Fast evaluation of polarizable forces.

Area of Science:

• Computational chemistry
• Molecular dynamics
• Biophysics

Background:

• Polarizable force fields are crucial for next-generation biomolecular simulations.
• Calculating induced atomic dipoles in polarizable models is computationally expensive.
• Current methods require solving large linear systems, hindering routine use.

Purpose of the Study:

• To develop computationally efficient methods for polarizable force fields.
• To reduce the cost of electrostatic energy and force calculations in biomolecular simulations.
• To enable routine application of polarizable models in molecular dynamics.

Main Methods:

• Efficient implementation of the particle-mesh Ewald method.
• Least-squares fitting for an accurate and robust predictor.
• Non-stationary iterative methods accelerated by a preconditioner.
• Comparison of self-consistent and extended Lagrangian approaches.

Main Results:

• Reduced computational cost from ~7.5x to <2x that of nonpolarizable models.
• Self-consistent approach with larger timesteps proved faster than extended Lagrangian.
• Energy drift is acceptably small with previous dipole moments or a strict convergence criterion.

Conclusions:

• Developed methods make polarizable force fields practical for routine biomolecular simulations.
• Computational efficiency is significantly improved, enabling wider adoption.
• The choice of initial guess and convergence criteria impacts energy drift.