Multidimensional van der Corput and sublevel set estimates

From MaRDI portal





Let \(Q^n=[0,1]^n\) denote the unit cube in \(\mathbb R^n\). The authors investigate the following questions: Given a function \(u(x)\), \(x\in Q^n\), which is oscillating rapidly. (i) How small must the set of points over which \(u\) takes on a prescribed range of values be? (ii) How rapidly must the integral \(I(\lambda) =\int_Q^n e^{i\lambda u(x)} dx \) decay as \(\lambda\to \infty\)? Precise answers to these questions in one variable are known, and precise results are known in several variables when one puts extra hypotheses, such as smoothness assumptions, on \(u\). Such added assumptions are undesirable in certain applications. The only hypotheses imposed on \(u\) here are that \(u\) is locally integrable and, for some multiindex \(\beta\) one has \(D^\beta u(x)\geq 1\) on \(Q^n\) in the sense of distributions. Under these conditions the authors show that there are a constant \(C\) and an exponent \(\varepsilon\) depending only on \(\beta\) and \(n\) such that for all \(\alpha\) the sublevel set estimate \[ |\{ x\in Q^n : |u(x)|\leq \alpha\}|\leq C\alpha^\varepsilon \] holds and the decay estimate \[ |I(\lambda)|\leq C|\lambda|^{-\varepsilon} \] holds. The authors show that the optimal decay with \(\varepsilon=1/|\beta|\) also holds when \(u\) satisfies some additional convexity criteria, but that, in general, the optimal \(\varepsilon\) is not known. Applications to oscillatory integrals and connections with combinatorial problems are also discussed.




Cited in
(84)








This page was built for publication: Multidimensional van der Corput and sublevel set estimates

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4257708)