Lifshits tails for random Schrödinger operators with nonsign definite potentials (Q929322)

Let \(H_0\) be a \({\mathbb Z}^d\)-periodic Schrödinger operator on \(L^2({\mathbb R}^d)\), so that it has a spectrum \(\Sigma_0\) consisting of bands of an absolutely continuous spectrum separared by gaps. Assume that \(\Sigma_0\) has a gap below \(0\) of length at least \(\delta >0\), i.e. that for some \(a> 0\), \(\Sigma_0 \cap [0,a) = [0,a)\) and \(\Sigma_0 \cap [-\delta,0) =\emptyset\). Consider the random Schrödinger operator \(H_{\lambda,w} =H_0 + \lambda \sum_{\gamma \in {\mathbb Z}^d} w_{\gamma} V(\cdot -\gamma)\) with \(\lambda\) being a positive coupling constant. Here \(V\) is a real-valued potential changing sign and having compact support, and \(\{w_{\gamma}\}_{\gamma \in {\mathbb Z}^d}\) is a collection of i. i. d. (independent identically distributed), bounded, nonnegative random variables whose essential support contains zero but does not reduce to this single point. Let \(E_{\lambda}\) be the spectral band edge of the almost sure spectrum \(\Sigma_{\lambda}\) of \(H_{\lambda,w}\), closest to \(0\), which under some suitable conditions is seen to be equal to \(0\) for \(\lambda\) small. Then the paper under review proves, assuming some further conditions on the single site potential \(V\) and the property of the spectrum \(\Sigma_0\) of \(H_0\), that, for all \(\lambda\) sufficiently small, the integrated density of states \(N_{\lambda}(E)\) for \(H_{\lambda,w}\) shows a Lifshits tail at \(E_{\lambda}\), namely, the difference \(N_{\lambda}(E)- N_{\lambda}(E_{\lambda})\) decays exponentially fast, as \(E\) tends to \(E_{\lambda}\) from above inside the spectrum \(\Sigma_{\lambda}\). The proof is done by using the ideas and techniques by \textit{F. Klopp} [Duke Math. J. 98, No. 2, 335--396 (1999; Zbl 1060.82509)], who treated the case where the potential \(V\) has definite sign.