Artificial Neural Networks as Propagators in Quantum Dynamics.

Area of Science:

• Quantum chemistry
• Computational physics
• Chemical dynamics

Background:

• Molecular simulations are crucial for understanding chemical and biological processes.
• Simulating quantum dynamics, especially with time-dependent potentials, presents significant computational challenges.
• Artificial neural networks (ANNs) offer potential for accelerating complex simulations.

Purpose of the Study:

• To implement artificial neural networks (ANNs) as propagators for solving the time-dependent Schrödinger equation.
• To develop ANNs capable of simulating quantum dynamics of systems with time-dependent potentials.
• To apply these ANN propagators to model nuclear quantum effects in proton transfer systems.

Main Methods:

• ANNs were trained to map wavepackets between time steps in the discrete variable representation.
• Iterative application of trained ANN propagators allowed for long-time scale simulations.
• The method was applied to one- and two-dimensional proton transfer systems exhibiting hydrogen tunneling.
• ANNs were trained for both specific time-independent and general time-dependent potentials.

Main Results:

• The developed ANN propagators successfully simulated quantum dynamics, including nuclear quantum effects.
• The approach demonstrated accuracy in modeling proton transfer systems.
• Hierarchical, multiple time step algorithms facilitated parallelization.
• The methodology proved extensible to higher dimensions.

Conclusions:

• ANNs provide an effective strategy for accelerating molecular simulations of quantum dynamics.
• This approach accurately captures nuclear quantum effects like hydrogen tunneling.
• The method is versatile and applicable to a broad range of chemical and biological processes.