Direct Quantum Dynamics Using Grid-Based Wave Function Propagation and Machine-Learned Potential Energy Surfaces.

Area of Science:

• Computational Chemistry
• Quantum Dynamics
• Theoretical Chemistry

Background:

• Accurate nuclear quantum dynamics simulations are crucial for understanding molecular behavior.
• Traditional methods often require computationally expensive, pre-fitted potential energy surfaces (PES).
• On-the-fly calculations offer a more efficient alternative but require robust PES construction methods.

Purpose of the Study:

• To introduce a novel direct dynamics method for nuclear quantum dynamics.
• To enable efficient and accurate PES calculation during wavepacket propagation.
• To demonstrate the method's applicability to proton transfer dynamics in complex molecular systems.

Main Methods:

• Utilized Gaussian process regression (GPR) for on-the-fly PES construction.
• Employed standard grid-based algorithms, including the multiconfiguration time-dependent Hartree (MCTDH) method.
• Integrated with ab initio electronic structure codes for PES generation in specific cases.

Main Results:

• Achieved excellent agreement between direct dynamics simulations and previous studies for malonaldehyde and salicylaldimine proton transfer.
• Demonstrated the method's effectiveness for 2- and 6-dimensional quantum dynamics simulations.
• Successfully performed simulations without the need for prior PES fitting.

Conclusions:

• The developed direct dynamics approach provides an accurate and efficient route for nuclear quantum dynamics.
• This method facilitates simulations of large, multi-dimensional molecular systems.
• The implementation in the Quantics package opens new possibilities for theoretical chemistry research.