ScalingStacks

Proof. [0341]

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

Proof.

The core of the proof consists in showing that

VK,ω(x)=sup{1Nlog||s||N​h(x)/N≥1,s∈Γ(X,LN) and supK||s||N​h≤1}.V_{K,\omega}(x)=\sup\left\{\frac{1}{N}\log||s||_{Nh}(x)\,/\,N\geq 1,s\in\Gamma(X,L^{N})\text{ and }\sup_{K}||s||_{Nh}\leq 1\right\}.

Note that for any of the sections ss involved in the supremum, φ:=N−1​log⁡‖s‖N​h\varphi:=N^{-1}\log||s||_{Nh} belongs to P​S​H​(X,ω)PSH(X,\omega) and satisfies φ≤0\varphi\leq 0 on KK. Therefore φ≤VK,ω\varphi\leq V_{K,\omega}.

Conversely fix x0∈Xx_{0}\in X and a<VK,ω​(x0)a<V_{K,\omega}(x_{0}). Fix φ∈P​S​H​(X,ω)\varphi\in PSH(X,\omega) such that supKφ≤0\sup_{K}\varphi\leq 0 and φ⁡(x0)>a\varphi(x_{0})>a. Regularizing φ\varphi (see Appendix) and translating, we can assume φ∈P​S​H​(X,ω)∩𝒞∞​(X)\varphi\in PSH(X,\omega)\cap{\mathcal{C}}^{\infty}(X), supKφ<0\sup_{K}\varphi<0 and φ⁡(x0)>a\varphi(x_{0})>a. Fix ε>0\varepsilon>0. Let B=B⁡(x0,r)B=B(x_{0},r) be a small ball on which φ>a\varphi>a. We choose BB so small that the oscillation of hh is smaller than ε\varepsilon on BB. Let χ\chi be a test function with compact support in BB and such that χ≡1\chi\equiv 1 in B⁡(x0,r/2)B(x_{0},r/2). We can assume w.l.o.g. that B⊂𝒰α0B\subset{\mathcal{U}}_{\alpha_{0}} for some α0\alpha_{0} but B∩𝒰β=∅B\cap{\mathcal{U}}_{\beta}=\emptyset for all β≠α0\beta\neq\alpha_{0}. This insures that χ\chi is a smooth section of LNL^{N} for all N≥1N\geq 1.

Let ψ1\psi_{1} be a smooth positive metric of LN1⊗KX∗L^{N_{1}}\otimes K_{X}^{*} on XX (this is possible if N1N_{1} is chosen large enough since LL is positive). Let ψ2\psi_{2} be a positive metric of LN2L^{N_{2}} on XX which is smooth in X∖{x0}X\setminus\{x_{0}\} and with Lelong number ν⁡(ψ2,x0)≥n=dimℂX\nu(\psi_{2},x_{0})\geq n=\dim_{\mathbb{C}}X (this is again possible if N2N_{2} is large enough, since LL is ample). Observe that ∂¯​χ\overline{\partial}\chi is a smooth ∂¯\overline{\partial}-closed (0,1)(0,1)-form with values in LNL^{N} (for all N≥1N\geq 1). Alternatively it is a smooth ∂¯\overline{\partial}-closed (n,1)(n,1)-form with values in LN⊗KX∗L^{N}\otimes K_{X}^{*}. Applying Hörmander’s L2L^{2}-estimates (see e.g. [15], chapter VIII) with weight ψN:=(N−N1−N2)​(φ+h)+ψ1+ψ2\psi_{N}:=(N-N_{1}-N_{2})(\varphi+h)+\psi_{1}+\psi_{2}, we find a smooth section ff of LNL^{N} such that ∂¯​f=∂¯​χ\overline{\partial}f=\overline{\partial}\chi and

∫X|f|2​e−2​(N−N1−N2)​(φ+h)−2​ψ1−2​ψ2​d​Vω≤C1​∫X|∂¯​χ|2​e−2​ψN​d​Vω.\int_{X}|f|^{2}e^{-2(N-N_{1}-N_{2})(\varphi+h)-2\psi_{1}-2\psi_{2}}dV_{\omega}\leq C_{1}\int_{X}|\overline{\partial}\chi|^{2}e^{-2\psi_{N}}dV_{\omega}.

Note that ∂¯​χ\overline{\partial}\chi has support in B∖B⁡(x0,r/2)B\setminus B(x_{0},r/2) where ψN\psi_{N} is smooth so that both integrals are finite. Since ν⁡(ψ2,x0)≥n\nu(\psi_{2},x_{0})\geq n, this forces f⁡(x0)=0f(x_{0})=0. The second integral is actually bounded from above by C2​e−2​N​(a−ε)C_{2}e^{-2N(a-\varepsilon)}, where C2C_{2} is independent of NN, since −φ<−a-\varphi<-a on BB and the oscillation of hh is smaller than ε\varepsilon on BB. Therefore s:=χ−f∈Γ⁡(X,LN)s:=\chi-f\in\Gamma(X,L^{N}) satisfies s⁡(x0)=1s(x_{0})=1 and

∫X|s|2​e−2​N​(φ+h)​d​Vω≤C3​e−2​N​(a−ε),\int_{X}|s|^{2}e^{-2N(\varphi+h)}dV_{\omega}\leq C_{3}e^{-2N(a-\varepsilon)},

where C3C_{3} is independent of NN. Now φ<0\varphi<0 in a neighborhood of KK, so the mean-value inequality applied to the subharmonic functions |sα|2|s_{\alpha}|^{2} yields for all xx in KK,

|s|2​e−2​N​h​(x)\displaystyle|s|^{2}e^{-2Nh}(x) ≤\displaystyle\leq Cδ​∫B⁡(x,δ)|s|2​(y)​e−2​N​[φ+h]​(y)​e2​N​[h⁡(y)−h⁡(x)+φ⁡(y)]​𝑑λ​(y)\displaystyle C_{\delta}\int_{B(x,\delta)}|s|^{2}(y)e^{-2N[\varphi+h](y)}e^{2N[h(y)-h(x)+\varphi(y)]}d\lambda(y)
≤\displaystyle\leq C4​e−2​N​(a−ε)\displaystyle C_{4}e^{-2N(a-\varepsilon)}

if δ\delta is so small that |supB⁡(x,δ)φ|>0|\sup_{B(x,\delta)}\varphi|>0 is bigger than the oscillation of hh on B⁡(x,δ)B(x,\delta). Therefore S:=C4−1/2eN⁡(a−ε)s∈Γ(X,LN)S:=C_{4}^{-1/2}e^{N(a-\varepsilon)}s\in\Gamma(X,L^{N}) satisfies supK‖S‖N​h≤1\sup_{K}||S||_{Nh}\leq 1 and N−1​log⁡‖S‖N​h​(x0)≥a−ε−log⁡C42​NN^{-1}\log||S||_{Nh}(x_{0})\geq a-\varepsilon-\frac{\log C_{4}}{2N}. Letting N→+∞N\rightarrow+\infty, ε→0\varepsilon\rightarrow 0 and a→VK,ω​(x0)a\rightarrow V_{K,\omega}(x_{0}) completes the proof of the equality.

To conclude observe that by rescaling one gets

−log⁡Tω​(K)=supXVK,ω\displaystyle-\log T_{\omega}(K)=\sup_{X}V_{K,\omega}
=\displaystyle= sup{1NsupXlog||S||N​h/N≥1,S∈Γ(X,LN) and supK||S||N​h=1}\displaystyle\!\!\!\!\sup\left\{\frac{1}{N}\sup_{X}\log||S||_{Nh}\,/\,N\geq 1,S\in\Gamma(X,L^{N})\text{ and }\sup_{K}||S||_{Nh}=1\right\}
=\displaystyle= sup{−1NsupKlog||S||N​h/N≥1,S∈Γ(X,LN) and supX||S||N​h=1}\displaystyle\!\!\!\!\sup\left\{-\frac{1}{N}\sup_{K}\log||S||_{Nh}\,/\,N\geq 1,S\in\Gamma(X,L^{N})\text{ and }\sup_{X}||S||_{Nh}=1\right\}
=\displaystyle= −log⁡Tω′​(K).\displaystyle-\log T_{\omega}^{\prime}(K).

∎

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