A conservative linearized difference scheme for the nonlinear fractional Schrödinger equation (Q2356075)

A linearized finite difference scheme is proposed for the nonlinear fractional Schrödinger equation \[ i u_t - (-\Delta)^{\frac{\alpha}{2}} u + \beta |u|^2 u = 0, \qquad a < x < b, \;\; 0 < t \leq T, \tag{1} \] with initial condition \(u(x,0)=u_0(x)\) and Dirichlet boundary conditions \(u(a,t) = u(b,t)=0\). Here \(- (-\Delta)^{\frac{\alpha}{2}}\) is the fractional Laplace operator, whose action can be computed with the Fourier transform, being equivalent to the Riesz fractional derivative. When \(\alpha = 2\), the standard cubic nonlinear Schrödinger equation is recovered. In this sense, Equation (1) constitutes a natural generalization to deal with fractional quantum phenomena. Although different numerical procedures have been designed over the years to solve this problem, and some of them do indeed preserve some invariants of the original equation, they generally require solving a nonlinear system with high accuracy at each step, and so they are very time consuming. By contrast, the finite difference method proposed here only requires solving a linear system whereas the norm of the numerical solution is preserved along the time integration interval. The scheme is thus unconditionally stable. It is proved rigorously that the scheme admits a unique solution and that it is convergent of order \(\mathcal{O}(\tau^2 + h^2)\) in the \(l^2\) norm. Here \(\tau\) represents the time step and \(h\) the space mesh size. The same results can be generalized to a system of coupled nonlinear fractional Schrödinger equations. The proposed scheme is illustrated on a pair of numerical examples. In the first one, the numerical accuracy of the method and the preservation of the norm are numerically checked, whereas in the second the effect of the fractional order \(\alpha\) is analyzed when simulating the collision of two soliton waves. For doing that, a pair of coupled nonlinear fractional Schrödinger equations is considered.