A transferable double exponential potential for condensed phase simulations of small molecules

Area of Science:

• Computational Chemistry
• Molecular Modeling
• Physical Chemistry

Background:

• The Lennard-Jones potential is a legacy function for non-bonded interactions in molecular modeling.
• Its persistence is due to computational limitations and the high cost of rewriting software and retraining force fields.

Purpose of the Study:

• Introduce Smirnoff-plugins, a flexible framework to extend the Open Force Field software stack.
• Enable the use of custom force field functional forms beyond the Lennard-Jones potential.
• Train and evaluate a novel transferable small molecule force field.

Main Methods:

• Developed Smirnoff-plugins to allow custom functional forms within the Open Force Field software.
• Utilized an automated infrastructure to train a new force field based on a double exponential form.
• Trained on over 1000 experimental condensed phase properties.

Main Results:

• The new double exponential force field demonstrates improved transfer free energies.
• Achieved acceptable conformational energetics, run times, and convergence properties.
• Outperformed state-of-the-art Lennard-Jones based force fields in key metrics.

Conclusions:

• Smirnoff-plugins provide a flexible solution for incorporating novel functional forms in molecular modeling.
• The developed double exponential force field offers a promising alternative to Lennard-Jones potentials.
• This work facilitates the development of more accurate and efficient molecular force fields.