Gang Tian Canonical Metrics in Kahler Geometry Notes taken by Meike Akveld Birkhiuser Verlag Basel - Boston - BerlinAuthor's address: Department of Mathematics Massachusetts Institute of Technology 2-246 TT Massachusetts Avenue Cambridge, MA 02139 USA 2000 Mathematical Subject Classification 35C55; 58G30; 81T30 ACIP catalogue record for this book is available from the ibrary of Congress, Washington D.C,, USA Publication Data Library of Congress Cataloging- in Tian, G. Canonical metrics in Kahler geometry / Gang Tian. P. cm. -- (Lectures in mathematics ETH Ziirich) Includes bibliographical references and index. ISBN 3-7643-6194-8 (ulk. paper). ~- ISBN 0-8176-6194-8 (alk. paper) 1. Kihlerian manifolds. 2. Geometry, Differential. 1. Title. IL Series. QA649.T46 2000 $15°.73 ~ de21 Deutsche Bibliothek Cataloging in Publication Data Tian, Gang: Canonical metrics in Kahler geometry / Gang Tian. - Basel ; Boston ; Berlin : Birkhauser, 2000 (Lectures in mathematics) ISBN 3-7643-6194-8 ISBN 3-7643-6194-8 Birkhauser Verlag, Basel — Boston — Berlin ‘This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, re-use of illustrations, recitation, broadcasting, reproduction on microfilms or in other ways, and storage in data banks. For any kind of use permission of the copyright owner must be obtained. ‘© 2000 Birkhiiuser Verlag, PO. Box 133, CH-4010 Basel, Switzerland Printed on acid-free paper produced from chlorine-free pulp. TCF o> Printed in Germany ISBN 3-7643-6194-8 987654321Contents 1 Introduction to Kahler manifolds 1.1 Kahler metrics 1 1.2 Curvature of Kahler metrics 4 2 Extremal Kabler metrics 2.1 The space of Kabler metrics... eee ee eee u 2.2 A brief review of Chern classes .... 2... 14 2.3. Uniformization of Kahler-Kinstein manifolds . 18 3. Calabi-Futaki invariants 3.1 Definition of Calabi-Futaki invariants. 6.2.2... 05 23 3.2. Localization formula for Calabi-Futaki invariants... 2... . 29 4 Scalar curvature as amoment map... 2.0... 200.000.0205 35 5 Ka&hler-Einstein metrics with non-positive scalar curvature 5.1 The Calabi-Yau Theorem .. 2... 2. ee ee 43 5.2 Kihler-Einstein metrics for manifolds with c,(M) <0 ...... 55 6 Kabler-Einstein metrics with positive scalar curvature 6.1 A-variational approach... 6.2... eee eee 58 6.2 Existence of Kihler-Einstein metrics 62 6.3 Examples 85 7 Applications and generalizations 7.1 A manifold without Kahler-Einstein metric ............ 89 7.2 K-energy and metrics of constant scalar curvature ....... . 93 7.3 Relation to stability ©... 6.2... eee eee 96 Bibliography ©. 2... ee ees 99Preface This monograph results from the author’s lectures at the ETH during the Spring Semester of 1997, when he was presenting a Nachdiplom course on Kahler- Einstein metrics in complex differential geometry. ‘There has heen fundamental progress in complex differential geometry in the last, two decades. The uniformization theory of canonical Kahler metrics has been established in higher dimensions. Many applications have been found. One manifestation of this is the use of Calabi-Yau spaces in the superstring theory. The aim of this monograph is to give an essentially self-contained introduc- tion to the theory of canonical Kahler metrics on complex manifolds. It is also the author’s hope to present the readers with some advanced topics in complex differential geometry which are hard to be found elsewhere. The topics include Calabi-Futaki invariants, Extremal Kahler metrics, the Calabi-Yau theorem on existence of Kahler Ricci-flat metrics, and recent progress on Kihler-Einstein metrics with positive scalar curvature. Applications of Kahler-Einstein metrics to the mniformization theory are also disenssed Readers with a good general knowledge in differential geometry and par- tial differential equations should be able to understand the materials in this monograph, I would like to thank the ETH for the opportunity to deliver the lectures in a very stimulating environment. In particular, I thank Meike Akveld for her patience and efficiency in taking notes of the lectures and producing the beautiful IATEX file. Without her efforts, this monograph could never have been as it is now. I would also like to thank Ms. Nini Wong for her endless pa- tience in proof-reading and correcting numerous typos in earlier versions of this monograph. Part of my work involved in this monograph was supported by National Sci- ence Foundation Grants DMS-9303999 and DMS-9802479, at Courant Institute of Mathematical Sciences and Massachusetts Institute of Technology. My re- search was also supported by a Simons Chair Fund at Massachusetts Institute of Technology. MIT, April 1999. Gang TianChapter 1 Introduction to Kahler manifolds 1.1 Kahler metrics Let M be a compact C® manifold. A Riemannian metric g on M is a smooth section of T*M @ T*M defining a positive definite symmetric bilinear form on TM for each x € M. In local coordinates 21,...,¢n, one has a natural local basis 32;,++-, a2 for TM, then g is represented by a smooth matrix-valued function {g;;}, where y-9(2 2 95 = 9\ 55 Bay) Note that {gi,} is positive definite. The pair (M,) is usually called a Rieman- nian manifold. Reralll that an almost complex structure .J on M is a bundle automorphism of the tangent bundle TM satisfying J? = — id. Definition 1.1. The Nijenhuis tensor N(J):TM x TM + TM is given by N(v,w) = [v,w] + J[Jv, w] + Jv, Jw] — (Jv, Ju] for v,w vector fields on M. An almost complex structure J on M is called integrable if there is a holo- morphie structure (that. is a set of charts with holomorphic transition finctions) such that J corresponds to the induced complex multiplication in TM x C. Clearly, any complex structure induces an integrable almost complex structure. The following theorem is due to Newlander and Nirenberg, sce for example Appendix 8 in [14]. Theorem 1.2 An almost compler structure is integrable if and only if N(J) =0. The hard part is to prove that N(J) = 0 implies integrability.2 Chapter 1. Introduction to Kahler manifolds We say that J is compatible with the metric g or g is a Hermitian metric if g(u,v) = g(Ju, Jv). We can then define a 2-form wy by w(u,v) = —g(u, Jv). Usually, we call such an wy the Kahler form of g. Example 1.3 Consider M = C” as a real manifold by identifying R2" with C” in the usual way, then the corresponding almost complex structure J is given by aa 8 a Be, ~ Ban Oa Let 21,...,%, be the canonical complex coordinates, 2; = 2 + V—Izrp4j;. Then dz; = dr; +V—Hdany;, 42; = day — V—Ideny; J Dens: Ot” and a (2 2), a 1, a a 357335 Vda) Gy 7 2(y tS) for j =1,...,n. We also have that (ez) avai and J( If g is the Euclidean metric, then its Kahler form is given by v-1 =. y= ye Ady = Lee Adtnti- We will denote by V the Levi-Civita connection of g, which is the unique torsion free connection which makes g parallel. Definition 1.4 A Kahler manifold (M,g,J) is a Riemannian manifold (M,9) together with a compatible almost complex structure J, such that VJ = 0, where V is the Levi-Civita connection of g. Let (M,g,J) be a Kahler manifold. We can extend the metric g C-linearly to TM @C. Recalling that T+°M and T°1M are the +i -cigenspaces of J, we see that g(u,v) = 0 for u,v € T!°M or u,v € T°'M (use compatibility of J with the metric). Define h(u,v) = g(u,5) for u,v € T1M, then this defines a Hermitian inner product on T'°M,1.1, Kahler metrics 3 If (M, J) is a complex manifold and J is compatible with the metric, then from the above, we get that -!) 22s 9\ G2; 85) ~ 9 05 3%) ~ and in local coordinates, we can write _. aa 95 =9| 9555)? SV gijdes A dz. cy On a Kahler manifold M, a Kahler metric is uniquely determined by its Kahler form. So we often denote a Kihler metric g by its Kahler form w,. Note that N(J) = 0 on M, so M is a complex manifold and dur, = 0, that is, wy is a closed form. Proposition 1.5 If the Nijenhuis tensor N(J) vanishes, then VJ = 0 if and only if dy =0. In particular, if (M, J) is a complex manifold and g is a Hermitian metric, then (M,g,J) is Kahler if dry = 0. Proof. For all u,v,w € TM, we have so we have Wg = durg(u,v,w) =u(wo(v, w)) + v(wg(w, u)) + w(w9(t,v)) — wo([u, 0], w) + wo ([w, u],v) — wo({v, w],u). Since w,(u,v) = g(Ju,v) for any u,v, we deduce from the above duty (u,v, w) = 9((Vud)v,w) + g((VoF)w,u) + (Vw u,v). Replacing u by Ju or v by Ju in the above, we obtain duty (Ju, v, w) = 9((Viut)v,w) + 9((Vod)w, Ju) + 9((Vod)Ju,»), dary (u, Jv, w) = (Vad) Jv, w) + o((VD)w, u) + o((Vovd)u, Jv). Summing up the above two and using the facts that P= =I and glJu,v) + glu, Je) — 9, we have dug (Ju, v,w) + dw9(u, Jv, w) =29((Vid)u, Jv) + g((Vud) Jv — (Ved) Ju + (Vaud )v — (Vaud), w) =29((Vi.J)u, Jv) — g(Ny(u,v),w). It follows that if dw, = 0 and Ny = 0, then VJ = 0. It is trivial to sce that VJ =0 implies du’, = 0.4 Chapter 1. Introduction to Kahler manifolds 1.2 Curvature of Kahler metrics Let (M,g, J) be a Kahler manifold and V be its Levi-Civita connection. We extend V in a C-linear way to '(TcM). Since M is also a complex manifold, we have local coordinates (21, ...,2n) and hence a local basis (32-,--+ , 2) for TeM. We define the Christoffel sym- bols Ff, by a 4.O £20 Vo Be; 8 Om tT Oe and a, Og A Ve %, "Sax, tT oag —. Because VJ = 0 and JZ. = V—1% and JE = implies vA(r 9 ark 2) =a(r ope =) 7 Oxy, 683, Tze 7 OR, a a /— ve ie T(r da, TS a) and therefore P¥, = 0, similarly, P, = P% = 0, so the only possible non-zero terms are Pf, and r= 18. Moreover, if 954 = 5 3%;) denote the metric tensor in local coordinates, then don 8 (8 B\_ a8) (,a@ @\ 4 Oa Oa! (sero) = (Va oa) “9 (sgeae) =P ior and hence = gi O9;e _ hk OOR Oa Oz," So if the Kahler metric is given by {g,j}, its connection V is given by ri, = gl Sit 8950 Oz Proposition 1.6 (Normal coordinates in Kahler case) Let M be a Kahler mani- fold with a real analytic Kéhler metric. Given x € M, there exist local complex ry coordinates (21,...,2m) unique modulo unitary linear transformations such that 94g(2) = 65 , dgig(a) = 0 and aig) =0 for alll > 0 andiy+---+in =1, and this also holds for its conjugate.1.2. Curvature of Kahler metrics 5 This proposition can either be proved by induction in an elementary way or for another elegant proof, see [16], page 286. So given a Kahler manifold (M,g) with its Kahler form = S YS gijdei A dz; ig and its compatible connection V, the Riemannian curvature tensor is defined by R(u, v)w = VuVyw — VoVuw — Viaje and can be extended in a C-linear way to TeM. Note that because J is parallel, that is VJ = 0, we have that R(u,v)Jw = IR(u, v)w. Defining R(u, v,w, 2) = 9(R(u,v)z,w), we can easily see that R(u,v, Jw, Jz) = R(u, v, w, 2) and because of the splitting To M =T1°M @T°M into the +i-eigenspaces of J, we can deduce that R(u,v,w,z) — 0 unless w and z are of different type. In local coordinates z1,..., 2m, this means that the only possibly non-vanish- ing terms are essentially (ee Raut = BR (Sa RS a _ pido 0 Veda; B25 9 De; Ox’ Using that we can deduce that P95 | ot 95 Ogi "dnd 57 Om Oa We define the Ricci curvature to be the trace of this, so we get . j a Rieys = Ry = 9 Rajat = 0,05 (log det 9,3) We will denote the Ricci curvature both by Ric and by Rx which should not cause any confusion. In complex coordinates we have found a nice expression6 Chapter 1. Introduction to Kahler manifolds for the Ricci curvature, but we need to check that it is still the same as that in the Riemannian case. So choose an orthonormal basis ¢1,...,€2n such that Jes = ny fori = 1,...,n and set us = J5(e:— /—LJe;), then {u;} is a unitary basis. It follows that Rui, Hs) = YO Rei, ts, uy, ) i => 5Rle — V-TWe;,e; + V-1es, uj, 4) 3 = Vi v=TR(e:, Jei,uj, 5) a = =P Rei, Jei,€3, Jes) 7 = YE RJei,e5,€4, Jes) + D> R(es,€4, Fei, Jej) 7 7 = DP Rei, ej, Jey,06) + D> R(eise55€5,€4) i 7 = YR, ens js €nsjres) + D R€:,e5, 6,68) 7 7 = Ric(ei, ei). Here we have used the first Bianchi identity for R: R(e;, Jei,e;, Je;) + R(Jei,e;,€:, Je;) + R(e;,e:, Je:, Je;) = 0 ‘This shows that the Ricci curvature defined above is the same as the one in Riemannian geometry. Recall that if |z| = |y| = 1 and 2 is perpendicular to y, then R(2x,y,y, x) is the sectional curvature of the plane spanned by x,y. Set now 2 — V71Je) and v = Su — Vly), then Definition 1.7 The bisectional curvature is defined to be R(u, i, v, 0) = R(x, yy, x) + R(x, Jy, Jy, 2). Definition 1.8 A Kahler manifold (M,g) is said to be of constant biscctional curvature if there exists a constant 0 such that in any local coordinates of M, Rajni — M95 9ni + 9:09K5)-1.2, Curvature of Kihler metrics 7 Example 1.9 Let M = CP" = {[zo :---: anli0 # (zo :-+-: zn) € C+} and let Uo = {[1, 21: +++: zn]} & C® be an open subset of CP". Set = @ 2 2 95 = 5:05, log(1 + |z1|? +++» + lznl?) or equivalently 5 ea em BA log(1 + lz? + --- + en!) T+ke2 (+le)? (% Adz %dz4 N2;dz5 ) Wg 2 2 We will check that w, is globally defined on CP”. Let Uy = {(wo,1,we,-+,Wn)} CCP" and check what happens to wy under a coordinate transformation on the overlap Up NU, = {Liza tees zn] = [wo 1s wes: wal}. There exists a non-zero constant ) such that Awo = 1, A= 21 and Aw; = % for i=2,...,n, 80 = 2, and therefore 2) = “4 for all i # 1 and 2: = 4. So we see that (1... wy = ~F-Adog(1 + 7 soot [enl?) “pve in) 1 a0 0g(1+ 2, +e a = eee + |w|?) — 88 log(|wo|?) = A aatog(t + ful), since wo is holomorphic. So w, is globally defined and the corresponding metric on CP" is called the Fubini Ste naturally on CP" as an isometry of g, and SU(n +1) acts on CP” transitively, Note that ™ (dz; A dz)" +P and hence Rieg = — 2 tog( 1 _ = (n+ log O OOa; PLT 4 apy ) = + 958 Chapter 1. Introduction to Kahler manifolds The following calculation P95 HF log(1 + |z/?) © Dende ° ~~ Oz4,0%102,02, |r=0 _1__ Het ~ 2 0%4,0%02%02; = 5 lle) ~ OxdH0m 1 71/10 = Babe (2D leH0 = 615 5m1 + SuSey shows us that CP” is a manifold of constant bisectional curvature and that Rajat = 9439xi + gitgnj, since the isometry group of g acts on CP transitively. Example 1.10 Let M = C” then w, = “"dz; A dz; is the flat metric and the bisectional curvature vanishes. ‘ Example 1.11 Let M = B" = {z € C%;|z| <1} and let var 88 log(1 — |z\?), Wy 2 then Rigt = —(9.39:0 + 910943) and (B",g) is Kahler manifold of constant bi- sectional curvature —1. Theorem 1.12 (Uniformization Theorem) If (M,g) is a complete Kihler mani- fold of constant bisectional curvature R347 = (9:3 9;40+ 911943) for some constant A, then its universal covering M is one of the above examples. Moreover, up to scaling, g pulls back to one of te metrics in the above examples. Proof. After scaling, we can distinguish three cases, = —1, 4 = 0 and A= 1. We will first prove the cases 1 < 0. Let (Mj,gy) be the Kahler manifold of constant bisectional curvature \ in the above examples. Consider the maps exp : ToMy > My and exp, : ToM —> M, where exp, and exp, are the expo- nential maps of g, and g, respectively, for any v € TyMa, expo(v) is the geodesic 7 of gq at time 1, satisfying +(0) = 0 and 7/(0) = v. We need completeness to guarantee the existence of geodesics. Identify both TpM, and T,M with R?" and define the map = exp, (expg) ~1.2. Curvature of Kabler metrics 9 We claim that ¢ is an isometry and once this is proved, we have proved the theorem. So we must consider dé. We will first look at dexpg(v). For w € R2", we have dexpg(v)(w) = © expo(ste + tw)) = 4(s,t) =: Xu(s) where we have introduced (s,#) = expg(s(v + tw)) for simplicity. We now check that X,,(s) is a Jacobi field. Because 7(s, t) is a geodesic for cach fixed t, we have that Vp 4(4,.) §47(8)t) = 0 and hence Vp Vp £7(s,t) = 0 and then a 0= Vx. Vg 5" _ oy oy oy oy =VaVxu 5 + Vix gigs ( a) i _ oy oy = VgsVgsXu~R( $2 Xe) as together with the initial conditions X,,(0) — 0 and X/,(0) = w. Now fix an orthonormal basis e,,...,¢2, for TyMy, such that e, = 32|9%\-1 and enyi = Je;. Parallel ee this basis along +, that is, V 9,e;(s) = 0 and e;(0) = ei. So we can write X,,(s) = > X‘(s)e;(s) and the Jacobi equation reads 2X4 e, -2(Fre 1) FLXI = 0 Os Ps ds’ ‘Taking the inner product with e;, we get a | \z a a) Xi = 0 +|2/ R(e1, 6), ¢:,€1)X? = 0. eal fa(B which is equivalent to PX4(s) es By the lemma below, the R(e1,€,€:,1) are completely determined by the bi- ed by wand \. I X(s) + = dexp,(v)(w) i is a Jacobi field along the corresponding geodesic 7 in Mf then X,, satisfies the same cquation as X,, does because the biscctional curvatures are the same and hence we can deduce that |X| = |Xw]|, so ¢ is an isometry. If 4 > 0 we know that the Ricci curvature is also positive and by Myers’ Theorem [5], we know that M is compact. Let Uo C CP” as before then we can show that ¢ is an isometry from Up onto its image in exactly the same way as before. Now because Up is dense in CP” and because M is compact we can extend ¢ to all of CP” so that # remains an isometry. Qo urvature, so X,, is uniquely det:10 Chapter 1. Introduction to Kahler manifolds Lemma 1.18 If the bisectional curvature is constant, then ife; Lei, Jer de Reweda={ he Fez oe Proof. As before, we write ui = J5(e;— V=1Je;). Then Rui, iti, Ue, th) = —R(l:, ei, ex, Jer) + V-1R(Gi, Jes, ex, €1)- Since the biseotional curvature is constant A, we have —R(€;, Jess ex, Sex) + V=TR(€5y Sess €bs€1) = { ei tees Comparing the real and imaginary parts of both sides, we get Qu, ifk=i =Rlex Jeane = { Nu, ifk xi. and R(ei, Jei, ex, €1) = 0. For any i # j, if we replace e; by yale + ;) in the above arguments, we get Rei, Jej, ex, Jer) — R(ei, Jej, ex, €2) — 0. It follows from the above that R(e1,€:,€1, Je) =0, R(er,€i,€1,€1) = 0,1 # i. So R(e1,e:)er = je; for some jt. If e; = Jey, then pr = —22, since Rei, Jer, Jer, €1) = R(uj, 1, v1, ti) = 2A. Now we may assume that e; | 1, Je,. Replace e), Je, by cose, + sinOJe1, —sin Oe, + cos6Je,, as above, we can deduce R(cos be; + sin 8Je1, e, J(cos Be + sin OJex), €;) = 0. This is the same as 4 R(cosber + sin 9.Je,,¢:, J(cos Ge; + sin 8Je1), €;) 6 0. Hence, R(e1,€:,€1,¢:) = R(Je1,¢:,Je1,e:), then by R(wi,%1,u,%1) = A, we have j: = —3. Then the lemma is proved. aoChapter 2 Extremal Kahler metrics 2.1 The space of Kahler metrics In this section, we introduce the Calabi functional on the space of Kahler met- rics. We will start with a simple lemma. Lemma 2.1 (93-Lemma) Let (M,g) be a Kahler manifold and let ¢1,¢2 € H*\(M,C) and suppose that gx is cohomologous to $2. Then there exists a function f € C°(M,R) such that $1 — ba = d0f. Proof. We know that ¢1 — ¢2 = dy for some y € H1(M,R) and that we can write y= Yh? + yO with Gy! = Ap) = 0, so we want to find functions f, f satisfying y®! = Of and y = OF. We will show how to do this for 4°? and by taking conjugates, one can do this also for ~+". Write y°! = , then locally, @=6;dz; and 5°06 =~—9'6;,. The following equation can be solved for u, Fu 50 = dtu = gi a8. ‘§ because fy, O*6w? = 0. So d*(— du) = 0 and we also have that 8(9 — du) = 0. We claim that 8(9 — du) = 0. Suppose this claim were true. Then writing yl = 9 — du+ Ou, we see that dyb®! = Adu, which proves the lemma. So it remains to prove the claim. Set ¢ = @ — du, then we have the identity 1 n(n—1) : (3) 86 AG Nw? = (Io0? ~ oro? Jug.12 Chapter 2, Extremal Kahler metrics Integrating this, we get os f laoruy = f 1 pPuoy — sae) I, 86 NBO Nw? = I, 7 owe - nine» I, a(a% ndnvy) - nent) [336 Abrus. ‘Therefore, 06 = 0, so the lemma is proved. u Note that this lemma is also true for any (p,q)-forms (compare [3]). Corollary 2.2 Given 2 € H™(M,C)N H?(M,R), define Ko = {all Kabler metrics w with [w] 0}, then Ka = {uy + 57066; 6 €C°(MR), [| bu =0} = (GE C*(MR): | buf = 0,094 vad > 0} In In what follows we will use results and definitions from Calabi (1980) in [4]. Definition 2.3 The scalar curvature of a metric w is defined to be 2 a Seay eect ow Definition 2.4 ‘The Calabi functional Ca is given by Cawy— 4 f sw)2o™ Vu where V = fy.0". Note that V only depends on the class of w and is independent of the partic- ular representative that we choose. Fix now a metric w = “J g3d2; A dz;2.1, The space of Kahler metrics 13 and calculate the Euler-Lagrange equations for the Calabi functional. Let wg =Wt F700, the then; Po 8(wig) = 5 ) Fe; eet (au + 5a 5) — gil. Fo oN Rg) +08). 3@_o = s(w) (05 And so we get Ca(wy) = Calo) + ef, as(w) (- A6— bah + 2 Cae) w" + Of). Here the last term under the integral comes from the fact that we have also changed the volume form and we have 5 a (++ vo86) leno = wt NABH — Agu”. Integration by parts gives us then the following Euler-Lagrange equation for the Calabi functional. —2A?s — 2(sRiz),¢ + A(s?) = 0. Here, for any covariant tensor ¢ given by {¢;,. j,} in local coordinates, we sipfiJq define es %in.iphy Fai = 3a; * — SETH $i... sdyntehin de? at these , j,i ate components of the covariant derivative V'°@ of @. Sim- ph ilarly, one can define ¢;,_. i,3,...3,,3 epresenting V°g. We know A(s?) = 2|Vs|? + 2sAs. So we can nse the second Rianchi Rijktm + Rijtm,k + Rigmit = 0 to reduce the above to A?s + |Vsl? + Rizsij = 0. This is still an equation of too high order, but now using that M is compact, we can reduce the equation to a second order one in s. We will show that the4 Chapter 2, Extremal Kahler metrics above equation is equivalent to sj; = 0, which again follows from taking and switching some derivatives and using the second Bianchi identity, namely ae + (Ragajsx)s = 0- 854 = And so we see that 0= [spat = foanie= [Is Definition 2.5 A Kahler metric w is called extremal if s(w);3 = 0. Therefore, 335 We can also give this a geometrical interpretation: Define a vector field on M by X' = g'Js;. Then w is extremal is equivalent to saying that OX = 0 (since the metric is parallel) and this is equivalent to saying that X is a holomorphic vector field. Corollary 2.6 If M has no non-zero holomorphic vector fields, then the extremal metrics are the metrics with constant scalar curvature. 2.2 A brief review of Chern classes Given a Kahler metric wy, we can define a matrix valued 2-form 2 = (9), which is actually of type (1,1), by = GP Riguiden A de. Then we have that det(I + 421) = 1 + t¢1(g) + #2¢2(g) + --- which has the following well-known properties (compare [3}): ~ doi(g) = 0 and [¢;] € H'*(M,C) H™*(M,R), [¢i(g)] is independent of g, ~ ¢(M)p is represented by ¢;(9). Tiere the o;(4f) deuvies the it” Cheru class of the manifold 1. be interested in c,(M) and c2(M) and will only check the second property for e(M). We will mainly Ji, v= . 1(9) = “p95 = “a Raden A di AD log det(g,2) Qn and hence én(a) ~ x60) =~" 00 tog SA). ur2.2. A brief review of Chern classes 15 Clearly, wos ($28 4) is a well-defined function. Moreover, if at dV =v.de Ad A+++ din Nn is any volume form on M, we can also represent c)(M) by —Ad log v:. Definition 2.7 We say that c:(M) > 0 (< 0) if c1(M) can be represented by a positive (negative) form. In local coordinates, this means that p= V—Mpyzdey, Nd where dyr is positive (negative) definite. We say that c\(M) = 0 if the first Chern class cy(M) is cohomologous to zero. Example 2.8 Let M = CP” and let _ Vale 2 wes = Y= dDlog(1 +S lel?) > 0, then we have n VaI\" nidzy A-++ A din whe = (4 —) SA. 2 (a Dla?) This implies that ntl Ric(wrs) = neers and so ¢(M)= th orsh where +[wps] is the positive generator of the cohomology H?(CP",Z) = Z. One can show that Ae Doan = e(My Here we mean 1 by squaring that we take the cup product of c,(M) with itself. L, Fae 204+ D om) — a(m?) AwRs? =0. As a side remark, we would like to mention here the concept of an index of a symplectic manifold (M,w). Recall that a symplectic manifold (M,w) is a differentiable manifold M with a non-degenerate, d-closed 2-form w. The index of M is r if c\(M) = r¢ for some primitive symplectic form ¢, that is, ¢ is not an integer multiple of another form. So the index of CP” is n + 1.16 Chapter 2. Extremal Kahler metrics For n > 1, one can prove that for an n-dimensional Kahler manifold M, r 2 was proved by §. Kobayashi and Ochiai (see [15]). A plausible conjecture is that the index of any compact symplectic manifold of complex dimension n is no more than n+ 1. It may also be true that if a compact symplectic manifold of complex dimension n. is of index n + 1, then it is CP". This conjecture is indeed true in the case that n = 2 and the underlying manifold is already diffeomorphic to CP?. This was a result of C. Taubes. Example 2.9 Let My = {z € CP"; f(z) = 0} where f is a homogeneous polynomial of degree d. It follows from the Implicit Function Theorem that My is a smooth manifold if and only if nnd { 5£ = 0) = (0.....0)} So assume this is the case. We have seen before that we can equip CP"™+? with the Fubini-Study metric so that 7- log( + |z|?). 95 ~ Be,0z, °° . We claim that Loa nto” e1(M;) = —— rs}. Set on Lge w= ton Sin ) and note that 1 is globally defined. We will now calculate wits. It suffices to do this on the open and dense subset von{ of zoh nay, Benn 80 we get = DEAD det(gqa)1<0,8sntt arises a dzy Ad3, A-- Ada Adi, A+ Adzy Aaj Av Aden gi A@in412.2. A brief review of Chern classes 17 =det(g.g)iza,acndea A dia Nis N dtm A dip +2Re(D(-1)"4# det(gag)in+2. AD logw%s = —ADlog - Blog| so m+1 WPS. 2.3 Uniformization of Kahler-Einstein manifolds In this section, we collect a few facts on Kahler-Einstein manifolds (compare [3], [27]). The main theorem gives a characterization of Kahler manifolds of constant curvature in terms of Chern numbers. First we introduce Definition 2.10 We say that g is a Kahler-Einstein metric if there exists a real constant such tat Ric(g) = dwg, where Ric(y) is the Ricci form defined in any local coordinates as v-1 . + > Ride; A dz;. is A Kahler manifold (M,g) is Kahler-Einstein if g is a Kahler-Einstein metric. For simplicity, we will denote a Kahler metric by its Kabler form. Y ey Lemna 2.11 The average uf ihe scatar curvature: deperds vnty vn [wg) and 9) c(M). Proof. Given any point x € M, we can choose coordinates 21,...,zn such that at 2, we have Wg = yo Saenz; and Rie(g) = SI Se Rides 0 da.2.3. Uniformization of Kéhler-Einstein manifolds 19 Tt follows that. at x, Ric(g) Awe =(n— oS 1 n 8(wy wy. 7 Slwa)we ) YE Ra dey Adz, A+++ Adin Adin So we get rex(M)[wg]"~* = s . oO * fs F Proposition 2.12 If mc1(M) = lw] and w is a Kahler metric with constant scalar curvature, then w is Kéhler-Einstein. Proof. We know that + = Ric(w) represents the first Chern class of M and that it is a (1, 1)-form. So by the 09-Lemma, we have that Ric(w) — Ww = 03f. Now taking the trace of this, we have that s(w) —n\ = Af, but note now that by the above lemma, s(w) = 7, so f is a harmonic function on a compact manifold and therefore f is constant. o Theorem 2.13 Given any Kéhler-Einstein manifold (M,g). Then (M,w) ~ CP",C” or B” if and only if (FE Pecan ean?) Here by ~, we mean isometric up to scaling. Proof. By Theorem 1.10, it suffices to prove that the last equality is equivalent to M being a manifold of constant bisectional curvature, that is, there exists a constant, \ such that Risqr = \(94j9x8-+ 914943). Recall that, we defined the Chern classes by taking {t/t N 5 det (TSW) = 14 to(0) + Pda(O) +, where © is defined in last section. The 4, represent the Chern classes of M, in particular, ¢; (9) = tr represents the first Chern class. Viewing © as a matrix valued 2-form and using the properties of trace and determinant, we see that only? = 26x(9) = (4) (0.9),20 Chapter 2. Extremal Kahler metrics So we have that 2 (cx(M)? — 2e2(M))|u)"~? = (=) f (QAM) Aw"2, Denote by R°(g) the traceless part of the curvature R(g). Because M is Kahler- Einstein, we have Ric(g) = Aw, and so in local coordinates, we get Reet = Rijet — nat aso + Gii9K3)- The tensor R° measures how much the metric deviates from having constant bisectional curvature. So it now suffices to show that R° = 0 is equivalent to the equality in the statement of the theorem. If we denote by | —| the norm given by the metric, then by direct computations, we have 2 2 0 2 vol n-2, => QAL — aaa yeeros --( 5 ) trQAM) Aug? + Aap. It follows that the Chern number 2 1 A eu(M)? = Be) fog)? = fap Af yaa We also know that f, Mult = re (M)? Jug)" and therefore ; (aan? - 222g) o,)"# = - [R?PPwr. I M 8 So we have shown that M is a manifold of constant bisectional curvature if and only if tex(M)? ~ 2U* Dogan? = More generally, we have shown that stor any K&hler-Einstein manifolds, not nec- essarily of constant bisectional curvature, the following Chern number inequality holds (cf. [28]). (atan? - 22x) wl"? <0. o Note that if A = +1, we have that [wg] = +mc:(M), but if \ = 0, we have more freedom in choosing [wg].2.3. Uniformization of Kahler-Einstein manifolds 21 Remark 2.14 Any Kahler manifold M with c:(M) > 0 is simply connected. It follows from Yau’s solution of the Calabi conjecture (see Chapter 5) that there is always a Kahler metric w with positive Ricci curvature on M. It follows then from Kobayashi (see [13]) that M is simply connected. This is because Rie(w) > 0 implies that the fundamental group of M is finite. But for a compact Kahler manifold with positive first Chern class, we know that h?° = 0 for q=1 Defining the holomorphic Euler characteristic by x(M) = S77_9(-1)7h™ a compact Kéhler manifold with c)(M) > 0 has holomorphic Euler characteristic equal to 1. If M is the k-fold covering of M, then x(f) = kx(M), but because both M and M have positive first Chern class, they have the same holomorphic Euler characteristic and therefore we have a one-fold covering. This then implies that the fundamental group of M has no proper subgroups of finite index. Since the group is finite, this implies that M is simply connected. Remark 2.15 In general, if (M,9) is a compact Kahler manifold with the Kahler form wy and constant scalar curvature s(w,), then n+2 7 (ea(a)* ent) bol? < Sart Co a and the equality holds if and only if g is of constant biscctional curvature.Chapter 3 Calabi-Futaki invariants 3.1 Definition of Calabi-Futaki invariants In this section, we will introduce a holomorphic invariant for Kahler manifolds, which was first done by Futaki for manifolds with positive first Chern class and then by Calabi and Futaki for general Kahler manifolds. We will take a slightly different approach from the original one taken. Definition 3.1. We define the Kahler cone Ka(M) to be the set of all cohomology classes [w| € H?(M,R)M H'(M,C) which can be represented by a Kahler metric. Let. (IM) be the space of all holomorphic vector fields on M. That is, all sections X € T(TeM), which can locally he written as X = X*22-, where the X? are holomorphic functions on M. We will define a complex valued function fu on Ka(M) x n(M) in the following way: Given any (w] € Ka(M) and any X €n(M), pick a Kahler metric wy € [w]. Let hy be the function defined by so ¥ f 3(g)w” = Aghy. This is well defined since integrating the left-hand side of this expression over M equais 0 and jig is unique up to addition of a constant. Using results from the last section, we can define _ emer Mya}? = Ie stole Define the Calabi-Futaki invariant fry by sullehX) = ff XUry dus.24 Chapter 3. Calabi-Futaki invariants Next. we will prove that. the Calabi-Futaki invariant is well defined and does not depend on the particular representative we choose. We will need various steps to prove this. First note that because wy is a closed form and because X is holomorphic, we have that 5(ixw,) = 0. It follows from the Hodge Theorem that we can find a smooth function 6x and a harmonic 1-form a such that ixw, —a—d0x. Clearly, da: = 0 and we also have that d*a = 0, so we see that fale X) =f) X¢go3 and therefore, without loss of generality, we may assume that a = 0. So from now on, we have that ixw, — —90x. Defining the auxiliary function F(g,X)=(n+ pat [ hgAgAxwy, M it remains to show that. this is independent of g. Note that F' does not change if we replace 8x by 8x +c or hy by hg +c for any constant c. In the following lemma, we will prove that F can be written without any terms involving hy. Lemma 3.2 FX) =D sega, (Mate + Reta) + (20x +9)!" j=0 \ — (AgOx — Rie(g) + (n — 25)(0x + wey") =a fos buy Proof. In order to prove this, we need the following binomial identities 1 ift p 0 ifk C.28 Chapter 3. Calabi-Futaki invariants In particular, there is a Kahler metric in w] with constant scalar curvature only if frs(w],-) = 0. Proof. We only need to prove the last statement, but this is clear because constant scalar curvature implies that h, = constant and hence fy =0 o Remark 3.4 ‘he proof of Theorem 3.3 here is different from the original ones by either Calabi or Futaki. This proof here is not simpler, however, it is more general, for example, let Z be a hermitian bundle with a metric h over M and let ¢ be a symmetric polynomial of degree n + 1. Then [oes + R(h),...,0x | R(A)) is independent of h if 50x = —ixR(h), where R(h) denotes the curvature form of h. To see this, one needs the Bott-Chern classes and a good reference is [2]. Corollary 3.5 If fu(lw],—) = 0, then any extremal Kéhler metric g in. (w] has constant scalar curvature. Proof. If g is extremal, then X = g'Js(g);32 is a holomorphic vector field. So we get beh) = fi a otayGeeag = I, @(s(q) — weap == f (610) -Wahgus =~ f(a) ~ wag and therefore s(g) = 1. a Corollary 8.6 If X = |¥,Z] € In(M),n(M)], then frs(|w], X) = 0 and therefore fur: Ka(M) x n(M)/[n(M),n(M)] > C. Proof. Re(Z) generates a one parameter family of holomorphic transformations 4 So that $2@ is still a Kahler metric that lies in the same class as g. Hence we have that F(¢jg,Y) = F(g,Y) and differentiating this with respect to t, we see that F(g, [Y,Re(Z)]) = 0 and similarly F(g,[Y,Im(Z)]) = 0. This proves the claim. oO3.2, Localization formula for Calabi-Futaki invariants 29 3.2 Localization formula for Calabi-Futaki invariants In this section, we state without proof a localization formula proved in [11]. The proof of this formula is similar to that of Bott’s residue formula for Chern numbers. We will also compute the Calabi-Futaki invariants for three explicit Kahler manifolds. Definition 3.7 A holomorphic vector field X is non-degenerate if the zero set of X is the disjoint union of smooth connected complex submanifolds {Zy},ca and if at each z © Zp, the linear map DX :T,M/T.Z, > T-M/T.Zy is non-degenerate, that is, det(DX|r, m/7,z,) # 0- In the case that Z, = {z} and X = X'2-, we have DX(z) = (5) scsenl” and so the non-degeneracy here means that axt at (FE) sessee #0 Furthermore, if we are given a Kahler metric wy, we can make the following identification T.M/T.2 = Nujzys where Niyjz, is the normal bundle to Z, with respect to the metric wy, and hence DX|r.ajr.z, = (VX)"|Wagey Given any Kahler class © € Ka(M), we will define a “trace” tro(X):{Z,} > C which is only well defined up to addition of constants. Fixing wy with [wg] = , we may, as above, assume that ixw, = —00x and we set tra(X)(Za) = 0x(Zy)- Because X|z, = 0, we see 60x =U and hence @x|z, is constant. Kor a ditferent metric wy: in the same Kahler class ©, we have wy = w, + 00d, 50 Oy = 0x — X(6) +6,30 Chapter 3. Calabi-Futaki invariants where ¢ is a constant independent. of d. It follows that 9 (Zx) = Ox(Zy) +6, since X vanishes on Z,. Therefore, the trace is well defined up to addition of constants. In the special case that Q = we,(M), the form wy is the curvature of a hermitian metric on the line bundle L = A"T!°M and we have the induced vector field X* on A"T!°M, which is the canonical lifting of X on M. Atzé€Z),L,=2zxC. If we denote by € the coordinate on C, then we can write the vector field X* as where a = tr(DX)(z). So in this case, we have tre,(m)(X)(z) = tr(DX)(z) = ye (2) Theorem 3.8 For any (9,X) € Ka(M) x n(M) with X being non-degenerate, we have fu(Q,X) = 0" I (tr(La(X)) + er(M)) (tra X)(Za) +2)” — Hh (trol X) (Za) + pO det(Ly (x)+4 = K. a) AEAy 2% Here Ly(X) = (VX)*|ujz, and Ky is the curvature form of the induced metric on N|ayjz, bY Wy. If Q = 7e1(M), then this theorem is duc to Futaki [10] and reads tr (hal X) bao fuefer(M4), X) = mile det(Ly(X) + 42K)” We first of all need to explain why this expression is well defined. For simplicity, set for the moment, L(X) = 2, e1(M) = y and Y=, = u, then the above yields am aT ren whore F is of the form fey. Here both f and g are analytic functions, so we can write down a power series for F, F(a,y,u) = >> asjea'yeuk, fu3.2. Localization formula for Calabi-Futaki invariants 31 This is an infinite series, but because only terms of a certain degree contribute to the integral, this becomes a finite expression. We will not prove the theorem (see [22| for the proof), but we will con- sider two special cases and then compute the Calabi-Futaki invariant for certain examples. Case 1 Assume that X only has isolated zeroes, so dim (Z,) = 0 and hence the only terms that will contribute to the integral are the degree 0 terms and therefore a LAX) tra XY — hy tra XH fra(Q,X) = 9") det(ExX)) » and if @ = rey(M), then this reduces to tr(Ly(X))"*" n+1 4 det(Ly(X)) fu(a(M),X) = In particular, if M has a Kihler-Einstein metric, then we know that fry = 0 and therefore this puts a constraint on the zero set of any holomorphic vector fields. Case 2 Assume that M is a complex surface (dim M = 2), then A = Ag UA, where Aj = {A € A;dim Z, = i}. Assume that Q = 7¢)(M), then tr(La(X)) ful. x)= FY aaa Fle )(2ea{M)(s)+2=29(2a)). To see this, we will consider the two terms separately. The first term follows from the case above. To understand the second term, we must do a little bit more work. Because A, consists of 1-dimensional submanifolds, we can omit the trace and determinant, so we have f tr(Ly(X) + e1.(M))8 2 det(Ly(X) + 41 Ky) — f LEs(X)+ a(n) Zs (Ly(X) + Ka) -[ Ly(X)3 + 3c1(M)L(X)? 25 Lx(X)(1+ et KaLa(X)-) -[ (za? +0042.) (- = I. (sera) EX) _ Yl 1,00)32 Chapter 3. Calabi-Futaki invariants ~ La(X)(3eu(M)(Zx) = xManiza)) = L4(X)(2e1(M)(Zy) +2~29(2a)), where we have omitted all terms that are not of a degree so that they can contribute to the integral. Furthermore, we used the fact that J S286 = xz.) = 7M 2) ~ Zs) = 6) (BS) +2 — 202s) Example 3.9 Let M = CP", then Aut(M) = SL(n +1,C)/~, where A~ AA for some non-zero complex number 3, then (M) = Lie(Aut(M)) = sl(n +.1,C) = {A € C™t*"41 tr A = O}. Because any A € sl(n + 1,C) can be written as A = BC — CB, we have that n(M) = [n(M),n(M)| and so from Corollary 3.5, we know that fy = 0. Example 3.10 Let M be the blow up of CP? in a point. Without loss of gen- erality, we may assume that we blow up in [1 : 0 : 0] because we can always move points around by the automorphism group of CP*. Topologically, we see that M ~ CP?#CP2, where the bar denotes the reversed orientation. Now CP? \ {{1: 0: 0]} = M \ Z and in local coordinates, we get M=CP?\ {(1:0: 0} U {IE sn] x (Ls: y]s€y = nr}, where E = {[€ : n] x [1 : 0: 0]} ~ CP? is the exceptional divisor. Note that E has self intersection EE Claim n(M) = {X € n(CP?); X((1: 0: 0]) = 0}. Proof. Let X be a vector field on CP? which vanishes at [1 : 0 : 0]. Then X induces a one-parameter subgroup ¢; on CP? which fixes [1:0 : 0]. Clearly lifts to a one-parameter subgroup on M so that. 44, defines a holomorphic vector field Y on M. Conversely, let ¥ be a holomorphic vector field on M and ¢; be its integral flow of holomorphic automorphisms of M. Since E has self intersection —1 and other holomorphic curves always have positive self intersection numbers, the flow 4, must fix H and hence can be descended to a flow of automorphisms of CP?, these automorphisms must fix [1 : 0 : 0]. Then the derivative of ¢¢ along t gives rise to the required vector field on CP?. This vector field vanishes at [1:0: 0] a We will now consider the flow ¢ : [L: a: y] — [1 : ete : ety] which is defined on Up. We can define it on all of CP? by taking the limit of ¢; ({1: Ax: Ay]) as A — oo. Since ¢; fix [1 : 0: 0], they lift to a one-parameter subgroup of automorphisms of M, still denoted by gy. One can then see that Fix(@r) = CP3, UB,3.2. Localization formula for Calabi-Futaki invariants 33 where CPL = {[0: x: yl}. Let X be the vector field associated to @ on M. Then the zero set of X is CP, UE, which is the union of two one-dimensional submanifolds. It is not hard to sce that L,(X) = —1 on CP}, and L)(X) =1 on E (¢; flows out of E and towards infinity). Therefore, we can now read off that putea Mt), X) — F2er(M)(B) +2 2e1(W)(CPL) — 2) = a (cx(M)(E) — e1(M)(CP3,)) = FOB) + ENE x(CPL) — CPL, nN CPL) =$@-1-2-1)=-3 so far #0 and therefore M does not. admit a Kahler-Finstein metric. Example 3.11 Let M = CP?#2CP? be the blow up of CP? in two points. Again without loss of generality, we may assume that we have blown up in [1 : 0: 0] and (0: 1: 0] with exceptional divisors E, and Ep, respectively. Analogously to the above example, we have that n(M) = {X € n(CP*); X({1:0: 0]) = X([0: 1: 0]) = 0}. We will consider the same flow on CP? as that in the last example. Let X be its associated vector field on M. Let I be the line through [1 : 0 : 0] and (0: 1; 0], then J is preserved by the flow. The picture on the following page shows that Fix(¢:) = £, UFU {IN By}, where F is the image of CP, after blowing up in [0:1 : 0]. Now at {1.9 Ea}, we have 1 0 L,(X) = tr ( ) =0. So again we can read off that 1 1 qafir(er(M), X) = 5 (2er(M)(E1) ~ 2e1(M)(F)) 2 = 32-1 -(2-0))= because FOF = 0 as blowing up CP" in a point reduces the intersection number by one. And once more we can deduce that fy # 0, so M does not admit a Kahler-Einstein metric.34 Chapter 3. Calabi-Futaki invariants cP? cP? foo blow up oT Remark 3.12 Note that ¢(M) > 0 if M is the blow up of CP? in m points in general position for m < 8. If m > 4, we have that n(M) = {0} and if m = 3, n(M) ~ C® and one can check that fy = 0. In fact, these M admit Kahler-Einstein metrics (see [23]).Chapter 4 Scalar curvature as a moment map Let (V,w) be a simply connected symplectic manifold and let G be a group acting on V preserving the symplectic form. Let g be the Lie algcbra of G which consists of all left-invariant vector fields on G. Then any v € g induces a one-parameter subgroup {d:} of G. Since G acts on V,@: induces a vector field X, on V. It is well known that there exists a map m, called moment map, m:V —g", satisfying — m is G-equivariant with respect to the co-adjoint action on g*, ~ for all v € g and all u € TV, w(u, Xp) = dm(u)(v). Our goal in this section is to prove that the scalar curvature of Kahler metrics is a moment map with respect to Symp(M), the group of symplectomorphisms of a compact Kahler manifold M. I learned this from [6]. In [8], Fijiki and Schumacher also studied moduli spaces of Kihler manifolds as quotients by symplectic diffeomorphism groups. Our presentation here follows closely the discussion by $.K. Donaldson in [6]. We should point out that though we as- sume for simplicity that M is simply-connected and Kiihlcr in the following discussions, all calculations are still valid with small modifications in the gon- eral case of symplectic manifolds. Also it is not necessary for the readers to go through this section in order to understand other scctions. Let M be a simply-connected, compact Kahler manifold and fix a Kahler I ={I:TM + TM; (Ju, Jv) = w(u,v), w(u, Ju) > 0 for u# 0} and let Jint C J be defined as Tmt = {J € J; J isintegrable}, which are both infinite-dimensional manifolds once appropriate norms are given. However, we will skip these technical difficultics. Let G be the group of exact36 Chapter 4, Scalar curvature as a moment map symplectomorphisms of w, G acts on .7 by (4,0) > b.Jo.” which still lies in J. Note that for @ € G, we get that djw = w, so if X is the vector field induced by ¢;, then £xw = 0 and by the Cartan formula for the Lie derivative, this implies that d(ixw) = 0. Therefore, we can find a function Hx such that ixw = dHx. By requiring that f,, Hxw” =0, such an Hy is unique, so the Lie algebra of G is ox (HEM oR | Hu" 0} = OF (a, Secondly, observe that TyJ ={A:TM > TM; AJ + JA = 0,w(JAu, v) + w(u, JAv) = 0}, in other words, JA is gj-symmetric, where g, is the metric induced by J, that is a(t, ¥) = w(u, Jv). Given A € Ty J, we can define j1a(u,v) = w(Au,v) and we see that pa is anti J-invariant, that is 4(Ju, Jv) = —p(u,v), and symmetric. The tangent space TJ can be characterized to be the space of all such j14’s. Finally for the orbit G(J) = {$.Jb1; ¢ € G}, we have T)G(J) = {LxJ;X € g} Let now D be the distribution given by Dy = {Lx J Lyx J;X € gh. Remark 4.1 In what follows we will always assume that the complex structure is integrable. We can also prove everything for arbitrary almost complex structures (see [6]), but we then need to take the Nijenhuis tensor into account and hence for simplicity we assume integrability. Claim Dy C TJ. Proof. It is clear that both LxJ and LyxJ satisfy AJ + JA = 0. Because X €@ it is also clear that LJ satisfies the second condition to lie in Ty, J, and hence it suffices to show that Lyx J does so, too. Now we have (Lux d)¥ = Lyx(JY) — JL£IxY = [JX, JY] — J[JX,Y] = [X,Y] + JIX, JY] =LxY + ILx(JY) =-Ly(J?Y) + JLx(JY) =—-(LxJ)(JY).Chapter 4, Scalar curvature as a moment map 37 It shows that (CrxJ)¥ = —(LxJ)(JY) and therefore, we have w(J(LyxJ)u, v) + w(u, I(LyxJ)v) = -w(J(LxJ) Iu, v) — wu, J(LxJ) Jv) = w((LxJ) Ju, Jv) + (Ju, (Lx J)J0) -0, so LyxJ ET J. o Remark 4.2 This also shows that Dy is a holomorphic distribution, this means that it is invariant with respect to the almost complex structure J on Ty J induced by J. Proposition 4.3 D is integrable. Proof. Fix J € Jint and let K = {Kahler metrics within a fixed Kahler class}. It suffices to construct an integral submanifold 7 of D through each Kahler metric w in K. Any Kahler metric in K near w is of the form we =w —dJdf =w + 2V—100f for some small f € C§°(M). ‘We now define a bundle S over K as follows: The fiber Sj over wy consists of all exact symplectomorphisms of wy. Note that cach wy is J-invariant and So=G. By a lemma of J. Moser, we can always find a diffeomorphism ws such that wyws =w, and so given @ € Sy we have that wy dubs € So. Define the function I : C3°(M) x So — J which associates to the pair (f,6) € C§°(M) x So the complex structure ¢(¥,1(J)). We must check that this is well defined opp (Ju, by (Jv) = W((UF")eT(Wp)e, (U7 ")eT(¥y)-0) = we (I(hp) att, T(hp)v) = ws ((by) eu, (p)-v) = w(t2) and since ¢ is a symplectomorphism, this shows that I’ is well defined. At a given point (f,4) € Cg°(M) x So, we will compute the derivative of J applied to (H), Hz). Here we use the exponential to identify the symplectomorphisms38 Chapter 4. Scalar curvature as a moment map with the Hamiltonian functions D1\¢5,9)(Hi, Ha) = 5 (contr Wyte, wn) 0 a tHe 1 tHe »-1 = (lee lop deTbraen)ees™O) =b (Slvr Jellvpsun)e) 9" + GLK Hg (Up TW ped" We will consider both terms separately. First note that OL ng Wp IV G=)Os? = Lo. xXug (OF): we now want to find a nice expression for @.X 1, 80 i4.xn,0(Y) = w(b-XuzY) = ¢'w(Xu,,0,'Y) =u(Xi,9;1¥) = dih(9;"Y) = o5'(Y)(H2) =¥((6") Ha) = d(Hz06"')(Y) and therefore beX Hy = Xtrp0p-- For the first term, we need to do a bit more work. Observe that Sophia eI bpst de = L0G odsol Up! FOr OR (pate = Lx, (UF) bye), where X; is the vector field ar” . ai (U7 Meee ) Again we want to find a nice expression for X1. This will show that the distii- bution is integrable. First of all t=0 (OF spun) e = Vener = Uf pun, (ws — td IdM) + td) =o+tbF yn, ddA.Chapter 4. Scalar curvature as a moment map 39 Since we can ignore sccond order terms in t, we have Lx,w = vpdJdHy = d(w7"(J))dvz Ah. As Lx,w = d(ix,w) and since M is simply connected, we get that ixyw =v; "(J)dvpAi + dh for some h. So we sce that on the one hand, w(X1 — Xn ¥) = 9X1 — Xn v7 (IY), where g is the metric corresponding to the complex structure ;'(J), on the other hand, w(Xi — Xn, ¥) = oF (J)(dbpth)(Y) = dys Hi (wy (J)Y) = 07 (JY (vp) = 9 V0}, YF (DY) and hence X; = Vv} Hi + Xn. It follows Dip) Hs He) = 64 Coojaiaxs (67 NID) ) 65" + Ex yoga (800740) which reads by the same argument as above DU\c49) (His Ha) = Lou ttcg-r 4%, 5-1 (Vj AI) + Lx ygayea (OCUp CIN: This shows that DI, g,¢)(Has Ha) € Deyn) and that DZ is surjective, hence the distribution is integrable. a Consider now the following two operators P:C03(M) > Ti J, Q:T7J > CP(M), where P represents the infinitesimal action of g on Jint and Q represents the derivative of the map that associates to J € Jin the scalar curvature of the metric gy induced by J. To achieve the goal that we stated at the beginning of this section, to relate the scalar curvature of a Kahler metric to the moment: map, we will prove that the following two L?-pairings are the same: (PCH), Ju) = (H,Q(H))-40 Chapter 4. Scalar curvature as a moment map Here H € C§°(M) and yp € T;.7. The pairing on the left-hand side equals 1 5 (PUD tas”, where we have used the inner product defined by gy, and the pairing on the right-hand side is the standard L? inner product on functions, that is [Hower In order to prove that the two pairings are the same, we will compute both. Let us first of all consider P. Since P is linear in H, we can assume that H has small compact support and so we can do the calculation in a local coordinate chart. Choose local coordinates 21,...,t2n compatible with J so that J = (Ji), 9d = (ik) = (wig Jp) aud PH) = (94 = PONG gee = da,’ Then we have on the other hand, a 0 ’ # (Xn ze) =o (1x03) = aj IXu)*, ad aH Xi, = digi —. = Shah Writing for the moment a LxyI = Big de, we see that : a\_ a a bani eed (ag) =e (795) 7B and 50 ; Bi = — Jf Ifa!” Hye — a"? Hs. Therefore, PH) = (-2}4ta 7 JH.)Chapter 4. Scalar curvature as a moment map 41 so that the pairing on Tj Jin reads (PUD. I) == f Hamageo" = — f Hy mao'os Now we will compute Q. Since Q(x) is a tensor, there are no problems with the choice of coordinates. Let J; be a family of almost complex structures with Jo = J and $Jr|e=0 = py. Let gt be the metric induced by J;. Then the Christoffel symbols of the metric g are as follows, ‘i 80015, WGiik — Ouse ri 8Gnte _ B95 ‘jk = 50 (Fes Oe, aa1 ) where (gi?) is the inverse of (gt,:3) and oo G09 ~ 9 \ Ba. Be, ) * At p © M, we can assume that as(P)= 65, day(p) =0 and Jte)=(7 9!) ee tants ean, Stee where g = go. Then T% ,, are of order ¢, and the curvature of g: at p is given by 7 o Oo 8@ Sy oO oo Riyal) = 0( Vp, Vate gor ~ Vat; Vater des) or} ory = Gat Hee — Het 2, Lik + Olt) Poni — Poet vy dx,dx d2,dx, | O¢» so og a. > — pik git (Pose Fou 2 (9) = 91° 9 (Fae wa) + O(2?). Differentiating it on t, we get d a 090) = —2tak Rig + Hakka — jg.kk- and Q(t) = %9(9)|t=0- Claim 1, Rijnj = 0 and pj; = 0. Proof. The first expression reads ix Rijkj = pik Ricix and since p is anti J- invariant and Ric is J-invariant, we can deduce that (yu, Ric) = 0. Furthermore, p being anti J-invariant implies that the trace of uz vanishes, so jj =0. 042 Chapter 4. Scalar curvature as a moment map The corollary tells us that Q(u) = s4jz,43 and hence we have the following pairing wb pet QW) = f HQ." Sf Hyrs sn”. Comparing this to the above pairing for P, we have shown that the two pairings are equal and we have found a relation between scalar curvature and moment maps. To be more precise, we have shown that the Calabi functional equals the square of the norm of the moment map m for the symplectic group G defined by scalar curvature as above. So we can deduce the following Corollary 4.4 9, is an extremal metric if and only if c(M)w"? ) Jem*( = and therefore nT {extremal metrics} /holomorphic isometries = m~1 (er) /G. wwChapter 5 Kahler-Einstein metrics with non-positive scalar curvature 5.1 The Calabi-Yau Theorem We have seen that the Ricci curvature represents the first Chern class. In this section, we will consider the converse problem, namely, given a Kahler class [w] ¢ H?(M,R) 9 H)1(M,C) on a compact Kahler manifold M and any form Q representing the first Chern class, can we find a metric w € [w] such that Ric(w) = ? This is known as the Calabi conjecture and it was solved by Yau in 1976. We will state it here as a theorem and refer to it as the Calabi-Yau Theorem. Theorem 5.1 (Calabi-Yau) Let M be a compact Kéhler manifold and let |w] € H?(M,R)U H™"1(M,C). Given any form Q representing ™c1(M), there exists a unique Kahler metric w € {w] such that Ric(w) = Q. Before proving this theorem, we will first discuss some corollaries and an exam- ple. Corollary 5.2 Any compact Kahler manifold with c1(M)x = ¢2(M) = 0 is flat (thie fe — Cnr /P). (this MCN /P). Proof. The Calabi-Yau Theorem tells us that ci(M) = 0 implies that there cxists a Ricci-flat metric (that is Ric(w) = @ = 0). By the Uniformization ‘Theorem (2.13), we know that Af = C” and therefore M = C"/T. o This shows that the flatness is characterized by the first two Chern classes. Corollary 5.3 If c(M) > 0, then M has a Kahler metric with positive Ricci curvature (this implies that M is simply connected, see [13]).44 Chapter 5. Kéhler-Einstein metrics with non-positive scalar curvature Proof. If (M) > 0, then there is a form © > 0 representing 1(M) and by the Calabi-Yau Theorem, we can find w such that 9 = Ric(w) > 0. o Remark 5.4 It is unknown whether this result is also true for symplectic mani- folds. Corollary 5.5 Assume that c:(M) = 0. Given any Kahler class |w], then (~]} Aut(M, {w]) = {0 : M — M; ois biholomorphic and o* is a finite-dimensional Lie group. Proof. By the Calabi-Yau Theorem, there exists a unique Ricci-flat metric w € [w]. Observe that o*w is still Ricci-flat and is also Kahler (since o is biholomorphic). By assumption [o*w] = 0° fw] = [2] and therefore, by uniqueness, ow=w. So g is contained in the isometry group Isom(w) of w and we have shown that Aut(M, [w]) = Isom(w) and the latter one is a finite-dimensional Lie group (see {14}). o Remark 5.6 There exists a compact K3 surface M (that is a complex surface with 7(M) = 0,c:(M) = 0) such that Aut(M) is an infinite discrete group. Hence, the extra assumption o*[w| = [w] is necessary. Example 5.7 Let M ¢c CP* bg defined by a quartic (that is degree 4) homo- geneous polynomial. Example 2.9 tells us that ¢,(M) = 0 and it follows from the Lefschetz Hyperplane Theorom that ™1(M) = 0. The Calabi-Yau Theorem says that M has a Ricci-flat metric, but M is not flat (if M were flat then M = C*/T and hence m(M) =T). A question that arises is whether it is possible to construct this metric in terms of classical functions. We will now give the proof of the Calabi conjecture due to Yau (see [27]). Proof. Choose any Kahler metric w € {w]. For convenience, we will drop the in w. Then, in local complex coordinates 21,..-2ny normalization factor w = gyda A dz; and _ Ric(w) = —O8 log det(g;3).5.1, The Calabi-Yau Theorem 45 Both the Ricci curvature and ©. represent the first Chern class and therefore the 08-lemma tells us that we can find f such that 2 Rie( where f is unique after normalizing to Le —1)w" =0. Note that f only depends on w and Q. ‘Again using the 0d-lemma, we know that any other metric in [w] is of the form w + 90¢. Suppose that we have found the right function ¢, so Ric(w + 844) = 2 = Rie(w) — a5f. This reads in local coordinates, ao Blog det (6 + 3205 ) = —00 log det(g;3) — 00 f. Although this is only locally defined, the following is globally defined det(gi3 + woos ) aDtos( age ) = 20. Therefore . det(9i5 + 32.55) efte det(9,3) which is equivalent to _ (w +059)" = ef a". In order to determine c, observe that [ fteqn = f (4 Aap)" = fumca }= f ao, Ju Ju iM Ju which follows from the Stokes Theorem. This implies that ¢ = 0. So we have shown that Ric(w + 06¢) = © implies (w+ 089)" = eu", which is known as the complex Monge-Ampére equation. In fact, it is true that Ric(w + 034) = <=> (w+ 009)" =efu™.46 Chapter 5. Kahler-Einstein metrics with non-positive scalar curvature Calabi proved in the 50’s the uniqueness part of his conjecture using the Maximum Principle. Assume that there are two metrics w, and w2 with Ric(w,) = Ric(w2) = Q, then w, = wt 09g, and. wy =w + 09g. Without, loss of generality we can assume that ¢2 = 0 , w2 =w and ¢1 = $ so that we have Ow ut =w" — (w + 056)" = (w— (w + 54)) A (uw? 4? Nay bee tw Aw? 4 wR) = —0$ A (ww! +"? Aw +--+ WP?) and multiplying by ¢ and integrating, we see that 0 -| GOI A (oP) 4 tea }) =| AGN ABA (WA? 4-0 be) IM IM Now because w; and w are Kahler metrics and w;,w2 > 0, we get that 1 Fy n—1 ni) 1 5 n-1 7 OPA DG A (w' te tut ")> 7 OO A 06 Aw > where V = fy, w”. Here we say that two top degree forms u = fdzi A--- Adin and ti = fdzy A--- Adi, satisfy u > a if f > f. So we have 1 ar 2y0 1 2m 0> Ff 26n86n~ = ay [are = py [vere and we can deduce that V¢ = 0, so ¢ = constant and therefore w) = w2. Hence we have proved the uniqueness part of the theorem and we now con- sider the existence part, so we want to solve (w +0¢)" = efw” for ¢. The proof will follow by the continuity method: Define f, = sf +c, for c, constants and 0 R;uisC*-smooth, sup 2 %@?— dy d(x, y)? [D*u(x) — D*u(y)] d(c,y)t We denote by Ci’ (M,R) the subspace consisting of all u in Cv (M,R) with the extra linear condition : [ two? = 0, k loll, = sup [D'v(2)| + sup Soe a4y48 Chapter 5. Kahler-Einstein metrics with non-positive scalar curvature whereas C’ (M, R)o denotes the subset of all u in C’" (MR) satisfying the extra non-linear condition fe =u? = 0. Note that fore —1u? = foe = elu" = [we =1)-(e = 1))u" =0 implies that o= / (essa an 1) ot = fv. +05)" —u2. By the above, it follows that we can define an operator @: 024(M,R) + C°3(M,R)o by 7 (w. + 0d)" (p) = log a This is well defined. We must prove that for jfé — frllo,4 sufficiently small, we can find a } with OY) = fe- fe and [Idlla.g < CIF — Sello,3- We will apply the Implicit Function Theorem and hence it suffices to check that 1 : D®|yao : Cy’? (M,R) > C3? (M,R) is invertible (note that the tangent space to C4 (M,R)o at u = 0 is Ch’? (M, R)). Now DA|gao(u) = Ayu and it follows from [12] that Au, 3 (M,R) = 08 (MR) is invertible. Therefore, (x) = f; — f is solvable for |t — s| sufficiently small and so we have shown that $ is open. It remains to prove that S is closed, that is, suppose that we have a sequence si; € S with lim;... 8; = Soo, then we must show that s., € S. Now the s; correspond to solutions #; of (1,,) which are unique up to constants. We must. prove that by taking a subsequence if necessary, there exist c; such that $; — c; converge to some goo in the C?:3-topology. Since (w + 08(¢; — 61)" = ew", we know that _ (© + 0500)” = ew”5.1. The Calabi-Yau Theorem 49 and hence. 85, € $. By the Arzela-Ascoli lemma, it suffices to show the following a priori estimate ||4; —ci{]3 < C. Let ¢ be any solution of (*),. We will establish an a priori estimate for ||¢||3- We will first state the following two facts that can be found in [1] — Let g be the metric corresponding to w = wg, then there exists a Green function G(x,y) on M x M \ A satisfying ¢ < ————, OS Glew) < Tey and for all ¢ € C%(M), =7 [oemred— 7 ff Aootnee. nor). ‘To see that we have positivity follows from the fact that we can always add a sufficiently large constant to G to ensure positivity. ~ (Sobolev inequality) For a fixed metric, there exist constants c1,c2 > 0 de- pending on (M,w) such that for all f € C1(M,R) a(} [uiter) 8 f ype sd f iwsper This is trne because on a compact: manifold, we can use a partition of unity to cut the manifold in pieces which are diffeomorphic to a domain in Euclidean space and there we can apply the standard Sobolev inequality. To show the desired estimate, we will procecd in three steps. In what follows we will always use C to denote a uniform constant, but this capital may mean many different constants and it should be clear from the context what is meant. Step 1 (C°-estimate) We know that ¢ is unique up to a constant, so we can choose the constant so that supj,¢ = —1 and because M is compact, we can assume that ¢ attains its supremum at zo so that we have -1= 600) =-7 flower F [ A6cnaCao, ye" where we have used that @ is always negative so @ = —|gJ. Since w + 036 determines a metric, by taking trace with respect to w, we have O 1 and w’ = w — 00¢-_, then we see as before (ef = 1) =(w')" — =~ BBG A (WY + WY? Nw te tu Aw? +0") where both w and w/ are metrics so w,w! > 0. For any p > 1, we multiply it by @”., and integrate over M to get i 5 NNM=1 gee pga 7 [2080-0 feta) 1 5, ae > [08 8 86-nw a2 1 5 o “fe 06 N5b— Aw? “Phe oF 04_Ag 7? Fb Aw? - at - wie 5 [oo A b0" 4p 2 =? | we Pum mere |! 2 Po 4 (Lf ig erun) * —@ fig Pun 2 sctmlo(t Loh) 9 [wt On the other hand, we have that -7 [9080 A (Wt $e teh) = zl, & (ef — Iu” 1. Putting all these together, we get wpm (a(y fee yep ie pan) <5 f een5.1. The Calabi- Yau Theorem 51 which after some rearranging reads (5 L eh =u") 7 << SOE) f apttyan, By raising both sides to the power a we obtain ld-lleorn,ar S (C(p + 1))74 [oles We will apply Moser’s iteration method. Choose po = 1 and define p; by ptl= 7 (p+), then isl Ue-linoes < [] (Clos +1) lols. i And if we now let i — oo, then p; + 1 — 00, so sup|¢-| = jim, Ilo-Ilbeva S Tce: +1) |l6-llnz < 00, j=0 which we can see by considering ele RolCOs4)PFE _ C%o szbyr(lowC Howley +0), So we have shown that sup |¢| = sup |¢_| < Clldl|z2- M M Furthermore, we have & [towne $f oa eur = 7 [oe - (o+ 086") av f,IvoPo" e(Ler-(L~)) where the last step follows from the Poincaré inequality, so we deduce from the above that Iv } lidllz2 < C(lidllzs + 1) and hence supyy |4| < C.52 Chapter 5. Kéhler-Einstein metrics with non-positive scalar curvature Step 2 (C?-estimate) First observe that, ||V?@l|co < max{n + Ad,n}. To see 2 this, note that the metric satisfies gz + ae; > 0 (positive definite, hence all the eigenvalues are positive) and therefore 4 Pe _, Fo \_ lls + 3es03; 9 and the above transforins lo Los M08) = para yg boutoe + AP + Ras (14 + AH). Set now C= infizs Rijug and observe that ba (Pee = 98)? Ras (- Ra “ ( T+ a) 2+ ba) + ben) 5 CL + bez - 1 60)? = 2 T+ 6a + ux) =o(it ea _ H(i :); dizn digg + OF + aq +40) > ; 4, _- ). this yields 1 1 (49)? Ty gpd Foe) ‘We need to apply one more trick to obtain the requested estimates. Namely, Al(e™?(n + Ad)) = e °F A (Ag) + 2V'e°Y"(n + Ads) + Ale) (n + Ad) =e A(Ag) — Ae (g')b:(AG); — Ae >#(9')"4:( Ad): = AePPA'G(n + AG) +e ™*(g')dig(n + Ad) > e#A'(AG) — €*(g')(n + AG)*(Ad)i(Ad)z de#A'9(n + Ad), which follows from the Schwarz Lemma applied to the middle two terms. We will write out one term here, the other goes in an analagous way, (re Bgi(n + AG)4)(e°2°(A9);(n + A9)-4) AF + C(n + Ad) 1+ Fe Putting all these together, we have Te A(e*(n + Ad) >e* (ar + O(n + Md) om -) —de~#A'o(n + Ad). Al(e*?(n + Ag) > —ce** — ce (n + Ad) +e Dr tn + Ae), where we have used the fact that ' oe i N= ao errs5.2, Kahler-Einstein metrics for manifolds with c,(M) <0 55 Now let us notice the following inequality 1 Fata) a \ >a = wT wet Divas 2(Hase erring Aa), which can be verified by taking the (n — 1)-th power of both sides. So the last term in the above can be estimated by MD ean td ze mre Ae Mn+ Ag) Setting now u = e~*¢(n+Ag) and recalling that ¢ < —1 and hence e~** > 1,we have finally obtained the following estimate Au > —c — cu + cgu™t. Assume that u achieves its maximum at zo, then at this point, A’u < 0 and therefore the maximum principle gives us an upper bound u(z9) < C which in its turns gives 0< (n+A¢)(2) < u(x) < C and hence we found a C?-estimate of ¢. Step 3 The final step will not be performed here, it is similar to step 2, but the argument will be longer and can be found in the appendix of [27]. This last step will give a bound for ||V°¢l|co. With this bound, we are done. This last bound will namely give us the equicontinuity of ¢; and together with the a priori estimate ||¢; — ci||3 < C’, the Arzela-Ascoli lemma now tells us that S is compact and in particular that S is closed. a 5.2 Kahler-Einstein metrics for manifolds with c:(M) < 0 In this section, we will tirst state a theorem proved independently by Aubin and Yau. Its proof is easier than that of the Calabi-Yau theorem in the last section. Then we will give a simple application of Kahler-Einstein metrics. We will end this section with an open question, which is a generalization of the application to the symplectic manifolds. Theorem 5.8 (Aubin, Yau) Let M be any compact Kéhler manifold with c1(M) <0, then there exists a unique Kéhler-Einstein metric w with Ric(w) = -w. Remark 5.9 The statement of this theorem is equivalent to the solvability of the following Monge-Ampére equation (wo + 059)" =e™oteu2 where Duy = Ric(wo) + wo. The plus sign here makes life a lot easier. In fact, we can use the Maximum Principle to get a priori C°-estimates for solutions of the above equation.56 Chapter 5. Kahler-Einstein metrics with non-positive scalar curvature Application (Yau, [28]) There exists a unique complex surface M homotopic to cP?, Proof. We already have the following information about the Euler class and the signature, both topological invariants xia) = [ela = CP) =3 7(0) = F(ou(M)? ~ 2ox(M)) = r(CP*) = 1, which follows from the Hirzebruch-signature formula and which implies that o(M)? = e(CP*)2 9. Since H?(CP?,Z) = Z, we have that H?(M,Z) = Z. Fix a generator w > 0 of H?(M,Z) so that we have c:(M) = Aw and since w* = 1, we have that \? = 9 so A= +3. Tf \ = 3, then c,(M) > 0 and this implies that there exists a holomorphic S? C M which implies in its turn that M = CP?. For this last implication see Chapter 4 of [11]. If \ = —3, then c(M) < 0 and there exists a Kahler-instein metric wx p. Now 3x(CP2) = 3c2(M) = c:(M)? and by the uniformization theorem, we deduce that M = D?/T, so ™(M) =T # {id}, which gives us a contradiction, so only \ = 3 is possible and that proves the application. u One is now automatically led to the following generalization: Does there exist a unique symplectic surface M homotopic to CP?? Analogously to the above, we can deal with the case \ = 3, but it is not clear how to show that A = —3 leads to a contradiction. Taubes showed that one can use the Seiberg-Witten theory to deal with the case that M is diffeomorphic to CP?, but the general case is still open.Chapter 6 Kahler-Einstein metrics with positive scalar curvature In this chapter, we will study Kahler-Einstein manifolds of positive scalar cur- vature. We will assume that M is a compact manifold and that c(M) > 0. Because of the latter, we can choose a metric w with w] = 7¢,(M). Then Ric(w) = w + Odh,,, where we can choose h = h,, in such a way that fe = lw" =0. Suppose now thal we can deform w within its Kabler class, so Unal we can find an we =w t+ 006 which is Kahler-Einstein, then we have the following w + 006 = Ric(w + 809) = —O0 log det(w + 0¢) —aB log + 280)" ue + Ric(w) = 00 10g +289 45 6h. w This leads us to (w+ 809)" = (*) Note that — log det(w + 88¢) is not globally defined, but by taking a quotient as above, we have a well-defined notion again. We have seen in Chapter 3 that this equation is in general not solvable and that for example the non-vanishing of the Calabi-Futaki invariants forms an58 Chapter 6. Kahler-Einstein metrics with positive scalar curvature obstruction to the solvability of (+). It is, however, clear that the solvability of (+) is equivalent to the existence of a Kihler-Einstein metric on M. The goal of this chapter is to provide sufficient and necessary conditions for the existence of Kahler-Einstein metrics on manifolds with positive first Chern class. 6.1 A variational approach In this section, we will take a variational approach and find a functional for which (*) is the Euler-Lagrange equation. First let us discuss some interesting functionals. Definition 6.1 Let w be any Kahler metric. Then the generalized energy is given by i+] - 5, ip yn ini Ja($) = FEE [on dens Aws where V = fypw" = [w]"([M]) and wy = w + 050. To illustrate this definition, let us compute J.,(¢) for n = 1,2. Ifn = 1, 1 a 1 Jul) = 557 [ 20080= 5p [ oie. It follows that J.,(¢) is positive on any non-constant function. In the case that n= 2, we have 1 a 2 < Jul) = gi f OND AUG + Fe ff DOnIOAw. At any point of M, we can choose local coordinates z;,.-. ,2n such that at z, we have w = Y>dz; Adz. Because w is positive definite, we can diagonalize w and 8¢ simultaneously so that $;; = $75; at z. Now the integrand of the first integral reads locally 89 NOG A wy = (Hides) A (Gadi) A (CL + aa)dei A dzi) 2 = (ldi?@ + 40) + ooPO + ud) Again we see that J,,(¢) is positive, if @ is not constant and w + 00¢ is positive. In general, whenever we is positive definite, J.,(¢) > 0. This can be easily secn from the above definition. Lemma 6.2 Let {¢,} be a smooth family of functions, then 4 Lléolen o=-y fa blu where $= do and d= S¢u|t=0.6.1. A variational approach 59 Proof. The proof is just a simple calculation, nl, d _ 1S itl sab Awt Anni! qlee) =-F dantl Jf, (24006 6.0 Aw + (n= 1 1)6006 Aw’ Awe? A 008) Lisl i90. i, ytoicl my Saat |, 2600n~ Awt n=l -Z fH in-i-1) [5005 Nut nuZr* A (a + 080—w) 0 J n=l. _ 1 i+l . Lan ‘ n-1-i any eri) | 5889 Aw, n=l. Lidl . SoAm A wit! A nic? Qa. rin i=) f $086 00% Awk 1oitt . i006 Awi Awt Ini =-5 OD f $089.00 Aum nol ai ana [ 006 Nwi Nw t Vojntl iM ¥ =i 5006 A wh? ty) b0OG Aw Awh-t} =-7 MW? bp Nw Vey MW? pAwt Aw pe = bs Ad Nw A jn-i-1 vo [686 wi Awe pec =-F G(009 + w —w) wt Awe vol, ( ) 5 iw yet =i oi Aw® tg b stl pgm it = 7X found +E he Aw 1 : <7 f sera. o If we define 1 d 1 + F6)= 10) -F [ono then FAG ao =F buh. This functional has many nice properties. Let us list some of them.60 Chapter 6. Kahler-Einstein metrics with positive scalar curvature Tet. us choose another metric, say w’ = w + Ody in the same Kahler class. In that case, we have we = w + 009 = w! + OOo — YW) = wh, _y. We will now see how F2(¢) depends on the choice of representatives of [w]. Claim 1 For w and w’ as above, we have FO(9) — FO(6 — 4) = FAY) and if we set FO(¢) = F°(w,w), this yields the following cocycle relation F(w,w9) — F® (wy, wg) = F(w,wy)- Proof. Consider now the following difference FL(0) ~ F(@~ $), where ~ is fixed. The above lemma gives us the following 4 (860) — F.(e-¥)) =0 for any family {,}. Therefore the difference is constant, in particular, choosing =, we see that this constant equals F2(q) and the claim is proved. o Let P(M,w) be the following space of functions {p € C*(M);w + 896 > 0}. Claim 2 The functional F° is convex on P(M,w), this means that for any 1, ¢2 in P(M,w), 1 PYG (1 + 2) <5 (Fon) + 2) - Proof. Put w’ = W4(d.4¢2)* Using Claim 1, we can deduce 1 1 FEH61 +4) 3 (HSU) + F063) =} (FaGler ~ on) + £2G 2 - &)) 1 1 yi “i ynet Tea LV [Hoe 6h 0) 1 1 . ‘ : _ ~ Fe ET Ley fi, 2101 68) Os ~ a) Re Ay aia 5=0 <0, the claim is proved. a6.1. A variational approach 61 The next two claims are obvious. Claim 3 For any constant \ > 0, F2,(A¢) = AF2(¢). Claim 4 If 7: M’ ++ M is a branched covering, then for any ¢ € P(M,w), Frey(n*¢) = F(9). ‘The functional, which has (*) as its Euler-Lagrange equation, is given as follows: FA(@) = FO) ~ oe f, oneut), Remark 6.3 If n = 1, (+) reads 1+A¢ = e*~¢. By the classical uniformization theorem for Riemann surfaces, M = $?, so we get 1 1 1 Ful0) = 5p [00° G [Ld —low gy [eee This is exactly the functional in L. Nirenberg’s problem of prescribing the Gauss curvature equation on $?. We also have the cocycle condition for F,: for w! =w + Ody), we have Fal) — Far( — h) = F(ab). To see this, we first observe that F..(¢) = F.(¢+C) for any constant C. So we can always normalize $ such that 1 heb, yn Bf octane Similarly we can choose 7% such that 7 feeten =n. Then F.(¢) = Fi(¢) and Fw) = FSW). Because Ric(w’) = —A0 log(w’)" = w! + ODhw, Ric(w) = —O0 logw” =w + 05h,62 Chapter 6. Kahler-Einstein metrics with positive scalar curvature we gct after subtracting the bottom from the top equation that wy we 20 oe ( ) = BIW + hur — hs), which then implies that 1 hur (O-W) ayn 2 f rb yn te! ws =1. | é! (w') ee Fu) Fu(@— 0) = F(¢~ 0), so the cocycle condition for F follows from Claim 1. Then Remark 6.4 The functional F,, is only the difference of two convex functionals, while in the case of c1(M) < 0, the corresponding functional is the sum of F® and another convex functional, so it is always convex. 6.2 Existence of Kahler-Einstein metrics In this section, we will prove an analytic criterion for the existence of Kiahler- Einstein metrics on Kahler manifolds with positive first Chern class. Definition 6.5 We suy F, is bounded from below if there exists ¢ = cw) > 0 such that F.(¢) 2 -e. F, is proper on P(M,w) if there exists an increasing function Bi R — [e(w), 00) satisfying limy_..o j(t) = 00, such that for any ¢ € P(M,w), F..(¢) 2 (Ju($))- F, is independent of w. More precisely, if w’ is another metric, then F,(¢) > H(Ju(¢)) implies Far(¢!) > !(Jur(¢"), where w! =w+ ddv, J =o-y and p! = —c for some constant ¢ depending only on w aud w’. Theorem 6.7 Assume that M has no non-trivial holomorphic vector fields, then M has a Kahler-Binstein metric if and only if F., is proper on P(M,w).6.2. Existence of Kahler-Einstein metrics 63 We also have the following generalization of the above theorom. We denote by Aut(M) the automorphism group of M. Let G be any maximal compact subgroup of Aut(M). If w is a G-invariant Kahler metric, then we define Pe(M,w) = {¢ € C°(M); w + 09g > 0, ¢ is G-invariant }. Definition 6.8 Let w be G-invariant. Then F is proper on P(M,w) if there evists an increasing function p :R — [c(w), 00) satisfying Jim p(2) = ©, such that for any ¢ € Po(M,w), F..(4) > ul Fo(o))- Remark 6.9 It follows from the cocycle condition of F that the properness on Pa(M,w) is independent of w. Theorem 6.10 M has a Kahler-Einstein metric if and only if F,, is proper on Po(M,w). We will prove Theorem 6.7. There are two directions to be proved and both will be done using the con- tinuity method. First we assume that F,, is proper. Consider (w+ 084)" = eh" (*:) for ¢ € P(M,w) and 0 tw, and the equality holds if and only if t = 1. Proof. The proof of this lemma follows from the following calculation Ric(w:) —dd logw? = —d5 log “ + Ric(w) = -d8(h, — toy) +w + Oh =wt t0d¢, =wt tu -w) = (1—t)w + tu, 2 tu because w & 6. Ii is clear that we bave equality if aud ouly if i= i. Qa Lemma 6.12 The first eigenvalue of A; satisfies (At) >t ift <1. Proof. Recall that we can characterize the first cigonvalue variationally by Sur Ob PoF (At) = (Ar) Suc Gut =0,640 fxg Pw and it is well known that there exists an eigenfunction u such that Aju = —Aru.6.2. Existence of Kahler-Einstein metrics 65 Set g = g and consider f 9 gM usustdV,, M where dV, is the volume form of g. At a given point, we can assume that 9i3 = 5:;, So that the integrand reads Ung Ua = (tea tz)5 — MEG UT = div(uijug) — ujguy iv (au;524) — angi, — Rjag toy ag Ua) — Myzity ~ Rajag tory iv (wigug — Ugquz) + uy;Uy — Risse; = div(uijuj — uguz) + (Au)? — Riswsuy = div(ugjug — ugguz) + (Aru)? — Rigusuz. This implies o< I, (cau? Ric(0u,du) a, and hence x f waV, > [ Ric(u, du)dV, >t [ |Oul2aV, = tds [ waV,, M M 'M M where we have used the previous lemma and the variational characterization of Ay, so we have that A; > #. Oo Remark 6.13 If t = 1, then Ric(w:) = w; and A; > 1. The equality holds if and only if f gi ony (dV, =0. IM This is equivalent to saying ujg = 0, which is in its turn equivalent to (9° u;)e = 0, since the metric is parallel. This then gives us a holomorphic vector field X= g)u; 2 7 8z;" Conversely, if X is a holomorphic vector field satisfying ixw = du, then w is an eigenfunction with cigenvalue 1. So we have established the follow- ing identification n(M) ~ { eigenfunctions of w; with eigenvalue 1}.66 Chapter 6. Kéhler-Einstein metrics with positive scalar curvature It follows from A. > t that Dy is invertible for 0 < t < 1 and by the Implicit Function Theorem, (#7) is solvable for |¢ — ¢’| sufficiently small, so we have shown that E is open in (0,1). One can also prove that E is open near + =0 by modifying the above arguments slightly. To prove that E is closed, we will proceed as before and we will try to find an a priori estimate |ldul|o2.4 0, such that n+Ad< ceb(suPar inka @) Remark 6.15 The proof of this lemma uses the arguments in the C?-estimate in Yau’s proof of the Calabi Conjecture. Note that this lemma implies IIdllc2 < cePls4Par d—infar 4) Here one needs to use the fact that w + 02¢ > 0. Exactly as in the Calabi-Yau Theorem, we will be able to prove the following lemma (the proof will be omitted) and it remains then to find a bound for the C°-norm. Lemma 6.16 There exists a constant C = C\(\dl|co), such that when o is a solution of (#,), we have Udllgag SC: Lemma 6.17 (Sobolev inequality) Let (X",ds”) be any compact Riemannian manifold (of complex dimension n) with Ric(ds?) > eds for some ¢ > 0 and Vol(ds?) > v > 0, then there exists a constant o = o(e,v) such that for all fec~(x), U nav)” sol f stav + f srav). Remark 6.18 The proof can be found in [17] and is based on a result by C. Croke. Note that this result is stronger than the usual Sobolev incquality because here the constant gis independent of the metric. This will turn out to be very useful because we will be comparing different metrics.6.2. Existence of Kahler-Einstein metrics 67 The following lemma tells us that to bound the C°-norm, it suffices to find a bound for J.,($) and once we have done that, we will have the required a priori estimate to prove closedness and hence the existence of a Kaibler-Einstein metric. Lemma 6.19 [f ¢ is a solution of (+1), then there exists a constant C = C(t) so that Iolo < C(L + F.(9)). Proof. Let wy =w+05¢ and define the functional I,,(), which is the difference of two averages of ¢ with respect to two Kahler metrics, and which satisfies the following estimate 10) =F fale" — 8) =F fear a en pee ug) =1f gen ml yl = 7 [200 dbn toss-up?) < (n+ 1)J.(4). So it suffices to prove the estimate in the lemma with I,(¢) in place of J.(¢). Uise, as hefore, the Green formula to get: 1 1 a= 5 f oo =f -aowiete, vou VI VI where G is chosen so that G > 0. Since —Ad < n, this yields 1 n wa) <% for +0, which gives us one half of the required estimate, namely, Claim \ o<-igroxe(s [ Cows + i). Proof. Denote by A; the Laplacian of wg, where this metric will also sometimes be denoted by w:, which should not cause any confusion. Then, because w+ 00¢ >0,68 Chapter 6. Kahler-Einstein metrics with positive scalar curvature we sce that _ w=wy —00¢6>0 and taking the trace of this latter expression with respect to wy, we get n— Aid = try, w > 0. Define now @_(x) — max{—¢(x), 1} > 1, so that 2 (n— Aed) > 0. Integrating this, we get oxi [ P.(n— Adu Vie =2/ oun ts f Ve Viduk Vida One VW Sug OO OS —8f vist n = ¢ foot F Ign PRO 1 Hof apes [ vet vl-6- 0 =7h,% % — veer fre Pup. Using the fact that ¢_ > 1 (and hence g?. < ¢?*?), this yields, < (p+)? +m 7 [leo Peps ME fatten. Observe that as before, when ¢ is a solution of (+,), we have Ric(wr) > tur. Since w and w; are in the same Kabler class, we get f ope foray. JM JM Using the Sobolev inequality in Lemma 6.17, we see 1 / es ")* wen [ ptm Lf ou < eee rf, | 7 Vodue Once more, Moser’s iteration will show us that sup g_ = lim ||d-Iler raw © Cllo-leaaer)- M pase6.2. Existence of Kihler-Einstein metrics 69 Recall that. 4; (u,) > t, so that. the Poincaré inequality reads i 1 e)ar< A f iverar Vda WIPE) AS ay fy O c . t for t <1. Let {gi} be a continuous family of solutions of (+), then this family is also smooth in £. Let + be a solution of (1), 80 (w + 5%)" = eh HN, Differentiate this expression with respect to ¢ dD. A (w + OBdr)"* = (=r = tee "uw", this implies that . . Argue = (br — thew? , which is the same as Aude = (64 ~ thr). This means that we can rewrite the above as Gi (toa tut00) =F f) onloe +ebnner _ -4(/, yet tet Joh, rete, Since we know that for all t, f che touyn = V, M differentiating with respect to t yields [ (de + threw" = 0, M so that we can simplify the above even further to § (10060) ~ 1060) =-$.( Ff owe) — ap fh aeetortoram6.2. Existence of Kahler-Einstein metrics 7 Multiplying it by t, we get. d d/(t F (ttl) ~ Jul60)) ~ alee) ~ Juloo=-F(F ff oust). Integrating this from 0 to t gives us * t Wo) Lolo) ~ [ Ualtn) ~ 1tbaas = =(% feat), b he which is equivalent, using the definition on J,, and J, to * 1 ~ [ 0.) Leenpas=e( 4.000) - % fou"). 0 M An easy manipulation shows that n-i n+l 1 Lob.) ~ Jel.) = 5 [20 Bens nar So, 7 and therefore, 1 Jolos)- 4 f uw" <0. slo) —% fe (206) few") is decreasing in t. We may assume that t is away from 0, since we know that (0) is solvable in any case. However, the above translates into 1 1 Jul) =F ff os” = alos) + 10e( f f, chet). In fact, the function Since 1 hobo yn =f ~(1-t)be phate [ -(1-t)o. Ve =| et tyra — | reletoryn = | @ oe V Iu VI iM Vv and a teal, Jm V together with the fact that the logarithm is a concave function, we get tox } j a) = toe( [ enue) M M un =f -A- thar [ne =! [ z= = foe72 Chapter 6. Kéahler-Einstein metrics with positive scalar curvature This leads to the following estimate Folds) <—loe( jf etmew") 1 > J oust. 1 Jato) Ff du" <0. We will now, finally, use the fact the F.,(4;) is a proper function, which is the assumption under which we started, and which means Fra(Ge) 2 w(Ju(oe)) for some increasing function ju, Furthermore, we have Lf cmurad f emtrutan This implies that ¢, must be negative somewhere and hence — infyy ¢; is posi- tive. So we have : 1 n O<—inf de < CCF [cont +e’). This follows Lf yom 7 [out se. Since ¢ € (0,1], we have, 1—t —— | buh 0 such that for any € P(M,w) which satisfy @ L Ar, we have F($) 2 Ju(#)® — 6, which is the same as cf cmb < Celo(d)- b Jar Ou" —Jul0)?, Vim Here A; is the space of cigenfunctions with eigenvalue 1 and by ¢ 1 Ai, we mean f gw" =0 IM for all y € Ay. Moreover, A(w) denotes the smallest eigenvalue of w which is greater than 1. Note that if n(M) = {0}, then gL A, is vacuous and the above inequality holds for any ¢. Remark 6.22 This theorem is a stronger statement than just showing properness for some increasing function, because we have the function jz here explicitly. If n = 1, then M = S? becanse of positivity of C,(M). If we further assume instead that @ is perpendicular to the eigenfunctions corresponding to the eigenvalue 1, then Aubin showed F.(¢) 2 €J..(¢) — C. Conjecture 6.23 (Tian) Under the assumption of Theorem 6.21, one can find € = e(n) and C =C(n, 1 ~ 1) for all n such that F..(6) > eJ..(d) — C. Before proving Theorem 6.21, we will first explain why we make the assump- tion on the algebra of holomorphic vector fields. Suppose that X € n(M) for X £0. Denote by ©, the flow corresponding to Re(X), then Bw =w + ddr. We will write w; for w + 03%, and as before we can choose ¢; such that 1 v I, choy = 1, d ; gp Flot) = Re(fu(X)), where fy is the Calabi-Futaki invariant. Claim74 Chapter 6. Kahler-Einstein metrics with positive scalar curvature Proof. First of all, we have Ric(w) —w = ddhw, 1 Ric(w) — fw = 958th... Because 4; is a diffeomorphism, we also have 1 shoe ef, Pi ho dtu = 1, 1 Pugh = wh 1 a Therefore, h, = 7h, which implies that hau, = Re(X)(hu.)- since We also have Rie(wp) — wy = Rie(w) — w — atog( 21) — 0861, which in its turn gives Phe = hha = toe(=E) - be Differentiating the ahove with respect to t, we get = Aube be, and therefore 1 . HE fads + Re() had)! w = Re(fu(X). a ‘The following corollary is an immediate consequence of this claim. Corollary 6.24 F., is bounded from below only if fur = 0. Proof. Suppose that fyy(X) # 0, then we can assume that Re(fu(X)) <0 (since if it was imaginary or positive we could change X to iX or —X). Then the claim shows that F.,(;) = tRe(f(X)) > —00 as t > oo. Qo6.2. Existence of Kéhler-Einstein metrics vi) Proof of Theorem 6.21. For simplicity, we assume that n(M) = {0}. We refer the readers to [24] for the general case. For any ¢ € P(M,w), put w’ = wy. Consider (w' + Bd)” = eho“ Mh!™ (++) We may assume that (**,) has a solution 41. We will try to use the continuity method backwards, that is starting at ¢1. So we set E = {t € [0,1]; (4%) is solvable for all s € [¢, 1]}. Clearly, 1 € E, so E is non-empty. Showing that E is open goes as before. We use that \; > 1 (since \, = 1 would imply that n(M) was non-trivial) and that w is a Kahler-Einstein metric. Then we can use the Implicit Function Theorem att=1. To show that E is closed comes down tu the following a priori estimate Hell c2.3 SC. Note that C always denotes a constant independent of t. It was shown in [25] that there are 6 > 0, which depends only on M, and C’, which may depend on w’, such that for any » € P(M,w'), if OUP Wal” < CF, It follows that if dt is a solution of (##;) for t < 6/2n, we have that supd, 0, we have76 Chapter 6. Kihler-Einstein metrics with positive scalar curvature 1? Me™ [ote haontbe 1) +1 < ee [ee we ay! y sagen 3/4 as ee y (beeen) ops toe — 1h) ) Using the Sobolev inequality, we can deduce from the above inequalities that Lf var)” 6/2n. Then by Lemma 6.19, we have elloag SOU + Jr(Ge)) SC(1L + n(Lar(b) — Jur()))- Lemma 6.25 Let {gi} be a smooth family of solutions of (xx,). Then the func- tion I(t) — Jur(04) #8 monotonically increasing in t. Proof. As before, we have § (Fal) ~ Jule) =F ff oededuat ap [Oude + tb debut where w = w' +00¢,. We can write de = 7%, civ, where 1; are the eigenfunc- tions of A, and Wo = 1 is the constant eigenfunction. So the above integrand then reads VilePQr Has, 1 which is non-negative since \; > t. a So we end up with Idllong < CO+ (ude) ~ Jalo0)) < ©. Therefore, we have shown that (#*,) is solvable for all t € 0,1] and here we have ¢, = —¢.6.2. Existence of Kahler-Einstein metrics 7 So we have found a smooth family of solutions {4+} defining a path in P(M,w). This path clearly depends on the base point g. From Lemma 6.25, we know that Iur(¢e) — Jur(¢t) is increasing with t. Using the cocycle relation for F,, after Remark 6.3, when we set ¢ — 0, B= o, w! = we, we get F..(4) = -F.,(—4) = -Fu(d1) 1 n 1 =o. andlor) ry fous! +toe(y fe Ory! ). but because of (##1), we know that 1 yt 1, maf n rhe eV da” so this disappears with the logarithm and therefore it follows from equation (b) that 1 FA) = [sled ~ ds DML > 0, where we used that the integrand is positive. So we can now state the following corollary. Corollary 6.26 (Ding & Tian) F.(d) > 0 for all @ with w + 59 > 0. This implies in particular the Moser-Trudinger-Onofri inequality on S?, which reads 1 + ty < ebr Sp2 Vode Ion he, 4m Joo where we have used that M = S? and hence V = 47. Here, however, we are after the extra J.,(@)°-term in the estimate. For t > 0, we have that 1 F.(6) > [ Us(@n) ~ Jo(oa)ds 2 (1 = t)(Lar (be) — Tur (du)). Now we want to show that Juv(4) ~ Jur(¢1) uniformly for t close to 1, where ~ means that they are comparable, which is at first sight a difficult requirement since 4; and ¢; are solutions of different equations. But if we manage to do this, we will have the following estimate F.(9) 2 (1— t)Jur(d1) — ©. The following lemma will be crucial but will not be proved here, see (24) for a proof,78 Chapter 6. Kihler-Einstein metrics with positive scalar curvature Lemma 6.27 (Smoothing lemma) Let & be any Kahler metric representing ci(M) > 0 with Ric(@) > (1 —e)@, where and Then there exists such that 4) |ldlico < 3ihalico 5 5) rarlled gory SCO + Mallee? where 8 = (n) and C = O(n, \i(&), ou)» Remark 6.28 By C?, we mean the Hélder norm which depends on the metric. The proof uses Hamilton’s Ricci flow Bu _ yg 4 5u)” a ~uths at 8 ae ut with boundary condition uly = 0. Sctting ¢ = w and & = & | ddu,, this proves the lemma. Remark 6.29 Notice that although ||ia|| might not be small, this lemma ensures that {{ho,|| is small as long as ¢ is sufficiently small. Also G’ is a continuous deformation of &. Because of the Kahler-Einstein property and the fact that we have a solution of («#,), we get that Oh. = Ric(we) — we = -d5D log(w' + 0561)" — w' — Bde = -ddloge*" ** — AG log(w') , = -00(1 td, so we can write fay = (1 the + ey where cy is determined by | (eves - ar =o M6.2. Existence of Kihler-Einstein metrics 79 This immediately implies the following bound on constants c+, lee] < (1 = #)[ldellco- Setting ¢ = 1—t, we get Ric(wr) > (1 — ewe = tur, and Wawelleo < 2e||delleo. Therefore, the above lemma provides us with a w} = w, + 0du, which satisfies 1) Iluelleo < 31 — 4) [I¢elloo, 2) Wrorll 0.8 guy) SCO + Ot)? lsulleo) (a - 9, where C = O(n, di (wh), ou)- ‘We now make the following observatiou: wl =w! + 08d1 =w + 00d — 00d = w. So in this case, we do not need to deform the metric, because at t = 1, we already have a Kahler-Finstein metric. Since w is a Kahler-Einstein metric, it now follows that there exists a unique 7, with w = w+ dy, and (wee we = eI (# 4 4) Similarly as above, but in addition using the Maximum Principle, we get b= G1 — Vem Me + Oe + Pts where ji; is determined by [ (em muetie er =<. iM Iuel S 10(1 — )Ildullco, It is easy to see and hence, we have that ue + ee + pe ~ (1 = E)Ildelleo- Our goal is to control the behavior of ¢:— 41 so that it remains to get # under control.80 Chapter 6. Kahler-Einstein metrics with positive scalar curvature The solvability of (#*,) and the Kahler-Einstcin property together imply that wo? = ele dP, which unfortunately does not give us any control over h,. This makes us study (* * *,) instead. Note that ¢ = 0 and as before, we will try to use the Implicit Function Theorem here. Consider the following operator &, : C?3(M,R) > C°2(M,R) defined by We Tog = ha = We would like to show that D®, is invertible at y = 0, so that we can deduce that &,(7) = 0 is solvable, which is equivalent to saying that —v solves (***;). In order to arrive at this conclusion, we will make the following two observations. ~ We know that n(M) = {0} and that this means that there are no eigenfunc- tions corresponding to the eigenvalue 1. So \,(w) > 1 which shows that the linearization of 4, at p = 0, given by D&|y-o(v) = —A.v — v, is invertible. ‘Therefore, the Implicit Function Theorem tells us that there exists a unique 2, such that &;(y) = 0 and {lylloa.4 < Ce if IIhuilloo.3 S$ € for some small €. Suppose we are given two metrics which are comparable, that is Lw <@ < Qu, then \1(@) > 2-1 Aq(w) and oz > 2"41¢,,. So the corresponding Poincaré and Sobolev constants are also comparable. Take now some C > C(n,2-"-!\,(w),2"+40,,) and assume that Ce < 1, then we can choose a to, which may still depend on ¢, such that (1 t0)®(1 + (1 = to)? [deo llé0)"*? = sup (1=#)9(1 + (1 ~ t)?Ibell@0)"? tostst é 4C4+1 Ciaim For any é € [ig, ij, we have jjPeliga.y < 4- Proof. Clearly, yi = 0 satisfies this. If there exists y%, such that 1 € [fo, I] with Ile, ll,2.4 = 4, then because of Remark 6.29, we get od wy SF [Poel for some small ¢ > 0, and since Fw Suh w+ OB < 2w,6.2. Existence of Kahler-Einstein metrics 81 the first of the above observations gives us the following contradiction C 1 IWerllong SOB max{to,1— g5}, we get IIéellco > (1 — 30(1 ~ t))||dr|leo — 1. This in its turn implies that Fuld) = Fu(- #1) = -Fa(-d1) = "(Uur (be) ~ der u))at = [vw ($1) ~ Jur(e) dt > min {1 anes a} (le (r.) ~ Jer(dug)) > min fi te. sf ( (dr) ~ Jur(r)) ~ 1. > —min n — tog Jeon) - Hence, F.(¢) > = * min {1 — to, ay I(b) - C and if 1 ~ t > gj, this reads Fu(¢) 2 Jud) — 2 an Thus, we are done. However, if 1 — t9 < gg, we need to do more. We then have Fu(b) — eJ.(9) _ © (1+ [igloo82 Chapter 6. Kéahler-Einstein metrics with positive scalar curvature where ¢’ may depend on € chosen in the first of the above two observations. If we have lidlloe S CL + Ju(9)), then it follows F.(d) 2 C'Ja(9) FF = C. This would prove the theorem, bul we do not want the assumption on the C°- norm of @ and in order to get rid of that, we need to do more work (due to X. Zhu and G. Tian). Lemma 6.31 There exists a constant C > 0 such that fort > 3, we have osem(de G1) SC(L + Jude 1), where osca f = supy f — inf as f. Proof. It suffices to prove the above estimate for I,, instead of J,, because we know how to bound J, in terms of J,. We know that nm w cher (de d1) 0 and that. 1 Ric(w1) > tor > swt, Ric(w) = w, which implies that we have uniform Sobolev constants for both w, and w. We also have that _ w+ 08(¢r -— 1) = we > 0, wt + 08(g1 — Ge) > 0, which imply respectively that Audi - $1) 2-7 Mule — 61) < — Set (dt — $1)4 = max{¢; — 41,0}, then Moser’s iteration method will provide us with the following bound sup( de — dr)+ SC (1+ IM(Ge ~ dr) +lle2) and we will obtain a similar estimate for (f: — 1)- = min{¢, — 1,0}. Using then the Poincaré inequality as before, we get the required bound by I, 0 Claim Setting w, = w’ + 08¢, we have 1 1 v [owt <0 and Jur($.) — 7 f, dua!” <0.6.2, Existence of Kahler-Einstein metrics 83 Proof. From (a), we have $ (—HtG6e)— Jt) = 3 fora) = Lalor) + Lo Integrating this as we did before, we see that Foust = tot 64) ~ Selb) + [Crt Gs) ~ ela) Since I.-() — Jur(da) increases with s, the integral on the right is bounded above by (Lar (Gt) — Juv (Ge))- Therefore, we have proved the first inequality claimed, namely, 1 in 7 I, uu <0. By (b), we have 5 (eee) ~ Ff oem!) = Ua) ~ Jarl) This yiclds upon integrating from 0 to t ( Jute % f uw" ") so Jalon ff os! <0. Since ¢ > 0, this means The claim is proved. oO Notice that Fulbn)— Falbs) = Jaleo) ~ 7 fou! elsfcnna). Jur(1) +e, bw” Using the concavity of logarithm, we have 1 _ sea (s [ot oear') < tp fet.84 Chapter 6. Kahler-Einstein metrics with positive scalar curvature Hence, it follows from the claim F(t) ~ Far(dr) Fur($e) — Far($) = Fi(@e ~ 1). On the other hand, by the last lemma and what we have already proved, we get that Fale — $1) 2 1 08cm (be — G1)" — 02, and consequently, (1 = t)Ju(9) 2 er osear(be ~ 1) — therefore, for t > to, we finally get Fad) = (1 = t)(Lur(be) — Jur(Ge)) > 15, (61) - 2(1 — t)osea (de — 61) A= * 1.16) - 201 - Heb (1 — NIG) +2)? Iv and the theorem now follows by choosing (1 — t) ~ (1 + Ju(o)) 1. a6.3. Examples 85 6.3 Examples In this section, we apply Theorem 6.10 to prove existence of Kabler-Einstein metrics on Fermat. hypersurfaces in complex projective space. It was proved that any Fermat hypersurface of degree p admits a Kihler-Einstein metric in (25] in the case of p > n, and in [20] in the case of n—1> p> "$4, where n is the complex dimension of the hypersurface. We will also discuss briefly the existence of Kihler-Einstein metrics on complex surfaces at the end of this section. A Fermat hypersurface of degree p and complex dimension n is defined as the zero locus M ={lzo: tenga] ECP" | 2h +--+ 2%, = OF CCP™. From Example 2.9, we know that if p < n+ 2, then c(M) > 0. We will confine ourselves to the cases: p< n+ 2. Consider the group Gy generated by Ou fay 009 Sega] > [20 F002 ez 20 Se] where ep =e"). Clearly M is invariant under Go. Let G be the maximal compact subgroup of the automorphism group of AutM containing Go. Choose a G-invariant metric w. We will show that F,, is proper on Pg(M,w). Let mi: M — CPP be the projection onto CPP = {[zot--+ tain 102 2a tt eayal} & CPX Note that this map is well defined because (Ore-:0:1:0:---:0) ¢ M. So any ¢ € Po(M,w) is of the form (n+2—p)mip for some y € P(M,wrg), where wyg is the Fubini-Study metric on CP”. Note that c1(M) = (n+2—p)[n*wrsl. f che bn oy, M We may normalize86 Chapter 6. Kahler-Einstein metrics with positive scalar curvature By the cocycle property of F°, we have Fo) = FSP) = Fenp2-pyrturs (9) — FS(u), where (n+2—p)t!wrs =w +00u might not be a Kahler metric anymore, but wand therefore F°(u) are bounded. Using the basic properties of F° described in Section 6.1, we have 0 n+2—p n+1 Fnt2—pyxiwrs(®) ~ aq Pinter (tee ; Since CP" has a canonical Kahler-Einstein metric (n + l)wrs, we know that Finstjwrs 18 bounded from below on P(CP™, (n+ 1)wes). This implies that 2- 1 nti P(g) 2 te woe F L, een 42 xu -6, where C’ always denotes a uniform constant. Consequently, HP (lf stb 0S ey Fl) ESP (Gf oaEn + 2-2)" Daas) ~ 6, since 77") {ws > cw for some positive constant c > 0, we have rt+2—Pi (lh fo atetpeehe—on) FO)> TF voe(= [| « Hye eho Ou") — C, We may assume that p > 1. Claim There is a uniform C > 0 such that sup¢ < e(1 + wa ene ae che “*s")). M V Iu Proof. It was proved in [25] that there are a > 0 and C > 0, such that for any o€ P(Mw), ‘7 a —a(d—supyy b) ym 2 [ eaommnti 0 such that 6(n +1) d(n+2-p)+p-1 a=6.3. Examples 87 Then by the Sobolev inequality, we have 1 [ cheer Jae ‘ 2syne2=p) ceomouel f erhe-oinnndi sn) (/, eeaia (he 2) etn) wT IM M Losing ap iM Then the claim follows. o It follows from this claim that F, is proper on Pg(M,w), and consequently, M admits a Kahler-Einstein metric. We already saw that if M has a Kahler-Einstcin metric, then the Calabi- Futaki invariant fy vanishes. It was proved in [23] that the converse is also true for complex surfaces. Theorem 6.32 Ifn = 2, then M has a Kahler Einstein metric if and only if the Calabi-Futaki invariant fy = 0. To prove this, basically all one needs to do is check that. F, is proper along the solutions of the corresponding Monge-Ampére equation. We refer the readers to [23] for its proof. ‘Theorem 6.10 can be also used to simplify the proof of this theorem for most complex surfaces by showing that F, is proper on Po(M,w). Let us illustrate it. By the classification theory of complex surfaces, we know that any surface M in the above theorem is of the form either CP! x CP! or the blow-up of CP? at k points (0 < k < 8). Clearly, CP! x CP! and CP? have homogeneous Kahler-Einstein metrics. We have shown before that the blow-up of CP? at one or two points does not admit any Kahler-Einstcin metrics, since they have non-vanishing Futaki invariants. It was shown in [26] that the blow-up of CP? at 3 or 4 points in general position has a Kahler-Einstein metric (k = 3 was also proved by Siu (21]). Theorem 6.10 can be used to prove the existence of Kahler-Einstcin metrics on any M with c:(M) > 0 which is a blow-up of CP? at k th The case th 6 eds the at k points with & — 5,7,8. The case that & — 6 still needs to use the proof in {23} at this moment. As an example, let us show existence of Kahler- Einstein metrics on any blow-up M of CP? at 7 points in general position (here c1(M) > 0). Let M be a blow-up of CP? at 7 points. Then it is a double covering a: M+ CP? with branch locus along a smooth quartic curve. Assume that G contains the deck transformation of 7. As before, w denotes a fixed, G-invariant Kahler88 Chapter 6. Kahler-Einstein metrics with positive scalar curvature metric with the Kahler class c:(M). Then for any ¢ € Po(M,w), *1.6 = 20. We normalize ¢ such that afte =1. Put. f = 1*w}.g/w?. It is non-negative and vanishes along the branch locus of a. Moreover, we have 1 -£2 4 tur< . 3 fi Ww Se< oo. Using the Hélder inequality, we deduce from this 1 4 1 ‘ ble eo(o [rere] Note that ¢ always denotes a uniform constant. By using Theorem 6.7 for CP?, we can show (compare the last example) that 1 1 an Fu($) 2 3108 G fie Seq hs) © It follows from the above inequalities that 1 Fu(y) = 7supy —e¢. 4m Therefore, F,, is proper on Po(M,w), and consequently by Theorem 6.10, M admits a Kihler-Einstein metric.Chapter 7 Applications and generalizations In this chapter, we will discuss some applications of theorems in previous chap- ters. We will also give some generalizations of previous results. 7.1 A manifold without Kahler-Einstein metric We will now consider the case n > 3 and here we will need to introduce some new concepts. Definition 7.1 A holomorphic degeneration of M is a fibration r: W"*! > D such that w 4(3) ~ M and x 1(t) is smooth for all t # 0. Here D is the unit disc in the complex plane Example 7.2 ‘Lhe easiest example of such a degeneration is the following: Let W = {(21, 22) € C?} and define (21, 22) = 2122 then w(t) = {z1z2 = t}. It is clear that this is smooth for t # 0 but not smooth at t = 0. Definition 7.3 A special degeneration is a degeneration that in addition satisfies: there exists a holomorphic vector field v on W such that 7.v = —t& (generating a one-parameter subgroup z ++ e~tz on D). Remark 7.4 The vector field ul,-1,) is not necessarily tangent. to 7—1(t) when t #0, but v}_-1(0) is, because 0 is a fixed point. Note that all fibers x(t) except 7~1(0) are biholomorphic to M in a special degeneration. In this section, for simplicity, we always assume that the central fiber 7~1(0) is smooth. We refer the readers to [24] for general cases where the central fiber may not be smooth. Example 7.5 (Trivial product) Let W = M x D so that v = v, — t2 for v, € n(M). Observe that x 1(t) ~ 1 1(s) when t,s are both non-zero. This is true90 Chapter 7. Applications and generalizations because v generates a complex onc-parameter group of automorphisms ¢; of W such that 7 0 ¢(w) = e~'n(w) for all w € W. So all the fibers over D are isomorphic. In general, 7~1(0) might not be isomorphic to +-1(t) for ¢ # 0 and we say in this case that the complex structure jumps at 7~1(0). Example 7.6 Let Mp, be the blow up of CP? in four points p1,...,pa, where p= : = (0:0: 1] are fixed. If p4 is not on any of the lines Pyp2, PiPs or Pzps, then the complex structure of M,, is independent of pa, since there is an automorphism of CP? fixing pi, po and ps and bringing ps to [1 : 1:1] in this case. Note that in this case, ¢1(.M) > 0, which is equivalent to the fact that there are no holomorphic CP!'s with self intersection number less than or equal to —2. If, however, for example ps € Paps, then the complex structure changes. To see this, let E be the line in M,, over Paps C CP?, that is we have blown up in three points on p3p3 and each time we blow up, we reduce the intersection number by 1, so E? = —2. Set now m:W= YU Maw > D teD sending M,, (+) to t. Assume p4(0) € p2(0)p3(0), then t = 0 corresponds to the jumping of complex structure. Remark 7.7 There is no complex surface M with c,(M) > 0 that admits a non- trivial special degeneration r : W — D such that ¢,(~1(0)) > 0. To see this, use the classification theory of surfaces. Theorem 7.8 If M has a Kahler-Einstein metric with c,(M) > 0, then for every special degeneration 7: W —» D we have that Re(fr-1(0)(¥le-1())) = 0 with equality if and only if W = M x D. The following example was missed by Fano in his classification of Fano 3- folds (three-dimensional complex manifolds with positive first Chern class) and was found by Iskovskih, The author learned Uhis description from S. Mukai (see [19]). It forms a counterexample to a folklore conjecture, which is none the less true in dimension two. Conjecture 7.9 One can always find a Kahler-Einstein metric on M if (M) = {0}.7.1. A manifold without Kahler-Einstein metric 91 Example 7.10 We will give two descriptions of the following manifold. First of all, consider S1(2,C)/I’, where I’ is the icosahedral group. The manifold M we are interested in is the compactification of the above quotient. Let Riz be the space of homogeneous polynomials of degree 12 in two variables. Then PRy2~ CP. Now SI(2,C) acts on Riz by composition and M is defined as Orbit of SL(2, Cf, where f is an appropriate, ['-invariant polynomial in Ris. ‘The following description is perhaps easier to understand. Consider the Grassmannian (4,7) consisting of four-dimensional subspaces of C7. Take now any 3-dimensional subspace P C A?C7 and define Xp = {E € G(4,7);1n(P) = 0} where mg : A2C7 > A?E+ is the orthogonal projection, then a dimension count tells us that we can expect that dim Xp = 3. We make the following two observations: © If Xp is smooth and non-degenerate (that is no vector field v € C7 vanishes in Xp), then c;(Xp) = c1(Q)|xp- Here Q is the universal quotient bundle on G(4,7) and therefore c(Q)|xp = 4a(G(4,7))|x, > 0. © n(M) = {v © sl(2,C); induced action by v preserves P}, where sl(2,C) = n(G(4,7)). Now let. Po be the span of uy = Ber Ae — 5e2 Nes + Gea A ea uz = Be, Ae7 — 2e2 Neg + €3 Nes Ug = €2 Aer — 3 Neg + ea Aes which is invariant under the irreducible representation of sl(2,C) on C7 given by fi 0 . (0 <1) 7 diag 8,2,1,0, —2,-3} (9) Socooce cocoooor cocooro oooorce 0 0 0 1 0 0 0 ocoHroscoceo ercoscoeo92 Chapter 7. Applications and generalizations 0000000 3000000 OL 0500000 GC ye 0060000 0006000 0000500 oO 00 6 32:0 Check that Xp, is smooth and non-degenerate. This implies that n(Xp,) = sl(2,C) is semi-simple and by Corollary 3.6, we know that this implies that xp, =0. We will now deform Py to P,=span{u;+ J > aise; Nex, = 1,2,3} G+bzi47 then (Xp,) = {0} for generic a and therefore by the above observation, we have o(t)(Pa) > Po, when Re(t) + 00, where o(t) denotes the diagonal matrix diag(e**, e*,e', lew at en 3ty for t € C. This follows that Xoityp, ~ Xp, and Xap, + XPp- So set mWa ( U Kar.) UXp, > D ‘Re(t)>0 sending X,()p, to et. So we have here a non-trivial degeneration (since Xp, # Xp,) and the Calabi-Futaki invariant vanishes (fx,, = 0) so in this case, Theorem 7.8 implies that Xp, has no Kihler-Einstein metric. ‘We will not prove Theorem 7.8 in this section (see for a proof (24]), but we will explain briefly some ideas for the proof, Fix any Kahler metric Q on W satisfying [Q|,-1¢¢] = c1(m1(t)) and let ©, be a biholomorphic map from M onto 7~!(t) for t #0 generated by the vector field v on W. If we set we = Unt) then one can check that ®ju, is in the class of c,(M), so Bi, = w + OdG. One can show 4 lim 7 Fol o1) = Re(Fa-1(0(0))- Because F,, is proper, this proves the theorem. To get the equality part, one needs to know more about the rate of convergence when t — 0.7.2, K-energy and metrics of constant scalar curvature 93 7.2. K-energy and metrics of constant scalar curvature In Chapter 4, we learned that the scalar curvature of Kahler metrics is a moment map for the action of symplectic diffeomorphisms on the space of almost complex structures compatible with a given symplectic form. In this section, we will introduce the K-cnergy, which was first defined by Mabuchi in [18], and discuss its basic properties. Let (M,w) be a compact Kahler manifold. As before, we denote by P(M,w) the space {@ € C*;w + 009 > 0}. For ¢ € P(M,w), we define its K-energy by rife. (6) =F [ Gx(s(we) — nwt A dt, 0 JM where {¢} is any path in P(M,w) with dy = 0 and $1 = b,ue = w + 006, and s(w) is its scalar curvature and ex(M) - [wh"* [we] Of course, in order for 1,,(¢) to be well defined, we need first to make sure that the integral on the right is independent of paths connecting 0 to ¢. In fact, integrating by parts, one can show the integral on the right is equal to +l) St 7 fi PRI) rw! Awnint 4 TE 7 cy fe ne i rar n+1e It follows that v,, is well defined. From the definition, one can casily show that v,, satisfies the cocycle condi- tion as F,, does, more preciscly, if w’ is another Kahler metric of the form w+ 0dv, then Yold) — Yar (b =) = HoH). If w is a Kahler metric with constant scalar curvature, then the K-energy yy can be expanded as 2 le) = 5 | 80" +00,94 Chapter 7. Applications and generalizations It follows that any Kahler metric of constant scalar curvature is a minimum of the K-energy, moreover, it is a strict minimum if there are no functions ¢ with i; = 0. Note that this last condition is satisfied if and only if n(M) = {0}. Now we Ict X be a holomorphic vector ficld and , be the integral curve of its real part Re(X). Notice that fw has the same Kahler class [w], so we have ¢ such that jw = w + 00¢,. Differentiating this, we get d(inexyw) = 0b. ‘This implies that inexyw = Od; +a, where a is a harmonic (0,1)-form. From the definition of the K-energy, we can then show Grelod =F [| bleleg) — mnhus, = 5 Re Ful lel Re). In particular, if the Calabi-Futaki invariant is non-zero, the K-energy v, is not bounded from below. We have seen that the K-cnergy shares many properties of F., in the case of Kahler-Einstein metrics. However, we expect more. Definition 7.11 We say that 1. is proper on a closed subset EC P(M,w) if there is an non-decreasing function (t) with limo A(t) = 00, such that v,($) > NJu()) for any $€ E, where J, is defined as before. As for the functional F,,, the properness of v,, is independent of initial metric w in an appropriate sense. In view of the previous results in the last two chapters, it is reasonable to expect that the following is true. Conjecture 7.12 © If M has a Kahler metric w of constant scalar curvature, then v. > 0, particularly, it is bounded from below; © Ifn(M) = {0}. then M has a Kahler metric of constant scalar curvature and Kéhler class (w] if and only if v,, is proper over P(M,w); @ Let G be a mazimal compact subgroup in Aut(M). Then M has a Kahler metric of constant scalar curvature and Kahler class |w] if and only if v, is proper over Po(M,w), where Pc(M,w) consists of G-invariant functions in P(M,w). Theorem 7.13 Let (M,w) be a compact Kéhler manifold. If either | or c1(M) = 0, then the above conjecture is true. +e1(M)7.2. K-energy and metrics of constant scalar curvature 95 Proof. This theorem is essentially contained in results of the last. two chapters. For the readers’ convenience, we outline its proof here. Under our assumptions, we have Ric(w) — yw = 09h, for some smooth function h,,. Then by a straightforward computation, we get a BY) nd yt nl =— f rel )err [ave a a i pyri Agr! + yf oinet ail Se [mines => [6(2)-3 +a fh hal haw” ~ Wh) ~ H(Lo(8) ~ Jal¢))- Here we have used mtg n-il 1a(9) — Jul) = af onwt aus (9) — 4o(#) Dntiv p ONY AWG It was shown in [25] that there is an a(M,w) > 0, depending only on M and w, such that for any ¢ € P(M,w), 1 vie #8) essen tafe un 7 ff e-Ald-supm #" < C, V du Vu where @ ix auy positive number which is less than a(M,w). By the convexity of the exponential function, we deduce from this 1 fwa(2)op>-2 [ o-mporg —c Combining this with the above inequality of v,,, we get ¥4(6) 2 Blo(d) — wu) — Ju(9)) - C. This implies that 1, is proper on P(M,w) when ys <0. In fact, evon if w— 1, us is still proper whenever a(M,w) > <27. Since M has a Kahler-Einstein metric when c)(M) < 0 (compare with Chapter 5), we have proved this theorem in the cases thai p< 0. Now we assume that jz = 1. It was proved in [7] that 1 R= M047 | heady ff how 1 JA, Lf phesyn = v8 Then this theorem can be easily deduced from Theorem 6.7 and 6.10. o Since , we have vl, agit £0.96 Chapter 7. Applications and generalizations Remark 7.14 If we assume that w is a Kihler-Einstein metric with Ric(w) = pw, then the arguments in the above proof yield ml f ioe( 2) ug - 1(106) - (8): This implies that if » <0, then Yu() 2 0, and the equality holds if and only if ¢ = 0 constant. Therefore, the Kahler- Einstein metric w attains the absolute minimum of the K-energy. The same is true even if = 1. This was proved by Bando and Mabuchi. 7.3 Relation to stability Let M be an algebraic manifold embedded in a complex projective space CP". Let H be the hyperplane line bundle over CP’. This bundle gives rise to a Kahler class [w], which is a positive multiple of c,(H|a)- In this section, we relate the properness of the K-energy v,, to the stability of the underlying polarized manifold (M,w). This is inspired by a conjecture of Yau on existence of Kahler-Einstein metrics with positive scalar curvature. ‘The stability of (M,w) is described in terms of the following algebraic family. Let G — SL(r,C) (r > 2). Let m: ¥ +> Z be a G equivariant holomorphic fibration between smooth varieties, such that (1) all fibers are connected subvarieties of complex dimension n and M is one of them; (2) there is a G-equivariant embedding of ¥ into Z x CP" for some N. Write L = x$H, where H is the hyperplane bundle over CP’. Furthermore, c1(L|a) is a positive multiple of [w]. Clearly, L is a G-equivariant line bundle over ¥, which is relatively ample over & aud has fixed topological type along smooth fibers. This iast property can be used to restate (2) in a more intrinsic way. Consider the virtual bundle €=(nt+1(K—K)@(L-L7))" —np(L- Ly), where K = Kx ® Kz! is the relative canonical bundle, and as before, e1(M) «ex (L1n)"? ex(L|m)" a7.3. Relation to stability 97 We define Lz to he the inverse of the determinant line bundle det(&. 7). By the Grothendick-Ricmann-Roch Theorem, we can compute er(Lz) = 2" Yar, ((r at Ver (K)er(L)” + myer (L)"*") . We also denote by Ly! the total space of the line bundle L7' over Z. Then G acts naturally on £;!. Recall that X. = 71(z) (z € Z) is weakly Mumford stable with respect to L, if the orbit G- 2 in £3' is closed, where Z is any non-zero vector in the fiber of Lz! over 2; if, in addition, the stabilizer G, of z is finite, then X, is Mumford stable. We also recall that X, is Mumford semistable, if the 0-section is not in the closure of G- 2. Clearly, this stability (resp. semistability) is independent of choices of 2. Theorem 7.15 Let 7: X ++ Z be as above. Assume that M = X, and v, is proper on P(M,w). Then M is Mumford stable with respect to L. This theorem was proved in Section 8 of [24] without explicitly stating it. We refer the readers to [24] for its proof. The arguments in Section 8 of [24] also shows that the converse to the statement of Theorem 7.14 is true under slightly stronger assumption on the family 7: 4 ++ Z, namely the fibers have no multiple components. Remark 7.16 More generally, if 7(M) is non-trivial, we may define the weak properness of v,, which means, roughly speaking, properness of v,, modulo action of Auty(M), the identity component of Aut(M). This weak properness will imply the weakly Mumford stability. However, we will not discuss this general case here. Now we discuss implications of the above theorem for the existence of Kahler- Einstein metrics with positive scalar curvature. Assume that (M,w) be a com- pact Kahler manifold with {w] = ci(M). In late 80's, Yau proposed Conjecture 7.17 M admits a Kéhler-Einstein metric if and only if it satisfies a certain stability condition in the sense of Mumford. Combining the above theorem with the theorem of last section, we have Theorem 7.18 /24/ Let (M,w) be and x: ¥ +» Z be as above. Assume that M = X, admits a Kahler-Einstein metric of positive scalar curvature. Then X, is weakly Mumford stable. If X. has no non-trivial holomorphic vector fields, then X45 is proper on P(M,w) and M is actually Mumford stable with respect to L. This answers Yau’s conjecture partially. Example 7.19 Let us apply the above theorem to proving again non-existence of Kahler-Einstein metrics in Example 7.10. We will adopt the notations in the previous sections.98 Chapter 7. Applications and generalizations Recall that W = G(4,7) consists of all 4-subspaces in C7. Let Q be its universal quotient bundle. Let 7; (é = 1,2) be the projection from W x G(3,H°(W,A2Q)) onto its ihfactor, and let $ be the universal bundle over G(3, H°(W, A2Q)). We define X = {(a, P) € W x G3, H°(W,02Q))} One can show that 2 is smooth. If L = det(Q), then e(L) is the positive generator of H?(W,Z). Consider the fibration m = mala: ¥ +4 Z, where Z={P€ G(3, H(W,A7Q)) | dime Xp = 3}. Its generic fibers are smooth Fano 3-folds. Using the Adjunction Formula, one can show e(K) = —aper(L) — 3x3er(S). Therefore, it follows that c1(Lz) = 16m, (12m3c1(S*)rfer(L)® — mfer(L)4). One can show that Lz is ample. By the definition of P,, one can show that none of G'- P, is closed in £z*. Therefore, none of generic Xp, admits Kahler-Einstein metrics.Bibliography T. Aubin Nonlinear Analysis on Manifolds. Monge-Ampére Equations, Springer-Verlag, Berlin-New York, 1982. R. Bott and S.S. Chern Hermitian vector bundles and the equidistribution of the zeroes of their holomorphic sections, Acta Math., 114, 1965, 71-112. A.L. Besse Einstein manifolds, Ergebnisse der Mathematik und ihrer Grenzgebiete; 3. Folge, Bd. 10, Springer-Verlag, 1987. E. Calabi Extremal Kahler metrics, Seminar on Diff. Geom., edited by Yau, Ann. of Math. Stud., 102, Princeton Univ. Press, 1982. J. Cheeger and D. Ebin Comparison theorems in Riemannian geometry, North-Holland Pub. Co., 1975. S.K. Donaldson Remarks on Gauge theory, Complex Geometry and 4- manifold Topology, Fields Medallists’ lectures, 384-403, World Sci. Ser. 20th Century Math., 5, World Sci. Publishing. W. Ding and G. Tian Kéhler-Einstein metrics and the generalized Futaki invariants, Invent. Math., 110 (1992), 315-335. A. Fijiki and G. Schumacher The moduli space of Kéhler structures on @ real compact symplectic manifold, Publ. Res. Inst. Math. Sci, 24, 1988, 141-168. A. Futaki An obstruction to the existence of Einstein-Kéhler metrics, Inv. Maih., 73 (1983), 437-443. A. Futaki Kahler-Einstein Metrics and Integral Invariants, Lecture Notes in Mathematics, 1314, Springer-Verlag. P. Griffiths & J. Harris Principles of algebraic geometry, Wiley Classics Library, New York, 1994. D. Gilbarg & N.S. Trudinger Elliptic partial differential equations of second order, Grundlehren der Mathematischen Wissenschaften, Springer-Verlag, Berlin-New York, 1983.20) 21) (22) [23] [24] [25] Bibliography S. Kobayashi, On compact Kahler manifolds with positive definite Ricci tensor, Annals of Mathematics 74, 1961. S. Kobayashi & K. Nomizu, Differential Geometry, Wiley-Interscience New York, 1963. S. Kobayashi & T. Ochiai, Characterizations of complex projective spaces and hyperquadrics, J. Math. Kyoto Univ. 13, 1973. M. Kontsevich, Mirror symmetry in dimension 8, Seminar Bourbaki, Vol. 94/95 No. 237, 1996 (page 286). P. Li, On the Sobolev constant and the p-spectrum of a compact Riemannian manifold, Ann. Sc. E.N.S. Paris 13, 1980. T. Mabuchi, K-energy maps integrating Futaki invariants, Tohoku Math. J., 38, 245 257 (1986). S. Mukai & H. Umemura, Minimal rational threefolds, Lecture notes in Math. 1016, Springer, Berlin-New York, 1983. A. Nadel Multiplier ideal sheaves and existence of Kahler-Einstein metrics of positive scalar curvature, Proc. Natl. Acad. Sci. USA, Vol. 86, No. 19, 1989. Y.T. Siu, The Existence of Kéhler-Einstein Metrics on Manifolds with Pos- itive Anticanonical Line Bundle and a Suitable Finite Symmetry Group, Ann. Math., 127, 1988, 585 627. G. Tian, Kahler-Einstein metrics on algebraic manifolds, Lecture notes in Math. 1646, Springer, Berlin-New York, 1997. G. Tian On Calabi’s conjecture for complex surfaces with positive first Chern class, Inv. Math, Vol. 101, No. 1, 1990, 101-172. G. Tian, Kéhler-Binstein metrics with positive scalar curvature, Inv. Math., 130 (1997), 1-39. G. Tian On Kéhler-Einstein metrics on certain Kahler Manifolds with C1(M) > 0, Invent. Math., 89 (1987), 225-246. G. Tian, G Kihler-Dinstein metrics on complex suifaces with ,(M) pos- itive, Comm. Math. Phy., 112, 1987. 8.7. Yan, On the Ricci curvature of a compact Kihler manifold and the complex Monge-Ampére equation I, Comm. Pure Appl. Math., 31, 1978. S.T. Yau On Calabi’s conjecture and some new results in algebraic geome- try, Proc. Nat. Acad. Sei. USA, 74, 1977, 1798-1799.Index almost complex structure, 1 automorphism group, 63, 85 hisectional curvature, 6 bounded from below, 62 Calabi conjecture, 43, 44 Calabi functional, 42 Calabi-Futaki, 57, 73, 87 Calabi-Futaki invariant, 23, 24, 26, 92 Calabi-Yau spaces, vii Calabi-Yau Theorem, 43, 44, 64 Chern class, 14 complex surfaces, 85, 87 constant bisectional curvature, 6, 8 constant scalar curvature, 93, 94 curvature, 30 determinant line bundle, 97 extremal, 14 extremal metrics, 14 Fermat hypersurface, 85 Fubini-Study metric, 85 G-equivariant embedding, 96 G-equivariant holomorphic fibration, 96 G-equivariant line bundle, 96 grinvariant, 63 generalized energy, 58 Green function, 49 Holder norm, 78 holomorphic, 25 holomorphic degeneration, 89 holomorphic vector field, 14, 23, 29, 31, 62, 65, 89, 94, 97 K-energy, 93, 94, 96 Kahler manifold, 2 Kahler metric, 3, 4 Kahlcr-Einstein, 18, 55 57, 63, 73, 85, 87 K&hler-Einstein manifold, 19 Kéhbler-Einstcin metric, 18, 33, 67, 72, 75, 79, 85-87, 90, 92, 97 maximal compact subgroup, 63 moment map, 35, 42 Monge-Ampére equation, 87 Moser-Trudinger-Onofri inequality, 77 Mumford semistable, 97 Mumford stable, 97 Nirenberg's problem, G1 Poincaré inequality, 82 positive scalar curvature, 57 proper, 62 properness, 72 Ricci flow, 78 Riemannian metric, 1 scalar curvature, 12, 35, 93 semistability, 97 Sobolev inequality, 66, 87 special degeneration. 89. 90 stability, 96, 97 symplectic manifold, 15, 35 Uniformaization Theorem, 8 virtual bundle, 96 weakly Mumford stable, 97