Molecular orbital propagation to accelerate self-consistent-field convergence in an ab initio molecular dynamics...

Area of Science:

• Computational chemistry
• Theoretical chemistry
• Molecular dynamics

Background:

• Direct ab initio molecular dynamics (AIMD) simulations require accurate initial guesses for self-consistent-field (SCF) calculations.
• Predicting these initial guesses efficiently is crucial for reducing computational cost.
• Existing methods may not always provide optimal starting points, impacting simulation speed and stability.

Purpose of the Study:

• To propose a novel and effective method for predicting initial guesses for SCF calculations in direct AIMD simulations.
• To enhance the efficiency and feasibility of AIMD simulations through improved initial guess prediction.
• To introduce a method that accounts for molecular orbital (MO) dynamics, including crossing and mixing.

Main Methods:

• The proposed method, LIMO (Lagrange Interpolation for MO propagation), utilizes Lagrange interpolation (LI) polynomial techniques.
• LIMO predicts initial MO coefficients for the next AIMD step using data from several preceding steps.
• Orbital invariant formulas are derived by considering MO crossing and/or mixing, and an optimal LI polynomial degree is determined.

Main Results:

• Numerical tests demonstrate that the LIMO method is effective in predicting initial MO coefficients.
• The method proves to be feasible for practical application in direct AIMD simulations.
• The proposed approach successfully addresses the challenge of generating accurate starting points for SCF calculations.

Conclusions:

• The LIMO method offers a significant advancement in predicting initial guesses for SCF calculations within AIMD.
• This technique enhances the computational efficiency of direct AIMD simulations.
• LIMO provides a robust and practical solution for accelerating molecular dynamics studies.