Search

In this paper, the regularity and finite difference methods for the two-dimensional delay fractional equations are considered. The analytic solution is derived by eigenvalue expansions and Laplace transformation. However, due to the derivative discontinuities resulting from the delay effect, the traditional L1-ADI scheme fails to achieve the optimal convergence order. To overcome this issue and improve the convergence order, a simple and cost-effective decomposition technique is introduced and a fitted L1-ADI scheme is proposed. The numerical tests are conducted to verify the theoretical result.

Exponential stability of system with time-varying delay is studied in this paper. By taking intermediate polynomials and slack variables, we propose novel matrix-refined weighted functions (MRWFs). Then, based on the MRWFs, we construct a suitable Lyapunov-Krasovskii functional (LKF) and propose an exponential stability condition. Finally, numerical examples verify that the stability condition proposed in this paper is more effective, which is less conservative than some past criteria.

We present a red–black skewed extrapolation cascadic multigrid (SkECMG) method to solve the Poisson equation in two dimensions based on the modified standard and skewed five-point finite difference discretization. With the help of the extrapolation technique, we develop a new extrapolation operator. Applying this proposed extrapolation operator for the second-order finite difference solutions on the current and previous coarse grid, we can design a fourth-order initial value for the iterative solver on the next finer grid. The red–black Gauss–Seidel method is adopted as a smoother which is conducive to parallel implementation. Moreover, we discuss a new sifting method as a stopping criterion to reduce the number of smoothing iterations. The numerical experiment is conducted on the square domain to verify that our SkECMG algorithm can achieve high efficiency and keep less cost simultaneously.

In this paper, a fast and memory-saving numerical scheme is presented for solving hidden-memory variable-order time-fractional diffusion equations based on the L1 method. Due to the nonlocality of fractional operators, the L1 method leads to a high computational complexity. To reduce the storage and computational cost, a modified exponential-sum-approximation method is utilized to approximate the convolution kernel involved in the fractional derivative. Additionally, one of the challenges faced during theoretical analysis is the loss of monotonicity of the temporal discretization coefficients caused by the hidden-memory variable order. A pioneering decomposition technique has been adopted to address this. The scheme has been theoretically proven to be convergent, and its effectiveness and accuracy have been confirmed through numerical examples.