ebook img

K-energy on polarized compactifications of Lie groups PDF

0.67 MB·
by  Yan Li
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview K-energy on polarized compactifications of Lie groups

K-ENERGY ON POLARIZED COMPACTIFICATIONS OF LIE GROUPS YANLI BINZHOU∗ XIAOHUAZHU∗∗ 7 1 0 Abstract. Inthispaper,westudyMabuchi’sK-energyonacompactification 2 M of a reductive Lie group G, which is a complexification of its maximal compact subgroup K. We give a criterion for the properness of K-energy on n thespaceofK×K-invariantK¨ahlerpotentials. Inparticular,itturnstogive a J analternativeproofofDelcroix’stheoremfortheexistenceofK¨ahler-Einstein metricsincaseofFanomanifoldsM. Wealsostudytheexistenceofminimizers 2 ofK-energyforgeneralK¨ahlerclassesofM. ] G D . h 1. Introduction t a ThefamousYau-Tian-Donaldson’sconjecturefortheexistenceofK¨ahler-Einstein m metricsonFanomanifoldsassertsthattheexistenceisequivalenttotheK-stability. [ The conjecture has been recently solved by Tian [25]. Chen, Donaldson and Sun 1 also give an alternative proof [8]. The notion of K-stability was first introduced by v Tian by using special degenerations [23] and then reformulated by Donaldson in 6 0 algebraic geometry via test-configurations [14]. For both special degenerations and 3 test-configurations, one has to study an infinite number of possible degenerations 0 of the manifold. A natural question is how to verify the K-stability by reducing it 0 . to a finite dimensional progress. The answer is known for Fano surfaces by Tian 1 0 [22] and for toric Fano manifolds by Wang and Zhu [29] (see also [30]). In fact, in 7 both cases the existence is equivalent to the vanishing of Futaki invariant. 1 More recently, Delcroix extends Wang-Zhu’s result to a polarized compactifica- : v tionM ofareductiveLiegroupGwithc (M)>0[12]. WecallM a(bi-equivariant) 1 i X compactification of G if it admits a holomorphic G×G action on M with an open r and dense orbit isomorphic to G as a G×G-homogeneous space. (M,L) is called a a polarized compactification of G if L is a G×G-linearized ample line bundle on M. For more examples besides the toric manifolds, see [4, 12, 13]. Let TC be a r-dimensional maximal complex torus of G with dimension n and M its group of characters. Assume that Φ is the root system of (G,TC) in M and Φ is a chosen set of positive roots. Let P be the polytope associated to (M,L), + and P the part of P defined by Φ . Denote by 2P its dilation at rate 2. Let + + + 2000 Mathematics Subject Classification. Primary: 53C25;Secondary: 53C55,58J05,19L10. Key words and phrases. K-energy,Liegroup,Fanomanifolds,K¨ahler-Einsteinmetrics. *PartiallysupportedbyNSFC11571018and11331001. **PartiallysupportedbyNSFCGrants11271022and11331001. 1 2 YANLI BINZHOU∗ XIAOHUAZHU∗∗ ρ = 1(cid:80) α and Ξ be the relative interior of the cone generated by Φ . Then 2 α∈Φ+ + Delcroix proved Theorem 1.1. Let M be a polarized compactification of G with c (M)>0. Then 1 M admits a Ka¨hler-Einstein metric if and only if (1.1) bar ∈4ρ+Ξ, (cid:82) yπ(y)dy where bar = 2P+ is the barycentre of 2P+ with respect to the weighted (cid:82) π(y)dy measure π(y)dy2aPn+d π(y)=(cid:81) (cid:104)α,y(cid:105)2. α∈Φ+ ItispointedbyDelcroixthat(1.1)impliesthattheFutakiinvariantvanishesfor holomorphic vector fields induced by G×G, but the inverse is not true in general. Thus one may ask if (1.1) is related to the K-stability and is determined by a generalized Futaki invariant for some test-configurations. In the present paper, we willanswerthisquestion. Infact,motivatedbythestudyontoricmanifolds[14],we investigatetheK-energyonthespaceofK×K-invariantK¨ahlerpotentialsthrough the reduced K-energy K(·) via Legendre transformation. We show that condition (1.1) comes from our formula of K(·) naturally when c (M) > 0 (cf. Proposition 1 3.1, Proposition 3.4). Moreover, we give an alternative proof of Theorem 1.1 by showing the properness of the K-energy (cf. Section 4). The K¨ahler-Ricci solitons case can be discussed similarly (cf. Section 5). The main purpose of this paper is to give a criterion for the properness of the K-energy on a general polarized compactification (M,L) of G as done on a toric manifold in [33]. We divide ∂(2P )∩∂(2P) into several pieces {F }d0 such that + A A=1 for any A, F lies on an (r−1)-dimensional hyperplane defined by (cid:104)y,u (cid:105) = λ A A A for some primitive u ∈ N, where N is the Z-dual of M. Define a cone by E = A A (cid:91)d0 {ty| t∈[0,1], y ∈F } for any A. It is clear that 2P = E . Let A + A A=1 2 (1.2) Λ = (1+(cid:104)2ρ,u (cid:105)). A λ A A Then the average of scalar curvature S¯ of ω ∈2πc (L) is given by1 0 1 (cid:80) (cid:82) n Λ πdy (1.3) S¯= A(cid:82) A EA . πdy 2P+ Define a weighted barycentre b(cid:102)ar of 2P+ by (cid:80) (cid:82) Λ yπdy (1.4) b(cid:102)ar = (cid:80)A A (cid:82)EA . Λ πdy A A EA Notethatbothbarandb(cid:102)arareinthedualspacea∗ofa,whereaisthenon-compact part of Lie algebra tC of TC. Denote by barss and b(cid:102)arss the projections of bar and b(cid:102)ar on the semisimple part a∗ of a∗, respectively. We prove ss 1(1.3)willbeverifiedattheendofSection2. K-ENERGY ON POLARIZED COMPACTIFICATIONS OF LIE GROUPS 3 Theorem 1.2. Let (M,L) be a polarized compactification of G with vanishing Futaki invariant, and ω ∈2πc (L) a K×K-inariant K¨ahler metric. Suppose that 0 1 the polytope 2P satisfies the following conditions, + (cid:16) (cid:17) (1.5) minΛA·b(cid:102)arss−4ρ ∈Ξ, A (cid:16) (cid:17) (1.6) b(cid:102)arss−barss ∈Ξ¯, (1.7) (n+1)·minΛ −S¯>0. A A Then the K-energy µ (·) is proper on H (ω ) modulo Z(G), where ω0 K×K 0 √ H (ω )={φ∈C∞(M) | ω =ω + −1∂∂¯φ>0 and φ is K×K-invariant } K×K 0 φ 0 and Z(G) is the centre of G. IncasethatM isFanoandL=K−1,thenS¯=nandΛ =1forallA. Wehave M A b(cid:102)ar =bar,thus(1.6),(1.7)areautomaticallysatisfied. Moreover,(1.1)isequivalent to the vanishing of Futaki invariant and (1.5) (cf. Corollary 3.3). Consequently, µ (·) is proper modulo the action of Z(G). Hence we get the an alternative proof ω0 for the sufficient part of Theorem 1.1 [10, 28]. Asmentionedabove,weproveTheorem1.2byusingthereducedK-energyK(·). One of the advantages of K(·) is that it can be defined on a complete space C˜ ∗ of convex functions on 2P . Following the argument in [34], we discuss the semi- + continuity property of K(·). As a consequence, we prove the following Theorem 1.3. K(u) is lower semi-continuous on C˜ . Furthermore, if µ (·) is ∗ ω0 proper on H (ω ) modulo Z(G), then there exists a minimizer of K(·) on C˜ . K×K 0 ∗ It is interesting to study the regularity of minimizers in Theorem 1.3. We guess that they are smooth in 2P if the dimension of the torus TC is less than two. In + case of toric surfaces, it is verified in [31, 32]. The paper is organized as following: In Section 2, we review some preliminaries on K×K-invariant metrics on M, and then we give a formula of scalar curvature of such metrics in terms of Legendre functions. The formula of K(·) is obtained in Section3. InSection4,weusetheideain[33]fortoricmanifoldstoproveTheorem 1.2,buttherearenewdifficultiesarisingfromenergyestimatesneartheWeylwalls to overcome. In Section 5, we focus on the Fano case, and prove the properness of modified K-energy provided a modified barycentre condition (5.2) (cf. Theorem 5.1). In Section 6, we prove Theorem 1.3. 2. Preliminaries In this section, we first recall some preliminaries for K ×K-invariant K¨ahler metrics on a polarized compactification (M,L) of G [11, 12, 13] and the associated Legendre functions, then we give a computation of scalar curvature in terms of Legendre functions. 4 YANLI BINZHOU∗ XIAOHUAZHU∗∗ 2.1. Polarized compactification. Let J be the complex structure of G and K be one of its maximal compact subgroup such that G=KC. Choose T a maximal torusofK. Denotebyg,k,tthecorrespondingLiealgebraofG, K, T,respectively. Then g=k⊕Jk. Set a=Jt and Lie algebra of Z(G) by z(g). We decompose a as a toric part and a semisimple part: a=a ⊕a , t ss where a :=z(g)∩a and a :=a∩[g,g]. Then for any x∈a, we have x=x +x t ss t ss with x ∈a and x ∈a . We extend the Killing form on a to a scalar product t t ss ss ss (cid:104)·,·(cid:105)onasuchthata isorthogonaltoa . Identifyaanditsduala∗ by(cid:104)·,·(cid:105). Then t ss a∗ also has an orthogonal decomposition a∗ =a∗⊕a∗ . t ss Denote by Φ and W the root system and Weyl group with respect to (G,TC), respectively. ChooseasystemofpositiverootsΦ . ThenitdefinesapositiveWeyl + chamber a ⊂a, and a positive Weyl chamber a∗ on a∗, where + + a∗ :={y| α(y):=(cid:104)α,y(cid:105)>0, ∀α∈Φ }, + + which is also called the relative interior Ξ of the cone generated by Φ . The Weyl + wall W is defined by W :={y| α(y)=0} for each α∈Φ . α α + 2.2. K ×K-invariant K¨ahler metrics. Let Z be the closure of TC in M. It is known that (Z,L| ) is a polarized toric manifold with a W-action, and L| is Z Z a W-linearized ample toric line bundle on Z [2, 3, 4, 12]. Let ω ∈ 2πc (L) be a 0 1 K×K-invariantK¨ahlerforminducedfrom(M,L)andP bethepolytopeassociated to(Z,L| ),whichisdefinedbythemomentmapassociatedtoω . ThenP isaW- Z 0 invariantdelzentpolytopeina∗. BytheK×K-invariance,foranyφ∈H (ω ), K×K 0 the restriction of ω on Z is a toric K¨ahler metric. It induces a smooth strictly φ convex function ψ on a, which is W-invariant [5]. By the KAK-decomposition ([21], Theorem 7.39), for any g ∈ G, there are k , k ∈ K and x ∈ a such that g = k exp(x)k . Here x is uniquely determined 1 2 1 2 up to a W-action. This means that x is unique in a¯ . Then we define a smooth + K×K-invariant function Ψ on G by Ψ(exp(·))=ψ(·): a→R. Clearly Ψ is well-defined since ψ is W-invariant. We usually call ψ the function associated to Ψ. It can be verified that Ψ is a K¨ahler potential on G such that √ ω = −1∂∂¯Ψ on G (cf. Lemma 2.2 below). The following KAK-integral formula can be found in [20], Proposition 5.28 (see also [19]) Proposition 2.1. Let dV be a Haar measure on G and dx the Lebesgue measure G on a. Then there exists a constant C > 0 such that for any K ×K-invariant, H K-ENERGY ON POLARIZED COMPACTIFICATIONS OF LIE GROUPS 5 dV -integrable function Ψ on G, G (cid:90) (cid:90) Ψ(g)dV =C J(x)ψ(x)dx, G H G a+ where J(x)=(cid:81) sinh2(α(x)). α∈Φ+ Next we recall the local holomorphic coordinates on G used in [12]. By the standard Cartan decomposition, we can decompose g as g=(t⊕a)⊕(⊕ V ), α∈Φ α where V = {X ∈ g| ad (X) = α(H)X, ∀H ∈ t⊕a}, the root space of complex α H dimension 1 with respect to α. By [18], one can choose X ∈V such that X = α α −α −ι(X ) and [X ,X ]=α∨, where ι is the Cartan involution and α∨ is a dual of α α −α α bytheKillingform. LetE :=X −X andE :=J(X +X ). Denoteby α α −α −α α −α k , k the real line spanned by E , E , respectively. Then we have the Cartan α −α α −α decomposition of k, (cid:0) (cid:1) k=t⊕ ⊕ (k ⊕k ) . α∈Φ+ α −α Choosearealbasis{E0,...,E0}oft. Then{E0,...,E0}togetherwith{E ,E } 1 r 1 r α −α α∈Φ+ forms a real basis of k, which is indexed by {E ,...,E }. {E ,...,E } can also be 1 n 1 n regarded as a complex basis of g. For any g ∈ G, we define local coordinates {zi } on a neighborhood of g by (g) i=1,...,n (zi )→exp(zi E )g. (g) (g) i It is easy to see that θi| = dzi | , where θi is the dual of E , which is a right- g (g) g i (cid:16) (cid:17) invariant holomorphic 1-form. Thus ∧n dzi ∧dzi¯ | is also a right-invariant i=1 (g) (g) g (n,n)-form, which defines a Haar measure dV . G The complex Hessian of the K × K-invariant function Ψ in the above local coordinates was computed by Delcroix as follows [12, Theorem 1.2]. Lemma 2.2. Let Ψ be a K ×K invariant function on G, and ψ the associated function on a. Let Φ ={α ,...,α }. Then for x∈a , the complex Hessian + (1) (n−r) + 2 matrix of Ψ in the above coordinates is diagonal by blocks, and equals to 1HessR(ψ)(x) 0 0  4  0 Mα(1)(x) 0  (2.1) HessC(Ψ)(exp(x))= 0 0 ... ... ,  ... ... ... 0    0 0 M (x) α(n−2r) where √ (cid:18) (cid:19) 1 cothα (x) −1 M (x)= (cid:104)α ,∇ψ(x)(cid:105) √(i) . α(i) 2 (i) − −1 cothα (x) (i) 6 YANLI BINZHOU∗ XIAOHUAZHU∗∗ By (2.1) in Lemma 2.2, we see that ψ is convex on a. The complex Monge- √ Amp´ere measure is given by ωφn =( −1∂∂¯Ψ)n =MAC(Ψ)dVG, where 1 1 (cid:89) (2.2) MAC(Ψ)(exp(x))= MAR(ψ)(x) (cid:104)α,∇ψ(x)(cid:105)2. 4r+p J(x) α∈Φ+ 2.3. Legendre functions. By the convexity of ψ on a, the gradient ∇ψ defines a diffeomorphism from a to the interior of the dilated polytope 2P2. Let P := + P ∩a¯∗, then by the W-invariance of ψ and P, the restriction of ∇ψ to a is a + + diffeomorphism to the interior of 2P . We note that one part of ∂(2P ) lies on + + ∂(2P) (which we call ”outer faces”) and the other part lies on Weyl walls {W }. α For simplicity, we may assume that 2P contains the origin O in its interior. Then 2P can be described as the intersection of l (y):=−ui y +λ >0, A˜=1,...,d, A˜ A˜ i A˜ where λ >0 and u are primitive vectors in N. A˜ A˜ Recall that Guillemin’s function of 2P is given by 1(cid:88) (2.3) u = l (y)logl (y). 0 2 A˜ A˜ A˜ Set C ={v| v is strictly convex, v−u ∈C∞(2P) and v is W-invariant} ∞,W 0 and C ={v | v ∈C }. ∞,+ |2P+ ∞,W By [17], the Legendre function u of ψ belongs to C . The inverse is also true. ∞,W This means that any u ∈ C corresponds to a K¨ahler potential in H (ω ) ∞,W K×K 0 (cf. [4, Proposition 3.2]). By a direct computation, we have (2.4) u =−1(cid:88)(logl (y)+1)ui , u = 1(cid:88)uiA˜ujA˜. 0,i 2 A˜ A˜ 0,ij 2 l (y) A˜ A˜ A˜ Note that uijν →0 as y →F , where (uij)=(u )−1 and ν =(ν ,...,ν ) is the 0 i A˜ 0 0,ij A˜ 1 r unit normal vector of face F = {y| l (y) = 0}. Similarly −uij ν → 2 (cid:104)y,ν (cid:105), A˜ A˜ 0,j i λA˜ A˜ where uij = ∂ui0j. Thus we get 0,k ∂yk Lemma 2.3. If u∈C , then for any A˜, as y →F , ∞,W A˜ 2 (2.5) uijν →0 and uijν → (cid:104)y,ν (cid:105), i ,j i λ A˜ A˜ where (uij)=(u )−1 and uij = ∂uij. ,ij ,k ∂yk 2Weremarkthatthemomentmapisgivenby 1∇ψ,whoseimageisP. 2 K-ENERGY ON POLARIZED COMPACTIFICATIONS OF LIE GROUPS 7 2.4. The scalar curvature. We compute the Ricci curvature of ω . Clearly it φ is also K ×K-invariant. As in Lemma 2.2, in the local coordinates in Sect. 2.2, Ric(ω ) can be expressed as φ −HessC(logdet(∂∂¯Ψ))(exp(x)) 1HessR(ψ˜)(x) 0 0  4  0 M˜α(1)(x) 0  =− 0 0 ... ...     ... ... ... 0    0 0 M˜ (x) α(n−2r) for any x∈a , where + ψ˜ = logdet(∇2ψ)+2 (cid:88) logα(∇ψ)+χ(x), α∈Φ+ (cid:88) χ(x) = −logJ(x)=−2 logsinhα(x), α∈Φ+ √ 1(cid:68) (cid:69)(cid:18)cothα(x) −1 (cid:19) M˜ (x) = α,∇ψ˜ √ . α 2 − −1 cothα(x) Then the scalar curvature (2.6) S(ω )| =tr(cid:16)∇2ψ)−1∇2ψ˜(cid:17)+ (cid:88) (cid:104)α,∇ψ˜(cid:105). φ exp(x) (cid:104)α,∇ψ(cid:105) α∈Φ+ By using the Legendre function u, we get Lemma 2.4.   S(ωφ)=−(cid:88)ui,ijj +4 (cid:88) ααi(uyi,)jj +4 (cid:88) αα(iyβ)jβu(iyj) −2 (cid:88) α(αiα(yj)u)i2j i,j α∈Φ+ α,β∈Φ+ α∈Φ+ (2.7) −(cid:88)u,ik ∂x∂i2∂χxk(cid:12)(cid:12)(cid:12)(cid:12) −2(cid:88) (cid:88) ∂∂xχi(cid:12)(cid:12)(cid:12)(cid:12) αα(yi), i,k x=∇u i α∈Φ+ x=∇u where y ∈2P , uij = ∂2uij and α are the components of α. + ,kl ∂yk∂yl i Proof. By the relations (∇2u)−1|y =(∇2ψ)|x=∇u, ∂xi∂∂x3ψj∂xk(cid:12)(cid:12)(cid:12)(cid:12)x = ∂∂xi (cid:0)ujk|y=∇ψ(cid:1)= uj,lkuli(cid:12)(cid:12)(cid:12)y=∇ψ, we have ∂∂xψ˜p(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)x=∇u =u,ijui,kjukp+2α(cid:88)∈Φ+ ααl(uyl)p + ∂∂xχp(cid:12)(cid:12)(cid:12)(cid:12)x=∇u, ∂x∂p2∂ψ˜xq(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)x=∇u =(u,ijui,kjukp),susq+2α(cid:88)∈Φ+(cid:18)ααl(uyl)p(cid:19),susq+ ∂x∂p2∂χxq(cid:12)(cid:12)(cid:12)(cid:12)x=∇u. Substituting them into (2.6), we obtain (2.7) immediately. (cid:3) 8 YANLI BINZHOU∗ XIAOHUAZHU∗∗ Note π(y)=(cid:81) (α(y))2. Since α∈Φ+ ∂π(y)=2π(y) (cid:88) αi , ∂y α(y) i α∈Φ+   (2.8) ∂2π (y)=π(y)4 (cid:88) αiβj −2 (cid:88) αiαj , ∂y ∂y α(y)β(y) (α(y))2 i j α,β∈Φ+ α∈Φ+ we can rewrite S as π π S(ω )=−uij −2uij ,i −uij ,ij φ ,ij ,j π π (2.9) −u,ik ∂x∂i2∂χxk(cid:12)(cid:12)(cid:12)(cid:12) − ∂∂xχi(cid:12)(cid:12)(cid:12)(cid:12) ππ,i. x=∇u x=∇u By Proposition 2.1, it follows (cid:90) (cid:90) (cid:90) (cid:89) Sωn =C Sdet(∇2ψ) (cid:104)α,∇ψ(cid:105)2dx=C Sπdy. φ H H M a+ α∈Φ+ 2P+ Since π ≡0 on each W , by integration by parts on (2.9), we get α (cid:90) Sπdy 2P+ (cid:90) (cid:90) (cid:90) =− uijν πdσ − uijπ dy− uijπ dy ,j i 0 ,j ,i ,ij ∂(2P+) 2P+ 2P+ (cid:90) ∂ (cid:18) ∂χ (cid:12)(cid:12) (cid:19) (cid:90) ∂χ (cid:12)(cid:12) − ∂y ∂xj(cid:12)(cid:12) πdy− ∂xj(cid:12)(cid:12) π,idy 2P+ i x=∇u 2P+ x=∇u (cid:88)(cid:90) (cid:18) 2 (cid:19) (cid:88) (cid:90) = (cid:104)y,ν (cid:105)+4(cid:104)ρ,ν (cid:105) πdσ = Λ (cid:104)y,ν (cid:105)πdσ λ A A 0 A A 0 A FA A A FA (cid:90) (cid:88) =n Λ πdy. A A EA HereweusedLemma2.3andthefactthat ∂χ(x)→−4ρ asx→∞. Ontheother ∂xi i hand, by Proposition 2.1, the volume of (M,ω ) is given by φ (cid:90) (cid:90) (cid:89) VM := ωφn = CH MAR(ψ) (cid:104)α,∇ψ(cid:105)2dx M a+ α∈Φ+ (cid:90) = C πdy. H 2P+ Hence, combining the above two relations, we get (1.3). 3. Reduction of the K-Energy Let(M,L)andω ∈2πc (L)beasbefore. DenotebyH(ω )thespaceofK¨ahler 0 1 0 potentials in [ω ]. Mabuchi’s K-energy is defined on H(ω ) by 0 0 1 (cid:90) 1(cid:90) (3.1) µ (φ)=− φ˙ (S(ω )−S¯)ωn ∧dt, ω0 V t φt φt M 0 M K-ENERGY ON POLARIZED COMPACTIFICATIONS OF LIE GROUPS 9 whereV =(cid:82) ωn,S¯istheaverageofS(ω )and{φ }isapathofK¨ahlerpotentials M M 0 0 t joining0andφinH(ω ). Inthissection,wegiveaformulaofµ (·)onH (ω ) 0 ω0 K×K 0 in terms of the Legendre function u. 3.1. Reduced K-energy. Define (cid:90) (cid:90) K(u)=(cid:88) Λ (cid:104)y,ν (cid:105)uπdσ − S¯uπdy A A 0 A FA 2P+ (cid:90) (cid:90) − logdet(u )πdy+ χ(∇u)πdy, ij 2P+ 2P+ (cid:80) where χ(x)=−logJ(x)=−2 logsinhα(x) for any x∈a. Then we have α∈Φ+ Proposition3.1. Letφ∈H (ω )andubetheLegendrefunctionofψ =ψ +φ. K×K 0 0 Then 1 µ (φ)= K(u)+const., ω0 V (cid:82) where V = πdy. 2P+ Proof. Note φ˙ =−u˙ . By (2.2), it is easy to see t t 1 (cid:90) 1(cid:90) (cid:90) S¯φ˙ ωn ∧dt= S¯uπdy. C t φt H 0 M 2P+ Then by (2.9), it suffices to compute the part 1 (cid:90) 1(cid:90) I := − φ˙ S(ω )ωn ∧dt C t φt φt H 0 M (cid:90) 1(cid:90) = − φ˙tS(ωφt)|exp(x)MAR(ψt)|x (cid:89) (cid:104)α,∇ψt(cid:105)2|xdx∧dt. 0 a+ α∈Φ+ (cid:90) 1(cid:90) = u˙ (−u ij )πdy∧dt t t,ij 0 2P+ (cid:90) 1(cid:90) (cid:90) 1(cid:90) + u˙ (−2u ijπ )dy∧dt+ u˙ (−uijπ )dy∧dt t t,j ,i t t ,ij 0 2P+ 0 2P+   (cid:90) 1(cid:90) ∂2χ (cid:88) ∂χ + u˙t−ut,ik∂xi∂xk|x=∇utπ− ∂xi|x=∇utπ,i dy∧dt. 0 2P+ α∈Φ+ By integration by parts, it follows (cid:90) 1(cid:90) (cid:90) 1(cid:90) I = u˙ (−u ijν )πdσ ∧dt+ u˙ u ijπdy∧dt t t,j i 0 t,i t,j 0 ∂(2P+) 0 2P+ (cid:90) 1(cid:90) (cid:90) 1(cid:90) − u˙ (u ijπ )dy∧dt− u˙ (uijπ )dy∧dt t t,j ,i t t ,ij 0 2P+ 0 2P+ (cid:90) 1(cid:90) ∂ (cid:32) ∂χ(cid:12)(cid:12) (cid:33) (3.2) − u˙t∂y ∂xi(cid:12)(cid:12) π dy∧dt. 0 2P+ i x=∇ut 10 YANLI BINZHOU∗ XIAOHUAZHU∗∗ Note that ∂χ(x) → −4ρ as x → ∞ in a and is away from Weyl walls, and π ∂xi i + vanishes quadratically along any Weyl wall. Then the last term in (3.2) becomes (cid:90)2P+χ(∇ut)πdy(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)10−(cid:90)01(cid:90)∂(2P+)u˙t ∂∂xχi(cid:12)(cid:12)(cid:12)(cid:12)x=∇utνiπdσ0∧dt (cid:90) (cid:90) (3.3) = χ(∇u)πdy+4 (cid:104)ρ,ν(cid:105)uπdσ +const. 0 2P+ ∂(2P+) On the other hand, by the second relation in Lemma 2.3, we have (cid:90) 1(cid:90) (cid:90) 1(cid:90) u˙ u ijπdy∧dt− u˙ (u ijπ )dy∧dt t,i t,j t t,j ,i 0 2P+ 0 2P+ (cid:90) 1(cid:90) (cid:90) 1(cid:90) = u˙ u ijν πdσ ∧dt− u˙ u ijπdy∧dt t,i t j 0 t,ij t 0 ∂(2P+) 0 2P+ (cid:90) 1(cid:90) (cid:90) 1(cid:90) + u˙ u ijπ dy∧dt− u˙ u ijν π dσ ∧dt t t ,ij t t j ,i 0 0 2P+ 0 ∂(2P+) (cid:90) 1(cid:90) d (cid:90) 1(cid:90) (3.4) =− [logdet(u )]πdy∧dt+ u˙ u ijπ dσ ∧dt. dt t,ij t t ,ij 0 0 2P+ 0 ∂(2P+) Thus combining (3.4) and (3.3), we get from (3.2), (cid:90) 1(cid:90) (cid:90) I = u˙ (−u ijν )πdσ ∧dt+4 (cid:104)ρ,ν(cid:105)uπdσ t t,j i 0 0 0 ∂(2P+) ∂(2P+) (cid:90) (cid:90) − logdet(u )πdy+ χ(∇u)πdy+const. ,ij 2P+ 2P+ By Lemma 2.3, we see (cid:90) (cid:88)(cid:90) 2 u˙ (−u ijν )πdσ = u˙ (cid:104)y,ν (cid:105)πdσ . t t,j j 0 tλ A 0 ∂(2P+) A FA A Hence, we obtain (cid:90) (cid:90) (cid:88) I = Λ (cid:104)y,ν (cid:105)uπdσ − [logdet(u )−χ(∇u)]πdy+const. A A 0 ij A FA 2P+ Recall that V =C ·V, the proof is finished. (cid:3) M H For convenience, we write K(u) as K(u)=L(u)+N(u), where (cid:90) (cid:90) (cid:90) (3.5) L(u)=(cid:88) Λ (cid:104)y,ν (cid:105)uπdσ − S¯uπdy− 4(cid:104)ρ,∇u(cid:105)πdy, A A 0 A FA 2P+ 2P+ (cid:90) (cid:90) (3.6) N(u)=− logdet(u )πdy+ [χ(∇u)+4(cid:104)ρ,∇u(cid:105)]πdy. ,ij 2P+ 2P+ By integration by parts, we can rewrite L(u) as (cid:90) (3.7) L(u)=(cid:88) (cid:2)(cid:104)Λ y−4ρ,∇u(cid:105)+(Λ n−S¯)u(cid:3)πdy, A A A EA

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.