Lyapunov exponent for the parabolic Anderson model with Lévy noise (Q2575170): Difference between revisions

Let \(\{Y_x\}_{x\in{\mathbb Z}^d}\) be independent copies of a Lévy process \(Y\) starting at 0, on a probability space \((\Omega, {\mathcal F}, Q)\). The authors investigate a.s. asymptotics of the Lyapunov exponents of the positive solutions of \[ du(t,x)=\kappa \Delta u(t,x)dt+ u(t-,x)dY_x(t),\quad u(0,x)=u_0(x),\quad x\in{\mathbb Z}^d,\;t>0, \] for small \(\kappa\), where \(\Delta\) is the discrete Laplacian over \({\mathbb Z}^d\). This work extends earlier results where \(Y\) was the Brownian motion. This equation has various interesting physical interpretations (e.g., the motion of an electron in a crystal with impurities which vary in time). Assume that the distribution of \(Y\) is determined by the characteristic function \[ \varphi(z)= -{\alpha^2\over 2}z^2+ \int_{{\mathbb R}\setminus \{0\}}(e^{izu}-1-izu)\rho (du), \] where the Lévy measure \(\rho\) is such that \(\rho((-\infty,1])=0\) and \(\int_1^\infty u\rho(du)<\infty\). Then there exists a constant \(\lambda(\kappa)\) such that for any bounded non-negative function \(u_0\) on \({\mathbb Z}^d\), \(u_0\not\equiv 0\), the solution \(u\) of the equation satisfies \(\lim_{t\to\infty}{1\over t}\log u(t,x)=\lambda(\kappa)\) \(Q\)-a.s., \(x\in{\mathbb Z}^d\). If, additionally, \(\int_{[-1,\infty)\setminus\{0\}}(\log(1+u))^2\rho(du) <\infty\), then (and it is the main result of the paper) \[ \lambda(\kappa)-\lambda(0)\sim {\sigma^2c_B^2\over 4\log(1/\kappa)} \] as \(\kappa\to 0\), where \[ \sigma^2=\alpha^2+\int_{(-1,\infty)\setminus\{0\}}(\log(1+u))^2\rho(du),\quad \lambda(0)=-{\alpha^2/2}+\int_{(-1,\infty)\setminus\{0\}}(\log(1+u)-u)\rho(du), \] and \(c_B\) is a constant independent of \(\rho\), related to a system of independent Brownian motions.