An analysis of the L1 scheme for the subdiffusion equation with nonsmooth data (Q2794702)

From MaRDI portal





scientific article; zbMATH DE number 6554289
Language Label Description Also known as
English
An analysis of the L1 scheme for the subdiffusion equation with nonsmooth data
scientific article; zbMATH DE number 6554289

    Statements

    An analysis of the L1 scheme for the subdiffusion equation with nonsmooth data (English)
    0 references
    0 references
    0 references
    0 references
    11 March 2016
    0 references
    fractional diffusion
    0 references
    L1 scheme
    0 references
    error estimates
    0 references
    space-time fractional diffusion
    0 references
    semidiscretization
    0 references
    initial boundary problem
    0 references
    numerical experiment
    0 references
    convergence
    0 references
    linear finite element
    0 references
    error estimate
    0 references
    The authors first consider an initial boundary problem in a bounded convex polygonal domain \(\Omega\in \mathbb R^d\), \(d=1,2,3\), where the time derivative is the left-sided Caputo fractional derivative of order \(\alpha\), \(0<\alpha<1\), and the space operator is the Laplacian along with homogeneouos Dirichlet boundary conditions. For the fractional derivative, the L1 scheme of \textit{Z.-Z. Sun} and \textit{X. Wu} [Appl. Numer. Math. 56, No. 2, 193--209 (2006; Zbl 1094.65083)] is used for which the local truncation error was shown to be \(O(\tau^{2-\alpha})\) in the case that the solution is twice continuously differentiable. The authors first show by a numerical experiment that this is not the order of convergence of the scheme, neither for non-smooth data nor for smooth ones. They then show by using linear finite elements in space and the resolvent of the semi-discretized problem that the \(L_2\)-error for \(L_2\) initial data is of order \(O(h^2t^{-\alpha})\). For the fully discretized problem on an equidistant time grid with step \(\tau\), applying the L1 scheme, they further show that the error for \(L_2\) initial data (compared to the semi-discretized case) is of the order \(O(\tau t^{-1})\) which improves to \(O(\tau t^{\alpha-1})\) for smooth data.NEWLINENEWLINE In short, for fixed time and \(h=O(\tau^{1/2})\), one has first-order convergence in both cases of smooth and non-smooth data. This is confirmed by a series of numerical experiments. Finally, an error estimate is obtained for an equation with the Laplacian replaced by a more general operator like the Riemann-Liouville fractional derivative of order \(\beta\in(\frac32,2)\). This estimate is also confirmed by numerical experiments.
    0 references

    Identifiers

    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references