Local convexity estimates for mean curvature flow (Q6112904): Difference between revisions

The author develops a local version of Huisken-Stampacchia iteration. As an application, he shows some local versions of the sharp asymptotic curvature pinching estimates for the mean curvature flow. More precisely, denote by \(A\) the second fundamental form, by \(H = \mathrm{tr}(A)\) the mean curvature, by \(\kappa_1\le\ldots\le\kappa_n\) the principal curvatures and by \(t\mapsto\mu_t\)the induced Riemannian measure along the flow, then under the following assumptions: \[\inf_{B_{\lambda R}\times\{0\}\cup\partial B_{\lambda R}\times(0,\frac{R^2}{2n})}\frac{\kappa_1+\cdots+\kappa_{m+1}}{|A|}\geq\alpha,\] \[\sup_{B_{\lambda R}\times\{0\}\cup B_{\lambda R}\setminus B_{(\lambda-\delta)R}\times(0,\frac{R^2}{2n})}H\leq\Theta R^{-1}, \] and \[\frac{R^2}{2n}\int_{B_{\lambda R}}H^qd\mu_0+\delta^{-2}\int^{\frac{R^2}{2n}}_0\int_{B_{\lambda R}\setminus B_{(\lambda-\delta) R}}H^qd\mu dt\leq(\Theta R^{-1})^q(VR)^{n+2},\] for constants \(\alpha, q, \Theta\), the author shows for any \(\epsilon>0, \theta\in\left(0,1\right)\) the pinching estimates: \[\kappa_1\geq-\varepsilon H-C_\varepsilon R^{-1},\] and \[\sum^n_{j=m+1}(\kappa_n-\kappa_j)\leq\sum^m_{j=1}\kappa_j+\varepsilon H+C_\varepsilon R^{-1},\] on cylindrical domains, where \(C_\epsilon=c(n,\alpha,q,\epsilon)\Theta\left[\frac{\Theta V}{1-\theta}\right]^\frac{2}{q}\). These local estimates do not depend on the quality of noncollapsing of the solution.