ScalingStacks

5. Comparison of capacities and applications [0345]

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

5. Comparison of capacities and applications

5.1. Josefson’s theorem

In this section we assume that ω\omega is Kähler and normalized by V​o​lω​(X)=1Vol_{\omega}(X)=1. We first prove an inequality relating TωT_{\omega} and C​a​pωCap_{\omega}. We do not know if a reverse inequality holds as it is the case in the local theory [2]. Then we prove (theorem 5.2) a quantitative version of Josefson’s theorem that every locally pluripolar set is actually P​S​H​(X,ω)PSH(X,\omega)-polar. In the local theory this result is due to El Mir [19]. We follow the approach of Alexander-Taylor [2].

Proposition 5.1.

Let KK be a compact subset of XX. If supXVK,ω≤1\sup_{X}V_{K,\omega}\leq 1 then C​a​pω​(K)=C​a​pω​(X)=1Cap_{\omega}(K)=Cap_{\omega}(X)=1. If supXVK,ω≥1\sup_{X}V_{K,\omega}\geq 1 then

0≤Tω(K)≤exp(−Capω(K)−1/n).0\leq T_{\omega}(K)\leq\exp(-Cap_{\omega}(K)^{-1/n}).
Proof.

Set MK=supXVK,ωM_{K}=\sup_{X}V_{K,\omega}. If MK=+∞M_{K}=+\infty then KK is P​S​H​(X,ω)PSH(X,\omega)-polar (theorem 3.2) and there is nothing to prove: Tω​(K)=C​a​pω​(K)=0T_{\omega}(K)=Cap_{\omega}(K)=0. So we assume in the sequel MK<+∞M_{K}<+\infty hence VK,ω∗∈P​S​H​(X,ω)V_{K,\omega}^{*}\in PSH(X,\omega). If MK≥1M_{K}\geq 1 then uK:=MK−1​VK,ω∗∈P​S​H​(X,ω)u_{K}:=M_{K}^{-1}V_{K,\omega}^{*}\in PSH(X,\omega) with 0≤uK≤10\leq u_{K}\leq 1 on XX. Since ωVK,ω∗≤MK​ωuK\omega_{V_{K,\omega}^{*}}\leq M_{K}\omega_{u_{K}}, we get

1MKn=1MKn​∫K(ωVK,ω∗)n≤∫K(ωuK)n≤C​a​pω​(K)\frac{1}{M_{K}^{n}}=\frac{1}{M_{K}^{n}}\int_{K}(\omega_{V_{K,\omega}^{*}})^{n}\leq\int_{K}(\omega_{u_{K}})^{n}\leq Cap_{\omega}(K)

whence Tω(K)≤exp(−Capω(K)−1/n)T_{\omega}(K)\leq\exp(-Cap_{\omega}(K)^{-1/n}).

If 0≤MK≤10\leq M_{K}\leq 1 then 0≤VK,ω∗≤10\leq V_{K,\omega}^{*}\leq 1 hence

1=∫K(ωVK,ω∗)n≤C​a​pω​(K)≤C​a​pω​(X)=1.1=\int_{K}(\omega_{V_{K,\omega}^{*}})^{n}\leq Cap_{\omega}(K)\leq Cap_{\omega}(X)=1.

∎

It follows from the previous proposition and corollary 2.8 that ω\omega-psh functions are quasicontinuous with respect to the capacity TωT_{\omega}.

Question. Is there -as in the local context [2]- a constant A>0A>0 s.t.

Tω(⋅)≥exp(−A/Capω(⋅))?T_{\omega}(\cdot)\geq\exp(-A/Cap_{\omega}(\cdot))\;?
Theorem 5.2.

Locally pluripolar sets are P​S​H​(X,ω)PSH(X,\omega)-polar.

Proof.

More precisely we are going to show the following: consider Ω\Omega an open subset of XX, v∈P​S​H−​(Ω)v\in PSH^{-}(\Omega) and P⊂{v=−∞}P\subset\{v=-\infty\}. Fix 0<ε<1/n0<\varepsilon<1/n and Vt:=VGt,ωV_{t}:=V_{G_{t},\omega} where Gt={x∈Ω/v(x)<−t}G_{t}=\{x\in\Omega\,/\,v(x)<-t\}. Then

φε​(x):=1ε​∫1+∞1t1+ε​[Vt​(x)−supXVt]​𝑑t\varphi_{\varepsilon}(x):=\frac{1}{\varepsilon}\int_{1}^{+\infty}\frac{1}{t^{1+\varepsilon}}[V_{t}(x)-\sup_{X}V_{t}]dt

is a ω\omega-psh function such that P⊂{φε=−∞}P\subset\{\varphi_{\varepsilon}=-\infty\}.

Indeed since GtG_{t} is open, we have Vt∈P​S​H​(X,ω)V_{t}\in PSH(X,\omega) and Vt=0V_{t}=0 on GtG_{t} (see proposition 3.6). Observe that φε\varphi_{\varepsilon} is a sum of negative ω\omega-psh functions hence it is either identically −∞-\infty or a well defined A​ωA\omega-psh function with A=ε−1​∫1+∞t−(1+ε)​𝑑t=1A=\varepsilon^{-1}\int_{1}^{+\infty}t^{-(1+\varepsilon)}dt=1. Recall that −C+supXVt≤∫XVt​ωn≤supXVt-C+\sup_{X}V_{t}\leq\int_{X}V_{t}\omega^{n}\leq\sup_{X}V_{t} (proposition 1.7). Therefore ∫Xφε​ωn≥−C\int_{X}\varphi_{\varepsilon}\omega^{n}\geq-C hence φε∈P​S​H​(X,ω)\varphi_{\varepsilon}\in PSH(X,\omega).

Fix x∈Ωx\in\Omega such that v⁡(x)<−1v(x)<-1. Observe that Vt−supXVt≤0V_{t}-\sup_{X}V_{t}\leq 0 with Vt​(x)=0V_{t}(x)=0 if x∈Gtx\in G_{t}, i.e. when |v⁡(x)|>t|v(x)|>t. Therefore

φε(x)≤−1ε∫1|v⁡(x)|supXVtt1+εdt.\varphi_{\varepsilon}(x)\leq-\frac{1}{\varepsilon}\int_{1}^{|v(x)|}\frac{\sup_{X}V_{t}}{t^{1+\varepsilon}}dt.

Recall now that C​a​pωCap_{\omega} is always dominated by C​a​pB​TCap_{BT} hence C​a​pω​(Gt)≤C1/t<1Cap_{\omega}(G_{t})\leq C_{1}/t<1 if tt is large enough. We infer from the previous proposition that

−supXVt≤−[Capω(Gt)]−1/n≤−C2t1/n,-\sup_{X}V_{t}\leq-[Cap_{\omega}(G_{t})]^{-1/n}\leq-C_{2}t^{1/n},

which yields

φε​(x)≤−C2ε​∫1|v⁡(x)|d​tt1+ε−1/n≤−C3​|v⁡(x)|1/n−ε+C4.\varphi_{\varepsilon}(x)\leq\frac{-C_{2}}{\varepsilon}\int_{1}^{|v(x)|}\frac{dt}{t^{1+\varepsilon-1/n}}\leq-C_{3}|v(x)|^{1/n-\varepsilon}+C_{4}.

Note that φε​(x)=−∞\varphi_{\varepsilon}(x)=-\infty whenever v⁡(x)=−∞v(x)=-\infty hence P⊂{φε=−∞}P\subset\{\varphi_{\varepsilon}=-\infty\}. ∎

5.2. Dynamical capacity estimates

Let f:ℂ​ℙn→ℂ​ℙnf:\mathbb{C}\mathbb{P}^{n}\rightarrow\mathbb{C}\mathbb{P}^{n} be an holomorphic endomorphism. We let ω\omega denote again the Fubini-Study Kähler form. Then f∗​ωf^{*}\omega is a smooth positive closed (1,1)(1,1)-form of mass λ=∫ℂ​ℙnf∗​ω∧ωn−1=:\lambda=\int_{\mathbb{C}\mathbb{P}^{n}}f^{*}\omega\wedge\omega^{n-1}=:the first algebraic degree of ff. Thus λ−1​f∗​ω=ω+d​dc​φ\lambda^{-1}f^{*}\omega=\omega+dd^{c}\varphi, where φ\varphi is a smooth ω\omega-psh function on ℂ​ℙn\mathbb{C}\mathbb{P}^{n}. Iterating this functional equation yields

1λj​(fj)∗​ω=ω+d​dc​gj,gj=∑l=0j−11λl​φ∘fl.\frac{1}{\lambda^{j}}(f^{j})^{*}\omega=\omega+dd^{c}g_{j},\;g_{j}=\sum_{l=0}^{j-1}\frac{1}{\lambda^{l}}\varphi\circ f^{l}.

We assume λ≥2\lambda\geq 2. Thus the sequence (gj)(g_{j}) uniformly converges on ℂ​ℙn\mathbb{C}\mathbb{P}^{n} towards a continuous function gf∈P​S​H​(X,ω)g_{f}\in PSH(X,\omega) called the Green function of ff. We refer the interested reader to [33] for a detailed study of the properties of the Green current Tf=ω+d​dc​gfT_{f}=\omega+dd^{c}g_{f}.

Dynamical volume estimates have revealed quite useful in establishing ergodic properties of the Green current TfT_{f} (see [20], [22] and references therein). We establish herebelow very simple dynamical capacity estimates and show how to derive from them dynamical volume estimates.

Proposition 5.3.

There exists 0<α<10<\alpha<1 such that for all Borel subsets KK of XX, for all j∈ℕj\in\mathbb{N},

[α​Tω​(K)]λj≤Tω​(fj​(K)).\left[\alpha T_{\omega}(K)\right]^{\lambda^{j}}\leq T_{\omega}(f^{j}(K)).
Proof.

This follows straightforwardly from proposition 3.9:

Tω​(fj​(K))≥T(fj)∗​ω​(K)=[Tλ−j​(fj)∗​ω​(K)]λj≥[α​Tω​(K)]λj,T_{\omega}(f^{j}(K))\geq T_{(f^{j})^{*}\omega}(K)=\left[T_{\lambda^{-j}(f^{j})^{*}\omega}(K)\right]^{\lambda^{j}}\geq\left[\alpha T_{\omega}(K)\right]^{\lambda^{j}},

where the first two inequalities follow from 3.9.4 and 3.9.2 and last one follows from 3.9.3 and the fact that λ−j​(fj)∗​ω=ω+d​dc​gj\lambda^{-j}(f^{j})^{*}\omega=\omega+dd^{c}g_{j}, where gjg_{j} is uniformly bounded. ∎

Corollary 5.4.

Let φ∈P​S​H​(X,ω)\varphi\in PSH(X,\omega). Then the sequence (λ−j​φ∘fj)(\lambda^{-j}\varphi\circ f^{j}) is relatively compact in L1​(ℂ​ℙn)L^{1}(\mathbb{C}\mathbb{P}^{n}).

Proof.

Set φj=λ−j​φ∘fj\varphi_{j}=\lambda^{-j}\varphi\circ f^{j}. Observe that φj\varphi_{j} is uniformly bounded from above and that φj+gj∈P​S​H​(ℂ​ℙn,ω)\varphi_{j}+g_{j}\in PSH(\mathbb{C}\mathbb{P}^{n},\omega). It follows from proposition 1.6 that either φj\varphi_{j} converges uniformly towards −∞-\infty or it is relatively compact in L1​(ℂ​ℙn)L^{1}(\mathbb{C}\mathbb{P}^{n}). It is sufficient to show that for A>0A>0 large enough, lim¯j→+∞​Tω​(φj<−A)<Tω​(X)=1\overline{\lim}_{j\rightarrow+\infty}T_{\omega}(\varphi_{j}<-A)<T_{\omega}(X)=1. Observe that fj(φj<−A)={φ<−Aλj}f^{j}(\varphi_{j}<-A)=\{\varphi<-A\lambda^{j}\}. Therefore

[α​Tω​(φj<−A)]λj≤Tω​(φ<−A​λj)≤C​exp⁡(−A​λj),\left[\alpha T_{\omega}(\varphi_{j}<-A)\right]^{\lambda^{j}}\leq T_{\omega}(\varphi<-A\lambda^{j})\leq C\exp(-A\lambda^{j}),

where the last inequality follows from proposition 3.8. We infer

lim¯j→+∞​Tω​(φj<−A)≤1α​exp⁡(−A)<1\overline{\lim}_{j\rightarrow+\infty}T_{\omega}(\varphi_{j}<-A)\leq\frac{1}{\alpha}\exp(-A)<1

for A>−log⁡αA>-\log\alpha large enough. ∎

Corollary 5.5.

Fix δ0>0\delta_{0}>0. There exists γ0∈]0,1[\gamma_{0}\in]0,1[ such that for all Borel subset KK of ℂ​ℙn\mathbb{C}\mathbb{P}^{n} with V​o​lω​(K)≥δ0Vol_{\omega}(K)\geq\delta_{0}, one has

V​o​lω​(fj​(K))≥γ0λj,∀j∈ℕ.Vol_{\omega}(f^{j}(K))\geq\gamma_{0}^{\lambda^{j}},\;\;\forall j\in\mathbb{N}.

In other words the volume of a given set can not decrease too fast under iteration. Such volume estimates are used in complex dynamics to prove fine convergence results towards the Green current TfT_{f} (see [20], [22]). One may hope that dynamical capacity estimates will allow to establish convergence results in higher codimension.

Proof.

By the change of variables formula one gets

V​o​lω​(fj​K)=∫fj​Kωn≥1dtj​∫K(fj)∗​ωn=1dtj​∫K|JF​S​(fj)|2​ωn,Vol_{\omega}(f^{j}K)=\int_{f^{j}K}\omega^{n}\geq\frac{1}{d_{t}^{j}}\int_{K}(f^{j})^{*}\omega^{n}=\frac{1}{d_{t}^{j}}\int_{K}|J_{FS}(f^{j})|^{2}\omega^{n},

where dt=λnd_{t}=\lambda^{n} denotes the topological degree of ff and JF​S​(f)J_{FS}(f) stands for the jacobian of ff with respect to the Fubini-Study volume form. Observe that log⁡|JF​S​(f)|=u−v\log|J_{FS}(f)|=u-v is a difference of two qpsh functions u,v∈P​S​H​(X,A​ω)u,v\in PSH(X,A\omega) for some A=A⁡(λ,n)A=A(\lambda,n). Moreover by the chain rule,

1λj​log⁡|JF​S​(fj)|=∑l=0j−11λj​log⁡|JF​S​(f)∘fl|.\frac{1}{\lambda^{j}}\log|J_{FS}(f^{j})|=\sum_{l=0}^{j-1}\frac{1}{\lambda^{j}}\log|J_{FS}(f)\circ f^{l}|.

Since λ−l​log⁡|JF​S​(f)∘fl|\lambda^{-l}\log|J_{FS}(f)\circ f^{l}| is relatively compact in L1​(ℂ​ℙn)L^{1}(\mathbb{C}\mathbb{P}^{n}) (previous corollary), the concavity of the log yields

1V​o​lω​(K)​∫K|JF​S​(fj)|2​ωn≥exp⁡(2​λjV​o​lω​(K)​∫K1λj​log⁡|JF​S​(fj)|​ωn)≥γλj.\frac{1}{Vol_{\omega}(K)}\int_{K}|J_{FS}(f^{j})|^{2}\omega^{n}\geq\exp\left(\frac{2\lambda^{j}}{Vol_{\omega}(K)}\int_{K}\frac{1}{\lambda^{j}}\log|J_{FS}(f^{j})|\omega^{n}\right)\geq\gamma^{\lambda^{j}}.

Decreasing slightly the value of γ\gamma if necessary, this yields the desired inequality. ∎

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