Toward Fully in Silico Melting Point Prediction Using Molecular Simulations

Area of Science:

• Computational Chemistry
• Materials Science
• Crystallography

Background:

• Melting point is a crucial compound property, with existing computational methods requiring experimental crystal structures, limiting predictive power.
• Crystal structure prediction (CSP) has advanced but faces challenges with closely-related polymorphs and the influence of thermal/kinetic factors on experimental structure identification.

Purpose of the Study:

• To develop a fully in silico method for accurate melting point prediction.
• To leverage crystal structure prediction (CSP) as a starting point for free energy-based melting point calculations.

Main Methods:

• Utilized free energy differences between crystal structures, which are often small.
• Employed crystal structures generated by CSP as input for free energy calculations.
• Tested the proposed method on two rigid molecules.

Main Results:

• Demonstrated that accurate melting point predictions can be achieved using CSP-derived crystal structures.
• Showcased that predicted crystal structures from CSP possess free energies close to experimental structures, enabling reliable calculations.

Conclusions:

• Accurate melting point prediction is feasible using a combination of CSP and free energy calculations.
• The developed method offers a fully in silico approach, bypassing the need for experimental crystal structure data.