Paperlane

no. 16 · Tuesday 17 November 2026

Fourier Ratio Bounds and Sensitivity to Sensor Placement

Nico Alberti, Christopher Housholder, Alex Iosevich, Steven J. Miller and Eyvindur A. Palsson

math.CA

Abstract

We study how the regularity of a continuous function controls the Fourier ratio of its sampled values, and how errors in sensor placement affect recovery from incomplete measurements. For a periodic function on the square sampled on an $N$ by $N$ grid, we show that Hölder regularity $C^{0,γ}$ with $0<γ<1$ leads to a bound of order $N^{1-γ}$, that Lipschitz regularity leads to the critical bound $\sqrt{1+\log N}$, and that fractional Sobolev regularity $W^{s,p}$ with $0<s<1$ and $sp>2$ leads to a bound of order $N^{1-s}$, where both polynomial exponents are sharp. A common translation of the grid removes the hypothesis on the sampled mass needed on a fixed grid and yields an estimate in terms of the $L^2$ norm of the gradient, and with two derivatives the bound is independent of the grid size. We extend the energy argument behind these results to compact Riemannian manifolds and prove that the critical square-root logarithm is necessary on every compact surface. Finally, we treat separate sensor displacements as measurement error and combine the analytic bounds with random sampling to obtain stable recovery of the sampled vector on the square and of bandlimited functions on compact manifolds, where the number of samples depends on the suprema of the chosen eigenfunctions.

fields
Classical Analysis and ODEs
pdf ↗ page arXiv:2611.19755 ↗