Phase classification using neural networks: application to supercooled, polymorphic core-softened mixtures

Abstract

Characterization of phases of soft matter systems is a challenge faced in many physical chemical problems. For polymorphic fluids it is an even greater challenge. Specifically, glass forming fluids, as water, can have, besides solid polymorphism, more than one liquid and glassy phases, and even a liquid–liquid critical point. In this sense, we apply a neural network algorithm to analyze the phase behavior of a mixture of core-softened fluids that interact through the continuous-shouldered well (CSW) potential, which have liquid polymorphism and liquid–liquid critical points, similar to water. We also apply the neural network to mixtures of CSW fluids and core-softened alcohols models. We combine and expand methods based on bond-orientational order parameters to study mixtures, applied to mixtures of hardcore fluids and to supercooled water, to include longer range coordination shells. With this, the trained neural network was able to properly predict the crystalline solid phases, the fluid phases and the amorphous phase for the pure CSW and CSW-alcohols mixtures with high efficiency. More than this, information about the phase populations, obtained from the network approach, can help verify if the phase transition is continuous or discontinuous, and also to interpret how the metastable amorphous region spreads along the stable high density fluid phase. These findings help to understand the behavior of supercooled polymorphic fluids and extend the comprehension of how amphiphilic solutes affect the phases behavior.

A DPD model of soft spheres with waterlike anomalies and poly(a)morphism

“…Previous works have successfully employed similar approaches to investigate the structural characteristics of complex systems. 23,78 To compare the structure and dynamics, the mean square displacement (MSD) is utilized:…”

Score-based denoising for atomic structure identification

“…As such, they are often application-and/or structure-specific, and are not always easy to generalize beyond their original scope of applicability. More recently, data-driven machine-learning (ML) approaches are being developed for performing ordered phase classification and sometimes defect detection [10][11][12][13][14][15][16][17], often employing existing tools such as Steinhardt order parameters [1] for featurization. While comparatively more straightforward to develop with modern ML pipelines, these emerging methods require considerable amounts of carefully curated training data and are often informed by material-specific physics and domain knowledge which limit transferability of the trained models.…”

Persistent Local Structural Defectiveness as an Early Time Predictor of Intermittent Glassy Relaxation Events in Supercooled Water

“…Another brushstroke for this phenomenological picture has been provided by the notion that glassy dynamics is locally intermittent: the relaxation of the different local regions proceeds by means of sparse rapid bursts of mobility embodied by the emergence of relatively compact clusters of mobile particles, which have been called as d-clusters. − However, the great conundrum that has been challenging us for decades is that, unexpectedly, the huge dynamical slowing down that characterizes the process of transformation of a liquid into a glass is not accompanied by any evident static counterpart: no significant structural change is noticeable since the system looks liquid-alike all along its way to the glass. − However, it is worth mentioning that it has been shown that in the case of supercooled water this behavior is true for the short-range order, but it is not the case for large-scale structures that differ between (supercooled) liquid and glassy water . Additionally, even when the dynamics of the different parts of a supercooled liquid sample indeed differ wildly from one other, the intuitively expected existence of a causal structural link responsible for this behavior has been so elusive as to be considered an article of faith, and its complete identification still remains lacking in spite of recent interesting advances. ,, Indeed, the difficulty to single out a specific static correlation as a proper structural predictor of the resulting dynamics together with the inability to determine the actual physical nature of structural defects, has recently fostered an intense use of machine learning strategies. ,− …”

“…Specifically, it has been found that low-density amorphous water (LDA) and high-density amorphous water (HDA) are genuine glassy states with an equilibrium counterpart at the thermodynamic conditions accessible to computer simulations, and thus, LDA-like and HDA-like structures are present in the supercooled liquid state. Additionally, the quest for the emergence of slow dynamics in glassy relaxation bears particular importance for liquid water, − since water has been shown to suffer a fragile-to-strong crossover upon supercooling, which is distinct from typical glass-formers and, additionally, water’s local structures have been shown to be correlated to mobility in several water models. , The latter represents a relevant piece of information since certain static quantities (including potential energy) had previously failed to exhibit strong correlations with dynamics when evaluated at the initial configuration. , Recently, the search for structural quantities that correlate directly with dynamics has promoted the use of indirect approaches with some investigations resorting to machine learning techniques. ,, Lately, an energetically based new structural order parameter has been introduced which has enabled a proper characterization of water structural defects (which were named as D molecules) . This is crucial in this context since previous indicators have greatly overestimated defective molecules, , and thus, its potentiality will be exploited in this work.…”

“…33,34 Recently, the search for structural quantities that correlate directly with dynamics has promoted the use of indirect approaches with some investigations resorting to machine learning techniques. 5,15,16 Lately, an energetically based new structural order parameter has been introduced which has enabled a proper characterization of water structural defects (which were named as D molecules). 35 This is crucial in this context since previous indicators have greatly overestimated defective molecules, 22,35 and thus, its potentiality will be exploited in this work.…”