A inequality in stability of mean curvature flow

70 Views Asked by At

$C_1,C_2$ are positive constants, and $\alpha\in(0,1)$ is a constant too.

If for any $\epsilon>0$, we have $$ \sum_{i<j}(k_i-k_j)^2 \le C_1 \epsilon^{2\alpha} \\ \sum k_i \ge C_2 \epsilon^\frac{\alpha}{2} $$ How to show $k_i>0$ ?

This is my guess, when $\epsilon$ is small enough, the first inequality means the $k_i$s have little difference. Although the RHS of second inequality approached zero also, but it is of low order respect to $\epsilon^{2\alpha}$.

This question is origin from
Lin, Longzhi; Sesum, Natasa, Blow-up of the mean curvature at the first singular time of the mean curvature flow, Calc. Var. Partial Differ. Equ. 55, No. 3, Paper No. 65, 16 p. (2016). ZBL1343.53064.

1

There are 1 best solutions below

0
On BEST ANSWER

OK, after your edit this makes more sense - if we normalize by setting $a_i = \epsilon^{-\alpha/2}k_i$ then we have $$\sum_i a_i \ge C_2,\;\;\sum_{i<j} (a_i-a_j)^2\le C_1 \epsilon^\alpha.$$

Now we are in good shape - decreasing $\epsilon$ will force the $a_i$ closer together without changing the lower bound. Letting $\hat a = \max_i a_i$, note that we have $$\hat a \ge \frac 1 n \sum_i^n a_i \ge \frac{C_2}n$$ and $$(a_i - \hat a)^2 \le \sum_{i<j}(a_i - a_j)^2\le C_1 \epsilon^\alpha;$$ so $$a_i \ge \hat a - |a_i-\hat a|\ge \frac{C_2}n-\sqrt{C_1 \epsilon^\alpha}.$$

Choosing $\epsilon$ small enough that $\sqrt{C_1 \epsilon^\alpha} < C_2/n$ completes the proof.