A machine learned potential for investigating single crystal to single crystal transformations in complex organic...

Area of Science:

• Solid-state chemistry
• Materials science
• Computational chemistry

Background:

• Organic molecular crystals exhibit complex packing dominated by weak non-covalent interactions.
• In situ rearrangement of these crystals under stimuli is difficult to study.
• Understanding single-crystal-to-single-crystal (SCSC) transformations is crucial for materials design.

Purpose of the Study:

• To investigate pressure-induced SCSC transformations in 2,4,5-triiodo-1H-imidazole polymorphs.
• To develop accurate computational models for simulating such transitions.
• To analyze the roles of halogen bonds, hydrogen bonds, and proton transfer.

Main Methods:

• Utilized machine learning potentials based on Density Functional Theory (DFT).
• Employed an active learning strategy starting from both stable crystal phases (α and β).
• Performed molecular dynamic simulations to observe bond dynamics.

Main Results:

• Developed a robust DFT-based machine learning potential describing stable phases and transition pathways.
• Accurately simulated anisotropic interactions and their role in the SCSC transition.
• Observed and analyzed detailed bond breaking and reforming during proton transfer.

Conclusions:

• The developed approach accurately models SCSC transitions in organic crystals with anisotropic interactions.
• This method provides insights into complex solid-state chemical changes, including proton transfer.
• The strategy is promising for simulating diverse SCSC transitions in molecular systems.