Table of Contents
Fetching ...

Adaptive Time-Step Semi-Implicit One-Step Taylor Scheme for Stiff Ordinary Differential Equations

S. Boscarino, E. Macca

TL;DR

The paper tackles stiff/non-stiff ODE integration by developing high-order Taylor-based semi-implicit and implicit one-step schemes. It derives first- and second-order SI-T-1, SI-T-2 and I-T-1, I-T-2 for the split system $U' = f(U) + g(U)$, employing explicit Jacobians where appropriate and implicit handling of the stiff part. A linear stability analysis shows $L$-stability and introduces stability regions to characterize performance, while an embedded time-step controller enables adaptive stepping. Numerical experiments on the Van der Pol problem demonstrate robustness, asymptotic preserving behavior, and competitive efficiency against ode15s and IMEX-RK(2,1), validating the practical value of the proposed methods for stiff ODEs and potentially PDEs via AP properties.

Abstract

In this study, we propose high-order implicit and semi-implicit schemes for solving ordinary differential equations (ODEs) based on Taylor series expansion. These methods are designed to handle stiff and non-stiff components within a unified framework, ensuring stability and accuracy. The schemes are derived and analyzed for their consistency and stability properties, showcasing their effectiveness in practical computational scenarios.

Adaptive Time-Step Semi-Implicit One-Step Taylor Scheme for Stiff Ordinary Differential Equations

TL;DR

The paper tackles stiff/non-stiff ODE integration by developing high-order Taylor-based semi-implicit and implicit one-step schemes. It derives first- and second-order SI-T-1, SI-T-2 and I-T-1, I-T-2 for the split system , employing explicit Jacobians where appropriate and implicit handling of the stiff part. A linear stability analysis shows -stability and introduces stability regions to characterize performance, while an embedded time-step controller enables adaptive stepping. Numerical experiments on the Van der Pol problem demonstrate robustness, asymptotic preserving behavior, and competitive efficiency against ode15s and IMEX-RK(2,1), validating the practical value of the proposed methods for stiff ODEs and potentially PDEs via AP properties.

Abstract

In this study, we propose high-order implicit and semi-implicit schemes for solving ordinary differential equations (ODEs) based on Taylor series expansion. These methods are designed to handle stiff and non-stiff components within a unified framework, ensuring stability and accuracy. The schemes are derived and analyzed for their consistency and stability properties, showcasing their effectiveness in practical computational scenarios.
Paper Structure (8 sections, 25 equations, 5 figures, 1 table)

This paper contains 8 sections, 25 equations, 5 figures, 1 table.

Figures (5)

  • Figure 1: Stability regions $S_1$, in the $(Re(w); Im(w))$ plane, of some second order schemes such as: IMEX-RK2 method; Implicit and Semi-Implicit Taylor method and Heun scheme.
  • Figure 2: Semi-implicit numerical $y-$solution for the Van der Pol system \ref{['VDP']} with well-prepared initial conditions \ref{['IC_wp']} obtained at time $t=3\mu$ with $\Delta t_0 = 0.1$. The zoom of the second challenging boundary layers $t \approx 1.6\mu$ are reported. The reference solutions have been computed with the ode15s matlab solver.
  • Figure 3: Implicit numerical $y-$solutions for the Van der Pol system \ref{['VDP']} with well-prepared initial conditions \ref{['IC_wp']} obtained at time $t=3\mu$ with $\Delta t_0 = 0.1$. The zoom of the second challenging boundary layers $t \approx 1.6\mu$ are reported. The reference solutions have been computed with the ode15s matlab solver.
  • Figure 4: Semi-implicit numerical $y-$solution for the Van der Pol system \ref{['VDP']} with unprepared initial conditions \ref{['IC_nowp']} obtained at time $t=3\mu$ with $\Delta t_0 = 0.1$. The zoom of the second challenging boundary layers $t \approx 1.6\mu$ are reported. The reference solutions have been computed with the ode15s matlab solver.
  • Figure 5: Implicit numerical $y-$solution for the Van der Pol system \ref{['VDP']} with unprepared initial conditions \ref{['IC_nowp']} obtained at time $t=3\mu$ with $\Delta t_0 = 0.1$. The zoom of the second challenging boundary layers $t \approx 1.6\mu$ are reported. The reference solutions have been computed with the ode15s matlab solver.