Curvature flows in the sphere (Q2351864): Difference between revisions

Let \(F\) be a smooth, symmetric, monotone and homogeneous of degree \(1\) curvature function defined in the positive cone in the \(n\)-dimensional Euclidean space. \(F\) is normalised to \(F(1,\dots,1)=1\). The functional \(\tilde F\) is defined to be the inverse of \(F\), that is \[ \tilde F(\kappa_1,\dots,\kappa_n)=\frac{1}{F(\kappa_1^{-1},\dots,\kappa_n^{-1})}. \] Let \(x\) be the hypersurface and \(\nu\) be the normal vector. The author consideres the curvature flow \[ \dot x=-F\nu \] starting from a smooth, strictly convex hypersurfaces \(M_0\) and the inverse curvature flow \[ \dot x=\tilde F^{-1}\nu. \] starting from \(\tilde M_0\), the polar of \(M_0\). He proves that 1) Both flows exist in the maximal time interval \([0,T^\ast)\) with finite time \(T^\ast\), the respective flow are polar sets of each other. 2) The curvature flow shrinks to a point \(x_0\) while the inverse flow converges to the equator of the hemisphere \(\mathcal H(-x_0)\), as \(t\rightarrow T^\ast\). 3) After suitable rescaling by the curvature solution with spherical initial hypersurface, the rescaled flows are \(C^\infty(\mathbb S^n)\) uniformly bounded with uniformly positive rescaled principal curvatures. 4) When \(F\) is strictly concave, or \(F=\frac{H}{n}\) where \(H\) is the mean curvature, the rescaled (inverse) curvature flows exponentially converge to the unit spheres with centre \(x_0\) (resp. \(-x_0\)) in \(C^\infty(\mathbb S^n)\). The example of strictly concave \(F\) is the \(k\)-th root of the elementary symmetric polynomials \(2\leq k\leq n\).