ScalingStacks

Proof. [021W]

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.

Let 0<δ<10<\delta<1 and K>1K>1 to be determined. We will compute Δϕ​(eδ​G​(|∇ϕ|2+K))\Delta_{\phi}(e^{\delta G}(|\nabla\phi|^{2}+K)). As before, for any point p∈B1p\in B_{1} we are considering, we can always do a unitary coordinate transform which makes ϕi​j¯​(p)=ϕi​i¯​(p)​δi​j\phi_{i\bar{j}}(p)=\phi_{i\bar{i}}(p)\delta_{ij}. Under this coordinate, we can compute:

(6.30) Δϕ​(eδ​G​(|∇ϕ|2+K))=eδ​G​(δ2​|∇ϕG|2−δ​R¯)​(|∇ϕ|2+K)+eδ​G​Δϕ​(|∇ϕ|2)+eδ​G​δ​Gi​(|∇ϕ|2)i¯+Gi¯​(|∇ϕ|2)iϕi​i¯.\begin{split}&\Delta_{\phi}(e^{\delta G}(|\nabla\phi|^{2}+K))=e^{\delta G}(\delta^{2}|\nabla_{\phi}G|^{2}-\delta\underline{R})(|\nabla\phi|^{2}+K)+e^{\delta G}\Delta_{\phi}(|\nabla\phi|^{2})\\ &+e^{\delta G}\delta\frac{G_{i}(|\nabla\phi|^{2})_{\bar{i}}+G_{\bar{i}}(|\nabla\phi|^{2})_{i}}{\phi_{i\bar{i}}}.\end{split}

Similar to the calculation in Theorem 2.1, we can find:

(6.31) Δϕ​(|∇ϕ|2)=|ϕi​j|2ϕi​i¯+Δ​ϕ+Gi​ϕi¯+Gi¯​ϕi\displaystyle\Delta_{\phi}(|\nabla\phi|^{2})=\frac{|\phi_{ij}|^{2}}{\phi_{i\bar{i}}}+\Delta\phi+G_{i}\phi_{\bar{i}}+G_{\bar{i}}\phi_{i}
(6.32) (|∇ϕ|2)i=∑jϕi​j​ϕj¯+ϕi​i¯​ϕi¯.\displaystyle(|\nabla\phi|^{2})_{i}=\sum_{j}\phi_{ij}\phi_{\bar{j}}+\phi_{i\bar{i}}\phi_{\bar{i}}.

Hence we obtain

(6.33) Δϕ​(eδ​G​(|∇ϕ|2+K))=eδ​Gϕi​i¯​|δ​ϕj​Gi+ϕi​j|2+K​eδ​G​(δ2​|∇ϕG|2−δ​R¯)−eδ​G​δ​R¯​|∇ϕ|2+eδ​G​Δ​ϕ+eδ​G​(1+δ)​(Gi​ϕi¯+Gi¯​ϕ)≥eδ​G​K​δ2​|∇ϕG|2+eδ​G​Δ​ϕ−δ​R¯​eδ​G​|∇ϕ|2−δ​R¯​K​eδ​G−12​K​δ2​eδ​G​|∇ϕG|2−12​(1+δ)2​eδ​G​|∇ϕ|2​Δ​ϕK​δ2.\begin{split}&\Delta_{\phi}(e^{\delta G}(|\nabla\phi|^{2}+K))=\frac{e^{\delta G}}{\phi_{i\bar{i}}}|\delta\phi_{j}G_{i}+\phi_{ij}|^{2}+Ke^{\delta G}(\delta^{2}|\nabla_{\phi}G|^{2}-\delta\underline{R})-e^{\delta G}\delta\underline{R}|\nabla\phi|^{2}\\ &+e^{\delta G}\Delta\phi+e^{\delta G}(1+\delta)(G_{i}\phi_{\bar{i}}+G_{\bar{i}}\phi)\geq e^{\delta G}K\delta^{2}|\nabla_{\phi}G|^{2}+e^{\delta G}\Delta\phi-\delta\underline{R}e^{\delta G}|\nabla\phi|^{2}\\ &-\delta\underline{R}Ke^{\delta G}-\frac{1}{2}K\delta^{2}e^{\delta G}|\nabla_{\phi}G|^{2}-\frac{1}{2}\frac{(1+\delta)^{2}e^{\delta G}|\nabla\phi|^{2}\Delta\phi}{K\delta^{2}}.\end{split}

Now we choose δ=18\delta=\frac{1}{8}, and we choose KK sufficiently large so that (1+δ)2​|∇ϕ|2K​δ2<1\frac{(1+\delta)^{2}|\nabla\phi|^{2}}{K\delta^{2}}<1. Hence we obtain from (6.33):

(6.34) Δϕ​(eδ​G​(|∇ϕ|2+K))≥12​eδ​G​Δ​ϕ−eδ​G​C6.5.\Delta_{\phi}(e^{\delta G}(|\nabla\phi|^{2}+K))\geq\frac{1}{2}e^{\delta G}\Delta\phi-e^{\delta G}C_{6.5}.

Here C9C_{9} depends only on R¯\underline{R} and ‖∇ϕ‖0||\nabla\phi||_{0}. Define η⁡(z)=(1−|z|2)−1\eta(z)=(1-|z|^{2})^{-1} for z∈B1z\in B_{1}. We show that |Δϕ​η|≤C6.6​η3​∑i1ϕi​i¯|\Delta_{\phi}\eta|\leq C_{6.6}\eta^{3}\sum_{i}\frac{1}{\phi_{i\bar{i}}}.Indeed,

Δϕ​η=ϕi​j¯​∂i​j¯(η)=ϕi​j¯​((1−|z|2)−2​δi​j+2​(1−|z|2)−3​z¯i​zj)=ϕi​j¯​η3​((1−|z|2)​δi​j+2​z¯i​zj).\begin{split}\Delta_{\phi}\eta&=\phi^{i\bar{j}}\partial_{i\bar{j}}(\eta)=\phi^{i\bar{j}}\bigg((1-|z|^{2})^{-2}\delta_{ij}+2(1-|z|^{2})^{-3}\bar{z}_{i}z_{j}\bigg)\\ &=\phi^{i\bar{j}}\eta^{3}\big((1-|z|^{2})\delta_{ij}+2\bar{z}_{i}z_{j}\big).\end{split}

From this the claim follows easily. Denote v=eδ​G​(|∇ϕ|2+K)v=e^{\delta G}(|\nabla\phi|^{2}+K). Suppose the function v−ηv-\eta achieves maximum at p∈B1p\in B_{1}. There are two possibilities:

Suppose v⁡(p)−η⁡(p)≤0v(p)-\eta(p)\leq 0, then we immediately conclude that

v⁡(z)≤η⁡(z)≤43, for any z∈B12.v(z)\leq\eta(z)\leq\frac{4}{3},\textrm{ for any $z\in B_{\frac{1}{2}}$.}

Then we are done.

Suppose otherwise v⁡(p)−η⁡(p)≥0v(p)-\eta(p)\geq 0, then we know at pp:

(6.35) 0≥Δϕ​(v−η)​(p)≥12​eδ​G​Δ​ϕ−eδ​G​C6.5−C6.6​η3​∑i1ϕi​i¯≥12​eδ​G​Δ​ϕ−C6.6​v3​e−G​Δ​ϕ−eδ​G​C6.5≥12​eδ​G​Δ​ϕ−C6.7​e−(1−3​δ)​G​Δ​ϕ−eδ​G​C6.5.\begin{split}0&\geq\Delta_{\phi}(v-\eta)(p)\geq\frac{1}{2}e^{\delta G}\Delta\phi-e^{\delta G}C_{6.5}-C_{6.6}\eta^{3}\sum_{i}\frac{1}{\phi_{i\bar{i}}}\\ &\geq\frac{1}{2}e^{\delta G}\Delta\phi-C_{6.6}v^{3}e^{-G}\Delta\phi-e^{\delta G}C_{6.5}\geq\frac{1}{2}e^{\delta G}\Delta\phi-C_{6.7}e^{-(1-3\delta)G}\Delta\phi-e^{\delta G}C_{6.5}.\end{split}

In the third inequality above, we used that ∑i1ϕi​i¯=e−G​Δ​ϕ\sum_{i}\frac{1}{\phi_{i\bar{i}}}=e^{-G}\Delta\phi, which is true only in dimension 2. Also we used that at pp, η≤v\eta\leq v.

Suppose at pp, we have 14​eδ​G≤C6.7​e−(1−3​δ)​G\frac{1}{4}e^{\delta G}\leq C_{6.7}e^{-(1-3\delta)G}, this immediately gives a bound for eGe^{G}, hence vv at pp. Then we are done.

Suppose otherwise, then we have at pp

(6.36) 0≥eδ​G​14​Δ​ϕ−eδ​G​C6.5≥eδ​G​(14​eG2−C6.5).0\geq e^{\delta G}\frac{1}{4}\Delta\phi-e^{\delta G}C_{6.5}\geq e^{\delta G}(\frac{1}{4}e^{\frac{G}{2}}-C_{6.5}).

Then we also get an estimate for eGe^{G} at pp. So we are done as well. ∎

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