Asymptotic Behaviour of Time Stepping Methods for Phase Field Models

Area of Science:

• Computational mathematics
• Numerical analysis
• Materials science

Background:

• Metastable dynamics in materials science are often modeled by the Allen-Cahn and Cahn-Hilliard equations.
• Efficient numerical methods are crucial for simulating these complex systems, especially for capturing long-term behavior.
• Adaptive time stepping is key to balancing accuracy and computational cost.

Purpose of the Study:

• To investigate and compare the performance of various adaptive time stepping methods for metastable dynamics.
• To formally predict the optimal time step sizes for different methods under specific stability conditions.
• To analyze the trade-offs between accuracy, stability, and computational efficiency.

Main Methods:

• Analysis of first and second-order time stepping methods in a semi-discrete setting.
• Formal prediction of time step sizes based on local truncation error and small length scale parameters.
• Introduction and application of the 'profile fidelity' concept to assess method stability.
• Computational studies to validate analytical predictions.

Main Results:

• Some energy stable and fully implicit methods require more time steps than anticipated.
• Certain popular energy stable methods underperform compared to standard schemes.
• Backward Euler method shows preserved energy decay and profile fidelity for larger time steps in Allen-Cahn dynamics than previously thought.
• Eyre-type methods often perform worse due to loss of profile fidelity.

Conclusions:

• The choice of time stepping method significantly impacts the efficiency and accuracy of simulating metastable dynamics.
• Profile fidelity is a critical factor in determining the suitability of a method for these problems.
• Standard methods like Backward Euler can be surprisingly effective under certain conditions, challenging existing assumptions.