What it is
The Gauss curvature K of a hypersurface in Rn+1 is the product of its principal curvatures. Gauss curvature flow moves a convex hypersurface inward with speed K, or more generally Kα. Unlike mean curvature flow it is fully nonlinear, and it degenerates where K=0: a flat side does not move until its edges have eaten into it.
- Origins. Firey (1974) proposed the flow in R3 as a model for the wearing of stones on a beach. Tso (1985) proved that convex hypersurfaces shrink to a point in finite time.
- Roundness. Andrews (1999) proved that convex surfaces in R3 become round as they shrink, settling Firey's conjecture. In higher dimensions, Guan and Ni found an entropy that shows the rescaled flow converges to a self-similar solution, and Andrews, Guan and Ni extended this to all powers α>1/(n+2). Brendle, Choi and Daskalopoulos then proved that the only such self-similar solutions are round spheres. At the critical power α=1/(n+2) the flow is affine invariant and the limits are ellipsoids.
- Flat sides. Daskalopoulos and Hamilton studied surfaces with flat sides, where the edge of the flat region is a free boundary.
Why it matters
The proofs use the same toolkit as Ricci flow: entropy functionals that decrease along the flow, Harnack inequalities, and the classification of self-similar solutions as the final step. Gauss curvature flow is also an analytic route to Minkowski problems: find a convex body whose Gauss curvature, as a function of the normal direction, is prescribed. Chou and Wang solved the classical problem with a logarithmic Gauss curvature flow, and normalized flows now attack the Lp and dual Minkowski problems of convex geometry.
Where it stands
The fate of convex bodies is understood for all powers. Open directions include anisotropic and non-homogeneous speeds, flows in curved ambient spaces, the regularity of flat sides in higher dimensions, and Minkowski-type problems for new curvature measures, which make up a large part of the feed below. See also anisotropic and nonlocal flows.