Compositional fluctuations and polymorph selection in crystallization of model soft colloids

TL;DR

This study uses simulations and machine learning to understand and control polymorph selection in colloidal crystallization, revealing mechanisms behind phase transitions and structural fluctuations.

Contribution

It introduces a novel approach combining thermodynamic tuning and topological data analysis to manipulate and analyze polymorph selection in colloids.

Findings

Polymorph transition from FCC to BCC via intermediate regimes.

Critical-like composition fluctuations during nucleation.

FCC- and BCC-like particle arrangements near triple points.

Abstract

Understanding polymorph selection in atomic and molecular systems and its control through thermodynamic conditions and external factors (such as seed characteristics) is fundamental to the design of targeted materials and holds great significance in materials sciences. In this work, using Monte Carlo simulations on the Gaussian Core Model and Hard-Core Yukawa colloidal systems, we investigated the control of polymorph selection and explored the underlying mechanisms by tuning thermodynamic parameters. We demonstrate that by carefully modifying the free energy landscape to render the globally stable face-centered cubic (FCC) phase metastable with respect to the body-centered cubic (BCC) phase, the polymorphic identity of particles transitions from FCC-dominated to BCC-dominated via an intermediate regime where both phases nucleate -- either selectively or competitively -- giving rise to a critical-like composition fluctuation of the growing solid-like cluster during the nucleation process. We further probed the critical solid-like cluster compositions, especially in the vicinity of the triple point where the three phases coexist, and observed an interpenetrating arrangement of FCC- and BCC-like particles rather than a commonly observed non-classical core-shell-like two-step nucleation scenario. In addition, we investigated the polymorph selection signatures encoded in local structural fluctuations of the metastable fluid using a machine learning approach based on structural descriptors derived from persistent homology, a topological data analysis method. We believe that the insights gained from this work have the potential to add to the ongoing efforts to control crystallization pathways to obtain the desired functional material.