GAtor: A First-Principles Genetic Algorithm for Molecular Crystal Structure Prediction.

Area of Science:

• Computational chemistry
• Materials science
• Crystallography

Background:

• Predicting molecular crystal structures is crucial for materials design.
• Accurate structure prediction requires efficient exploration of complex energy landscapes.

Purpose of the Study:

• To introduce GAtor, a parallel genetic algorithm for molecular crystal structure prediction.
• To enhance structure prediction by employing evolutionary niching and machine learning for population clustering.

Main Methods:

• Utilizing a first-principles genetic algorithm (GAtor) interfaced with FHI-aims for density functional theory (DFT) calculations.
• Implementing evolutionary niching via machine learning-based clustering to promote diverse sampling of the potential energy surface.
• Employing a hierarchy of DFT functionals and dispersion methods for re-optimization and re-ranking of candidate structures.

Main Results:

• GAtor successfully predicted experimentally observed and other low-energy structures for four diverse blind test targets.
• A novel Z'=2 crystal structure with P1̅ symmetry was identified as the top-ranked structure for Target II (5-cyano-3-hydroxythiophene).
• The method demonstrates effectiveness in exploring challenging regions of the potential energy landscape.

Conclusions:

• GAtor is a powerful tool for molecular crystal structure prediction, capable of finding known and novel structures.
• Evolutionary niching and machine learning integration enhance the efficiency and robustness of crystal structure prediction.
• The developed approach aids in discovering new materials with desired properties.