A multifidelity Monte Carlo approach for simulating the diffusion coefficient of water. I. Forward problem

Area of Science:

• Computational chemistry
• Statistical mechanics
• Molecular dynamics simulations

Background:

• Simulations of the diffusion coefficient in liquid water are sensitive to simulation box size effects.
• Uncertainties in force field parameters (Lennard-Jones, partial charges) impact simulation accuracy.
• Developing efficient simulation methods is crucial for understanding water's dynamic properties.

Purpose of the Study:

• To present a multifidelity Monte Carlo framework for molecular dynamics simulations.
• To investigate computational speed-ups by leveraging model hierarchy based on simulation box size.
• To propagate uncertainties in force field parameters through simulations.

Main Methods:

• Constructing a model hierarchy based on simulation box size for liquid water.
• Representing water as a rigid three-point model.
• Propagating uncertainties in Lennard-Jones parameters and partial charges.
• Conducting large-scale studies with multiple computational models and box shapes.

Main Results:

• Achieved computational speed-ups of up to one order of magnitude.
• Demonstrated that combining high- and low-fidelity models offers speed-up compared to single high-fidelity models.
• Showcased the effectiveness of the multifidelity framework for molecular dynamics simulations.

Conclusions:

• The multifidelity Monte Carlo framework significantly accelerates molecular dynamics simulations of water diffusion.
• Leveraging model hierarchy and parameter uncertainties is key to computational efficiency.
• This approach provides a practical method for enhancing the simulation of liquid water dynamics.