DPG method with optimal test functions for a fractional advection diffusion equation (Q1676908): Difference between revisions

In this paper, the authors consider the following one-dimensional fractional advection diffusion problem \[ \begin{aligned} -D D^{\alpha-2}Du +b Du + cu & = f,\quad \text{on}\,\,I:=(0,1),\\ u(0) & =u(1)=0. \end{aligned}\tag{1} \] Here, \(D\) denotes the single spatial derivative and \(D^{\alpha-2}\), for \(\alpha \in (1,2)\), represents a Riemann-Liouville fractional integral operator of order \(\alpha-2\); the coefficients \(b\) and \(c\) satisfy certain assumptions. It is known that fractional diffusion problems like (1) are difficult to solve numerically and the so-called discontinuous Petrov-Galerkin (DPG) method can provide robust discretizations for such problems. First, the authors rewrite (1) as a first-order system and propose an ultra-weak variational formulation. Then, the DPG method with optimal test functions is applied to yield a discrete formulation, which is symmetric and positive definite. By analyzing the associated bilinear form as well as the adjoint problem, the well-posedness of both the continuous and discrete formulations is proven. Furthermore, an error estimate is obtained and the quasi-optimal convergence of the discrete scheme is established. Finally, several numerical experiments are presented to illustrate convergence orders for the DPG method on uniform meshes and on adaptively refined meshes.