ScalingStacks

Proof. [021Q]

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.

(of Proposition 6.1)First we want to get boundedness of GG, using the second equation. We can write the second equation as

(6.3) det(ϕα​β¯)​ϕi​j¯​∂i​j¯G=eG​G.\det(\phi_{\alpha\bar{\beta}})\phi^{i\bar{j}}\partial_{i\bar{j}}G=e^{G}G.

which is equivalent to:

(6.4) R​e​(∂i(det(ϕα​β¯)​ϕi​j¯​∂j¯G))=eG​G.Re\big(\partial_{i}(\det(\phi_{\alpha\bar{\beta}})\phi^{i\bar{j}}\partial_{\bar{j}}G)\big)=e^{G}G.

Denote ai​j¯=det(ϕα​β¯)​ϕi​j¯a_{i\bar{j}}=\det(\phi_{\alpha\bar{\beta}})\phi^{i\bar{j}}, which is a hermitian matrix, then for some constant cn>0c_{n}>0, we have

(6.5) 1cn​(∑i1ϕi​i¯)n−1​I≤ai​j¯≤cn​(Δ​ϕ)n−1​I.\frac{1}{c_{n}\big(\sum_{i}\frac{1}{\phi_{i\bar{i}}}\big)^{n-1}}I\leq a_{i\bar{j}}\leq c_{n}(\Delta\phi)^{n-1}I.

The left hand side of (6.4) is a real elliptic operator in divergence form, which satisfies an ellipticity condition same as (6.5). We wish to apply Lemma 6.3 to the equation (6.4). Using (6.5), we can take λ=(Δ​ϕ+∑i1ϕi​i¯)n−1\lambda=\big(\Delta\phi+\sum_{i}\frac{1}{\phi_{i\bar{i}}}\big)^{n-1}, and f=eG​Gf=e^{G}G. In order to apply Lemma 6.3, we need to show (Δ​ϕ)n−1,(∑i1ϕi​i¯)n−1∈Lp​(B1)(\Delta\phi)^{n-1},\,(\sum_{i}\frac{1}{\phi_{i\bar{i}}})^{n-1}\in L^{p}(B_{1}), and eG​G∈Lp/2​(B1)e^{G}G\in L^{p/2}(B_{1}) for some p>3​np>3n. The desired integrability for Δ​ϕ\Delta\phi and ∑i1ϕi​i¯\sum_{i}\frac{1}{\phi_{i\bar{i}}} is clear from assumption, while for eG​Ge^{G}G, since eG≤(Δ​ϕ)ne^{G}\leq(\Delta\phi)^{n}, we just need to make sure (Δ​ϕ)n∈Lp′(\Delta\phi)^{n}\in L^{p^{\prime}} for some p′>3​n2p^{\prime}>\frac{3n}{2}. This is again clear from our assumption on Δ​ϕ\Delta\phi. So we can apply Lemma 6.3 to conclude GG is bounded(with the said dependence) on any interior ball of B1B_{1}. In the following we assume GG is bounded on B1B_{1} without loss of generality.

The estimate for Δ​ϕ\Delta\phi is really similar to our calculation in section 4, so we will be suitably brief here.

Choose any point pp and we can do a unitary coordinate transform so that ϕi​j¯​(p)=ϕi​i¯​(p)​δi​j\phi_{i\bar{j}}(p)=\phi_{i\bar{i}}(p)\delta_{ij}. We can compute

(6.6) Δϕ​(|∇ϕG|2)=1ϕi​i¯​ϕα​α¯​|ϕi​α−∑pϕi​α​p¯​Gpϕp​p¯|2+|Gp​i¯|2ϕi​i¯​ϕp​p¯−Gq​p¯​Gp​Gq¯ϕp​p¯​ϕq​q¯.\Delta_{\phi}(|\nabla_{\phi}G|^{2})=\frac{1}{\phi_{i\bar{i}}\phi_{\alpha\bar{\alpha}}}|\phi_{i\alpha}-\sum_{p}\frac{\phi_{i\alpha\bar{p}}G_{p}}{\phi_{p\bar{p}}}|^{2}+\frac{|G_{p\bar{i}}|^{2}}{\phi_{i\bar{i}}\phi_{p\bar{p}}}-\frac{G_{q\bar{p}}G_{p}G_{\bar{q}}}{\phi_{p\bar{p}}\phi_{q\bar{q}}}.

Here |∇ϕG|2=ϕp​q¯​Gp​Gq¯|\nabla_{\phi}G|^{2}=\phi^{p\bar{q}}G_{p}G_{\bar{q}}.

(6.7) Δϕ​(e12​G​|∇ϕG|2)=Δϕ​(e12​G)​|∇ϕG|2+e12​G​Δϕ​(|∇ϕG|2)+12​e12​G​Gi​(|∇ϕG|2)i¯+Gi¯​(|∇ϕG|2)iϕi​i¯.\Delta_{\phi}(e^{\frac{1}{2}G}|\nabla_{\phi}G|^{2})=\Delta_{\phi}(e^{\frac{1}{2}G})|\nabla_{\phi}G|^{2}+e^{\frac{1}{2}G}\Delta_{\phi}(|\nabla_{\phi}G|^{2})+\frac{1}{2}e^{\frac{1}{2}G}\frac{G_{i}(|\nabla_{\phi}G|^{2})_{\bar{i}}+G_{\bar{i}}(|\nabla_{\phi}G|^{2})_{i}}{\phi_{i\bar{i}}}.

One can also compute

(6.8) Gi​(|∇ϕG|2)i¯ϕi​i¯=Gi​Gpϕi​i¯​ϕp​p¯​(Gp¯​i¯−∑tϕt​q¯​i¯​Gt¯ϕt​t¯)+Gp​i¯​Gp¯​Giϕp​p¯​ϕi​i¯.\frac{G_{i}(|\nabla_{\phi}G|^{2})_{\bar{i}}}{\phi_{i\bar{i}}}=\frac{G_{i}G_{p}}{\phi_{i\bar{i}}\phi_{p\bar{p}}}(G_{\bar{p}\bar{i}}-\sum_{t}\frac{\phi_{t\bar{q}\bar{i}}G_{\bar{t}}}{\phi_{t\bar{t}}})+\frac{G_{p\bar{i}}G_{\bar{p}}G_{i}}{\phi_{p\bar{p}}\phi_{i\bar{i}}}.

Combining (6.6), (6.7), (6.8), we obtain

(6.9) Δϕ​(e12​G​|∇ϕG|2)​e−12​G=−12​R¯​|∇ϕG|2+1ϕi​i¯​ϕα​α¯​|Gi​α−∑pϕi​α​p¯​Gpϕp​p¯−12​ϕi​ϕα|2+|Gp​i¯|2ϕi​i¯​ϕp​p¯.\Delta_{\phi}(e^{\frac{1}{2}G}|\nabla_{\phi}G|^{2})e^{-\frac{1}{2}G}=-\frac{1}{2}\underline{R}|\nabla_{\phi}G|^{2}+\frac{1}{\phi_{i\bar{i}}\phi_{\alpha\bar{\alpha}}}|G_{i\alpha}-\sum_{p}\frac{\phi_{i\alpha\bar{p}}G_{p}}{\phi_{p\bar{p}}}-\frac{1}{2}\phi_{i}\phi_{\alpha}|^{2}+\frac{|G_{p\bar{i}}|^{2}}{\phi_{i\bar{i}}\phi_{p\bar{p}}}.

Also we can compute

(6.10) Δϕ​(Δ​ϕ)=|ϕi​j¯​p|2ϕi​i¯​ϕj​j¯+Δ​G.\Delta_{\phi}(\Delta\phi)=\frac{|\phi_{i\bar{j}p}|^{2}}{\phi_{i\bar{i}}\phi_{j\bar{j}}}+\Delta G.

Hence

(6.11) Δϕ(e12​G​|∇ϕG|2+Δ​ϕ)≥−12​R¯​e12​G​|∇ϕG|2+|Gp​i¯|2​e12​Gϕi​i¯​ϕp​p¯+Δ​G≥−12​R¯​e12​G​|∇ϕG|2+e12​G​Gi​i¯2ϕi​i¯2−12​Gi​i¯2​e12​Gϕi​i¯2−12​(Δ​ϕ)2​e−12​G.\begin{split}\Delta_{\phi}&(e^{\frac{1}{2}G}|\nabla_{\phi}G|^{2}+\Delta\phi)\geq-\frac{1}{2}\underline{R}e^{\frac{1}{2}G}|\nabla_{\phi}G|^{2}+\frac{|G_{p\bar{i}}|^{2}e^{\frac{1}{2}G}}{\phi_{i\bar{i}}\phi_{p\bar{p}}}+\Delta G\\ &\geq-\frac{1}{2}\underline{R}e^{\frac{1}{2}G}|\nabla_{\phi}G|^{2}+\frac{e^{\frac{1}{2}G}G_{i\bar{i}}^{2}}{\phi_{i\bar{i}}^{2}}-\frac{1}{2}\frac{G_{i\bar{i}}^{2}e^{\frac{1}{2}G}}{\phi_{i\bar{i}}^{2}}-\frac{1}{2}(\Delta\phi)^{2}e^{-\frac{1}{2}G}.\end{split}

In the last inequality above, we noticed

Δ​G=∑iGi​i¯≤12​Gi​i¯2​e12​Gϕi​i¯2+12​∑iϕi​i¯2​e−12​G≤12​Gi​i¯2​e12​Gϕi​i¯2+12​(Δ​ϕ)2​e−12​G.\Delta G=\sum_{i}G_{i\bar{i}}\leq\frac{1}{2}\frac{G_{i\bar{i}}^{2}e^{\frac{1}{2}G}}{\phi_{i\bar{i}}^{2}}+\frac{1}{2}\sum_{i}\phi_{i\bar{i}}^{2}e^{-\frac{1}{2}G}\leq\frac{1}{2}\frac{G_{i\bar{i}}^{2}e^{\frac{1}{2}G}}{\phi_{i\bar{i}}^{2}}+\frac{1}{2}(\Delta\phi)^{2}e^{-\frac{1}{2}G}.

Denote u=e12​G​|∇ϕG|2+Δ​ϕu=e^{\frac{1}{2}G}|\nabla_{\phi}G|^{2}+\Delta\phi, then we know

(6.12) Δϕ​(u)≥−(12​R¯+12​Δ​ϕ​e−12​G)​u.\Delta_{\phi}(u)\geq-(\frac{1}{2}\underline{R}+\frac{1}{2}\Delta\phi e^{-\frac{1}{2}G})u.

Denote f=12​R¯+12​Δ​ϕ​e−12​Gf=\frac{1}{2}\underline{R}+\frac{1}{2}\Delta\phi e^{-\frac{1}{2}G}. Recall that we now already know GG is bounded. Our assumption implies f∈Lpf\in L^{p} for some p>3​np>3n. Hence we may invoke Lemma 6.3 to get the desired result. ∎

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