ScalingStacks

Proof. [020T]

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.

This step is relatively easy. Let p∈Mp\in M, we may choose a local normal coordinate in a neighborhood of pp, such that

(2.1) gi​j¯​(p)=δi​j,∇gi​j¯​(p)=0,and​φi​j¯​(p)=φi​i¯​(p)​δi​j.g_{i\bar{j}}(p)=\delta_{ij},\;\;\nabla g_{i\bar{j}}(p)=0,\;\;{\rm and}\;\;\varphi_{i\bar{j}}(p)=\varphi_{i\bar{i}}(p)\delta_{ij}.

In this paper, we will always work under this coordinate unless specified otherwise. Choose the constant C2.1C_{2.1} to be C2.1=2​maxM​Ri​i¯+2​|R¯|n+1C_{2.1}=2\displaystyle\max_{M}R_{i\bar{i}}+\frac{2|\underline{R}|}{n}+1. Under this coordinate, we can calculate:

(2.2) Δφ​(F+C2.1​φ)=−R¯+Ri​i¯1+φi​i¯+C2.1​φi​i¯1+φi​i¯=−R¯+C2.1​n−C2.1−Ri​i¯1+φi​i¯≤−R¯+C2.1​n−n​C2.12​e−Fn≤2​C2.1​n−n​C2.12​e−Fn.\begin{split}\Delta_{\varphi}(F+C_{2.1}\varphi)&=-\underline{R}+\frac{R_{i\bar{i}}}{1+\varphi_{i\bar{i}}}+C_{2.1}\frac{\varphi_{i\bar{i}}}{1+\varphi_{i\bar{i}}}\\ &=-\underline{R}+C_{2.1}n-\frac{C_{2.1}-R_{i\bar{i}}}{1+\varphi_{i\bar{i}}}\leq-\underline{R}+C_{2.1}n-\frac{nC_{2.1}}{2}e^{-\frac{F}{n}}\\ &\leq 2C_{2.1}n-\frac{nC_{2.1}}{2}e^{-\frac{F}{n}}.\end{split}

In the second line above, we used the arithemetic-geometric inequality:

1n​∑i11+φi​i¯≥Πi​(1+φi​i¯)−1n=e−Fn.\frac{1}{n}\sum_{i}\frac{1}{1+\varphi_{i\bar{i}}}\geq\Pi_{i}(1+\varphi_{i\bar{i}})^{-\frac{1}{n}}=e^{-\frac{F}{n}}.

Now let p0p_{0} be such that the function F+C2.1​φF+C_{2.1}\varphi achieves minimum at p0p_{0}, then from (2.2), we see

(2.3) 0≤2​C2.1​n−n​C2.12​e−Fn​(p0).0\leq 2C_{2.1}n-\frac{nC_{2.1}}{2}e^{-\frac{F}{n}}(p_{0}).

This gives a lower bound for FF, depending only on C0C^{0} bound for φ\varphi.

∎

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