ScalingStacks

A priori $L^{\infty}$-estimates for degenerate complex Monge-Amp\`ere equations

Eyssidieux, P. · Guedj, V. · Zeriahi, A.

Original paper

Original official author HTML, exact retained edition. Historical TeX conversion verdicts remain unchanged. Cited-edition alignment and mathematical self-containment are not assessed.

A priori L∞L^{\infty}-estimates for degenerate complex Monge-Ampère equations

P. Eyssidieux    V. Guedj    A. Zeriahi

Abstract : We study families of complex Monge-Ampère equations, focusing on the case where the cohomology classes degenerate to a non big class. We establish uniform a priori L∞L^{\infty}-estimates for the normalized solutions, generalizing the recent work of S. Kolodziej and G. Tian. This has interesting consequences in the study of the Kähler-Ricci flow.

[02G7]

1 Introduction

Let π:X⟶Y\pi:X\longrightarrow Y be a non degenerate holomorphic mapping between compact Kähler manifolds such that n:=d​i​mℂ​X≥m:=d​i​mℂ​Yn:=dim_{\mathbb{C}}X\geq m:=dim_{\mathbb{C}}Y. Let ωX\omega_{X}, ωY\omega_{Y} Kähler forms on XX and YY respectively. Let F:X⟶ℝ+F:X\longrightarrow\mathbb{R}^{+} be a non negative function such that F∈Lp​(X)F\in L^{p}(X) for some p>1p>1.

Set ωt:=π∗​(ωY)+t​ωX\omega_{t}:=\pi^{*}(\omega_{Y})+t\omega_{X}, t>0t>0. We consider the following family of complex Monge-Ampère equations

(⋆)t {(ωt+d​dc​φt)n=ct​tn−m​F​ωXnmaxX⁡φt=0= 0\left\{\begin{array}[]{ll}(\omega_{t}+dd^{c}\varphi_{t})^{n}&=\ c_{t}t^{n-m}F\omega_{X}^{n}\\ \max_{X}\varphi_{t}=0&=\ 0\\ \end{array}\right.

where φt\varphi_{t} is ωt−\omega_{t}-plurisubharmonic on XX and ct>0c_{t}>0 is a constant given by

ct​tn−m​∫XF​ωXn=∫Xωtn.c_{t}t^{n-m}\int_{X}F\omega_{X}^{n}=\int_{X}\omega_{t}^{n}.

It follows from the seminal work of S.T. Yau [Y] and S. Kolodziej [K 1], [K 2] that the equation (⋆)t(\star)_{t} admits a unique continuous solution. (Observe that for t∈]0,1]t\in]0,1], ωt\omega_{t} is a Kähler form).

Our aim here is to understand what happens when t→0+t\to 0^{+}, motivated by recent geometrical developpments [ST], [KT]. When n=mn=m, the cohomology class ω0{\omega_{0}} is big and semi-ample and this problem has been adressed by several authors recently (see [CN], [EGZ], [TZ], [To]).

We focus here on the case m<nm<n. This situation is motivated by the study of the Kähler-Ricci flow on manifolds XX of intermediate Kodaira dimension 1≤k​o​d​(X)≤n−11\leq kod(X)\leq n-1. When n=2n=2 this has been studied by J.Song and G.Tian [ST].

In a very recent and interesting paper [KT], S. Kolodziej and G. Tian were able to show, under a technical geometric assumption on the fibration π\pi, that the solutions (φt)(\varphi_{t}) are uniformly bounded on XX when t↘0+t\searrow 0^{+}.

The purpose of this note is to (re)prove this result without any technical assumption and with a different method: we actually follow the strategy introduced by S. Kolodziej in [K] and further developped in [EGZ], [BGZ].

THEOREM. There exists a uniform constant M=M⁡(π,‖F‖p)>0M=M(\pi,\|F\|_{p})>0 such that the solutions to the Monge-Ampère equations (⋆)t(\star)_{t} satisfy

∥φt∥L∞​(X)≤M,∀t∈]0,1].\|\varphi_{t}\|_{L^{\infty}(X)}\ \leq\ M,\ \forall t\in]0,1].

It follows from our result that Theorems 1 and 2 in [KT] hold without any technical assumption on the fibration (see condition 0.2 in [KT]).

This result has been announced by J-P. Demailly and N. Pali [DP].

[02G8]

2 Proof of the theorem

[02G9]

2.1 Preliminary remarks

Uniform control of ctc_{t}. Observe that ω0k=0\omega_{0}^{k}=0 for m<k≤nm<k\leq n, hence for all t∈]0,1]t\in]0,1],

ωtn=∑k=1m(nk)​tn−k​ω0k∧ωXn−k.\omega_{t}^{n}=\sum_{k=1}^{m}{n\choose k}t^{n-k}{\omega_{0}}^{k}\wedge{\omega_{X}}^{n-k}.

Note that ]0,1]∋t⟼tm−nωtn]0,1]\ni t\longmapsto t^{m-n}\omega_{t}^{n} is increasing (hence decreases as t↘0+t\searrow 0^{+}) and satisfies for t∈]0,1]t\in]0,1]

(1) (nm)ω0m∧ωXn−m∫Xω0m∧ωXn−m≤ωtntn−m​∫Xω0m∧ωXn−m≤ω1n∫Xω0m∧ωXn−m.{n\choose m}\qquad\frac{\omega_{0}^{m}\wedge\omega_{X}^{n-m}}{\int_{X}\omega_{0}^{m}\wedge\omega_{X}^{n-m}}\leq\frac{\omega_{t}^{n}}{t^{n-m}\int_{X}\omega_{0}^{m}\wedge\omega_{X}^{n-m}}\leq\frac{\omega_{1}^{n}}{\int_{X}\omega_{0}^{m}\wedge\omega_{X}^{n-m}}.

In particular t⟼ctt\longmapsto c_{t} is increasing in t∈]0,1]t\in]0,1] and

0<(nm)∫Xω0m∧ωXn−m∫XF​ωXn=:c0≤ct≤c1.0<{n\choose m}\frac{\int_{X}\omega_{0}^{m}\wedge\omega_{X}^{n-m}}{\int_{X}F\omega_{X}^{n}}=:c_{0}\ \ \leq\ \ c_{t}\leq\ \ c_{1}.

Uniform control of densities. Let JπJ_{\pi} denote the (modulus square) of the Jacobian of the mapping π\pi, defined through

ω0m∧ωXn−m=Jπ​ωXn.\omega_{0}^{m}\wedge\omega_{X}^{n-m}=J_{\pi}\omega_{X}^{n}.

Let us rewrite the equation (⋆)t(\star)_{t} as follows

(ωt+d​dc​φt)n=ft​ωtn,(\omega_{t}+dd^{c}\varphi_{t})^{n}=f_{t}\omega_{t}^{n},

where for t∈]0,1]t\in]0,1]

0≤ft:=ct​tn−m​F​ωXnωtn≤c1​FJπ.0\leq f_{t}:=c_{t}t^{n-m}F\frac{\omega_{X}^{n}}{\omega_{t}^{n}}\leq c_{1}\frac{F}{J_{\pi}}.

Observe that

∫Xft​ωtn=ct​tn−m​∫XF​ωtn=∫Xωtn=:V​o​lωt​(X),\int_{X}f_{t}\omega_{t}^{n}=c_{t}t^{n-m}\int_{X}F\omega_{t}^{n}=\int_{X}\omega_{t}^{n}=:Vol_{\omega_{t}}(X),

hence (ft)(f_{t}) is uniformly bounded in L1​(ωt/Vt)L^{1}(\omega_{t}/V_{t}), Vt:=V​o​lωt​(X)V_{t}:=Vol_{\omega_{t}}(X). We actually need a slightly stronger information.

[02GA]
Lemma 2.1

There exists p′>1p^{\prime}>1 and a constant C=C⁡(π,‖F‖Lp​(X))>0C=C(\pi,\|F\|_{L^{p}(X)})>0 such that for all t∈]0,1]t\in]0,1]

∫Xftp′​ωtn≤C​V​o​lωt​(X).\int_{X}f_{t}^{p^{\prime}}\omega_{t}^{n}\leq C\,Vol_{\omega_{t}}(X).

Proof of the lemma. Set Vt:=V​o​lωt=∫XωtnV_{t}:=Vol_{\omega_{t}}=\int_{X}\omega_{t}^{n} and observe that

0≤ft​ωtnVt≤c1​F​ωXn∫Xω0m∧ωXn−m=C2​F​ωXn,0\leq f_{t}\frac{\omega_{t}^{n}}{V_{t}}\leq c_{1}F\frac{\omega_{X}^{n}}{\int_{X}\omega_{0}^{m}\wedge\omega_{X}^{n-m}}=C_{2}F\omega_{X}^{n},

where C2:=c1​∫XJπ​ωXnC_{2}:=c_{1}\int_{X}J_{\pi}\omega_{X}^{n}.

This shows that the densities ftf_{t} are uniformly in L1L^{1} w.r.t. the normalized volume fomrs ωtn/Vt\omega_{t}^{n}/\penalty V_{t}.

Since JπJ_{\pi} is locally given as the square of the modulus of a holomorphic function which does not vanish identically, there exists α∈]0,1[\alpha\in]0,1[ such that Jπ−α∈L1​(X)J_{\pi}^{-\alpha}\in L^{1}(X). Fix β∈]0,α[\beta\in]0,\alpha[ satisfying the condition β/p+β/α=1\beta/\penalty p+\beta/\penalty\alpha=1. It follows from Hölder’s inequality that

∫Xftβ​ωXn≤(∫XFp​ωXn)β/p​(∫XJπ−α​ωXn)β/α.\int_{X}f_{t}^{\beta}\omega_{X}^{n}\leq\Bigl(\int_{X}F^{p}\omega_{X}^{n}\Bigr)^{\beta/\penalty p}\Bigl(\int_{X}J_{\pi}^{-\alpha}\omega_{X}^{n}\Bigr)^{\beta/\penalty\alpha}.

Setting ε:=β/q\varepsilon:=\beta/\penalty q and using Hölder’s inequality again , we obtain

∫Xft1+ε​ωtnVt≤C2​∫Xftε​F​ωXn.\int_{X}f_{t}^{1+\varepsilon}\frac{\omega_{t}^{n}}{V_{t}}\leq C_{2}\int_{X}f_{t}^{\varepsilon}F\omega_{X}^{n}.

Now applying again Hölder inequality we get

∫Xft1+ε​ωtnVt≤C2​(∫Xftβ​ωX)1/q​‖F‖Lp​(X).\int_{X}f_{t}^{1+\varepsilon}\frac{\omega_{t}^{n}}{V_{t}}\leq C_{2}\Bigl(\int_{X}f_{t}^{\beta}\omega_{X}\Bigr)^{1/\penalty q}\|F\|_{L^{p}(X)}.

Therefore denoting by p′:=1+εp^{\prime}:=1+\varepsilon, we have the following uniform estimate

∫Xftp′ωtnVt≤C(π,∥F∥Lp​(X)),∀t∈]0,1],\int_{X}f_{t}^{p^{\prime}}\frac{\omega_{t}^{n}}{V_{t}}\leq C(\pi,\|F\|_{L^{p}(X)}),\forall t\in]0,1],

where

C⁡(π,‖F‖Lp​(X)):=C2​(∫XJπ−α​ωXn)β/α​q​‖F‖Lp​(X)1+β/q.C(\pi,\|F\|_{L^{p}(X)}):=C_{2}\Bigl(\int_{X}J_{\pi}^{-\alpha}\omega_{X}^{n}\Bigr)^{\beta/\penalty\alpha q}\|F\|_{L^{p}(X)}^{1+\beta/\penalty q}.

▶\blacktriangleright

[02GB]

2.2 Uniform domination by capacity

We now show that the measure μt:=ft​ωtn/V​o​lωt\mu_{t}:=f_{t}\omega_{t}^{n}/\penalty Vol_{\omega_{t}} are uniformly strongly dominated by the normalized capacity Capωt/V​o​lωt​(X).\mathrm{Cap}_{\omega_{t}}/\penalty Vol_{\omega_{t}}(X). It actually follows from a carefull reading of the no parameter proof given in [EGZ], [BGZ].

[02GC]
Lemma 2.2

There exists a constant C0=C0​(π,‖F‖Lp​(ωXn))>0C_{0}=C_{0}(\pi,\|F\|_{L^{p}(\omega_{X}^{n})})>0 such that for any compact set K⊂XK\subset X and t∈]0,1]t\in]0,1],

μt​(K)≤C0n​(Capωt​(K)V​o​lωt​(X))2.\mu_{t}(K)\leq C_{0}^{n}\left(\frac{\mathrm{Cap}_{\omega_{t}}(K)}{Vol_{\omega_{t}}(X)}\right)^{2}.

Proof: Fix a compact set K⊂XK\subset X. Set Vt:=V​o​lωt​(X)V_{t}:=Vol_{\omega_{t}}(X). Hölder’s inequality yields

μt​(K)≤(∫Xftp′​ωtnVt)1/p′​(∫KωtnVt)1/q′.\mu_{t}(K)\leq\left(\int_{X}f_{t}^{p^{\prime}}\frac{\omega_{t}^{n}}{V_{t}}\right)^{1/\penalty p^{\prime}}\left(\int_{K}\frac{\omega_{t}^{n}}{V_{t}}\right)^{1/\penalty q^{\prime}}.

It remains to dominate uniformly the normalized volume forms ωtn/Vt\omega_{t}^{n}/\penalty V_{t} by the normalized capacities Capωt/Vt\mathrm{Cap}_{\omega_{t}}/\penalty V_{t}. Fix σ>0\sigma>0 and observe that for any t∈]0,1]t\in]0,1],

∫KωtnVt≤∫Xe−σ⁡(VK,ωt−maxX⁡VK,ωt)​ωtnVt​Tωt​(K)σ,\int_{K}\frac{\omega_{t}^{n}}{V_{t}}\leq\int_{X}e^{-\sigma(V_{K,\omega_{t}}-\max_{X}V_{K,\omega_{t}})}\frac{\omega_{t}^{n}}{V_{t}}T_{\omega_{t}}(K)^{\sigma},

where

VK,ωt:=sup{ψ∈P​S​H​(X,ωt);ψ≤0,on​K}V_{K,\omega_{t}}:=\sup\{\psi\in PSH(X,\omega_{t});\psi\leq 0,\ \mathrm{on}\ K\}

is the ωt−\omega_{t}-extremal function of KK and Tωt(K):=exp(−supXVK,ωt)T_{\omega_{t}}(K):=\exp(-\sup_{X}V_{K,\omega_{t}}) is the associated ωt−\omega_{t}-capacity of KK (see [GZ 1] for their properties).

Observe that ωtn/Vt≤c1​ω1n\omega_{t}^{n}/\penalty V_{t}\leq c_{1}\omega_{1}^{n} and ωt≤ω1\omega_{t}\leq\omega_{1}, hence the family of functions VK,ωt−maxX⁡VK,ωtV_{K,\omega_{t}}-\max_{X}V_{K,\omega_{t}} is a normalized family of ω1−\omega_{1}-psh functions. Thus there exists σ>0\sigma>0 which depends only on (X,ω1)(X,\omega_{1}) and a constant B=B⁡(σ,X,ω1)B=B(\sigma,X,\omega_{1}) such that ([Z])

∫Xe−σ⁡(VK,ωt−maxX⁡VK,ωt)ωtnVt≤B,∀t∈]0,1].\int_{X}e^{-\sigma(V_{K,\omega_{t}}-\max_{X}V_{K,\omega_{t}})}\frac{\omega_{t}^{n}}{V_{t}}\leq B,\forall t\in]0,1].

The Alexander-Taylor comparison theorem (see Theorem 7.1 in [GZ 1]) now yields for a constant C3=C3​(π,‖F‖Lp​(X))C_{3}=C_{3}(\pi,\|F\|_{L^{p}(X)})

μt(K)≤C3exp[−σ(VtCapωt​(K))1/n],∀t∈]0,1].\mu_{t}(K)\leq C_{3}\exp\left[-\sigma\left(\frac{V_{t}}{\mathrm{Cap}_{\omega_{t}}(K)}\right)^{1/\penalty n}\right],\forall t\in]0,1].

We infer that there is a constant C4=C4​(π,‖F‖Lp​(X))C_{4}=C_{4}(\pi,\|F\|_{L^{p}(X)}) such that

(2) μt(K)≤C4(Capωt​(K)Vt)2,∀t∈]0,1].\mu_{t}(K)\leq C_{4}\left(\frac{\mathrm{Cap}_{\omega_{t}}(K)}{V_{t}}\right)^{2},\forall t\in]0,1].
[02GD]

2.3 Uniform normalization

The comparison principle (see [K] ,[EGZ]) yields for any s>0s>0 and τ∈[0,1]\tau\in[0,1]

τnCapωt({φt≤−s−τ})Vt≤∫{φt≤−s}(ωt+d​dc​φt)nVt.\tau^{n}\frac{\mathrm{Cap}_{\omega_{t}}(\{\varphi_{t}\leq-s-\tau\})}{V_{t}}\leq\int_{\{\varphi_{t}\leq-s\}}\frac{(\omega_{t}+dd^{c}\varphi_{t})^{n}}{V_{t}}.

It is now an exercise to derive from this inequality an a priori L∞−L^{\infty}-estimate,

‖φt‖L∞​(X)≤C5+s0​(ωt),\|\varphi_{t}\|_{L^{\infty}(X)}\leq C_{5}+\ {s}_{0}(\omega_{t}),

where s0​(ωt){s}_{0}(\omega_{t}) (see [EGZ],[BGZ]) is the smallest number s>0s>0 satisfying the condition enC0nCapωt({ψ≤−s})/Vt≤1e^{n}C_{0}^{n}\mathrm{Cap}_{\omega_{t}}(\{\psi\leq-s\})/V_{t}\leq 1 for all ψ∈P​S​H​(X,ωt)\psi\in PSH(X,\omega_{t}) such that supXψ=0\sup_{X}\psi=0. Recall from ([GZ 1], Prop. 3.6) that

Capωt({ψ≤−s−τ})Vt≤1s​(∫X(−ψ)​ωtnVt+n).\frac{\mathrm{Cap}_{\omega_{t}}(\{\psi\leq-s-\tau\})}{V_{t}}\leq\frac{1}{s}\left(\int_{X}(-\psi)\frac{\omega_{t}^{n}}{V_{t}}+n\right).

Since ωtnVt≤C1​ω1n\frac{\omega_{t}^{n}}{V_{t}}\leq C_{1}\omega_{1}^{n}, it follows that

Capωt({ψ≤−s−τ})Vt≤1s​(C1​∫X(−ψ)​ω1n+n).\frac{\mathrm{Cap}_{\omega_{t}}(\{\psi\leq-s-\tau\})}{V_{t}}\leq\frac{1}{s}\left(C_{1}\int_{X}(-\psi)\omega_{1}^{n}+n\right).

Since ψ\psi is ω1−\omega_{1}-psh and normalized, we know that there is a constant A=A⁡(X,ω1)>0A=A(X,\omega_{1})>0 such that C1​∫X(−ψ)​ω1n≤AC_{1}\int_{X}(-\psi){\omega_{1}^{n}}\leq A for any such ψ\psi. Therefore s0​(ωt)≤s0:=en​C0n​(A+n)s_{0}(\omega_{t})\leq s_{0}:=e^{n}C_{0}^{n}(A+n) for any t∈]0,1]t\in]0,1]. Finally we obtain the required uniform estimate for all t∈]0,1]t\in]0,1].

References

  • [BT1] E. BEDFORD & B. A. TAYLOR: The Dirichlet problem for a complex Monge-Ampère equation. Invent. Math. 37 (1976), no. 1, 1–44.
  • [BGZ] S.BENELKOURCHI & V.GUEDJ & A.ZERIAHI: A priori estimates for weak solutions of complex Monge-Ampère equations. Ann. Sc. N. Sup. Pisa, to appear.
  • [CN] P. CASCINI & G. LA NAVE: Kähler-Ricci Flow and the Minimal Model Program for Projective Varieties. Preprint, arXiv:math/06.03.064.
  • [DP] J-P. DEMAILLY & N.PALI: Degenerate complex Monge-Ampère equations over compact Kähler manifolds. Preprint, arXiv:math/0710.5109.
  • [EGZ] P.EYSSIDIEUX & V.GUEDJ & A.ZERIAHI: Singular Kähler-Einstein metrics. Preprint arXiv:math/06.03.431.
  • [GZ 1] V.GUEDJ & A.ZERIAHI: Intrinsic capacities on compact Kähler manifolds, J. Geom. Anal. 15-4 (2005), 607-639.
  • [GZ 2] V.GUEDJ & A.ZERIAHI: The weighted Monge-Ampère energy of quasiplurisubharmonic functions, J. Funct. Anal. 250 (2007), 442-482.
  • [K] S.KOLODZIEJ: The complex Monge-Ampère equation. Acta Math. 180 (1998), no. 1, 69–117.
  • [KT] S. KOLODZIEJ & G.TIAN: A uniform L∞−L^{\infty}-estimate for complex Monge-Ampère equations. Preprint arXiv:math/07.10.1144.
  • [ST] J. SONG & G.TIAN: The Kähler-Ricci flow on surfaces of positive Kodaira dimension,. Preprint, arXiv:math/06.02.150.
  • [KT] G. TIAN & Z.ZHANG: On the Kähler-Ricci flow on projective manifolds of general type, Chinese Ann. Math. B 27 (2) (2006), 179-192.
  • [To] V. TOSATTI: Limits of Calabi-Yau metrics when the Kähler class degenerates. Preprint arXiv:math/07.10.4579.
  • [Y] S.T.YAU: On the Ricci curvature of a compact Kähler manifold and the complex Monge-Ampère equation. I. Comm. Pure Appl. Math. 31 (1978), no. 3, 339–411.
  • [Z] A.ZERIAHI: Volume and capacity of sublevel sets of a Lelong class of plurisubharmonic functions. Indiana Univ. Math. J. 50 (2001), no. 1, 671-703.

Original mathematics by the credited authors. Source-backed reader collection; mathematical self-containment is not assessed.