Abstract:A reaction-diffusion problem with a Caputo time derivative is considered. An integral discretization scheme on a graded mesh along with a decomposition of the exact solution is proposed. The truncation error estimate of the discretization scheme is derived by using the remainder formula of the linear interpolation and some inequality estimate techniques. It is proved that the scheme is second-order convergent by applying a difference analogue of Gronwall's inequality, which exhibits an enhancement in the convergence rate compared with the L1 schemes. Numerical experiments are presented to support the theoretical result.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is to improve the convergence rate of the numerical solution of the time - fractional - order diffusion equation. Specifically, the author focuses on the reaction - diffusion problem with Caputo time derivative:
\[
D_t^\alpha u(x,t) + Lu(x,t) = f(x,t), \quad (x,t) \in Q := (0,l) \times (0,T],
\]
where
\[
Lu(x,t) := -p \frac{\partial^2 u}{\partial x^2}(x,t) + c(x)u(x,t),
\]
and the initial and boundary conditions are:
\[
u(x,0) = \varphi(x), \quad x \in [0,l],
\]
\[
u(0,t) = u(l,t) = 0, \quad t \in (0,T].
\]
Here \( D_t^\alpha \) represents the Caputo fractional - order derivative, \( 0 < \alpha < 1 \), \( p \) is a positive constant, \( c \in C[0,l] \) and \( c \geq 0 \), \( f \in C(\bar{Q}) \), \( \varphi \in C[0,l] \).
Existing literatures (such as [3, 10, 11]) have proposed different numerical methods to solve the above problem, but their convergence orders are limited. For example, in [10], a finite - difference scheme based on the standard L1 approximation of the time - fractional - order derivative and the central - difference space discretization was proposed, and its convergence order is \( O(M^{-2}+N^{-\min\{2 - \alpha,r\alpha\}}) \), where \( M \) and \( N \) are the space and time discretization parameters respectively, and \( r\geq1 \) is the mesh grading parameter. And in [3, 11], enhanced convergence orders \( O(M^{-2}+N^{-\min\{2 - \alpha,2r\alpha\}}) \) were obtained by using the pre - conditioned L1 scheme respectively.
In order to further improve the convergence rate, this paper proposes and analyzes an integral discretization scheme on a graded mesh and decomposes the exact solution. The truncation error estimate is derived by using the remainder formula of linear interpolation and some inequality estimation techniques. Finally, it is proved by using the discrete form of Gronwall's inequality that this scheme has second - order convergence \( O(M^{-2}+N^{-2}) \), thus significantly improving the convergence order.
Therefore, the main objective of this paper is to improve the convergence order of existing methods by proposing a new numerical method, so as to solve the time - fractional - order diffusion equation more efficiently.