On a new class of BDF and IMEX schemes for parabolic type equations

Fukeng Huang,Jie Shen
2024-05-01
Abstract:When applying the classical multistep schemes for solving differential equations, one often faces the dilemma that smaller time steps are needed with higher-order schemes, making it impractical to use high-order schemes for stiff problems. We construct in this paper a new class of BDF and implicit-explicit (IMEX) schemes for parabolic type equations based on the Taylor expansions at time $t^{n+\beta}$ with $\beta > 1$ being a tunable parameter. These new schemes, with a suitable $\beta$, allow larger time steps at higher-order for stiff problems than that is allowed with a usual higher-order scheme. For parabolic type equations, we identify an explicit uniform multiplier for the new second- to fourth-order schemes, and conduct rigorously stability and error analysis by using the energy argument. We also present ample numerical examples to validate our findings.
Numerical Analysis
What problem does this paper attempt to address?
The problem that this paper attempts to solve is the dilemma faced when using classical multi - step methods (such as BDF and IMEX schemes) in solving differential equations. That is, for stiff problems, in order to maintain stability, a smaller time step needs to be used, which makes high - order methods impractical in practical applications. Specifically, as the order of accuracy of BDF or IMEX - type schemes increases, their stability regions usually decrease, which means that for stiff problems, a smaller time step must be used, making high - order schemes practically infeasible for many complex nonlinear systems. Therefore, a natural question arises: Is it possible to develop higher - order multi - step methods so that their stability regions are comparable to or even larger than those of low - order classical BDF or IMEX schemes? To solve this problem, the paper proposes a new class of BDF and implicit - explicit (IMEX) schemes. These schemes are based on Taylor expansions at the time point \(t_{n + \beta}\), where \(\beta> 1\) is an adjustable parameter. By choosing an appropriate \(\beta\), these new schemes can use a larger time step in high - order cases, thus improving computational efficiency. In addition, the paper also conducts strict stability and error analyses on these new schemes and provides a large number of numerical examples to verify the theoretical results.