( (exp {-C inf q,}+ 1)-' independent of E. Hence, for some
geodesic ball B(x, R ) of radius R = C1,(-inf cp,)(exp {-C inf q,}+ l)-', (P, is
4
not greater then inf cpE. (Here C13 is a positive constant independent of E ,
and R is less than the injectivity radius of M.) Over this geodesic ball, the
integral of ((s('+ E)'+' exp {-Ncp,} is not less than

(5.28)
C,, exp {-4N inf cp,}
I,"
ra(k+l)dr

>- c
14
- ,
exp {-4N inf cp,}[CI3(-inf cp,)(exp {-C inf q,}+ l)-l]ak+n+l
ak+a
where C,4 and a are positive constants independent of E .
By choosing N large enough, inequality (5.28) and the estimate of

I, ((sI2+ E)~+' exp { - N q , } provide an estimate of -inf cpE independent of

and inequality (5.9) gives an upper estimate of rn + Acp, independent of E.
E

Let us now show that I(cpE)i;kk(2 has an upper bound independent of E on

compact subsets of the complement of the divisor of s. In fact, we can follow
c
the arguments of Section 2 to define S, to be g',"g',"g'k'(cpE)ii((P~)~~~. Let p
be a non-negative smooth function in M whose support is disjoint from the
divisor of s. Then over the support of p, the metric tensor cL,,(gij+
a2cp,/dzi dZJ) d z i @ d Z j is uniformly bounded from above and below by some
positive multiple of gii d z ' @ dZj, the constants being independent of E .
(This is because we have an upper estimate of rn + Acp, and, over the compact
set under consideration, the volume form has a fixed positive volume form as
a lower bound.)
From inequality (2.10), we can find positive constants C15 and C16
independent of E such that
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
372 SHING-TUNG YAU

Because the new metric is uniformly equivalent to the old one on the support
of p, the right-hand side of (5.30) can be estimated by a constant independent
of E . Since lsI>O on the support of p, we can therefore find an estimate of

I, pS, independent of E. Since the function p is chosen quite arbitrary, we

see that we have found an L,-estimate of S, over any compact subset of M
which is disjoint from the divisor of s.
Let B ( R )= {(z, , - xi
, 2,) 1 ( z i / ' 5R } be a coordinate chart in the com-
plement of the divisor of s. Then we claim that we can estimate S,(O) in terms
of the L,-norm of S, over the coordinate ball. First of all, we note that, using
the computations of Section 3, we know that there are positive constants C1,
and C , , independent of E and cp,, such that the inequality

(5.31)

holds on B ( R ) .
By choosing C , , larger if necessary, we may assume that the function
5, = S, + C,, AcpE+ C18(CiIzi ('+ 1) is ppitive. We can now solve the Dirichlet
pro)lem to find a smooth function S : = s , on the boundary of B ( R ) and
sE
A:Sz = 0. By the maximum principle, S, I > 0 in B ( R ) .
By using the fact that A: is the normalized Laplacian of a Kahler metric,
we see that A: can be written as

where gk,r= g,,+d'cp,/dz' a,? and (g'Li) is the inverse matrix of (gl.,;). Since the
metric z1,'
g',,idz_'@d.T' is uniformly equivalent to I,,, 8,; d z ' @ d Z J on B ( R ) ,
we know that s,is a solution of a uniform elliptic equation of divergence
form whose elliplicity is estimated.
Using Moser's Harnack inequality [ 1, we conclude that there is a
positive constant CI9 independent of cp, and E such that

(5.31')

We are going to prove that the right-hand side of (5.31') can in turn be
estimated by I
3B(R)
8. To achieve this, let u be a non-decreasing C'-function
defined on the real line such that u ( t > = O for tSO. Then, for 7< R, we can
define JI on the real line to be +(s) = ta(7- t ) dt. It is easy to check that
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ON THE RICCI CURVATURE OF A COMPACT KAHLER MANIFOLD 373

$ ( r ) = $((Ci I Z ~ ~ ~ is
) *in’ ~C’,’(B(R))
) and that $ ( r ) vanishes outside a compact
subset of the interior of B ( R ) .
By direct computation, we have

Since $ ( r ) -has compact support in the interior of B ( R ) , we can multiply

(5.32) by gE det (g&) and integrate with respect to the Euclidean volume
form dE to obtain

Since u Z 0 , and u‘1-0, it follows immediately from (5.33) that

Therefore, by the uniform ellipticity of A’, we can find a positive constant

C,, independent of u, T, qE such that

(5.35) & u (-
~r ) dE 5 c,, I,,., &’(T - r ) dE .
Letting T tend to R in (5.351, we see that (5.35) is also valid when T is
replaced by R. Now choose u SO that u(t)= 1 when t Z E and u’(t)S1 / for
~
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
374 SHING-TUNG YAU

all t. Then (5.35) implies

Letting F +0 in (5.36), we see that d E can be estimated from above

. ,

I
by aB(R)
s,.
Since $ I dB(R)= 9, 1 aB(R) and g, > O , we conclude from (5.31’)
that there is a positive constant CZ1independent of cpe and E such that

(5.37)

Since C,, can be chosen to be independent of R when B ( R ) lies in a fixed

coordinate chart in the complement of the divisor of s, we can integrate
(5.37) to find an estimate of s E ( 0 )in terms of the L’-norm of $ over a ball
in the coordinate chart. Combining this with the previous L’-estimate of sE,
we conclude that we have an estimate of S, on every compact subset of the
complement of the divisor of s.
Therefore we have a uniform Lipschitz estimate of (~p,)~; over any
compact subset of the complement of the divisor of s. It is now easy to
estimate the higher derivatives of cpE. In fact, we can fix a local coordinate
system in M so that (5.3) holds with cp replaced by rp,. Then we differentiate
(5.3) with respect to a/azk. We see that arpE/azksatisfies an equation of the
form

(5.38)
(3”2’.)
A: -r; = 7[log (C,((sl2+ E
a:
) exp
~ { F } det (gi;))].

In our coordinate system, the operator A, is an elliptic operator whose

coefficients are the coefficients of the inverse matrix of (gir + acpF/dziafj).
Since we have Lipschitz estimates of these coefficients over compact subsets
of M which are disjoint from the divisor of s, the Schauder estimate (see [13])
shows that all higher derivatives of rp, can be estimated over these sets.
By letting E + O , we can now conclude that {cp,} has a subsequence
converging to a solution rp of (5.11) such that cp is smooth outside of the
divisor of s and { l c p i ~ l } is bounded for all i, j.

THEOREM 3. Let L be a holomorphic line bundle over a compact Kahler

manifold M. Let s be a holomorphic section of L. Let Cii gii dzi@ dZi be the
Kahler metric of M. Then, for any k 2 0 and any smooth function F with
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ON THE RICCI CURVATURE OF A COMPACT KAHLER MANIFOLD 375

I, exp { F } = Vol ( M ) , we can find a solution cp of the equation

= ( s I z k exp {F)det (gir)

with the following properties: Outside the divisor of s, cp is smooth and, ozler M ,
{Icplrl} is bounded for all i, j. Furthermore, any function Jr satisfying the above
properties must be equal to cp plus a constant.
Proof: We have only to prove the last statement. Let A' be the Laplacian
of the metric XI,, (g,;+a2cp/dz' dT')dz'@ddT'. We claim that, if f is a function
such that {Ifi;l} is bounded over M for all i, j, then

(5.39)

Indeed, using our previous notation, we have

(5.40)
I (A:f)(lslz+ E ) exp
~ { F }= 0 .

If we let c(g,),r be the (i, T)-th cofactor of the matrix (giT+a2cp,/aziazi),

then (5.40) is equivalent to

Since c(g:)ii and a2f/azi 321 are bounded independent of E, we can use the
Lebesque dominated convergence theorem to obtain (5.39) from (5.41).
Now let $ be another solution of (5.1) satisfying the properties mentioned
in the theorem. Then we have

(5.42) det (gii + (pi;) = det (giT + $ i ~ )

Rewriting (5.42), we have

(5.43) det(gir+cpiT+($-cp)iT)det(g,T+cpiT)-l=1 .

Using the arithmetic-geometric mean inequality, we have

1
(5.44) -[m+A'($-cp)]Zl.
m
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
376 SHING-TUNG YAU

Therefore, $-cp is subharmonic with respect to the operator A'. Since \$irl
and Jcpiil are both bounded, I(4-cp)gl is also bounded over M and $-cp is a
C'-function over M. By adding a constant to 9-9, we may assume that
$ - cp 2 0. Then applying (5.39) to ($- cp)', we obtain

Since ($ - c p ) 2 0 and A'($ - cp) 2 0, we conclude from (5.45) that O f ( $ - cp) =0

and $-cp is a constant.

6. Complex Monge-Amphe Equation with More General

Right-Hand Side
In this section, we study equations of the form

= exp {F(x, cp)} det (gii) ,

where gii dz'@ddzJ is a Kahler metric defined on M, and F is a smooth

function defined on M x R with aFlat 2 0.
We shall look for real-valued smooth functions cp satisfying (6.1) such that
1(g,;+d2cp/dz' a,?') d z ' @ d z J defines another Kahler metric. If such a function
cp exists, then integrating (6.1), the integral of the right-hand side is equal to
the volume of M. Therefore, a certain compatibility condition must be
satisfied for the solvability of (6.1). The condition is that there exists a smooth
function $ defined on M such that

Relying on (6.2), we are going to use an iteration method to solve (6.1).

First of all, let us prove the uniqueness of the solution of (6.1).
LEMMA.Let cp and yS be two smooth solutions of (6.1) such that both

and

define Kahler metrics on M . Then cp-4 is a constant.

Proof of Lemma 1: Using the fact that both cp and Cp satisfy (6.1), we
can immediately derive the following equation:

Let A, be the normalized Laplacian of the metric

Then it follows from the arithmetic-geometric mean inequality and (6.3) that
we have the inequality

(6.4) rn+A,(cp-Cp)Zrnexp .

By the mean valued theorem we have

where t(x) is a number between inf (cp(x), Cp(x)) and sup (cp(x), Cp(x)).
Since aF/at 2 0, we can combine (6.4) with (6.5) to conclude that whenever
cp(x)- Cp(x) is strictIy positive, A,(cp - S;)(x) is non-negative.
If supxEM(cp - Cp)(x)> 0, then, since M is compact and cp - Cp is smooth,
the standard proof of the maximal principle shows that cp-Cp is a constant
function. Therefore we conclude that either cp-Cp is a constant function or
cp(x)S +(x) for all x E M. Clearly we can interchange the roles of cp and Cp in
the above assertion. Combining these two inequalities, we can deduce
Lemma 1.
After this preparation, we proceed to solve (6.1). By Theorem 1, we can
find a smooth function cpo such that C,,i(gi~+a2cpo/az' a i l ) d z ' @ d f J defines a
Kahler metric and

If we define
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
378 SHING-TUNG YAIJ

and

then clearly both Cpo and Go satisfy equation (6.6) with cpo replaced by Cpo
and yo, respectively. Furthermore,

(6.9) G o b ) 2 $(XI 2 G o ( 4

for all x on M.
The set A = {(x, t) I x E M, Cpo(x)2 t I &(x)} is a compact subset of M x R .
Hence we can define

aF
(6.10) k= SUP -(~,t)+1>0
(x, t ) E A at

Using this definition, we are going to prescribe our iteration scheme as

follows. For each i 2 1, we define Cpi and Gi as the smooth solutions of the
following equations:

(6.11) det =exp{k@,+F(x, @,-J- k@i-l}det(&p),

(6.12) (
det ~ ~ + +az) =az
exp(ky,+F(x, $i-l)-k@i-l}det(g+),

so that both

define Kahler metrics.

To make sure that such functions (pi and Gi exist, we need the following

LEMMA2 . Let M be a compact Kahler manifold with Kahler metric

g i i dz'@ d2'. Let F ( x ) be any smooth function defined on M. Then, for any
constant k > 0, there exists a unique smooth function cp such that

(
det gg + - = exp { kcp + F } det ( g i r )
82ZJ)
and ,&j gii dzi @ dz" defines a Kahler metric.
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ON THE RICCI CURVATURE OF A COMPACT KAHLER MANIFOLD 319

Proof of Lemma 2: As in Theorem 1, we can use the continuation

method where the one parameter family of equations is

The fact that k > O makes the proof of Lemma 2 much easier. In fact, the
estimate of supM Icpl can be derived easily by the maximum principle as
follows. When cp achieves its maximum at a point xo, we must have

Hence exp {kcp(xo)+tF(x,,)}S 1. This implies immediately sup cp = cp(xo)S

-(t/k)F(xo). Similarly one can draw an estimate of inf,cp. The uniqueness
part of Lemma 2 follows easily from Lemma 1 noting that k # 0.
Lemma 2 justifies our construction of the iteration scheme. We claim that,
for all i 2 0,

For i = O , we use the fact that & satisfies equation (6.6) and inequality
(6.9) to conclude that

At the point where Cpl-Cpo achieves its maximum,

)C,det ( g @ a2
+(Po
_p).
az az
Hence it follows from (6.14) that sup, (Cpl - Cpo)5
0. Similarly one can also
prove that sup, (fo - GI>5 0.
To show that CplSCpl,we divide (6.11) by (6.12) to obtain

(6.15)
) det (gp+=
=exp{k&- k&+F(x, &,)-F(x, Go)- k(cpo-$o))
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
380 SHING-TUNG YAU

Since Cpo2=Go and k is defined by (6.10), it follows that

Therefore,

At the point where Cpl-'yl achieves its minimum, the right-hand side of
(6.17) is non-negative and hence Cpl Z G1.
Suppose now that we have proved
- -
(6.18) , - -=z cpo
c p n ~ C p j - l ~ ; j ~ @ .-" cp,-1= -

for all j < i. We claim that (6.18) remains valid when j is replaced by i. In
fact, applying (6.11) twice, we have

(6.19)

= exp { kCpi - kCpi-, + F(x, Cpi-l)- F(x, - k(Cpi-l - .

Since Cpo2(piP22
CpiPl Z Go and k is defined by (6.10), we see that

(
)det gap 2 e x p { k @ ,- kc&-,}

Hence- the -maximal principle shows that @ i - l Z @ i .Similarly one can show
that @ i - l 5 Cpi.
To prove that Cpi 2 &, one just divides (6.11) by (6.12) and obtains

= exp {k(pi- kGi + F(x, CpiPl) - F ( x , Cpi-l) - kCpi_l + .

Using CpiPIZ@i-l, one can repeat the above argument to show that 'pi2&.
Therefore the inequality (6.13) is valid in general and both Cpi and Cpi are
uniformly bounded.
Our next step is to find a uniform estimate of a2Cpi/az"82". As in Section
2, it suffices to estimate A&.
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ON THE RICCI CURVATURE OF A COMPACT KAHLER MANIFOLD 38 1

Let A, be the normalized Laplacian operator associated with the metric

+a2Cpi/dza a Z p ) dz"@ddzP. Let C be any positive constant such that
c,,p(g,p
C+infi+l(Rinr)> 1. Then by the same computation as in (2.18), we have

Since SUP, has been estimated, it follows from Schauder's estimate (cf.
[13]) that, for some constant C1 independent of i,

(6.23)
M

As in (2.20),

(6.24)
1 2 (rn +
l+(@,),~
(exp {-m-k- 1 (qi +-m -11
- qi-l) F(x, qi-l)]

where C, is a positive constant independent of i.

Noting again that SUP, lqil has been estimated, it follows from (6.22),
(6.23) and (6.24) that there are positive constants C,, C,, C, and c 6 ,
independent of i, such that
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
382 SHING-TUNG YAU

At the point where exp {-C@,}(m +A@,) achieves its maximum, the right-hand
side must be non-positive and so

C6 m +sup A@,
( M

(6.26)
iexp [-mm- 1 sup (pi1x" C,
M
m +sup
M

+ C, (
m +sup ACpipl
M )
+ C,
(
m +sup
M

Choose a constant C, such that, for all x Z 0 ,

(6.27) x l+l'(m-l) +
2 (C, 2 C, + 2 C,) C; ' exp [ - sup cpi
my1 M I x- c, .
Then (6.26) implies that we can find a positive constant C,, independent of i,
such that

(6.28) (m+supA@,
M 5 - m+supA@ipl)+C,.
M

By iteration, (6.28) gives

(6.29)
m+supA@,S(;)' m+supA@, + 2 C 8 .
M ( M I
Therefore we have found uniform estimates for a"(pi/aza aZp. To find uniform
estimate of aZ@,/azadi@ azy, let

(6.30)

where (g',"') is the inverse matrix of (gar+az@,/az=35'). By a computation

similar to that of (3.12), we have

(6.30') Ai(Si+ C9A@i)2CloSi- C 1 1 J s a - C12,

where C,, Clo, Cll and C,, are positive constants independent of i.
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ON THE RICCI CURVATURE OF A COMPACT KAHLER MANIFOLD 383

Since [A@,[has been estimated, it follows from the maximum principle that

(6.31) sup S. 5%
M '-G o
d s d F l +&+ ClO
C9SUP \ A & \ .
M

It should be noted that in (6.30), we can choose Clo to be arbitrarily large if

we are allowed to increase C, and C1,. In particular, we may assume that
C I 1 / C l o S ~Then
. (6.31) shows that

(6.32) sup S. <-sup

1 Si-l+4c"+4C9 sup IA+il.
M '-3 M Cl 0 M

By iteration, we can find a uniform estimate of Si and hence a uniform

estimate of a2Cpi/azma z p azy.
Letting i tend to infinity in (6.11), we can then obtain a solution of (6.1).
The Schauder estimate guarantees the solution to be smooth.

THEOREM 4. Let M be a compact Kahler manifold with Kahler metric

gi; d z i @ dZi. Let F be a smooth function defined on M X R with aF/at I0.
r
Suppose that, for some smooth function II, defined on M, J exp {F(x, II,(x))} =
M
Vol(M). Then there exists a smooth function cp on M such that

= exp {F(x,cp(x))}det (gii)

and (gii+ a'cplaz' a i i ) dz'@ dZJ defines a Kahler metric. Furthermore, any
other smooth function satisfying the same property differs from cp by only a
constant.

As a consequence of Theorem 4, one deduces that, on a Kahler manifold

with ample canonical line bundle, one can find a (unique) Kahler-Einstein
metric whose Ricci tensor is the negative of the metric tensor.
In fact, by hypothesis, the negative of the first Chern class of M is
represented by some positive (1, 1)-form a Ci,i gi; dz' A di'. Take this form
as our Kahler form. Then by Chern's theorem [6], the closed ( 1 , 1)-form
-a? log det ( g i r ) also represents the first Chern class of M. Hence we can find
a smooth function f such that

(6.33)
as log det <gii)= 4 21gijdzi
i,i
A dZj +a;if.
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
384 SHING-TUNG YAU

Now by Theorem 4, we can solve the equation

(6.34) =exp{cp-f}det (sir)

so that (gii+d2cp/dzi82') d z ' @ & defines a Kahler metric. By (6.33) and

(6.34) we have

(6.35)

This last equation simply says that the Ricci tensor of the metric C,,,(g,r+
a2&z' a?) dz'@,dZ' is the negative of the metric tensor itself. Hence we
have found the metric that we are looking for.
To see that the metric is unique (that is to say, that the metric depends
only on the complex structure of M ) we notice that if I,,, &-dz'@ddzl is
another such metric, then its Kahler form must represent the negative of the
first Chern class of M. Hence we can find a smooth function (Ir defined on M
such that glr = g,, +d2(Ir/az' 1 3 2 . On the other hand, the fact that the Ricci
tensor of s l y dz' @ d2' is the negative of the metric tensor can be written as

= -aalogdet (gil)+aaf --a&,

where the last equation follows from (6.33).

Equation (6.36) shows that the function

is a pluriharmonic function. Since we are on a compact manifold, this function

must reduce to a constant exp { c } . Therefore,

(6.37) = exp {$ + c - f ) det (giT) .

The function $ + c then satisfies an equation of the form of (6.34). Lemma 1

shows that c p - $ is a constant. Hence,

THEOREM 5. Let M be a Kahler manifold with ample canonical line

bundle. Then there is a Kahler-Einstein metric whose Ricci tensor is the
negative of the metric tensor. Furthermore, a metric of this form is unique and
depends onty on the complex structure of M.

7. Degenerate Complex Monge-Ampere Equation with General

Right-Hand Side

In this section, we combine the main results of the last two sections. Let L
be a line bundle over a compact Kahler manifold M. Let s be a (non-trivial)
holomorphic section of L. Suppose L is equipped with a Hermitian metric so
that the function 1sI2 is globally defined on M. Then, for k 2 0 , we study
equations of the form

(7.1) =
exp { F ( x , c p ) } det ( s i r ) ,

where F is a smooth function defined on M x R with aFlat Z 0.

As in Section 6, we look for solutions cp of (7.1) such that (cpii(is
uniformly bounded on M and cp is smooth outside the divisor of s. (In fact, cp
is going to be the uniform limit of a sequence of smooth functions
(cp,} so that det (giT+a2’pa/azi32’) det (gii)-’ converges uniformly to
exp {F(x, cp)}. Clearly a necessary condition is obtained by integrating
(7.1). The result is that there is a function t,b whose partial derivatives are
uniformly bounded on M so that

(7.2)
I, exp {F(x, 9)) = Vol (M) .

With this sole assumption on the existence of $, we proceed to solve (7.1).

We approximate (7.1) by the following non-degenerate elliptic equation:

(7.3) = C,(lsI2+E)’ exp {F(x, c p ) } det (gi;) ,

where E > O is a smooth constant and

(7.4)

Here we choose {&} to be a sequence of smooth functions such that (CI,

converges to + uniformly on M and that sup, Jd*ICl,/dz' 82'1 is uniformly
bounded on every coordinate chart.
According to Theorem 4, we can find smooth-solutions qE of (7.3) such
that 1 i . j (gii+d2q,/dz' 35') dz'@ddi' defines a metric. By examining the proof
of Theorem 4, we can also assume that we have an estimate of SUP, (cp,( in
the following way.
Let qE and & be two smooth solutions of the equation

(7.5) = C,(lsI2+E ) exp

~ {F(x, &)} det (gii)

such that

Then the arguments of Theorem 4 show that

On the other hand, for the unique solution of (7.5) with

I, cp = 0, we

can find an estimate of SUP, 191 which is independent of E. (This is seen in

the proof of Theorem 3. Note that boundedness of AJI, is needed.) In
particular, we can choose Cp, to be bounded from above by a constant
independent of E , and & to be bounded from below by a constant
independent of E .
Hence according to (7.7), we can find a smooth solution cpE of (7.3)
whose supremum norm is bounded from above by a constant independent
of E and whose matrix (giF+d2cpe/dzi 6'2') is positive definite.
Let us now proceed to estimate AqE from above. Let As be the
normalized Laplacian of the Kahler metric (gi~-d2cpEldzi6'9)dz' @ d2'.
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ON THE RICCI CURVATURE OF A COMPACT KAHLER MANIFOLD 387

Then, as in (5.8), we have

(7.8)

Choose C so that C+infi.l ( R , n i ) 2 $ C h l . Then noting (5.7) and the fact

that sup, [ c p B ( is uniformly bounded, we can find positive constants C1, C , ,
C3 and C, independent of E such that
A’(exp {- Ccp,}(m + AqP,))2 - C , - C,( rn + Acp, )
(7.9)
- C3 IVq, I + C,(m + AqE)m/(m-l) .
On the other hand, by Schauder’s estimate and the estimate of
supM ]cp,I, there are positive constants C, and C, such that

Combining (7.9) and (7.10) and the estimate of sup, lcp&l, it becomes clear
that we can apply the maximum principle to find an upper estimate of
m + Acp,. Therefore, since Ci,i (giT+ a’ cpE/aziafj) dz‘ 8d f i defines a Kahler
metric, we have uniform estimates of (a2cp,/azia f J ( on every coordinate
chart of M.
Using the uniform estimate of a2cp,/azi a,?, one can now follow the
arguments of Section 5 to provide higher derivative estimates of qE on
compact subsets of the complement of the divisor of s. Letting E -+ 0, we
have then proved the following theorem.

THEOREM 6. Let L be a holomorphic line bundle over a compact Kahler

manifold M whose Kahler metric is given by Ci,i girdzi8ddii. Let s be a
holomorphic section of L. Let F be a smooth function defined on M x R such
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
388 SHING-TUNG YAU

that aFlatZ0. Suppose, for some function I/J whose partial derivatives
la2$/azi aziJ are bounded on every coordinate chart of M, we have

I, lslZkexp { F ( x , I/J(x))=Vol(M). Then we can find a solution cp of the

equation det (gir+a2cp/azia,?) = exp { F ( x , cp(x)} with the following prop-

erties. Outside the divisor of s, cp is smooth and (gir+a2cplazi aZi) dzi@ddii
defines a Kiihler metric. Over M, {Iqiil} is bounded for all i, j. Furthermore,
any solution satisfying the above properties must be equal to cp plus a constant.
Proof: We have only to prove the last statement. Let A' be the
(gii+ 32cplaz 82') dz 63dz'. Then we
normalized Laplacian of the metric I,,,
claim that if f is a C'-function on M such that, for all i, j , la2f/dzi aFi( is
bounded on every coordinate chart of M, then

(7.11) A'(f2) lS12k exp M x , cpb)} =0 .

I
{x f(x)>o)

Indeed, let A: be the normalized Laplacian of the Kahler metric

(g,r+d2qE/dz''82') dz'Bd2'. Then, for all 6 > O such that the boundary
I
of {x f ( x ) L S} is a C'-manifold, we know that by Stoke's theorem,

&(f2)(ls12+ exp {F(x, cp,(x)}

(x lfcx,zs

can be expressed in terms of the boundary integral of 2faflan. Here alan

is the normal of the sets {x l f ( x ) = S} taken with respect to our metric
xl,f (g,+8*q,/az' 8.2') dz'@ddzf. Since { z I f ( x ) >0} is an increasing union of
the sets {x I f(x)Z6}, we know that the measure of {x 16 > f(x)>O} taken
with respect to ( I s ~ ~ + E ) ~exp{F(x, cp,(x)} d V tends to zero as S tends to
zero. Applying the co-area formula (see [ll]) to the set {x 16 > f(x)>$S},
it is easy to see that we can find a sequence (6,) such that the product of
8, with the codimension one volume of {x I f(x) = S , } tends to zero as 8,
tends to zero. Considering this together with the previous fact, we conclude
that

A:(fz)(ls12+ E)k exp ~ x cp,(x)~

,
{ z If(x)LGJ

tends to zero as Si tends to zero. Hence we have

Letting E +0 as in Theorem 3, we see that (7.11) follows from (7.12).

Suppose now that @ is another solution of (7.1) with all the properties
described in the theorem. Then

(7.13) det (gir+ cpi7+ (4 - (P)~T)

det (giT+ = exp { F ( x , @)- F(x, cp)}

Consider the set f l = { x I Cp(x)-q(x)>O); if it is nonempty, then the

arithmetic-geometric mean inequality shows that the inequality

(7.14) A’(@ - cp) 2 rn exp [F(x, @(x)) - F(x, cp(x))]

holds on R. (Note that aF/at is used here.)

Applying (7.11) to the function (4 - cp)’, we conclude immediately that

(7.15)

=b A’(@- cp)’ lslZkexp {F(x, cp(x)} = 0

Combining (7.14) and (7.15), we see that V’(@ - cp) = 0 on fl and @ - cp is a

I
constant on each component of R = { x @(x)-cp(x)>O}. Since -cp is con- +
tinuous, this is possible only if fl is empty or Q = M. In the first case,
@(x)S cp(x) for all x E M. In the second case, @ - cp is a constant. By reversing
the role of @ and cp, we conclude easily that, in any case, @ - cp is a constant.

8. Complex Monge-Ampere Equations with Meromorphic

Right-Hand Side
Let L , and L, be two holomorphic line bundles over a compact Kahler
manifold M. Let s1 and s2 be two (non-trivial) holomorphic sections of L , and
Lz that are equipped with Hermitian metrics so that we have globally defined
functions lsllz and lszlz on M. Then, for klBO and k z B O , we shall study
equations of the form

where gijd z ’ @ dT’ is a Kahler metric and F is a smooth function such that
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
390 SHING-TUNG YAU

As before we approximate (8.1) by the equation

where

In order to prove that the solutions q E of (8.3) converge on the

complement of the divisors of s1 and s Z r we consider the expression
( I S ~ J ~ + S ) exp{-Cq,}(rn+Acp,)
~ with p 2 0 .
Letting A; be the normalized Laplacian of the metric (gii+
a2cp,/azi82') dz'addzj, we compute the Laplacian of the above expression as
follows:
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ON THE RICCI CURVATURE OF A COMPACT KAHLERMANIFOLD 391

By applying the same reasoning as in (2.15), we have

As in (5.15), we have a positive constant C1 which is independent of E such

that

(8.7)

Combining (8.6) and (8.7) and computing as before, we can find positive
constants C, and C3 which are independent of E such that

Clearly, for any fixed p, we can choose C large enough so that

With this choice of C, we consider the point where (ls212+ E ) exp

~ {-Cq,}X
(rn +AqE) achieves its maximum. At such a point, the left-hand side of (8.8) is
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
392 SHING-TUNG YAU

non-positive and so is the right-hand side of (8.8). Therefore, at this point,

It follows easily from (8.9) that, for some positive constant C, independent of
E . we have

It is easy to see from (8.2) that k 2 < 1 . Hence the third term in the
right-hand side of (8.10) will be the dominating term. If we choose p to be
(rn-l)/rn+(k,/rn)+Cq with q 2 0 in (8.10), we see that, for some positive
constant C, independent of E ,

We are going to estimate the supremum norm of 1cp.I. First of all, we

recall that when we solve (8.3), we can normalize our solution cp,
.. so that
cpE = 0. Then, using the inequality Acp, + rn > 0, we have an upper estimate
of Hence it remains to find a lower estimate of cpE.
cpE.
Integrating (8.8) with respect to the volume from (ls2I2+ ~ ) - ~ z ( l s ~ 1E~ )+~ ,
x exp {F)dV, we have

- pC, + inf Ririi

,#I 1
exp {inf F }

(Is2/' + E)~-(m-2)k,/(m-l)

x (m + AL(P,)~/(~-') exp {- Ccp,1

ON THE RICCI CURVATURE OF A COMPACT K ~ ~ H L EMANIFOLD
R 393

5 C, exp {sup F}
I, (ls2I2+~ ) ~ - ~ z ( l s , l ~E +) ~ exp
, {-Ccp,}

+ mC exp {sup F} ((s2I2+ E ) ~ - ~ Z ( ( S ~ E( ~) ~+ exp

, {-Ccp,}(m + Acp,)

Using (5.7), we can find a positive constant c

6 which is independent of E
such that

Hence, for any positive constant L,

+ LC, exp (sup F) (ls2I2+E exp {-Ccp,}

) ~ - ~ ~ ( ) sE ~) ~) ,~ +
JM
394 SHING-TUNG YAU

s-" (Js2I2+ E ) P - ~ ~ Z ( ( S J ~ +~ ) exp {-Ccp,}

~ ~ 1

For any p Z 0 , we choose a constant C large enough so that

Then we choose S so that
C21, C3(C-pC1+infi.,Rinr)~~C3C.

(rn - 1)6m"m-"CC;' exp {sup F - inf F } = .

Substituting (8.15) into (8.12) and keeping (8.14) in mind, we see that we can
find positive constants C, and c g which are independent of E and C for
which

In order to make use of (8.16) to derive an integral estimate exp{-Cq,},

we shall assume that the integral
I I ~ ~ 1 is- finite.
~ ~ ~ Then

= 4, we shall prove in the Appendix B that, €or some positive

C,(p - k2 3- C7)-l
a by choosing p so that

canstant C,,independent of 8 and C,

Since we have an estimate of Icpel, we can use the method of Section 5

to find an estimate of

which is independent of E.
Note that in (8.11) and (8.17) we may choose two different values of C as
long as they are larger than some fixed constant. From (8.11) and the
supremum estimate of cp&, we derive that, for any q 2 0,

5 C5(Cm-l+ 1)exp { C sup cp,}(sup (Is2I2 + ~ exp { - ~ p , } ) ~

) q ,
where C is any positive constant so that

From (8.17), we conclude that when ( p - k , ) ( p - k 2 + C 7 ) = i and N is a

large constant, we can find a positive constant C,, independent of E such that

(ls2I2+E)P-k2(Jsl12+ exp {-Ncp,}

E ) ~ ,

(8.19)
= (l~,1~+ ~ ) ~ 7 ( l s , l ~E +) ~ exp
, {-iVq,}S C,, .
Jh4

Recall that C, is independent of N also.

Using (8.18) and (8.19) we shall show that, for any q > O ,
(lszlz+ E ) exp{-cpE}
~ has an upper bound which is independent of E when E
tends to zero.
In fact, if this assertion were not true, we could find a sequence of
numbers { E ~ } and a sequence of points {xi} in M such that E + O and
(Isz12 (xi)+ E ~ exp
) ~{-cp,,)(x,) = sup, (lszl2+ E ~ exp
) ~{-cp,,} tends to infinity.
Without loss of generality, we may assume that {xi} converges to a point xo in
M.
Suppose the sequence { E ; ~Isi\’(xi)} is bounded independent of i. Then,
clearly, ~fexp{-cp,,}(x,) tends to infinity when xi tends to xo. On the other
hand, using inequality (8.18) and the L’-estimate of cp,, we can apply the
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
396 SHING-TUNG YAU

Schauder estimate to find a constant C , , independent of i such that

It follows from (8.20) that there is a constant CI4independent of i such

that
SUP (Wag (Is2IZ+ Ei)' - vo,,II
M
(8.21)
5 c,4 - ( m -1)Im-(k,/m)-Cq sup (ls212+ ei)q exp {-cp,,}

Clearly we may assume that SUP, [ q log ()s2I2 + E ~ ) - qC,]2 0. Then proceed-
ing as in Section 2, we can now conclude that for some constant eI4,
independent of i,

(8.22)
(ls2I2+ E J ~exp {-cp,,}
I"'
In (8.19), the number N can be made as large as we want (where C12
depends on N ) . Therefore, (8.22) shows that SUP, ((s2I2+ E ~ exp ) ~{-cp,,} is
bounded from above independently of i.
The above contradiction shows that, by passing to a subsequence, we may
assume that limi.+mE ; ~lsz12 (xi)= w.
For each x i , let B ( x i , S i )be a geodesic ball around xi such that, €or each
x EB(xi, ail,
(8.23) 1 lS2l2 (Xi) 2 IS2l2 (x) 24 lS2l2 (Xi) .
Let C,*be the supremum of /V Is2I2 I on M. Then one can easily prove that
we may assume
1
(8.24) s.' 2--2c15 lS2lZ ( X i ) .
We only need inequalities (8.23) and (8.24) with suitably large CIS.Hence we
may assume that ai is smaller than the injectivity radius of M.
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ON THE RICCI CURVATURE OF A COMPACT KAHLER MANIFOLD 397

It is easy to derive from (8.18) that, over the ball B ( x i ,& ) ,

By applying the Schauder estimate on the ball B( x i ,$), we can find a

constant c 1 6 which is independent of i such that

(8.26)

Since we have an L’-estimate of

1. IcpE,), it follows from (8.26) that there

is a constant C 1 7 ,independent of i, such that

(8.27)

Since s1 is holomorphic, one can find positive constants 1 and CI8,

independent of i, such that, for any (small) r > O and X E M ,

(8.28)
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
398 SHING-TUNG YAU

As before, we may assume that sup, [q log (Is,\” + ei) - qs,J 2 0. Then proceed-
ing as in Section 2, we can now conclude from (8.23), (8.27) and (8.28) that
there is a constant CI9,independent of i, such that

(I~Zl’(xi))~7(5 Is21’(xi)+ Ei)-”” exp {-~p,,}

(ls2I2+ ~ i ) ‘
I”’

Since limi, EL’ Is21’(xi)=03 and the right-hand side of (8.29) can be
estimated even when N is large, we see that the quantity sup, (Is21z+ EJ‘
X exp {-cp,,} can be estimated by a constant independent of i.
In conclusion we have proved that, for any q > 0, -cpE + q log ((s2l2+ E ) is
bounded from above by a constant independent of E . In particular, -cp, is
uniformly bounded over compact subsets of the complement of the divisor of
sz. From (8.18) and the supremum estimate of cpE, we see that both 191 . and
IAcp,l are uniformly bounded over compact subsets of the complement of the
divisor of s,. The arguments of Section 5 now show that one can find uniform
estimates of { ( c p , ) i T k } over compact subsets of the complement of the divisors
of s1 and s2. Hence we have proved the following theorem.

THEOREM 7. Let L , and L, be two holomorphic line bundles over a

compact Kahler manifold M whose Kahler metric is given by giTdzi@dZi.
Let s1 and s2 be two holomorphic sections of L1 and L ,, respectively. Let F be a
smooth function defined on M such that
I, Is112kl Is21-2k2exp { F }=Vol ( M ) ,

where k, and k, are two non-negative integers. Suppose that, for m =dim M,
<*. Then we can solve the equation
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ON THE RICCI CURVATURE OF A COMPACT KAHLER MANIFOLD 399

so that
(i) cp is smooth outside the divisors of s1 and s2 with sup, cp <w,
(ii) (qIi) is a bounded matrix outside the divisor of s2 and, for any q>O,
2(m-l)m-'+Zk2m-'+q
IS21 Acp is bounded on M,
(iii) for any q > 0 , the function cp - q log ls2I2 is bounded from below,
(iv) the matrix (gIr+a2cp/az'a 2 ) is positive definite outside the complement
of the divisors of s1 and s2.

Furthermore, if we assume that < for some q > 0 ,

( I ~ ~ 1 ~ ) ( ~ - ~ ) ~ - ' - ~w2 ~ - ' - ~

then any two solutions of the equation which has the above properties (i), (ii)
and (iv) must differ from each other by a constant. If we also know that
(IS212)(1-m)m-k2m-'-q is integrable over every analytic disc of M , then the unique
solution cp is bounded from below on M.
Proof: We have only to prove the last sentence. Suppose t,h is another
solution of the equation which possess properties (i), (ii) and (iv) stated in the
theorem. Then clearly we have

O n the other hand, we can rewrite det (giT+d2+/'lazi 13.2~)as

Therefore the arithmetic-geometric mean inequality shows that

(8.31) m+A:(+-cpE)Z m [ C i 1Is112k1IS^(^+ E ) - ~ , (Is212+ ~ ) ~ 2 ] " ~ ,

where A: is the normalized Laplacian of the metric

Let k be any constant. We claim that, over the domain a,&

=
{x E I 4 4 x ) - PE (x) > k),
(8.32)
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
400 SHING-TUNG YAU

In fact, let S > O be a positive constant. Then the domain f l E , k , S =

1
{x J/(x)-cp,(x)> k - S log Is2]’} is disjoint from the divisor of s2 by property
T ) bounded on f l . , k , S and we can apply
(i). Hence both (Jli7) and ( ( ( P ~ ) ~are
the divergence theorem on to find

Outside the divisor of s2, A; log lsz12 is simply the trace of the first Chern
form of L, with respect to the metric

Hence for some constant C2n,independent of 6, we have

Using property (ii) and the assumption that IS^^^)(^-^)^-^-^^^-'- q < w , we

can find a constant C21,independent of S, such that

(8.35)

It follows easily from (8.33), (8.34) and (8.35) that

(8.36) lim
s-0
J (J/-q,-k)A:($-cpE)50.
ne,k.S

Using the definition of flZB,k,S, $- cp. - k is bounded

we see that, over aE,k,&,
by a constant independent of S when S is small. The function ($-cp,-k)
X &($ - cp,) is therefore uniformly integrable and we can apply Lebesque’s
dominated convergence theorem to prove (8.32).
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ON THE RICCI CURVATURE OF A COMPACT KAHLER MANIFOLD 40 1

LO.

When E tends to zero, the integral on the left-hand side tends to zero. Let
K be any compact set of the complement of the divisor of s2, then by (8.31)
and (8.37), we have

X ~ S ~ ) - ~ ~ ~ ( ~ S +
~ (b ~~+( E
$ -) ~~ ~~ )] }~ =’ o~ .
(Note that (8.31) is used strongly here.) Let flk ={x I t(r(x)-cp(x)> k } . Then
(8.38) gives

(8.39)

where A’ is the normalized Laplacian of the metric

As in (8.31), we know that A’(JI-cp)ZO and hence we conclude from (8.39)

that A‘(+ - rp) = 0 on nk. When the arithmetic-geometric mean inequality
becomes equality, the eigenvalues of the corresponding matrix are equal.
Therefore, $1~7 = cpi; everywhere on a,. Since k is arbitrary, JIii= ( p i ~
everywhere on the complement of the divisor of sz. One can now prove that
Q - JI = constant by simply letting first E tend to zero and then 6 tend to zero

in (8.33). This shows that

IVY$ - q)I2< -lime-,o

which is equal to zero by (8.37).

6.. ( 4- pE- k ) A:($- Q,)

It remains to prove that the unique solution Q is bounded from below.

From (8.11) and the estimate of (ls2I2+ E ) exp

~ {-cp,} for any q > 0, we know
that

(8.40) SUP [(m+ Aqe)((s21'+ E 1 5c z z

) ~ + ( ~ - ~ ) ~ - ' + ~ Z ~ - '
M

for any 4'0, where C,, is a constant independent of E .

Let x be any point on the divisor of s2. Let D, be an analytic disc passing
through x such that s2 is not zero on aD,. Then Icpel is uniformly bounded on
aD, when E tends to zero.
Since [gir+(cpe)ir] is positive definite, it follows from (8.40) that when we
restrict qE to D,, the absolute value of its Laplacian is estimated by
(Isz12 + E ) - q - ( m - l ) m - l - k , m - ' . On the other hand, the assumption guarantees the
integrability of (Is2I2 + E ) - ~-(m-l)m-l-kzm-l over D, for some ij > 0.
We can now apply Green's formula to 0,to prove that supDxIcpEl can be
estimated from above by a quantity involving supaDxIpEJand

Therefore, by choosing q < Lj, we obtain an estimate of IqE1 on 0,.

Since x is
arbitrary, we can conclude the boundedness of cp.

9. The General Case

In this section, we write down a general theorem.
Let M be a compact Kahler manifold with Kahler metric gii dz'8.5'.
Let t l , r2, * , I ~ be~ non-zero
9 + ~ non-negative functions defined on M with
the following properties. For each ti, we can find holomorphic sections
sl,s2, * * , sI of some Hermitian holomorphic line bundles such that ti =
c=, I ~ where ki 2 0 for each j .
~ 1 ~ ~

Then we consider the equation

j ,

where F is a smooth function defined on M x R with aF/at LO.

As in the previous section, we assume that we can find a smooth function
i,lI defined on M such that

(9.2)
I, tltz . . t,, . - . t;;+,,,F(x, 4) = Vol (M) .
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ON THE RICCI CURVATURE OF A COMPACT KAHLER MANIFOLD 403
If m =dim M, we also assume that

(9.3)

and, for some q > O ,

is integrable over M and over every analytic disk of M.

The arguments of Section 8 can easily be modified to prove the solvability
of (9.1) when either F is independent of cp or F(x, cp) = exp {kcp} with k > 0.
The arguments of Section 6 can then be modified to solve (9.1) in general.
(The required comparison theorem can be treated as in the end of Section 8.)

THEOREM 8. Let M be a compact Kahler manifold. Suppose that, in

equation (9.1), the ti are functions satisfying the above mentioned properties.
Then we can find a solution cp of (9.1) such that
(i) cp is smooth outside the zeros of the ti and supM Icp(<m,
(ii) SUP, (AV)(tnl+l* * tnl+nJq+l/m ( ( Alog tn,+l * * * tn,+nzl + i ) ( l - m ) / m is
bounded,
(iii) outside the zeros of the ti, Ci,j(gir+dzcp/dzidZi) dzi@dZi defines a
Kahler metric.
Furthermore, any solution of (9.1) satisfying the above three properties differs
from cp by a constant.

Appendix A
In this appendix, we carry out the computation needed in Section 3. In
other words, we prove formula (3.2) here.
From the commutation formula of Section 1, we have
404 SHING-TUNG YAU

Since Proposition 2.1 shows that the metric tensor Ca,pg&,-dz" @ d Z P is

uniformly equivalent to gap dz* @ dfP, we see from (3.2) that

Similarly, one can also prove that

Let us now write (2.5) as

Then differentiating, we have

10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
405
ON THE RICCI CURVATURE OF A COMPACT KAHLER MANIFOLD
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Diagonalizing and simplifying, we obtain
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ON THE RICCI CURVATURE OF A COMPACT KAHLER MANIFOLD 407

Appendix B
In this appendix, we carry out the proof of (8.17) promised in Section 8.
Integrating by parts in (8.161, we have

-&I- k,) I
M
(~s2~2++)P-kz-1((sl~2++E)kiexp(-C(p,)V
l~,1~.V(p,

- kl e)P-kz(lsl12+&Ik1-' exp {-C(p,)V (s1I2 * V(p,

(ls2I2+ E)P-k2(ls112+ &Ik, exp {-CCp,) IVcp,12

(B.1)
+ ( p - k,)C,C-'
I, &Iki exp {-Ccp,)
(Is212+ e)P-kz-Z(lsl12+ IV ls2I2 1'

+ klC,C-' ((s2I2+ e)P-k2-1(lsl12+ exp {-Ccp,) V Is1l2 V lszlz

9
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
SHING-TUNG YAU
408
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
409

To estimate the right-hand side of (B .3), we use (5.7) again to find a positive
Since we have assumed that C 7 ( p y k2+C,)-l=$, we know that
qm KAHLER MANIFOLD

(p - k2)(p - k2+ C,)-’ = and (B.2) shows tHf at

ON THE RICCI CURVATURE OF A COMITZ

constant C, independent of E such ti .hat

have
Integrating by parts in (B.4), we
4
410 SH ING-TUNG YAU

Multiplying (€3.5) by akl and sub stituting the resulting inequality into (B.3),
we obtain

of
Since we assume
E such that
I, IsJZmk2<m, p ve can find a constant Cl0 independent

Then by Holder’s inequality, we have

Added in proof (February 21, 1978): Some de VeloPments on the Problem

after the completion of this paper may be recoi d e d here. In Section 2, I
offered two proofs of Calabi’s conjecture, one €or c OmPlex dimension two and
one for the general case. After I finished my proc lf, I was kindly invited by
10970312, 1978, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/cpa.3160310304 by Fudan University, Wiley Online Library on [19/08/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ON THE RICCI CURVATURE O F A COMPACr KAHLER MANIFOLD 41 1

Professor E. Calabi to Philadelphia to present a talk in 1976 in which

Professor J. Kazdan was present.. After a few months, J. Kazdan was able to
simplify my two-dimensional proof. This simplification will also appear in this
issue. Then Professor J. Kazdan was invited by Professor T. Aubin to present
in Paris my two-dimensional proof and his simplification. A t the end of 1977,
Professor Aubin and Professor J. P. Bourguignon, who was kindly giving a
talk on my paper in the Bourbaki seminar, independently generalized my
two-dimensional proof to higher dimensions. I wish to thank these colleagues
for their interest in my wotik and for giving another complete proof of the
general case.

Bibliography
[l] Aubin, T., Mitriques Riemianniennes et courbure, J. Diff. Geom. 4, 1970, pp. 383-424.
[2] Aubin, T., Equations du type Monge-Ampere sur les uariitis Kuhleriennes compactes, C.R.
Acad. Sc. Paris, 283, 1976, pp. 119-121.
[3] Calabi, E., The space of Kiihler metrics, Proc. Internat. Congress Math. Amsterdam, 1954,
Vol. 2, pp. 206-207.
[4] Calabi, E., On Kiihler manifolds with oanishing canonical class, Algebraic Geometry and
Topology, A symposium in honor of S. Lefschatz, Princeton Univ. Press, Princeton, 1955,
pp. 78-89.
[ 5 ] Calabi, E., Improper affine hyperspheres and generalization of a theorem of K . Jiirgens, Mich.
Math. J. 5, 1958, pp. 105-126.
[6] Chern, S. S., Chavacteristic classes of Hermitian manifolds, Ann. of Math. 47, 1946, pp.
85-121.
[7] Chern, S. S., Complex Manifold Without Potential Theory, Princeton, N.J., van Nostrand,
1967.
[8] Cheng, S. Y. and Yau, S . T., On the regularity of the Monge-Ampire equation
det (a*u/ax, ax,) = F(x, u ) , Comm. Pure Appl. Math., Vol. 30, 1977, pp. 41-68.
[9] Cheng, S . Y. and Yau, S. T., On the regularity of the solution of the n-dimensional
Minkowski problem, Comm. Pure Appl. Math., Vol. 29, 197'6, pp. 495-516.
[ 101 de Rham,G., Vari&% Diflerentiables;Formes, Courants, Formes Harmoniques, Paris, Hermann,
1960.
[ll] Federer, H., Geometric Measure Theory, Springer-Verlag, 1969.
[12] Moser, J., A new proof of De Giorgi's theorem concerning the regularity problem for elliptic
differential equations, Comm. Pure Appl. Math., Vol. 13, 1963, pp. 457-468.
[ 131 Morrey, C. B., Multiple Integrals in the Calculus of Variation, Springer-Verlag, 1966.
[14] Schwartz, J., Nonlinear Functional Analysis, N.Y., Gordon and Breach, 1968.
[15] Yau, S. T., On Calabi's conjecture and some new results in algebraic geometry, Nat. Acad.
Sci. U.S.A., 74, 1977, pp. 1798-1799.

Received June, 1977

Ricci-flat Kähler Metrics on Manifolds
No ratings yet
Ricci-flat Kähler Metrics on Manifolds
6 pages
L∞ Estimates for Monge-Ampère Equations
No ratings yet
L∞ Estimates for Monge-Ampère Equations
26 pages
Kahler Manifolds
No ratings yet
Kahler Manifolds
183 pages
Lectures on Kähler Manifolds
No ratings yet
Lectures on Kähler Manifolds
183 pages
Kahler Elliptic Notes
No ratings yet
Kahler Elliptic Notes
140 pages
Existence of Solutions to Gursky-Streets Equation
No ratings yet
Existence of Solutions to Gursky-Streets Equation
28 pages
Kähler Metrics with Positive Curvature
No ratings yet
Kähler Metrics with Positive Curvature
17 pages
Exponential Decay of Bergman Kernels
No ratings yet
Exponential Decay of Bergman Kernels
27 pages
Introduction to Kähler Manifolds
No ratings yet
Introduction to Kähler Manifolds
55 pages
Tian, Canonical Metrics in Kahler Geometry PDF
No ratings yet
Tian, Canonical Metrics in Kahler Geometry PDF
107 pages
Auxiliary Monge-Ampere Equations in Geometric Analysis
No ratings yet
Auxiliary Monge-Ampere Equations in Geometric Analysis
38 pages
Lect 18
No ratings yet
Lect 18
2 pages
A Uniformization Theorem of Complete Noncompact Kahler Surfaces With Positive Bisectional Curvature
No ratings yet
A Uniformization Theorem of Complete Noncompact Kahler Surfaces With Positive Bisectional Curvature
57 pages
Sufficient Bigness Criterion For Differences of Two Nef Classes
No ratings yet
Sufficient Bigness Criterion For Differences of Two Nef Classes
7 pages
Lecture Notes in Mathematics: Editors
No ratings yet
Lecture Notes in Mathematics: Editors
314 pages
Cohomology and Vanishing Theorems
No ratings yet
Cohomology and Vanishing Theorems
114 pages
Almost Kaehler Manifolds
No ratings yet
Almost Kaehler Manifolds
8 pages
Kawa Lecture Notes On The K Ahler-Ricci Flow: Abstract
No ratings yet
Kawa Lecture Notes On The K Ahler-Ricci Flow: Abstract
85 pages
L∞ Estimates for Non-Linear PDEs
No ratings yet
L∞ Estimates for Non-Linear PDEs
34 pages
Generalized Monge-Ampère on Almost Kähler Surfaces
No ratings yet
Generalized Monge-Ampère on Almost Kähler Surfaces
21 pages
Non-Uniqueness in Scalar Curvature PDE
No ratings yet
Non-Uniqueness in Scalar Curvature PDE
30 pages
Laplace Comparison On Kahler Ricci Flow
No ratings yet
Laplace Comparison On Kahler Ricci Flow
45 pages
BICMR ComplexGeometry
No ratings yet
BICMR ComplexGeometry
81 pages
Trivial Solutions in Harmonic-Divgrad PDEs
No ratings yet
Trivial Solutions in Harmonic-Divgrad PDEs
8 pages
Locally Conformal Kahler Geometry
No ratings yet
Locally Conformal Kahler Geometry
346 pages
Complex Manifolds and Geometry Insights
No ratings yet
Complex Manifolds and Geometry Insights
175 pages
Diameter Estimates for Kähler-Ricci Flow
No ratings yet
Diameter Estimates for Kähler-Ricci Flow
22 pages
Existence and Uniqueness of ?1-Invariant Kähler-Ricci Solitons
No ratings yet
Existence and Uniqueness of ?1-Invariant Kähler-Ricci Solitons
36 pages
Comm Pure Appl Math - July 1974 - Folland - Estimates For The Documentclass Article Pagestyle Empty Begin Document
No ratings yet
Comm Pure Appl Math - July 1974 - Folland - Estimates For The Documentclass Article Pagestyle Empty Begin Document
94 pages
Geometric Analysis of Riemannian Manifolds
No ratings yet
Geometric Analysis of Riemannian Manifolds
53 pages
Prescribing Scalar Curvature on AH Manifolds
No ratings yet
Prescribing Scalar Curvature on AH Manifolds
31 pages
Buckling of A Thin Elastic Plate
No ratings yet
Buckling of A Thin Elastic Plate
6 pages
Coarsely Holomorphic Curves in Symplectic Topology
No ratings yet
Coarsely Holomorphic Curves in Symplectic Topology
15 pages
Applications of Partial Differential Equations To Problems in Geometry
No ratings yet
Applications of Partial Differential Equations To Problems in Geometry
137 pages
ANDERSON 1989 Complete Ricci-Flat Kahler-Manifolds of Infinite Topological Type-1
No ratings yet
ANDERSON 1989 Complete Ricci-Flat Kahler-Manifolds of Infinite Topological Type-1
6 pages
Riemannian Manifolds with Density Insights
No ratings yet
Riemannian Manifolds with Density Insights
7 pages
Kahler Geometry 2020-2
No ratings yet
Kahler Geometry 2020-2
83 pages
Holomorphic Vector Fields on Hermitian Manifolds
No ratings yet
Holomorphic Vector Fields on Hermitian Manifolds
41 pages
Demailly
No ratings yet
Demailly
11 pages
Complex Manifolds
100% (2)
Complex Manifolds
204 pages
Some Aspects of The Hodge Conjecture: Claire Voisin
No ratings yet
Some Aspects of The Hodge Conjecture: Claire Voisin
36 pages
Chern-Calabi Flow on Hermitian Manifolds
No ratings yet
Chern-Calabi Flow on Hermitian Manifolds
15 pages
Geometric Analysis by Zhiqin Lu
No ratings yet
Geometric Analysis by Zhiqin Lu
44 pages
Generalized Conditions for PDEs
No ratings yet
Generalized Conditions for PDEs
15 pages
Complex Manifolds
No ratings yet
Complex Manifolds
113 pages
Space of Kähler Metrics Analysis
No ratings yet
Space of Kähler Metrics Analysis
46 pages
Complex Hypersurfaces in Projective Space
No ratings yet
Complex Hypersurfaces in Projective Space
4 pages
602 Notes1
No ratings yet
602 Notes1
97 pages
M' manifoldsRIEMANNIAN-tanno
No ratings yet
M' manifoldsRIEMANNIAN-tanno
10 pages
Differential Geometry Concepts and Problems
No ratings yet
Differential Geometry Concepts and Problems
38 pages
1989 - Expansiveness, Hyperbolicity and Hausdorff Dimension
No ratings yet
1989 - Expansiveness, Hyperbolicity and Hausdorff Dimension
14 pages
K Ahler Manifolds and Transcendental Techniques in Algebraic Geometry
No ratings yet
K Ahler Manifolds and Transcendental Techniques in Algebraic Geometry
34 pages
10.1515 - Crelle 2015 0103
No ratings yet
10.1515 - Crelle 2015 0103
35 pages
10.1515 - Crelle 2013 0004
No ratings yet
10.1515 - Crelle 2013 0004
16 pages
Cohomology of Flat Line Bundles on Lie Groups
No ratings yet
Cohomology of Flat Line Bundles on Lie Groups
17 pages
Classification of Born Lie Algebras
No ratings yet
Classification of Born Lie Algebras
22 pages
(Chris A.M. Peters, Joseph H. M. Steenbrink) Mixed Hodge Structures
100% (2)
(Chris A.M. Peters, Joseph H. M. Steenbrink) Mixed Hodge Structures
467 pages
Introduction To Complex Manifolds 1st Edition John M. Lee
No ratings yet
Introduction To Complex Manifolds 1st Edition John M. Lee
407 pages
From Holomorphic Functions To Complex Manifolds
100% (4)
From Holomorphic Functions To Complex Manifolds
406 pages
Recent Advances in Ricci Solitons
No ratings yet
Recent Advances in Ricci Solitons
32 pages
Canonical Metrics on Teichmüller Spaces
No ratings yet
Canonical Metrics on Teichmüller Spaces
108 pages
M2 GEO Program2025 26
No ratings yet
M2 GEO Program2025 26
6 pages
DDVV Conjecture For Riemannian Maps From Quaternionic Space Forms
No ratings yet
DDVV Conjecture For Riemannian Maps From Quaternionic Space Forms
9 pages
Complex Geometry Course Overview
No ratings yet
Complex Geometry Course Overview
73 pages
Geometry Topology and Physics 2nd Ed Edition Mikio Nakahara
No ratings yet
Geometry Topology and Physics 2nd Ed Edition Mikio Nakahara
394 pages
Bochner Flat Tangent Bundles Study
No ratings yet
Bochner Flat Tangent Bundles Study
12 pages
Quantum Gravity & Dark Energy
No ratings yet
Quantum Gravity & Dark Energy
10 pages
Pluripotential Theory: Giorgio Patrizio Zbigniew Błocki François Berteloot Jean-Pierre Demailly
No ratings yet
Pluripotential Theory: Giorgio Patrizio Zbigniew Błocki François Berteloot Jean-Pierre Demailly
328 pages
Introduction To Complex Manifolds John M Lee
No ratings yet
Introduction To Complex Manifolds John M Lee
439 pages
Ioannis Bakas - Renormalization Group Equations and Geometric Flows
No ratings yet
Ioannis Bakas - Renormalization Group Equations and Geometric Flows
10 pages
Harmonic Functions on Surfaces
No ratings yet
Harmonic Functions on Surfaces
11 pages
教材推荐（第8章代数几何) -复旦
No ratings yet
教材推荐（第8章代数几何) -复旦
7 pages
Period Mappings and Period Domains: Editorial Board
No ratings yet
Period Mappings and Period Domains: Editorial Board
576 pages
Lecture Notes On Chern Simons Witten Theory
100% (2)
Lecture Notes On Chern Simons Witten Theory
214 pages
Stable Forms and Special Metrics Analysis
No ratings yet
Stable Forms and Special Metrics Analysis
31 pages
Differential Geometry and Its Application
No ratings yet
Differential Geometry and Its Application
4 pages
Mean Curvature of Direct Image Bundles: Kuang-Ru Wu
No ratings yet
Mean Curvature of Direct Image Bundles: Kuang-Ru Wu
23 pages
Hodge Theory 1st Edition Eduardo Cattani
No ratings yet
Hodge Theory 1st Edition Eduardo Cattani
463 pages
Phase Transition
No ratings yet
Phase Transition
177 pages
Toric Varieties and Residues: Alexey Shchuplev
No ratings yet
Toric Varieties and Residues: Alexey Shchuplev
70 pages
Apanasov. .Geometry - And.topology - Of.complex - Hyperbolic.and - CR Manifolds. (1997) - (Sharethefiles - Com)
No ratings yet
Apanasov. .Geometry - And.topology - Of.complex - Hyperbolic.and - CR Manifolds. (1997) - (Sharethefiles - Com)
34 pages
BSMF 2008 136 2 243 0
No ratings yet
BSMF 2008 136 2 243 0
19 pages