FFCASP: A Massively Parallel Crystal Structure Prediction Algorithm.

Area of Science:

• Crystallography
• Materials Science
• Computational Chemistry

Background:

• Predicting crystal structures is crucial for understanding material properties.
• Existing methods face challenges with large unit cells and complex intermolecular interactions.

Purpose of the Study:

• To develop and evaluate a novel, efficient algorithm for crystal structure prediction.
• To assess the algorithm's performance on diverse molecular crystals with varying intermolecular forces.

Main Methods:

• Developed Fast and Flexible CrystAl Structure Predictor (FFCASP), a massively parallel global optimizer combining particle swarm and simulated annealing.
• Applied FFCASP to predict polymorphs of cytosine, coumarin, and pyrazinamide using DFT-SAPT and GAFF force fields.
• Utilized data mining and machine learning to analyze generated structures and identify common features.

Main Results:

• FFCASP successfully reproduced known experimental crystal structures for cytosine, coumarin, and pyrazinamide.
• The algorithm generated over 20,000 unique crystal structures, including novel low-energy polymorphs.
• FFCASP demonstrated efficiency in handling systems with over 200 atoms and optimizing ~100 parameters.

Conclusions:

• FFCASP is a powerful and versatile tool for crystal structure prediction.
• The algorithm's success in locating known and novel structures highlights its potential for discovering the polymorphic nature of other molecular crystals, including pharmaceuticals.