Accelerating Ab Initio Path Integral Simulations via Imaginary Multiple-Timestepping.

Area of Science:

• Quantum Chemistry
• Computational Physics
• Molecular Dynamics

Background:

• Path integral simulations are crucial for modeling quantum molecular motion.
• Accurate simulations often require significant computational resources.
• Multiple-timestep techniques have shown promise in classical molecular dynamics.

Purpose of the Study:

• To develop and validate computationally efficient multiple-timestep schemes for ab initio equilibrium path integral simulations.
• To explore the effectiveness of 1-electron and 2-electron truncations within these schemes.
• To assess the potential for achieving quantum mechanical simulation costs comparable to classical methods.

Main Methods:

• Implementation of multiple-timestep schemes in imaginary time for path integral simulations.
• Utilizing different levels of electronic structure theory for various path integral replicas.
• Testing 1-electron (atomic-orbital basis set) and 2-electron (electron correlation) truncations.
• Validation using analytic potentials and ab initio molecular examples.

Main Results:

• Demonstrated effectiveness of both 1-electron and 2-electron truncations.
• Observed computational speedups of 1.6-4.0x with a 4-fold reduction in time slices.
• Feasibility of up to an 8-fold reduction in time slices in certain cases.
• Successful testing of structural distributions and thermodynamic averages.

Conclusions:

• Multiple-timestep schemes offer significant computational efficiency for ab initio path integral simulations.
• The developed methods allow for quantum mechanical simulations at costs approaching classical simulations.
• The framework is adaptable to various quantum chemistry methods, including DFT and MP2 theory.