Search

Machine learning (ML) strategies are opening the door to faster computer simulations, allowing us to simulate more realistic colloidal systems. Since the interactions in colloidal systems are often highly many-body, stemming from e.g. depletion and steric interactions, one of the challenges for these algorithms is capturing the many-body nature of these interactions. In this paper, we introduce a new ML-based strategy for fitting many-body interactions in colloidal systems where the many-body interaction is highly local. To this end, we develop Voronoi-based descriptors for capturing the local environment and fit the effective potential using a simple neural network. To test this algorithm, we consider a simple two-dimensional model for a colloid-polymer mixture, where the colloid-colloid interaction and colloid-polymer interactions are hard-disk like, while the polymers themselves interact as ideal gas particles. We find that a Voronoi-based description is sufficient to accurately capture the many-body nature of this system. Moreover, we find that the Pearson correlation function alone is insufficient to determine the predictive power of the network emphasizing the importance of additional metrics when assessing the quality of ML-based potentials.

The relationship between structure and dynamics in glassy fluids remains an intriguing open question. Recent work has shown impressive advances in our ability to predict local dynamics using structural features, most notably due to the use of advanced machine learning techniques. Here we explore whether a simple linear regression algorithm combined with intelligently chosen structural order parameters can reach the accuracy of the current, most advanced machine learning approaches for predicting dynamic propensity. To do this we introduce a method to pinpoint the cage state of the initial configuration -- i.e. the configuration consisting of the average particle positions when particle rearrangement is forbidden. We find that, in comparison to both the initial state and the inherent state, the structure of the cage state is highly predictive of the long-time dynamics of the system. Moreover, by combining the cage state information with the initial state, we are able to predict dynamic propensities with unprecedentedly high accuracy over a broad regime of time scales, including the caging regime.

In the quest to understand how structure and dynamics are connected in glasses, a number of machine learning based methods have been developed that predict dynamics in supercooled liquids. These methods include both increasingly complex machine learning techniques, and increasingly sophisticated descriptors used to describe the environment around particles. In many cases, both the chosen machine learning technique and choice of structural descriptors are varied simultaneously, making it hard to quantitatively compare the performance of different machine learning approaches. Here, we use three different machine learning algorithms -- linear regression, neural networks, and GNNs -- to predict the dynamic propensity of a glassy binary hard-sphere mixture using as structural input a recursive set of order parameters recently introduced by Boattini et al. [Phys. Rev. Lett. 127, 088007 (2021)]. As we show, when these advanced descriptors are used, all three methods predict the dynamics with nearly equal accuracy. However, the linear regression is orders of magnitude faster to train making it by far the method of choice.

Charged colloidal particles - both on the nano and micron scales - have been instrumental in enhancing our understanding of both atomic and colloidal crystals. These systems can be straightforwardly realized in the lab, and tuned to self-assemble into body-centered cubic (BCC) and face-centered cubic (FCC) crystals. While these crystals will always exhibit a finite number of point defects, including vacancies and interstitials - which can dramatically impact their material properties - their existence is usually ignored in scientific studies. Here, we use computer simulations and free-energy calculations to characterize vacancies and interstitials in both FCC and BCC crystals of point-Yukawa particles. We show that, in the BCC phase, defects are surprisingly more common than in the FCC phase, and the interstitials manifest as so-called crowdions: an exotic one-dimensional defect proposed to exist in atomic BCC crystals. Our results open the door to directly observing these elusive defects in the lab., Comment: 8 pages, 4 figures