Sharp estimation of the almost-sure Lyapunov exponent for the Anderson model in continuous space (Q2498925): Difference between revisions

This article is devoted to study the almost-sure large-time exponential behavior of the continuous Anderson model, i.e., the solution of the stochastic partial differential equation \[ u(t,x)=1+\int_0^1 \kappa \Delta_x u(s,x) \,ds + \int_0^1 u(s,x) W(ds,x), \] where \(x \in \mathbb R\) and \(W\) is a centered Gaussian field on \(\mathbb R_+ \times \mathbb R\) that is Brownian in time. Moreover, \(W\) can be not homogeneous in space. The authors prove that the almost-sure Lyapunov exponent defined by \(\lim_{t \to \infty} t^{-1} \log u(t,x)\), if it exists, is nonrandom, uniformly positive and does not depend on \(x\). In the proof, they make use of Malliavin calculus and they do not use the Gaussian properties of \(W\). They are also able to find upper and lower bounds for \(\limsup_{t \to \infty} t^{-1} \log u(t,x)\) and \(\liminf_{t \to \infty} t^{-1} \log u(t,x)\), respectively, as functions of the diffusion constant \(\kappa\) which depend on the regularity of \(W\) in space. The methods used are an extension of techniques used only before in a discrete space and they are simpler than those used in all previous results in continuous spaces, but notice that they are powerful since they do not require homogeneity. Moreover, the bounds are sharper than the previous ones and are optimal when the spatial regularity of \(W\) is in a logarithmic scale.