The continuous Anderson Hamiltonian in d 3

From MaRDI portal
Publication:2324754



Abstract: We construct the continuous Anderson hamiltonian on (−L,L)d driven by a white noise and endowed with either Dirichlet or periodic boundary conditions. Our construction holds in any dimension dle3 and relies on the theory of regularity structures: it yields a self-adjoint operator in with pure point spectrum. In dge2, a renormalisation of the operator by means of infinite constants is required to compensate for ill-defined products involving functionals of the white noise. We also obtain left tail estimates on the distributions of the eigenvalues: in particular, for d=3 these estimates show that the eigenvalues do not have exponential moments.


Let \(L\in(0,\infty)\), \(d\in\{1,2,3\}\) and let \(\xi\) be a white noise on \((-L,L)^d\). The author constructs a random operator \(\mathcal H\) that can be formally written as \(-\Delta+\xi\) with either the homogeneous Dirichlet or the periodic boundary conditions. The operator \(\mathcal H\) is self-adjoint in \(L^2((-L,L)^d)\) and has a pure point spectrum. Consequently, left tail estimates on the distributions of its eigenvalues are obtained. The construction is non-trivial since \(\mathcal Hf\) does not belong to \(L^2\) for smooth functions \(f\). So, the author considers operators \[ \mathcal H_\varepsilon=-\Delta+\xi_\varepsilon+C_\varepsilon \] where \(\xi_\varepsilon\) is a mollification of \(\xi\) (i.e., a convolution of \(\xi\) and a smooth symmetric compactly supported mollifier) and \(C_\varepsilon\) is a suitable constant. It is proved that the increasing sequence of eigenvalues \(\{\lambda_{\varepsilon,n}:n\ge 1\}\) of \(\mathcal H_\varepsilon\) converge in probability to the eigenvalues \(\{\lambda_{n}\}\) of \(\mathcal H\). As for the eigenvectors \(\{\varphi_{\varepsilon,n}\}\) and \(\{\varphi_n\}\) of \(\mathcal H_\varepsilon\) and \(\mathcal H\) respectively, if \(m_i\) is the multiplicity of \(\lambda_i\) and \(M_n=m_1+\dots +m_n\), then, for any \(n\in\mathbb N\), the unit ball of the linear span of \(\{\varphi_{\varepsilon,j}:j\le M_n\}\) converges to the the unit ball of the linear span of \(\{\varphi_j:j\le M_n\}\) in the Hausdorff metric in probability. The second result asserts that, for any \(n\in\mathbb N\), there exist \(0<a<b<\infty\) such that \[ e^{-ax^{2-d/2}}\le\mathbb P\,(\lambda_n<-x)\le e^{-bx^{2-d/2}} \] for \(x\) large enough. The construction relies on the theory of Hairer's regularity structures.




Cited in
(35)








This page was built for publication: The continuous Anderson Hamiltonian in \(d \leq 3\)

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2324754)