A time dependent model problem with the Riesz or the Riemann-Liouville fractional differential operator of order 1<2 is considered. By penalizing the primary variable with a term of order h1−α, optimal rates of convergence are obtained for the semi-discrete minimal dissipation Local Discontinuous Galerkin (mdLDG) method. Using a von Neumann analysis, stability conditions proportional to hα are derived for the forward Euler method and both fractional operators in one dimensional domains. The CFL condition is numerically studied with respect to the approximation degree and the stabilization parameter. Our analysis and computations carried out using explicit high order strong stability preserving Runge-Kutta schemes reveal that the proposed penalization term is suitable for high order approximations and explicit time advancing schemes when α is close to one. A series of numerical experiments in 1D and 2D problems are presented to validate our theoretical results and those not covered by the theory.
Castillo, P., Gomez, S. (2019). Optimal stabilization and time step constraints for the forward Euler-Local Discontinuous Galerkin method applied to fractional diffusion equations. JOURNAL OF COMPUTATIONAL PHYSICS, 394, 503-521 [10.1016/j.jcp.2019.06.005].
Optimal stabilization and time step constraints for the forward Euler-Local Discontinuous Galerkin method applied to fractional diffusion equations
Sergio Gomez
2019
Abstract
A time dependent model problem with the Riesz or the Riemann-Liouville fractional differential operator of order 1<2 is considered. By penalizing the primary variable with a term of order h1−α, optimal rates of convergence are obtained for the semi-discrete minimal dissipation Local Discontinuous Galerkin (mdLDG) method. Using a von Neumann analysis, stability conditions proportional to hα are derived for the forward Euler method and both fractional operators in one dimensional domains. The CFL condition is numerically studied with respect to the approximation degree and the stabilization parameter. Our analysis and computations carried out using explicit high order strong stability preserving Runge-Kutta schemes reveal that the proposed penalization term is suitable for high order approximations and explicit time advancing schemes when α is close to one. A series of numerical experiments in 1D and 2D problems are presented to validate our theoretical results and those not covered by the theory.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.