A neural network-based four-body potential energy surface for parahydrogen

Area of Science:

• Computational Chemistry
• Quantum Chemistry
• Molecular Interactions

Background:

• Accurate potential energy surfaces (PES) are essential for molecular simulations.
• Existing models for parahydrogen (para-H2) interactions may lack precision, especially for many-body effects.
• Understanding para-H2 interactions is vital in fields like condensed matter physics and astrophysics.

Purpose of the Study:

• To develop an accurate isotropic ab initio four-body interaction potential energy surface (PES) for para-H2.
• To enhance the predictive power of molecular simulations involving para-H2.
• To provide a more reliable computational tool for studying para-H2 systems.

Main Methods:

• Electronic structure calculations using coupled-cluster theory with single, double, and perturbative triple excitations.
• Utilizing an augmented correlation-consistent double zeta basis set with midbond functions.
• Constructing the PES with a multilayer perceptron and applying energy rescaling for improved accuracy.
• Adjusting long-range interaction energies to match empirical four-body dispersion interactions.

Main Results:

• A novel isotropic ab initio four-body PES for para-H2 has been generated.
• The four-body interaction energy is found to be net repulsive at short intermolecular separations.
• The developed PES, combined with existing two- and three-body potentials, is expected to improve agreement with experimental data.

Conclusions:

• The new four-body PES offers a significant advancement in modeling para-H2 interactions.
• This work provides a more accurate computational framework for studying para-H2.
• The improved PES is anticipated to lead to more precise predictions in simulations involving para-H2.