ScalingStacks

1. Weak solutions to Monge-Ampère equations [02D7]

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

Complete original source context · Original author HTML

1. Weak solutions to Monge-Ampère equations

For any Kähler form Ω\Omega on XX and for all ε>0\varepsilon>0, the form ωε:=ω+ε​Ω\omega_{\varepsilon}:=\omega+\varepsilon\Omega is again Kähler on XX. We are going to extend several results that are known to hold true when ω\omega is Kähler to the more general setting of semi-positive and big forms.

Recall that the set of ω\omega-plurisubharmonic functions (ω\omega-psh for short) is

PSH(X,ω):={φ∈L1(X,ℝ∪{−∞})/ddcφ≥−ω and φ is u.s.c.}.PSH(X,\omega):=\{\varphi\in L^{1}(X,\mathbb{R}\cup\{-\infty\})\,/\,dd^{c}\varphi\geq-\omega\text{ and }\varphi\text{ is }\text{u.s.c.}\}.

We refer the reader to [GZ 1] for basic properties of ω\omega-psh functions. The following subclass has been extensively studied in [GZ 2]:

Definition 1.1.

We let ℰ1​(X,ω){\mathcal{E}}^{1}(X,\omega) denote the set of ω\omega-psh functions with finite self-energy: this is the set of functions φ∈P​S​H​(X,ω)\varphi\in PSH(X,\omega) for which there exists a sequence φj∈P​S​H​(X,ω)∩L∞​(X)\varphi_{j}\in PSH(X,\omega)\cap L^{\infty}(X) such that

φj↘φ​ and ​supj∫X(−φj)​(ω+d​dc​φj)n<+∞.\varphi_{j}\searrow\varphi\;\;\text{ and }\;\;\sup_{j}\int_{X}(-\varphi_{j})(\omega+dd^{c}\varphi_{j})^{n}<+\infty.

This class of functions is studied in [GZ 2] when ω\omega is a Kähler form. We leave it to the reader to check that the basic properties of this class of functions proved in [GZ 2] when ω\omega is Kähler apply with no modification to the case where ω\omega is merely semi-positive and big. In particular the complex Monge-Ampère operator (ω+d​dc​φ)n(\omega+dd^{c}\varphi)^{n} is well-defined for φ∈ℰ1​(X,ω)\varphi\in{\mathcal{E}}^{1}(X,\omega), and it is continuous on decreasing sequences of functions in ℰ1​(X,ω){\mathcal{E}}^{1}(X,\omega).

We shall need a slightly more general continuity result, which takes into account the dependence in ω\omega:

Proposition 1.2.

Fix Ω\Omega a Kähler form on XX, and let (εj)(\varepsilon_{j}) be a sequence of positive real numbers decreasing to zero. Let φj∈ℰ1​(X,ω+εj​Ω)\varphi_{j}\in{\mathcal{E}}^{1}(X,\omega+\varepsilon_{j}\Omega) be a sequence of functions which decrease pointwise towards φ\varphi, and such that

supj≥1∫X|φj|​(ω+εj​Ω+d​dc​φj)n<+∞.\sup_{j\geq 1}\int_{X}|\varphi_{j}|(\omega+\varepsilon_{j}\Omega+dd^{c}\varphi_{j})^{n}<+\infty.

Then φ∈ℰ1​(X,ω)\varphi\in{\mathcal{E}}^{1}(X,\omega), and (ω+εj​Ω+d​dc​φj)n→(ω+d​dc​φ)n(\omega+\varepsilon_{j}\Omega+dd^{c}\varphi_{j})^{n}\rightarrow(\omega+dd^{c}\varphi)^{n}.

Proof.

Set ωj:=ω+εj​Ω\omega_{j}:=\omega+\varepsilon_{j}\Omega. We can assume w.l.o.g. that φj≤φ≤0\varphi_{j}\leq\varphi\leq 0. Set

φK:=max⁡(φ,−K)∈P​S​H​(X,ω)​ and ​φjK:=max⁡(φj,−K)∈P​S​H​(X,ωj).\varphi^{K}:=\max(\varphi,-K)\in PSH(X,\omega)\;\text{ and }\;\varphi_{j}^{K}:=\max(\varphi_{j},-K)\in PSH(X,\omega_{j}).

Observe that, KK being fixed, (φjK)j(\varphi_{j}^{K})_{j} is uniformly bounded and decreases towards φK\varphi^{K} as jj goes to infinity. Therefore (ωj+d​dc​φjK)n→(ω+d​dc​φK)n(\omega_{j}+dd^{c}\varphi_{j}^{K})^{n}\rightarrow(\omega+dd^{c}\varphi^{K})^{n}, by a classical result of E.Bedford and A.Taylor [BT 82]. Moreover the sequence of positive measures (−φjK)​(ωj+d​dc​φjK)n(-\varphi_{j}^{K})(\omega_{j}+dd^{c}\varphi_{j}^{K})^{n} has uniformly bounded mass, since by lemma 7.2 in [GZ 2],

0≤∫X(−φjK)​(ωj+d​dc​φjK)n≤2n​∫X(−φj)​(ωj+d​dc​φj)n≤2n​M,0\leq\int_{X}(-\varphi_{j}^{K})(\omega_{j}+dd^{c}\varphi_{j}^{K})^{n}\leq 2^{n}\int_{X}(-\varphi_{j})(\omega_{j}+dd^{c}\varphi_{j})^{n}\leq 2^{n}M,

where M:=supj∫X(−φj)​(ωj+d​dc​φj)n<+∞M:=\sup_{j}\int_{X}(-\varphi_{j})(\omega_{j}+dd^{c}\varphi_{j})^{n}<+\infty.

Since φj\varphi_{j} is u.s.c., a standard argument yields that any cluster point ν\nu of the sequence (−φjK)​(ωj+d​dc​φjK)n(-\varphi_{j}^{K})(\omega_{j}+dd^{c}\varphi_{j}^{K})^{n} satisfies 0≤(−φK)​(ω+d​dc​φK)n≤ν0\leq(-\varphi^{K})(\omega+dd^{c}\varphi^{K})^{n}\leq\nu. In particular

0≤∫X(−φK)​(ω+d​dc​φK)n≤lim infj→+∞∫(−φjK)​(ωj+d​dc​φjK)n≤2n​M0\leq\int_{X}(-\varphi^{K})(\omega+dd^{c}\varphi^{K})^{n}\leq\liminf_{j\rightarrow+\infty}\int(-\varphi_{j}^{K})(\omega_{j}+dd^{c}\varphi_{j}^{K})^{n}\leq 2^{n}M

is bounded from above uniformly with respect to KK. Since φK\varphi^{K} decreases towards φ\varphi, this shows φ∈ℰ1​(X,ω)\varphi\in{\mathcal{E}}^{1}(X,\omega).

It remains to show that (ωj+d​dc​φj)n→(ω+d​dc​φ)n(\omega_{j}+dd^{c}\varphi_{j})^{n}\rightarrow(\omega+dd^{c}\varphi)^{n}. Since (ωj+d​dc​φjK)n→(ω+d​dc​φK)n(\omega_{j}+dd^{c}\varphi_{j}^{K})^{n}\rightarrow(\omega+dd^{c}\varphi^{K})^{n} for any fixed KK, it is enough to get an upper bound on the mass of (ωj+d​dc​φjK)n(\omega_{j}+dd^{c}\varphi_{j}^{K})^{n} in (φj≤−K)(\varphi_{j}\leq-K) which is uniform in jj. This follows from Chebyshev inequality, namely

∫(φj≤−K)(ωj+d​dc​φj)n≤1K​∫X(−φj)​(ωj+d​dc​φj)n≤MK.\int_{(\varphi_{j}\leq-K)}(\omega_{j}+dd^{c}\varphi_{j})^{n}\leq\frac{1}{K}\int_{X}(-\varphi_{j})(\omega_{j}+dd^{c}\varphi_{j})^{n}\leq\frac{M}{K}.

This yields the desired result. ∎

The Monge-Ampère capacity C​a​pω​(⋅)Cap_{\omega}(\cdot) has been studied in [GZ 1],

Capω(K):=sup{∫Kωun/u∈PSH(X,ω),0≤u≤1},Cap_{\omega}(K):=\sup\left\{\int_{K}\omega_{u}^{n}\,/\,u\in PSH(X,\omega),0\leq u\leq 1\right\},

where KK is a Borel subset of XX. Here – and in the sequel – we use the notation ωu:=ω+d​dc​u≥0\omega_{u}:=\omega+dd^{c}u\geq 0. In this article we are interested in measures which are dominated by the Monge-Ampère capacity in the following way:

Definition 1.3.

A probability measure μ\mu on XX satisfies condition ℋ⁡(α,A,ω){\mathcal{H}}(\alpha,A,\omega) if for all Borel subset KK of XX,

μ⁡(K)≤A​C​a​pω​(K)1+α.\mu(K)\leq ACap_{\omega}(K)^{1+\alpha}.

It has been shown by S.Kolodziej that when ω\omega is Kähler, a probability measure μ\mu which satisfies ℋ⁡(α,A,ω){\mathcal{H}}(\alpha,A,\omega) can be written as the Monge-Ampère measure of some continuous ω\omega-psh function. This is still true when ω\omega is merely semi-positive and big, and the proof will occupy us until the end of section 2. We start by observing – following [GZ 2] – that μ\mu is the Monge-Ampère of a function φ\varphi which is not too singular.

Proposition 1.4.

Let μ\mu be a probability measure on XX which satisfies condition ℋ⁡(α,A,ω){\mathcal{H}}(\alpha,A,\omega). Then there exists a unique function φ∈ℰ1​(X,ω)\varphi\in{\mathcal{E}}^{1}(X,\omega) s.t.

μ=(ω+d​dc​φ)n​ and ​supXφ=−1.\mu=(\omega+dd^{c}\varphi)^{n}\;\text{ and }\;\sup_{X}\varphi=-1.
Proof.

Fix Ω\Omega a Kähler form on XX, and set ωj:=ω+εj​Ω\omega_{j}:=\omega+\varepsilon_{j}\Omega, where εj>0\varepsilon_{j}>0 decreases to 00. We start by showing that ℰ1​(X,ωj)⊂L1​(μ){\mathcal{E}}^{1}(X,\omega_{j})\subset L^{1}(\mu).

Fix φ∈ℰ1​(X,ωj)\varphi\in{\mathcal{E}}^{1}(X,\omega_{j}). We can assume without loss of generality that supXφ=−1\sup_{X}\varphi=-1. It follows from propositions 3.6 and 2.7 in [GZ 1] that there exists a constant C=C⁡(ω,Ω)>0C=C(\omega,\Omega)>0 independent of jj such that C​a​pωj​(φ<−t)≤C/tCap_{\omega_{j}}(\varphi<-t)\leq C/t for all t>0t>0. Since C​a​pω​(⋅)≤C​a​pωj​(⋅)Cap_{\omega}(\cdot)\leq Cap_{\omega_{j}}(\cdot), the measure μ\mu satisfies ℋ⁡(α,A,ωj){\mathcal{H}}(\alpha,A,\omega_{j}). We infer

(1) 0≤∫X(−φ)​𝑑μ=∫t=1+∞μ⁡(φ<−t)​𝑑t≤A​C1+αα<+∞,0\leq\int_{X}(-\varphi)d\mu=\int_{t=1}^{+\infty}\mu(\varphi<-t)dt\leq\frac{AC^{1+\alpha}}{\alpha}<+\infty,

with an upper-bound which is independent of jj.

The main result in [GZ 2] guarantees in this case that there exists a unique function φj∈ℰ1​(X,ωj)\varphi_{j}\in{\mathcal{E}}^{1}(X,\omega_{j}) such that

(ωj+d​dc​φj)n=λj​μ​ and ​supXφj=−1,(\omega_{j}+dd^{c}\varphi_{j})^{n}=\lambda_{j}\mu\,\text{ and }\;\sup_{X}\varphi_{j}=-1,

where λj=∫X(ω+εj​Ω)n>1\lambda_{j}=\int_{X}(\omega+\varepsilon_{j}\Omega)^{n}>1 decreases to 1 as jj goes to infinity.

The normalization supXφj=−1\sup_{X}\varphi_{j}=-1 implies that the sequence (φj)(\varphi_{j}) is relatively compact in L1​(X)L^{1}(X) (see proposition 2.7 in [GZ 1]). Let φ\varphi be a cluster point of (φj)(\varphi_{j}). Relabelling if neccessary, we assume φj→φ\varphi_{j}\rightarrow\varphi in L1​(X)L^{1}(X). Note that φ∈P​S​H​(X,ω)\varphi\in PSH(X,\omega) and supXφ=−1\sup_{X}\varphi=-1 (by Hartogs’ lemma, see proposition 2.7, [GZ 1]). We are going to show that φ∈ℰ1​(X,ω)\varphi\in{\mathcal{E}}^{1}(X,\omega) and ωφn=μ\omega_{\varphi}^{n}=\mu.

Set Φj:=(supl≥jφl)∗\Phi_{j}:=(\sup_{l\geq j}\varphi_{l})^{*}, where u∗u^{*} denotes the upper-semi-continuous regularization of uu. Then Φj∈P​S​H​(X,ωj)\Phi_{j}\in PSH(X,\omega_{j}) with Φj≥φj\Phi_{j}\geq\varphi_{j}, hence Φj∈ℰ1​(X,ωj)\Phi_{j}\in{\mathcal{E}}^{1}(X,\omega_{j}) (see proposition 3.2 in [GZ 2]), and Φj\Phi_{j} decreases towards φ\varphi. For l≥jl\geq j, we have

(ωj+d​dc​φl)n≥(ωl+d​dc​φl)n=λl​μ≥μ.(\omega_{j}+dd^{c}\varphi_{l})^{n}\geq(\omega_{l}+dd^{c}\varphi_{l})^{n}=\lambda_{l}\mu\geq\mu.

It follows therefore from an inequality due to J.-P.Demailly [Dem 1] that (ωj+d​dc​Φj)n≥μ(\omega_{j}+dd^{c}\Phi_{j})^{n}\geq\mu. Now by (1) and lemma 7.2 in [GZ 2],

0≤∫X(−Φj)​(ωj+d​dc​Φj)n≤2n​∫X(−φj)​(ωj+d​dc​φj)n=2n​λj​∫X(−φj)​𝑑μ,0\leq\int_{X}(-\Phi_{j})(\omega_{j}+dd^{c}\Phi_{j})^{n}\leq 2^{n}\int_{X}(-\varphi_{j})(\omega_{j}+dd^{c}\varphi_{j})^{n}=2^{n}\lambda_{j}\int_{X}(-\varphi_{j})d\mu,

is uniformly bounded with respect to jj thanks to (1).

We infer from proposition 1.2 that φ∈ℰ1​(X,ω)\varphi\in{\mathcal{E}}^{1}(X,\omega) and (ωj+d​dc​Φj)n→(ω+d​dc​φ)n(\omega_{j}+dd^{c}\Phi_{j})^{n}\rightarrow(\omega+dd^{c}\varphi)^{n}. Thus (ω+d​dc​φ)n≥μ(\omega+dd^{c}\varphi)^{n}\geq\mu, but these are two probability measures, whence μ=(ω+d​dc​φ)n\mu=(\omega+dd^{c}\varphi)^{n}. The uniqueness of φ\varphi follows from Theorem 7.4, [GZ 2]. ∎

Remark 1.5.

It follows from the work of S.Kolodziej [K 1,2,3] that the φj\varphi_{j}’s are actually continuous functions. Proving however that φ=limφj\varphi=\lim\varphi_{j} is continuous is a difficult task and is the goal of the next section.

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