On large deviations for the parabolic Anderson model (Q2380772): Difference between revisions

Consider on a probability space \((\Omega, {\mathcal F}, {\mathbf P})\), a field \(\{W_x:~x\in {\mathbb Z}^d\}\) of i.i.d. one-dimensional Brownian motions. Let \(\Delta\) denote the discrete Laplacian and let \(\kappa\) be a positive constant. Then consider the solution \(u(x,t)\) to the parabolic Anderson equation: \[ \frac{\partial u}{\partial t}(x,t)=\kappa \Delta u(x,t)+ u(x,t)\partial W_x,~~x\in {\mathbb Z}^d,~~t>0,~~~u(x,0)\equiv 1, \] where \(\partial\) denotes the Stratonovich differential. The focus of this article is on the different behavior of large deviations of random functionals associated with the parabolic Anderson model above the mean versus large deviations below the mean. The functionals the authors treat are the solution \(u(x,t)\) (to the spatially discrete parabolic Anderson model) and a functional \(A_n\) which is used in analysing the a.s. Lyapunov exponent for \(u(x,t)\). Both satisfy a ``law of large numbers'', with \[ \lim_{t\to \infty}\frac{1}{t} \log u(x,t)=\lambda(\kappa) \] and \[ \lim_{n\to \infty}\frac{1}{n} A_n=\alpha. \] The authors then think of \(\lambda(\kappa) t\) and \(\alpha\) as being the mean of the respective quantities \(\log u(x,t)\) and \(A_n\). Typically, the large deviations for such functionals exhibits a strong asymmetry; large deviations above the mean take on a different order of magnitude from large deviations below the mean. The authors develop robust techniques to quantify and explain the differences.