Adiabatic/nonadiabatic state-to-state reactive scattering dynamics implemented on graphics processing units

Area of Science:

• Computational Chemistry
• Quantum Dynamics
• Chemical Physics

Background:

• Accurate simulation of atom-diatom reactive scattering is crucial for understanding chemical reaction dynamics.
• Traditional computational methods can be computationally intensive, limiting the scope of achievable simulations.
• Leveraging parallel processing, such as with graphics processing units (GPUs), offers a pathway to accelerate these calculations.

Purpose of the Study:

• To develop and implement an efficient time-dependent wavepacket code utilizing graphics processing units (GPUs) for atom-diatom state-to-state reactive scattering.
• To enhance the computational efficiency of quantum dynamics simulations for chemical reactions.
• To validate the developed GPU code by comparing its performance and accuracy against established methods.

Main Methods:

• Development of a GPU-accelerated time-dependent wavepacket code.
• Implementation of the split-operator method for wavepacket propagation entirely on GPUs.
• Introduction of an approximate rotational operator in the Hamiltonian to minimize GPU communication while preserving accuracy.

Main Results:

• The GPU-accelerated code successfully calculates differential cross sections for the H + H2 reaction.
• State-resolved reaction probabilities for non-adiabatic O((3)P,(1)D) + H2 transitions are accurately determined.
• Significant global speedups of up to 44.80 are achieved compared to serial CPU computations, demonstrating substantial performance gains.

Conclusions:

• The developed GPU-based time-dependent wavepacket code provides a highly efficient platform for studying atom-diatom reactive scattering.
• The computational efficiency gains enable more complex quantum dynamics simulations in chemical physics.
• The optimized approach balances computational speed with the accuracy required for state-to-state reaction dynamics.