No label defined (Q3973384)

[For the entire collection see Zbl 0725.00019.] This paper describes a model for laser distortion in turbulence that consists of a random Schrödinger equation: \[ {\partial\over \partial z}V(x,y,z)={i\over 2k}\nabla^ 2V(x,y,z)+ikn_ 1(x,y,z)V(x,y,z), \] where \(\nabla^ 2=(\partial^ 2/\partial x^ 2)+(\partial^ 2/\partial y^ 2)\) and \(n_ 1(x,y,z)\) is a Gaussian random field representing index of refraction fluctuations. This model represents a bilinear system driven by a random noise input \(n_ 1\). A bilinear system model is defined for \(V_ z=V(\cdot,\cdot,z)\in L_ 2[R^ 2]\) as \[ {d\over dz}V_ z=AV_ z+B(V_ z,N_ z), \] where \(B\) is a bilinear operator and \(N(\cdot)\) is an \(L_ 2[(0,T); L_ 2(R^ 2)]\) valued white noise, using the Hilbert space valued white noise theory of Balakrishnan. Approximate solutions using Trotter-Kato product forms are shown to be random variables and converge in mean square. The corresponding Itô integral representations are calculated and also shown to converge in mean square. The Markov approximation, or assumption of a white noise input, is verified by considering the input \(N\) to be an Ornstein-Uhlenbeck process and taking the limit (weak) as \(N\) approaches a white noise, i.e. the input ''bandwidth'' is expanded to a white noise. The product form approximations are shown to converge weakly as \(N\) approaches the white noise limit. Simulation results show this convergence is rapid.