Oscillation of Fourier integrals with a spectral gap. (Q1429961): Difference between revisions

Let \(f\) be a function on the real line \({\mathbb R}\). The number of sign changes \(s(r, f)\) of \(f\) on an open interval \((0; r)\), \(r\in {\mathbb R}\), is defined as the minimal degree of a polynomial \(p\) such that the restriction of \(p f\) on \((0; r)\) is a nonnegative function. A function \(\omega\geq 0\) on \({\mathbb R}\) with the property \(\int \omega(x)(1+x^2)^{-1} \,dx<\infty\) is called Beurling-Malliavin weight. The authors prove the following conjecture of F. Logan: Suppose that a real function \(f\), \(f\not \equiv 0\), has a spectral gap, that is its Fourier transform is zero on an interval \((-a; a)\), \(a>0\). If \(\int e^{-\lambda \omega(x)}| f(x)| \,dx<\infty\) for some \(\lambda>0\), and Beurling-Malliavin weight, then \[ \liminf_{r\to\infty}s(r; f)/ r\geq a/\pi. \] They also show that the growth restriction \(\int e^{-\lambda \omega(x)}| f(x)|\, dx<\infty\) is best possible (for sufficiently regular weights). For functions and distributions \(f\) of faster growth, Logan's conjecture does not hold, and for this class a weaker result estimating the number of sign changes from below in terms of an averaged lower density are given.