11 July 2019 in Chemical Science
Identifying local structure in molecular simulations is of utmost importance. The most common existing approach to identify local structure is to calculate some geometrical quantity referred to as an order parameter. In simple cases order parameters are physically intuitive and trivial to develop (e.g., ion-pair distance), however in most cases, order parameter development becomes a much more difficult endeavor (e.g., crystal structure identification). Using ideas from computer vision, we adapt a specific type of neural network called a PointNet to identify local structural environments in molecular simulations. A primary challenge in applying machine learning techniques to simulation is selecting the appropriate input features. This challenge is system-specific and requires significant human input and intuition. In contrast, our approach is a generic framework that requires no system-specific feature engineering and operates on the raw output of the simulations, i.e., atomic positions. We demonstrate the method on crystal structure identification in Lennard-Jones (four different phases), water (eight different phases), and mesophase (six different phases) systems. The method achieves as high as 99.5% accuracy in crystal structure identification. The method is applicable to heterogeneous nucleation and it can even predict the crystal phases of atoms near external interfaces. We demonstrate the versatility of our approach by using our method to identify surface hydrophobicity based solely upon positions and orientations of surrounding water molecules. Our results suggest the approach will be broadly applicable to many types of local structure in simulations.
17 August 2021 in Journal of Chemical Information and Modeling
1 March 2019 in Journal of Chemical Information and Modeling
28 November 2017 in The Journal of Chemical Physics
The mechanism of nucleation of clathrate hydrates of a water-soluble guest molecule is rigorously investigated with molecular dynamics (MD) simulations. Results from forward flux sampling, committor probability analysis, and twenty straightforward MD trajectories were combined to create a comprehensive understanding of the nucleation mechanism. Seven different classes of order parameters with a total of 33 individual variants were studied. We rank and evaluate the efficacy of prospective reaction coordinate models built from these order parameters and linear combinations thereof. Order parameters based upon water structuring provide a better approximation of the reaction coordinate than those based upon guest structuring. Our calculations suggest that the transition state is characterized by 2–3 partial, face-sharing 512 cages that form a structural motif observed in the structure II crystal. Further simulations show that once formed, this structure significantly affects the ordering of vicinal guest molecules, likely leading to hydrate nucleation. Our results contribute to the current understanding of the water–guest interplay involved in hydrate nucleation and have relevance to hydrate-based technologies that use water-soluble guest molecules (e.g., tetrahydrofuran) in mixed hydrate systems.
23 January 2021 in AIChE Journal
1 July 2020 in The Journal of Chemical Physics
Accurate molecular models of pure alkali halides are a prerequisite for developing transferable models of molten salts that can predict the properties of complex salt mixtures, such as those including dissolved actinide species and metal ions. Predicting the melting point of a substance represents a rigorous test of model quality. To this end, we compute the melting points of the alkali chlorides for a popular non-polarizable and polarizable model. Neither model yields more accurate predictions of the melting points across the entire family of alkali chlorides. Further calculations suggest that this may be because neither model simultaneously represents both the solid and liquid phases with sufficient accuracy across all four alkali chlorides. We find that the deviation from experiment in the model enthalpy of melting may be a good indicator of the deviation from experiment in the model melting temperature. Since the enthalpy of melting is easier to calculate in simulation than melting temperature, it may be a useful quantity to target when developing new force fields for molten salts.