Topological bounds for Fourier coefficients and applications to torsion (Q1694690)

For a bounded, convex domain \(\Omega\subset{\mathbb R}^2\), the torsion function is the unique solution of \(-\Delta u=1\) which vanishes on \(\partial\Omega\). This is the prototype of a solution to an elliptic equation for studying level sets. One important question is how eccentric level sets may become locally. There is a unique global maximum point \(x_0\), and close to that point, the eccentricity is determined by the eigenvalues \(\lambda_1,\lambda_2\) of the Hessian \(D^2u(x_0)\). Both are negative, and their sum is \(-1\). In a first theorem, it is proven that there exist universal constants \(c_1,c_2>0\) such that \[\lambda_{\max}(D^2u(x_0))\le -c_1\exp\Big(-c_2\frac{\mbox{diam}(\Omega)}{\mbox{inrad}(\Omega)}\Big).\] This is mainly interesting in the case where \(\Omega\) is not strictly convex, in which case a better estimate is proven by more classical methods. It turns out that a new sort of ``topological'' bounds for Fourier coefficients is needed to prove the first theorem. The result here is that for a \(2\pi\)-periodic function \(f\) having \(n\) sign changes, the sum \(|a_1|+\ldots+|a_{n/2}|+|b_0|+|b_1|+\ldots+|b_{n/2}|\) can be estimated from below by \(c_n\|f\|_{L^1}^{n+1}/\|f\|_{L^\infty}^n\), with \(c_n\) depending on \(n\) only.