Machine Learning Molecular Dynamics Research Articles

Large-scale machine-learning molecular dynamics simulation of primary radiation damage in tungsten

Large-scale machine-learning molecular dynamics simulation of primary radiation damage in tungsten

Discovery of multi-functional polyimides through high-throughput screening using explainable machine learning

Polyimides have been widely used in modern industries because of their excellent mechanical and thermal properties, e.g., high-temperature fuel cells, displays, and aerospace composites. However, it usually takes decades of experimental efforts to develop a successful product. Aiming to expedite the discovery of high-performance polyimides, we utilize computational methods of machine learning (ML) and molecular dynamics (MD) simulations. Our study provides compelling evidence for the effectiveness of a data-driven approach in discovering novel polyimides. We first build a comprehensive library of more than 8 million hypothetical polyimides based on the polycondensation of existing dianhydride and diamine/diisocyanate molecules. Then we establish multiple ML models for the thermal and mechanical properties of polyimides based on their experimentally reported values, including glass transition temperature, Young’s modulus, and tensile yield strength. The obtained ML models demonstrate excellent predictive performance in identifying the key chemical substructures influencing the thermal and mechanical properties of polyimides. The use of explainable machine learning describes the effect of chemical substructures on individual properties, from which human experts can understand the cause of the ML model decision. Applying the well-trained ML models, we obtain property predictions of the 8 million hypothetical polyimides. Then, we screen the whole hypothetical dataset and identify three (3) best-performing novel polyimides that have better-combined properties than existing ones through Pareto frontier analysis. For an easy query of the discovered high-performing polyimides, we also create an online platform https://polyimide-explorer.herokuapp.com/ that embeds the developed ML model with interactive visualization. Furthermore, we validate the ML predictions through all-atom MD simulations and examine their synthesizability. The MD simulations are in good agreement with the ML predictions and the three novel polyimides are predicted to be easy to synthesize via Schuffenhauer’s synthetic accessibility score. Following the proposed ML guidance, we successfully synthesized a novel polyimide and the experimentally obtained high glass transition/thermal decomposition temperature demonstrated its excellent thermal stability. Our study demonstrates an efficient way to expedite the discovery of novel polymers using ML prediction and MD validation. The high-throughput screening of a large computational dataset can serve as a general approach for new material discovery in other polymeric material exploration problems, such as organic photovoltaics, polymer membranes, and dielectrics.

Understanding the solvation structures of glyme-based electrolytes by machine learning molecular dynamics

Glyme-based electrolytes are of great interest for rechargeable lithium metal batteries due to their high stability, low vapor pressure, and non-flammability. Understanding the solvation structures of these electrolytes at the atomic level will facilitate the design of new electrolytes with novel properties. Recently, classical molecular dynamics (CMD) and ab initio molecular dynamics (AIMD) have been applied to investigate electrolytes with complex solvation structures. On one hand, CMD may not provide reliable results as it requires complex parameterization to ensure the accuracy of the classical force field. On the other hand, the time scale of AIMD is limited by the high cost of ab initio calculations, which causes that solvation structures from AIMD simulations depend on the initial configurations. In order to solve the dilemma, machine learning method is applied to accelerate AIMD, and its time scale can be extended dramatically. In this work, we present a computational study on the solvation structures of triglyme (G3) based electrolytes by using machine learning molecular dynamics (MLMD). Firstly, we investigate the effects of density functionals on the accuracy of machine learning potential (MLP), and find that PBE-D3 shows better accuracy compared to BLYP-D3. Then, the densities of electrolytes with different concentration of LiTFSI are computed with MLMD, which shows good agreement with experiments. By analyzing the solvation structures of 1 ns MLMD trajectories, we found that Li+ prefers to coordinate with a G3 and a TFSI− in equimolar electrolytes. Our work demonstrates the significance of long-time scale MLMD simulations for clarifying the chemistry of non-ideal electrolytes.

Switching of Redox Levels Leads to High Reductive Stability in Water-in-Salt Electrolytes.

Developing nonflammable electrolytes with wide electrochemical windows is of great importance for energy storage devices. Water-in-salt electrolytes (WiSE) have attracted great interests due to their widely opened electrochemical windows and high stability. Previous theoretical investigations have revealed changes in solvation shell of water molecules result in opening of HOMO-LUMO gaps of water, leading to the formation of an anion-derived solid-electrolyte-interphase (SEI) in WiSE. However, how solvation structures affect electrochemical windows at atomic level is still a puzzle, which hinders optimization and design of aqueous electrolytes. Herein, machine learning molecular dynamics and free energy calculation method are applied to compute redox potentials of anions of Li-salts and water of aqueous electrolytes at a range of salt concentrations. Furthermore, an analysis based on local solvation structures is employed to demonstrate the structure-property relations. Our calculation shows that the hydrogen evolution reaction is impeded in WiSE due to switching of the order of redox levels of the anion and H2O, leading to formation of SEI and high reductive stability. Level switching is caused by the special solvation environments of isolated water molecules. Our work provides new insight into the electrochemistry of aqueous electrolytes which would benefit the electrolyte design in energy storage devices.

Distinct anharmonic characteristics of phonon-driven lattice thermal conductivity and thermal expansion in bulk MoSe2 and WSe2

Machine-learning molecular dynamics simulations pave the way to completely treat the anharmonicity of phonons. Low-energy anharmonic modes in transition-metal dichalcogenides drive the thermal and transport properties.

Atomistic insights into the mechanical anisotropy and fragility of monolayer fullerene networks using quantum mechanical calculations and machine-learning molecular dynamics simulations

In this work, we comprehensively study the mechanical properties of the newly synthesized monolayer quasi-hexagonal-phase fullerene (qHPF) membrane [Hou et al., 2022] under uniaxial tension by using quantum mechanical density-functional-theory (DFT) calculations and molecular dynamics (MD) simulations with a machine-learned neuroevolution potential (NEP). The elastic properties and fracture behaviors of monolayer qHPF are found to be strongly anisotropic due to the different properties between the inter-fullerene C–C single bonds and [2 + 2] cycloaddition bonds. Moreover, the tensile strength and fracture strain of monolayer qHPF are much smaller than those of any other existing two-dimensional (2D) carbon crystals. The very small tensile strength or fracture strain is ascribed to the inhomogeneous deformation of the stretched monolayer qHPF, which originates from the stiffness difference between the soft inter-fullerene bonds and the rigid intra-fullerene bonds. Compared with DFT calculations at the ground state, the NEP-based extensive MD simulations predict a much smaller tensile strength and fracture strain for monolayer qHPF due to their consideration of the effects of temperature and membrane size. Our work not only reveals the underlying mechanism of the fracture behaviors of monolayer fullerene networks from an atomistic perspective, but also shows the effectiveness and accuracy of the NEP approach in determining the mechanical properties of 2D materials in the realistic situations.

Dynamical Study of Adsorbate-Induced Restructuring Kinetics in Bimetallic Catalysts Using the PdAu(111) Model System.

Dynamic restructuring of bimetallic catalysts plays a crucial role in their catalytic activity and selectivity. In particular, catalyst pretreatment with species such as carbon monoxide and oxygen has been shown to be an effective strategy for tuning the surface composition and morphology. Mechanistic and kinetic understanding of such restructuring is fundamental to the chemistry and engineering of surface active sites but has remained challenging due to the large structural, chemical, and temporal degrees of freedom. Here, we combine time-resolved temperature-programmed infrared reflection absorption spectroscopy, ab initio thermodynamics, and machine-learning molecular dynamics to uncover previously unidentified timescale and kinetic parameters of in situ restructuring in Pd/Au(111), a highly relevant model system for dilute Pd-in-Au nanoparticle catalysts. The key innovation lies in utilizing CO not only as a chemically sensitive probe of surface Pd but also as an agent that induces restructuring of the surface. Upon annealing in vacuum, as-deposited Pd islands became encapsulated by Au and partially dissolved into the subsurface, leaving behind isolated Pd monomers on the surface. Subsequent exposure to 0.1 mbar CO enabled Pd monomers to repopulate the surface up to 373 K, above which complete Pd dissolution occurred by 473 K, with apparent activation energies of 0.14 and 0.48 eV, respectively. These restructuring processes occurred over the span of ∼1000 s at a given temperature. Such a minute-timescale dynamics not only elucidates the fluxional nature of alloy catalysts but also presents an opportunity to fine-tune the surface under moderate temperature and pressure conditions.

Machine learning potentials of kaolinite based on the potential energy surfaces of GGA and meta-GGA density functional theory

Machine learning molecular dynamics (MLMD) is a promising method for predicting material properties with high accuracy and low computational costs. Two kaolinite machine learning potentials (MLPs) were created based on a dataset that consists of the results of the first-principles calculations with the Perdew–Burke–Ernzerhof (PBE) generalized gradient approximation (GGA) and the strongly constrained and appropriately normed (SCAN) mata-GGA functionals. The structural and mechanical properties of kaolinite were evaluated by the MLMD simulations with the MLPs. The results were compared with those obtained by the density functional theory simulations, classical molecular dynamics simulations, and experiments. The MLMD with the MLP based on SCAN performed well for an accurate evaluation of these kaolinite properties. The vibrational density of states of kaolinite was evaluated using the MLPs, and the results were compared with inelastic neutron scattering experiment data. The MLP based on SCAN agreed well with the experimental results. Remarkably, the MLP successfully reproduced the spectral shape in the low-wavenumber, where evaluation requires a long-time simulation with high accuracy.

Elucidation of Cu-Zn Surface Alloying on Cu(997) by Machine-Learning Molecular Dynamics.

The Cu-Zn surface alloy has been extensively involved in the investigation of the true active site of Cu/ZnO/Al2O3, the industrial catalyst for methanol synthesis which remains under controversy. The challenge lies in capturing the interplay between the surface and reaction under operating conditions, which can be overcome given that the explicit dynamics of the system is known. To provide a better understanding of the dynamic of Cu-Zn surface at the atomic level, the structure and the formation process of the Cu-Zn surface alloy on Cu(997) were investigated by machine-learning molecular dynamics (MD). Gaussian process regression aided with on-the-fly learning was employed to build the force field used in the MD. The simulation reveals atomistic details of the alloying process, that is, the incorporation of deposited Zn adatoms to the Cu substrate. The surface alloying is found to start at upper and lower terraces near the step edge, which emphasize the role of steps and kinks in the alloying. The incorporation of Zn at the middle terrace was found at the later stage of the simulation. The rationalization of alloying behavior was performed based on statistics and barriers of various elementary events that occur during the simulation. It was observed that the alloying scheme at the upper terrace is dominated by the confinement of Zn step adatoms by other adatoms, highlighting the importance of step fluctuations in the alloying process. On the other hand, the alloying scheme at the lower terrace is dominated by direct exchange between the Zn step adatom and the Cu atom underneath. The alloying at the middle terrace is dominated by the wave deposition mechanism and deep confinement of Zn adatoms. The short propagation of alloyed Zn in the middle terrace was observed to proceed by means of indirect exchange instead of local exchange as proposed in the previous scanning tunneling microscopy (STM) observation. The comparison of migration rate and activation energies to the result of STM observation is also made. We have found that at a certain distance from the surface, the STM tip significantly affects the elementary events such as vacancy formation and direct exchange.

Machine learning molecular dynamics simulations toward exploration of high-temperature properties of nuclear fuel materials: case study of thorium dioxide

Predicting materials properties of nuclear fuel compounds is a challenging task in materials science. Their thermodynamical behaviors around and above the operational temperature are essential for the design of nuclear reactors. However, they are not easy to measure, because the target temperature range is too high to perform various standard experiments safely and accurately. Moreover, theoretical methods such as first-principles calculations also suffer from the computational limitations in calculating thermodynamical properties due to their high calculation-costs and complicated electronic structures stemming from f-orbital occupations of valence electrons in actinide elements. Here, we demonstrate, for the first time, machine-learning molecular-dynamics to theoretically explore high-temperature thermodynamical properties of a nuclear fuel material, thorium dioxide. The target compound satisfies first-principles calculation accuracy because f-electron occupation coincidentally diminishes and the scheme meets sampling sufficiency because it works at the computational cost of classical molecular-dynamics levels. We prepare a set of training data using first-principles molecular dynamics with small number of atoms, which cannot directly evaluate thermodynamical properties but captures essential atomistic dynamics at the high temperature range. Then, we construct a machine-learning molecular-dynamics potential and carry out large-scale molecular-dynamics calculations. Consequently, we successfully access two kinds of thermodynamic phase transitions, namely the melting and the anomalous lambda transition induced by large diffusions of oxygen atoms. Furthermore, we quantitatively reproduce various experimental data in the best agreement manner by selecting a density functional scheme known as SCAN. Our results suggest that the present scale-up simulation-scheme using machine-learning techniques opens up a new pathway on theoretical studies of not only nuclear fuel compounds, but also a variety of similar materials that contain both heavy and light elements, like thorium dioxide.

On the anomalous diffusion of proton in Y-doped BaZrO3 perovskite oxide

Proton transport in perovskite oxides (ABO3) is a complex dynamical process involving hydroxide ion rotation and proton transfer. Combining ab initio molecular dynamics (AIMD) and machine-learning molecular dynamics (MLMD) simulations of proton dynamics in Y-doped BaZrO3 perovskite oxide, we systematically illustrate the anomalous characteristics of proton motion. At short times, proton performs superdiffusion through the OH wag and proton transfer. At intermediate times, the larger rotation angle results in the larger plateau value. At longer times, proton diffusion first becomes subdiffusive due to the hydroxide ion rotations then gradually turns to normal diffusion. A theoretical model is also offered to quantitatively describe this abnormal proton diffusion. In all, this work provides a more complete understanding of proton dynamics in perovskite oxides.

The chemical origin of temperature-dependent lithium-ion concerted diffusion in sulfide solid electrolyte Li10GeP2S12

Solid-state batteries have received increasing attention in scientific and industrial communities, which benefits from the intrinsically safe solid electrolytes (SEs). Although much effort has been devoted to designing SEs with high ionic conductivities, it is extremely difficult to fully understand the ionic diffusion mechanisms in SEs through conventional experimental and theoretical methods. Herein, the temperature-dependent concerted diffusion mechanism of ions in SEs is explored through machine-learning molecular dynamics, taking Li10GeP2S12 as a prototype. Weaker diffusion anisotropy, more disordered Li distributions, and shorter residence time are observed at a higher temperature. Arrhenius-type temperature dependence is maintained within a wide temperature range, which is attributed to the linear temperature dependence of jump frequencies of various concerted diffusion modes. These results provide a theoretical framework to understand the ionic diffusion mechanisms in SEs and deepen the understanding of the chemical origin of temperature-dependent concerted diffusions in SEs.

Dynamics-Based Peptide–MHC Binding Optimization by a Convolutional Variational Autoencoder: A Use-Case Model for CASTELO

An unsolved challenge in the development of antigen-specific immunotherapies is determining the optimal antigens to target. Comprehension of antigen-major histocompatibility complex (MHC) binding is paramount toward achieving this goal. Here, we apply CASTELO, a combined machine learning-molecular dynamics (ML-MD) approach, to identify per-residue antigen binding contributions and then design novel antigens of increased MHC-II binding affinity for a type 1 diabetes-implicated system. We build upon a small-molecule lead optimization algorithm by training a convolutional variational autoencoder (CVAE) on MD trajectories of 48 different systems across four antigens and four HLA serotypes. We develop several new machine learning metrics including a structure-based anchor residue classification model as well as cluster comparison scores. ML-MD predictions agree well with experimental binding results and free energy perturbation-predicted binding affinities. Moreover, ML-MD metrics are independent of traditional MD stability metrics such as contact area and root-mean-square fluctuations (RMSF), which do not reflect binding affinity data. Our work supports the role of structure-based deep learning techniques in antigen-specific immunotherapy design.

Targeted design of advanced electrocatalysts by machine learning

Exploring the production and application of clean energy has always been the core of sustainable development. As a clean and sustainable technology, electrocatalysis has been receiving widespread attention. It is crucial to achieve efficient, stable and cheap electrocatalysts. However, the traditional “trial and error” method is time-consuming, laborious and costly. In recent years, with the significant increase in computing power, computations have played an important role in electrocatalyst design. Nevertheless, it is still difficult to search for advanced electrocatalysts in the vast chemical space through traditional density functional theory (DFT) computations. Fortunately, the development of machine learning and interdisciplinary integration will inject new impetus into targeted design of electrocatalysts. Machine learning is able to predict electrochemical performances with an accuracy close to DFT. Here we provide an overview of the application of machine learning in electrocatalyst design, including the prediction of structure, thermodynamic properties and kinetic barriers. We also discuss the potential of explicit solvent model combined with machine learning molecular dynamics in this field. Finally, the favorable circumstances and challenges are outlined for the future development of machine learning in electrocatalysis. The studies on electrochemical processes by machine learning will further realize targeted design of high-efficiency electrocatalysts.

Evolution of Metastable Structures at Bimetallic Surfaces from Microscopy and Machine-Learning Molecular Dynamics

The restructuring of interfaces plays a crucial role in materials science and heterogeneous catalysis. Bimetallic systems, in particular, often adopt very different compositions and morphologies at surfaces compared to the bulk. For the first time, we reveal a detailed atomistic picture of long-time scale restructuring of Pd deposited on Ag using microscopy, spectroscopy, and novel simulation methods. By developing and performing accelerated machine-learning molecular dynamics followed by an automated analysis method, we discover and characterize previously unidentified surface restructuring mechanisms in an unbiased fashion, including Pd-Ag place exchange and Ag pop-out as well as step ascent and descent. Remarkably, layer-by-layer dissolution of Pd into Ag is always preceded by an encapsulation of Pd islands by Ag, resulting in a significant migration of Ag out of the surface and a formation of extensive vacancy pits within a period of microseconds. These metastable structures are of vital catalytic importance, as Ag-encapsulated Pd remains much more accessible to reactants than bulk-dissolved Pd. Our approach is broadly applicable to complex multimetallic systems and enables the previously intractable mechanistic investigation of restructuring dynamics at atomic resolution.

Efficient sampling of equilibrium states

Machine Learning Molecular dynamics or Monte Carlo methods can be used to sample equilibrium states, but these methods become computationally expensive for complex systems, where the transition from one equilibrium state to another may only occur through rare events. Noe et al. used neural networks and deep learning to generate distributions of independent soft condensed-matter samples at equilibrium (see the Perspective by Tuckerman). Supervised training is used to construct invertible transformations between the coordinates of the complex system of interest and simple Gaussian coordinates of the same dimensionality. Thus, configurations can be sampled in this simpler coordinate system and then transformed back into the complex one using the correct statistical weighting. Science , this issue p. [eaaw1147][1]; see also p. [982][2] [1]: /lookup/doi/10.1126/science.aaw1147 [2]: /lookup/doi/10.1126/science.aay2568