On Variational Inequalities for Averages over Hypersurfaces

Abstract

This paper establishes sharp variational estimates for families of averaging operators over smooth hypersurfaces in ${\mathbb{R}}^{n+1}$. For a hypersurface with $k$ non-zero principal curvatures at a given point, we prove that the $r$-variation operator $V_r(A_{t}f:t>0)$ is bounded on $L^p$ for all exponents satisfying $(k+1)/k<p<r(k+1)$, with operator norm depending explicitly on a translation parameter $D$. In the special case $k=n$ (i.e., non-vanishing Gaussian curvature), we show that the range $p<r(n+1)$ is optimal, thereby recovering and extending classical results for the sphere. Furthermore, for convex hypersurfaces of finite type (in the sense of Iosevich-Sawyer), we obtain variational bounds in the range $p_0<p<2r$, where $p_0$ is the contact index of the associated hypersurface. The proofs combine long-short variational decompositions, stationary phase analysis, vector-valued singular integrals, and local smoothing estimates for Fourier integral operators. These results unify and generalize earlier work on maximal functions and variational inequalities for averages over hypersurfaces.