Assessing Polymorph Stability and Phase Transitions at Finite Temperature: Integrating Crystal Structure Prediction, Lattice Dynamics, and Molecular Dynamics

GB Correa and S Konstantinopoulos and BI Tan and Y Zhang and FW Tavares and CS Adjiman and EJ Maginn, JOURNAL OF CHEMICAL THEORY AND COMPUTATION, 21, 12197-12213 (2025).

DOI: 10.1021/acs.jctc.5c01387

Determining finite-temperature polymorph stability and phase transitions remains a major challenge in crystal structure prediction (CSP). While static lattice energy methods, the zeroth-order CSP, may offer some initial insights, they neglect vibrational and entropic contributions that can influence stability under real-world conditions. Molecular dynamics (MD) approaches, on the other hand, can be used for free energy analysis across wide temperature ranges, but their computational cost limits their applicability to a small number of putative crystal structures. This work introduces a multistage protocol that combines CSP, harmonic approximation lattice dynamics (HA-LD), and MD to assess a larger number of structures than is possible with MD alone. Using tetracyanoethylene (TCNE) as a test case, we first carry out a zeroth- order CSP and then apply the relatively cheap HA-LD method to prescreen the 100 lowest-energy structures. Five thermodynamically relevant candidates are identified and advanced to detailed free energy analysis using the more expensive pseudosupercritical path (PSCP) method in MD. MD-PSCP captures the anharmonic and thermal expansion effects neglected in HA-LD, enabling excellent predictions of observed phase stability. The cubic-monoclinic enantiotropic transition of TCNE is predicted at 322 K, within the experimental range of 292-326 K, and the melting temperature of the monoclinic form is estimated with a deviation of less than 10 K from the experimental value. Our simulations also indicate a potentially more stable hypothetical form, whose viability requires further investigation. By combining CSP for structure generation, HA-LD for efficient prescreening, and MD-PSCP for enhanced predictive accuracy, our workflow overcomes the limitations of each method alone. This approach improves finite-temperature polymorph prediction and offers a practical framework for materials and pharmaceutical applications, where thermal effects are critical.

Return to Publications page