Predicting Thermodynamic Properties of Alkanes by High-Throughput Force Field Simulation and Machine Learning

Area of Science:

• Computational chemistry
• Materials science
• Chemical engineering

Background:

• Accurate thermodynamic properties are crucial for chemical process design and material development.
• Experimental measurements are often costly and environmentally unsustainable.
• Previous molecular simulations were limited to specific molecule classes due to inconsistent force fields.

Purpose of the Study:

• To develop a high-throughput force field simulation (HT-FFS) procedure for calculating thermodynamic properties of diverse molecules.
• To validate the HT-FFS procedure against experimental data for alkanes.
• To explore the application of generated data for machine learning models.

Main Methods:

• Developed a high-throughput force field simulation (HT-FFS) procedure.
• Combined a general force field with a validated simulation protocol.
• Applied HT-FFS to calculate liquid densities, heats of vaporization, heat capacities, vapor-liquid equilibrium curves, critical temperatures, critical densities, and surface tensions for alkanes.

Main Results:

• HT-FFS predictions showed high accuracy and precision compared to experimental data for a wide range of alkanes.
• The procedure successfully calculated multiple thermodynamic properties.
• Generated a substantial dataset suitable for machine learning applications.

Conclusions:

• The HT-FFS procedure is a valid and efficient approach to supplement experimental thermodynamic property measurements.
• The developed method enables the study of a larger number of molecules than previously possible.
• Machine learning models can be developed from HT-FFS data to expand predictive capabilities.