Proton-Transfer Mechanisms at the Water-ZnO Interface: The Role of Presolvation

Mechanism of Fe(II) Chemisorption on Hematite(001) Revealed by Reactive Neural Network Potential Molecular Dynamics

Atomic-scale understanding of important geochemical processes including sorption, dissolution, nucleation, and crystal growth is difficult to obtain from experimental measurements alone and would benefit from strong continuous progress in molecular simulation. To this end, we present a reactive neural network potential-based molecular dynamics approach to simulate the interaction of aqueous ions on mineral surfaces in contact with liquid water, taking Fe(II) on hematite(001) as a model system. We show that a single neural network potential predicts rate constants for water exchange for aqueous Fe(II) and for the exergonic chemisorption of aqueous Fe(II) on hematite(001) in good agreement with experimental observations. The neural network potential developed herein allows one to converge free energy profiles and transmission coefficients at density functional theory-level accuracy outperforming state-of-the-art classical force field potentials. This suggests that machine learning potential molecular dynamics should become the method of choice for atomistic studies of geochemical processes.

Data-Efficient Active Learning for Thermodynamic Integration: Acidity Constants of BiVO4 in Water

The protonation state of molecules and surfaces is pivotal in various disciplines, including (electro-)catalysis, geochemistry, biochemistry, and pharmaceutics. Accurately and efficiently determining acidity constants is critical yet challenging, particularly when explicitly considering the electronic structure, thermal fluctuations, anharmonic vibrations, and solvation effects. In this research, we employ thermodynamic integration accelerated by committee Neural Network potentials, training a single machine learning model that accurately describes the relevant protonated, deprotonated, and intermediate states. We investigate two deprotonation reactions at the BiVO4 (010)-water interface, a promising candidate for efficient photocatalytic water splitting. Our results illustrate the convergence of the required ensemble averages over simulation time and of the final acidity constant as a function of the Kirkwood coupling parameter. We demonstrate that simulation times on the order of nanoseconds are required for statistical convergence. This time scale is currently unachievable with explicit ab-initio molecular dynamics simulations at the hybrid DFT level of theory. In contrast, our machine learning workflow only requires a few hundred DFT single point calculations for training and testing. Exploiting the extended time scales accessible, we furthermore asses the effect of commonly applied bias potentials. Thus, our study significantly advances calculating free energy differences with ab-initio accuracy.

Water-Mediated Proton Hopping Mechanisms at the SnO2(110)/H2O Interface from Ab Initio Deep Potential Molecular Dynamics

The interfacial proton transfer (PT) reaction on the metal oxide surface is an important step in many chemical processes including photoelectrocatalytic water splitting, dehydrogenation, and hydrogen storage. The investigation of the PT process, in terms of thermodynamics and kinetics, has received considerable attention, but the individual free energy barriers and solvent effects for different PT pathways on rutile oxide are still lacking. Here, by applying a combination of ab initio and deep potential molecular dynamics methods, we have studied interfacial PT mechanisms by selecting the rutile SnO2(110)/H2O interface as an example of an oxide with the characteristic of frequently interfacial PT processes. Three types of PT pathways among the interfacial groups are found, i.e., proton transfer from terminal adsorbed water to bridge oxygen directly (surface-PT) or via a solvent water (mediated-PT), and proton hopping between two terminal groups (adlayer PT). Our simulations reveal that the terminal water in mediated-PT prefers to point toward the solution and forms a shorter H-bond with the assisted solvent water, leading to the lowest energy barrier and the fastest relative PT rate. In particular, it is found that the full solvation environment plays a crucial role in water-mediated proton conduction, while having little effect on direct PT reactions. The PT mechanisms on aqueous rutile oxide interfaces are also discussed by comparing an oxide series composed of SnO2, TiO2, and IrO2. Consequently, this work provides valuable insights into the ability of a deep neural network to reproduce the ab initio potential energy surface, as well as the PT mechanisms at such oxide/liquid interfaces, which can help understand the important chemical processes in electrochemistry, photoelectrocatalysis, colloid science, and geochemistry.

Hybrid Quantum Mechanical, Molecular Mechanical, and Machine Learning Potential for Computing Aqueous-Phase Adsorption Free Energies on Metal Surfaces

Performing reliable computer simulations of elementary processes occurring at metal-water interfaces is pivotal for novel catalyst design in sustainable energy applications. Computational catalyst design hinges on the ability to reliably and efficiently compute the potential energy surface (PES) of the system. Due to the large system sizes needed for studying processes at liquid water-metal interfaces, these systems can currently not be described using density functional theory (DFT). In this work, we used a hybrid quantum mechanical, molecular mechanical, and machine learning potential for studying the adsorption behavior of phenol, atomic hydrogen, 2-butanol, and 2-butanone on the (0001) facet of Ru under reducing conditions when Ru is not oxidized. Specifically, we describe the adsorbate and the surrounding metal atoms at the DFT level of theory. Here, we also considered the electrostatic field effect of the water molecules on adsorbate-metal interactions. Next, for the water-water and water-adsorbate interactions, we used established classical force fields. Finally, for the water-Ru surface interaction, for which no reliable force fields have been published, we used Behler-Parrinello high-dimensional neural network potentials (HDNNPs). Employing this setup, we used our explicit solvation for metal surface (eSMS) approach to compute the aqueous-phase effect on the low-coverage adsorption of selected molecules and atoms on the (0001) facet of Ru. In agreement with previous experimental and computational studies of oxygenated molecules over transition metal facets, we found that liquid water destabilizes the tested adsorbates on Ru(0001). Interestingly, our findings indicate that adsorbates on Ru are less affected by the presence of an aqueous phase than on other transition metals (e.g., Pt), highlighting the necessity of experimental investigations of Ru-based catalytic systems in liquid water.

Molecular insights into the water dissociation and proton dynamics at the β-TaON (100)/water interface

Understanding the dynamic nature of the semiconductor-water interface is crucial for developing efficient photoelectrochemical water splitting catalysts, as it governs reactivity through charge and mass transport. In this study, we employ ab initio molecular dynamics simulations to investigate the structural and dynamical properties of water at the β-TaON (100) surface. We observed that a well-defined interface is established through the spontaneous dissociation of water and the reorganization of surface chemical bonds. This leads to the formation of a partially hydroxylated surface, accompanied by a strong network of hydrogen bonds at the TaON-water interface. Consequently, various proton transport routes, including the proton transfer through "low-barrier hydrogen bond" path, become active across the interface, dramatically increasing the overall rate of the proton hopping at the interface. Based on our findings, we propose that the observed high photocatalytic activity of TaON-based semiconductors could be attributed to the spontaneous water dissociation and the resulting high proton transfer rate at the interface.

Long-range proton and hydroxide ion transfer dynamics at the water/CeO2 interface in the nanosecond regime: reactive molecular dynamics simulations and kinetic analysis

The structural properties, dynamical behaviors, and ion transport phenomena at the interface between water and cerium oxide are investigated by reactive molecular dynamics (MD) simulations employing neural network potentials (NNPs). The NNPs are trained to reproduce density functional theory (DFT) results, and DFT-based MD (DFT-MD) simulations with enhanced sampling techniques and refinement schemes are employed to efficiently and systematically acquire training data that include diverse hydrogen-bonding configurations caused by proton hopping events. The water interfaces with two low-index surfaces of (111) and (110) are explored with these NNPs, and the structure and long-range proton and hydroxide ion transfer dynamics are examined with unprecedented system sizes and long simulation times. Various types of proton hopping events at the interface are categorized and analyzed in detail. Furthermore, in order to decipher the proton and hydroxide ion transport phenomena along the surface, a counting analysis based on the semi-Markov process is formulated and applied to the MD trajectories to obtain reaction rates by considering the transport as stochastic jump processes. Through this model, the coupling between hopping events, vibrational motions, and hydrogen bond networks at the interface are quantitatively examined, and the high activity and ion transport phenomena at the water/CeO2 interface are unequivocally revealed in the nanosecond regime.

Electric Double Layer Effects in Electrocatalysis: Insights from Ab Initio Simulation and Hierarchical Continuum Modeling

Structures of the electric double layer (EDL) at electrocatalytic interfaces, which are modulated by the material properties, the electrolyte characteristics (e.g., the pH, the types and concentrations of ions), and the electrode potential, play crucial roles in the reaction kinetics. Understanding the EDL effects in electrocatalysis has attracted substantial research interest in recent years. However, the intrinsic relationships between the specific EDL structures and electrocatalytic kinetics remain poorly understood, especially on the atomic scale. In this Perspective, we briefly review the recent advances in deciphering the EDL effects mainly in hydrogen and oxygen electrocatalysis through a multiscale approach, spanning from the atomistic scale simulated by ab initio methods to the macroscale by a hierarchical approach. We highlight the importance of resolving the local reaction environment, especially the local hydrogen bond network, in understanding EDL effects. Finally, some of the remaining challenges are outlined, and an outlook for future developments in these exciting frontiers is provided.

Computational Study of H2O Adsorption, Hydrolysis, and Water Splitting on (ZnO)3 Nanoclusters Deposited on Graphene and Graphene Oxides

Graphene and graphene oxide (GO)-based metal oxides could play an important role in using metal oxide like zinc oxide (ZnO) as photocatalysts to split water. The π conjugation structure of GO shows greater electron mobility and could enhance the photocatalytic performance of the bare ZnO catalyst by increasing the electron-hole separation. In this work, we use density functional theory (DFT) with the B3LYP exchange functional and DGDZVP2 basis set to study the impact of adsorbing (ZnO)3 nanoparticles on graphene and four different GO models (GO1, GO2, GO4, and GO5) on the hydration and hydrolysis of water that precedes water splitting to produce H2 and O2 atoms in the gas phase and compare them with our previous studies on the bare catalyst in the absence of the substrate. The potential energy curves and activation energies are similar, but the triplet states are lower in energy than the singlet states in contrast to the bare (ZnO)3 catalyst. We extend our calculations to water splitting from the hydrolyzed (ZnO)3 on GO1 (GO1-(ZnO)3). The triplet state energy remains lower than the singlet state energy, and hydrogen production precedes the formation of oxygen, but there is no energy inter-crossing during the formation of O2 that occurs in the absence of a GO1 substrate. Although the hydrolysis reaction pathway follows similar steps in both the bare and GO1-(ZnO)3, water splitting with (ZnO)3 absorbed on the GO1 substrate skips two steps as it proceeds toward the production of the second H2. The production of two hydrogen molecules precedes oxygen formation during water splitting, and the first Zn-H bond formation step is the rate-determining step. The ZnO trimer deposited on GO systems could be potentially attractive nanocatalysts for water splitting.

Exploring Configurations of Nanocrystal Ligands Using Machine-Learned Force Fields

Semiconducting nanocrystals passivated with organic ligands have emerged as a powerful platform for light harvesting, light-driven chemical reactions, and sensing. Due to their complexity and size, little structural information is available from experiments, making these systems challenging to model computationally. Here, we develop a machine-learned force field trained on DFT data and use it to investigate the surface chemistry of a PbS nanocrystal interfaced with acetate ligands. In doing so, we go beyond considering individual local minimum energy geometries and, importantly, circumvent a precarious issue associated with the assumption of a single assigned atomic partial charge for each element in a nanocrystal, independent of its structural position. We demonstrate that the carboxylate ligands passivate the metal-rich surfaces by adopting a very wide range of "tilted-bridge" and "bridge" geometries and investigate the corresponding ligand IR spectrum. This work illustrates the potential of machine-learned force fields to transform computational modeling of these materials.

Neural network potentials for accelerated metadynamics of oxygen reduction kinetics at Au-water interfaces

The application of ab initio molecular dynamics (AIMD) for the explicit modeling of reactions at solid-liquid interfaces in electrochemical energy conversion systems like batteries and fuel cells can provide new understandings towards reaction mechanisms. However, its prohibitive computational cost severely restricts the time- and length-scales of AIMD. Equivariant graph neural network (GNN) based accurate surrogate potentials can accelerate the speed of performing molecular dynamics after learning on representative structures in a data efficient manner. In this study, we combined uncertainty-aware GNN potentials and enhanced sampling to investigate the reactive process of the oxygen reduction reaction (ORR) at an Au(100)-water interface. By using a well-established active learning framework based on CUR matrix decomposition, we can evenly sample equilibrium structures from MD simulations and non-equilibrium reaction intermediates that are rarely visited during the reaction. The trained GNNs have shown exceptional performance in terms of force prediction accuracy, the ability to reproduce structural properties, and low uncertainties when performing MD and metadynamics simulations. Furthermore, the collective variables employed in this work enabled the automatic search of reaction pathways and provide a detailed understanding towards the ORR reaction mechanism on Au(100). Our simulations identified the associative reaction mechanism without the presence of *O and a low reaction barrier of 0.3 eV, which is in agreement with experimental findings. The methodology employed in this study can pave the way for modeling complex chemical reactions at electrochemical interfaces with an explicit solvent under ambient conditions.

Machine Learning Assisted Simulations of Electrochemical Interfaces: Recent Progress and Challenges

The electrochemical interface, where the adsorption of reactants and electrocatalytic reactions take place, has long been a focus of attention. Some of the important processes on it tend to possess relatively slow kinetic characteristics, which are usually beyond the scope of ab initio molecular dynamics. The newly emerging technique, machine learning methods, provides an alternative approach to achieve thousands of atoms and nanosecond time scale while ensuring precision and efficiency. In this Perspective, we summarize in detail the recent progress and achievements made by the introduction of machine learning to simulate electrochemical interfaces, and focus on the limitations of current machine learning models, such as accurate descriptions of long-range electrostatic interactions and the kinetics of the electrochemical reactions occurring at the interface. Finally, we further point out the future directions for machine learning to expand in the field of electrochemical interfaces.

A first-principles exploration of the conformational space of sodiated pyranose assisted by neural network potentials

Sampling the conformational space of monosaccharides using the first-principles methods is important and as a database of local minima provides a solid base for interpreting experimental measurements such as infrared photo-dissociation (IRPD) spectroscopy or collision-induced dissociation (CID). IRPD emphasizes low-energy conformers and CID can distinguish conformers with distinct reaction pathways. A typical computational approach is to engage empirical or semi-empirical methods to sample the conformational space first, and only selected minima are reoptimized at first-principles levels. In this work, we propose a computational scheme to explore the configurational space of 12 types of sodiated pyranoses with the assistance of a neural network potential (NNP). We demonstrated that it is possible to train an NNP based on the density functional calculations extracted from a previous study on sodiated glucose (Glc), galactose (Gal), and mannose (Man). This NNP yields a better description of the other five types of aldohexoses than the four types of ketohexoses. We further show that such a discrepancy in the accuracy of NNP can be resolved by an active learning scheme where the NNP model is engaged in generating the data and has itself updated. Through this iterative process, we can locate more than 17 000 distinct local minima at the B3LYP/6-311+G(d,p) level and an NNP with an accuracy of 1 kJ mol^-1 was created, which can be used for further studies.

Predicting Frequency from the External Chemical Environment: OH Vibrations on Hydrated and Hydroxylated Surfaces

Robust correlation curves are essential to decipher structural information from IR-vibrational spectra. However, for surface-adsorbed water and hydroxides, few such correlations have been presented in the literature. In this paper, OH vibrational frequencies are correlated against 12 structural descriptors representing the quantum mechanical or geometrical environment, focusing on those external to the vibrating molecule. A nonbiased fitting procedure based on Gaussian process regression (GPR) was used alongside simple analytical functional forms. The training data consist of 217 structurally unique OH groups from 38 water/metal oxide interface systems for MgO, CaO and CeO₂, all optimized at the DFT level, and the fully anharmonic and uncoupled OH vibrational signatures were calculated. Among our results, we find the following: (i) The intermolecular R(H···O) hydrogen bond distance is particularly strong, indicating the primary cause of the frequency shift. (ii) Similarly, the electric field along the H-bond vector is also a good descriptor. (iii) Highly detailed machine learning descriptors (ACSF, SOAP) are less intuitive but were found to be more capable descriptors. (iv) Combinations of geometric and QM descriptors give the best predictions, supplying complementary information.

Crystal Nucleation and Growth of Inorganic Ionic Materials from Aqueous Solution: Selected Recent Developments, and Implications

In this review article, selected, latest theoretical, and experimental developments in the field of nucleation and crystal growth of inorganic materials from aqueous solution are highlighted, with a focus on literature after 2015 and on non-classical pathways. A key point is to emphasize the so far underappreciated role of water and solvent entropy in crystallization at all stages from solution speciation through to the final crystal. While drawing on examples from current inorganic materials where non-classical behavior has been proposed, the potential of these approaches to be adapted to a wide-range of systems is also discussed, while considering the broader implications of the current re-assessment of pathways for crystallization. Various techniques that are suitable for the exploration of crystallization pathways in aqueous solution, from nucleation to crystal growth are summarized, and a flow chart for the assignment of specific theories based on experimental observations is proposed.

Nanosecond solvation dynamics of the hematite/liquid water interface at hybrid DFT accuracy using committee neural network potentials

Metal oxide/water interfaces play an important role in biology, catalysis, energy storage and photocatalytic water splitting. The atomistic structure at these interfaces is often difficult to characterize by experimental techniques, whilst results from ab initio molecular dynamics simulations tend to be uncertain due to the limited length and time scales accessible. In this work, we train a committee neural network potential to simulate the hematite/water interface at the hybrid DFT level of theory to reach the nanosecond timescale and systems containing more than 3000 atoms. The NNP enables us to converge dynamical properties, not possible with brute-force ab initio molecular dynamics. Our simulations uncover a rich solvation dynamics at the hematite/water interface spanning three different time scales: picosecond H-bond dynamics between surface hydroxyls and the first water layer, in-plane/out-of-plane tilt motion of surface hydroxyls on the 10 ps time scale, and diffusion of water molecules from the oxide surface characterized by a mean residence lifetime of about 60 ps. Calculation of vibrational spectra confirm that H-bonds between surface hydroxyls and first layer water molecules are stronger than H-bonds in bulk water. Our study showcases how state of the art machine learning approaches can routinely be utilized to explore the structural dynamics at transition metal oxide interfaces with complex electronic structure. It foreshadows that c-NNPs are a promising tool to tackle the sampling problem in ab initio electrochemistry with explicit solvent molecules.

Accelerated Deep Learning Dynamics for Atomic Layer Deposition of Al(Me)3 and Water on OH/Si(111)

Knowledge of the detailed mechanism behind the atomic layer deposition (ALD) can greatly facilitate the optimization of the manufacturing process. Computational modeling can potentially foster the understanding; however, the presently available capabilities of the accurate ab initio computational techniques preclude their application to modeling surface processes occurring on a long time scale, such as ALD. Although the situation can be greatly improved using machine learning (ML), this technique requires an enormous amount of data for training datasets. Here, we propose an iterative protocol for optimizing ML training datasets and apply ML-assisted ab initio calculations to model surface reactions occurring during the Al(Me)₃/H₂O ALD process on the OH-terminated Si (111) surface. The protocol uses a recently developed low-dimensional projection technique (TDUS), greatly reducing the amount of information required to achieve high accuracy (ca. 1 kcal/mol or less) of the developed ML models. The resulting free energy landscapes reveal fine details of various aspects of the target ALD process, such as the surface proton transfer, zwitterionic surface configurations, elimination-addition/addition-elimination, and S_N2 reactions as well as the role of the surface entropic and temperature effects. Simulations of adsorption dynamics predict that the maximum physisorption rate of ca. 70% is achieved at the incidence velocity u_rms of the reactants in the range of 15-20 Å/ps. Hence, the proposed protocol furnishes a very effective tool to study complex chemical reaction dynamics at a much reduced computational cost.

Applying Classical, Ab Initio, and Machine-Learning Molecular Dynamics Simulations to the Liquid Electrolyte for Rechargeable Batteries

Rechargeable batteries have become indispensable implements in our daily life and are considered a promising technology to construct sustainable energy systems in the future. The liquid electrolyte is one of the most important parts of a battery and is extremely critical in stabilizing the electrode-electrolyte interfaces and constructing safe and long-life-span batteries. Tremendous efforts have been devoted to developing new electrolyte solvents, salts, additives, and recipes, where molecular dynamics (MD) simulations play an increasingly important role in exploring electrolyte structures, physicochemical properties such as ionic conductivity, and interfacial reaction mechanisms. This review affords an overview of applying MD simulations in the study of liquid electrolytes for rechargeable batteries. First, the fundamentals and recent theoretical progress in three-class MD simulations are summarized, including classical, ab initio, and machine-learning MD simulations (section 2). Next, the application of MD simulations to the exploration of liquid electrolytes, including probing bulk and interfacial structures (section 3), deriving macroscopic properties such as ionic conductivity and dielectric constant of electrolytes (section 4), and revealing the electrode-electrolyte interfacial reaction mechanisms (section 5), are sequentially presented. Finally, a general conclusion and an insightful perspective on current challenges and future directions in applying MD simulations to liquid electrolytes are provided. Machine-learning technologies are highlighted to figure out these challenging issues facing MD simulations and electrolyte research and promote the rational design of advanced electrolytes for next-generation rechargeable batteries.

Unveiling the Atomic Structure and Growth Dynamics of One-Dimensional Water on ZnO(10-10)

The adsorption and organization state of water on the metal oxide surface is of critical importance for wide fields where interface chemistry dominates. On the technically important ZnO(10-10) surface, we found water assembles into an one-dimensional (1D) chain structure at submonolayer coverage instead of the well-known half-dissociated two-dimensional (2D) island. With a combination of high resolution scanning tunneling microscopy (STM) and density functional theory (DFT) calculations, we clearly distinguished the single and double water chains, which are composed of dissociated monomers and half-dissociated dimers, respectively. Moreover, we unambiguously determined that single water molecules dissociate spontaneously before agglomerating into ordered phase, which is contrary to the proposition of previous studies. These results have deepened our understandings of the adsorbed water species on the ZnO surface, which may bring new insights into the mechanisms of water-stimulated surface reactions.

Insights into lithium manganese oxide-water interfaces using machine learning potentials

Unraveling the atomistic and the electronic structure of solid-liquid interfaces is the key to the design of new materials for many important applications, from heterogeneous catalysis to battery technology. Density functional theory (DFT) calculations can, in principle, provide a reliable description of such interfaces, but the high computational costs severely restrict the accessible time and length scales. Here, we report machine learning-driven simulations of various interfaces between water and lithium manganese oxide (Li_xMn₂O₄), an important electrode material in lithium ion batteries and a catalyst for the oxygen evolution reaction. We employ a high-dimensional neural network potential to compute the energies and forces several orders of magnitude faster than DFT without loss in accuracy. In addition, a high-dimensional neural network for spin prediction is utilized to analyze the electronic structure of the manganese ions. Combining these methods, a series of interfaces is investigated by large-scale molecular dynamics. The simulations allow us to gain insights into a variety of properties, such as the dissociation of water molecules, proton transfer processes, and hydrogen bonds, as well as the geometric and electronic structure of the solid surfaces, including the manganese oxidation state distribution, Jahn-Teller distortions, and electron hopping.

Dynamics of Heterogeneous Catalytic Processes at Operando Conditions

The rational design of high-performance catalysts is hindered by the lack of knowledge of the structures of active sites and the reaction pathways under reaction conditions, which can be ideally addressed by an in situ/operando characterization. Besides the experimental insights, a theoretical investigation that simulates reaction conditions-so-called operando modeling-is necessary for a plausible understanding of a working catalyst system at the atomic scale. However, there is still a huge gap between the current widely used computational model and the concept of operando modeling, which should be achieved through multiscale computational modeling. This Perspective describes various modeling approaches and machine learning techniques that step toward operando modeling, followed by selected experimental examples that present an operando understanding in the thermo- and electrocatalytic processes. At last, the remaining challenges in this area are outlined.

Implicit Solvation Methods for Catalysis at Electrified Interfaces

Implicit solvation is an effective, highly coarse-grained approach in atomic-scale simulations to account for a surrounding liquid electrolyte on the level of a continuous polarizable medium. Originating in molecular chemistry with finite solutes, implicit solvation techniques are now increasingly used in the context of first-principles modeling of electrochemistry and electrocatalysis at extended (often metallic) electrodes. The prevalent ansatz to model the latter electrodes and the reactive surface chemistry at them through slabs in periodic boundary condition supercells brings its specific challenges. Foremost this concerns the difficulty of describing the entire double layer forming at the electrified solid–liquid interface (SLI) within supercell sizes tractable by commonly employed density functional theory (DFT). We review liquid solvation methodology from this specific application angle, highlighting in particular its use in the widespread ab initio thermodynamics approach to surface catalysis. Notably, implicit solvation can be employed to mimic a polarization of the electrode’s electronic density under the applied potential and the concomitant capacitive charging of the entire double layer beyond the limitations of the employed DFT supercell. Most critical for continuing advances of this effective methodology for the SLI context is the lack of pertinent (experimental or high-level theoretical) reference data needed for parametrization.

Training algorithm matters for the performance of neural network potential: A case study of Adam and the Kalman filter optimizers

One hidden yet important issue for developing neural network potentials (NNPs) is the choice of training algorithm. In this article, we compare the performance of two popular training algorithms, the adaptive moment estimation algorithm (Adam) and the extended Kalman filter algorithm (EKF), using the Behler-Parrinello neural network and two publicly accessible datasets of liquid water [Morawietz et al., Proc. Natl. Acad. Sci. U. S. A. 113, 8368-8373, (2016) and Cheng et al., Proc. Natl. Acad. Sci. U. S. A. 116, 1110-1115, (2019)]. This is achieved by implementing EKF in TensorFlow. It is found that NNPs trained with EKF are more transferable and less sensitive to the value of the learning rate, as compared to Adam. In both cases, error metrics of the validation set do not always serve as a good indicator for the actual performance of NNPs. Instead, we show that their performance correlates well with a Fisher information based similarity measure.

Hydrogen Coupling on Platinum Using Artificial Neural Network Potentials and DFT

To date, the understanding of reactions at solid-liquid interfaces has proven challenging, mainly because of the inaccessible nature of such systems to current experimental techniques with atomic resolution. This has meant that many important features, including free energy barriers and the atomistic structure of intermediates, remain unknown. To tackle these issues, we construct and utilize a high-dimensional neural network (HDNN) potential for the simulation of hydrogen evolution at the HCl(aq)/Pt(111) interface, taking into consideration the influence of adsorbate-adsorbate, adsorbate-solvent interactions, and ion solvation explicitly. Long time scale MD simulations reveal coadsorbed H_ad/H₂O_ad on the surface. The free energy profiles for the Tafel and Heyrovsky type hydrogen coupling are extracted using umbrella sampling. It is found that the preferential mechanism can change depending on the surface coverage, highlighting the dual mechanistic nature for HER on Pt(111). Our work demonstrates the importance of controlling the solvent-substrate interactions in developing catalysts beyond Pt.

Recent Progress toward Ab Initio Modeling of Electrocatalysis

Electrode potential is the key factor for controlling electrocatalytic reactions at electrochemical interfaces, and moreover, it is also known that the pH and solutes (e.g., cations) of the solution have prominent effects on electrocatalysis. Understanding these effects requires microscopic information on the electrochemical interfaces, in which theoretical simulations can play an important role. This Perspective summarizes the recent progress in method development for modeling electrochemical interfaces, including different methods for describing the electrolytes at the interfaces and different schemes for charging up the electrode surfaces. In the final section, we provide an outlook for future development in modeling methods and their applications to electrocatalysis.

Photo-induced hydrophilicity at the ZnO(112̄0) surface: an evolutionary algorithm-aided density functional theory study

Evolutionary algorithm-aided density functional theory calculations were utilized to determine the stable adsorption structures of H₂O at ZnO(112̄0) extensively under different coverages. By decomposing the adsorption energetics, we illustrate that H₂O dissociation to a large extent is actually hampered by the barrier for induced distortion of the ZnO surface, and once the surface becomes less difficult to be distorted it will exhibit higher hydrophilicity or even superhydrophilicity. Specifically, photo-stimulation modelling suggests that the surface Zn-O bonds can be weakened by photo-excitation, and the layer of fully dissociated H₂O can be then facilitated to form. Accordingly, a novel mechanism for photo-induced superhydrophilicity is proposed.

Identification of a Grotthuss proton hopping mechanism at protonated polyhedral oligomeric silsesquioxane (POSS) - water interface

The attachment and dissociation of a proton from a water molecule and the proton transfers at solid-liquid interfaces play vital roles in numerous biological, chemical processes and for the development of sustainable functional materials for energy harvesting and conversion applications. Using first-principles computational methodologies, we investigated the protonated forms of polyhedral oligomeric silsesquioxane (POSS-H⁺) interacting with water clusters (W_n, where n = 1-6) as a model to quantify the proton conducting and localization ability at solid-liquid interfaces. Successive addition of explicit water molecules to POSS-H⁺ shows that the assistance of at least three water molecules is required to dissociate the proton from POSS with the formation of an Eigen cation (H₉O₄⁺), whereas the presence of a fourth water molecule highly favors the formation of a Zundel ion (H₅O₂⁺). Reaction pathway and energy barrier analysis reveal that the formation of the Eigen cation requires significantly higher energy than the Zundel features. This confirms that the Zundel ion is destabilized and promptly converts in to Eigen ion at this interface. Moreover, we identified a Grotthuss-type mechanism for the proton transfer through a water chain close to the interface, where symmetrical and unsymmetrical arrangements of water molecules around H⁺ of protonated POSS-H⁺ are involved in the conduction of proton through water wires where successive Eigen-to-Zundel and Zundel-to-Eigen transformations are observed in quick succession.

Photoinduced Desorption Dynamics of CO from Pd(111): A Neural Network Approach

Modeling the ultrafast photoinduced dynamics and reactivity of adsorbates on metals requires including the effect of the laser-excited electrons and, in many cases, also the effect of the highly excited surface lattice. Although the recent ab initio molecular dynamics with electronic friction and thermostats, (T_e,T_l)-AIMDEF [AlducinM.;Phys. Rev. Lett.2019, 123, 246802]31922860, enables such complex modeling, its computational cost may limit its applicability. Here, we use the new embedded atom neural network (EANN) method [ZhangY.;J. Phys. Chem. Lett.2019, 10, 496231397157] to develop an accurate and extremely complex potential energy surface (PES) that allows us a detailed and reliable description of the photoinduced desorption of CO from the Pd(111) surface with a coverage of 0.75 monolayer. Molecular dynamics simulations performed on this EANN-PES reproduce the (T_e,T_l)-AIMDEF results with a remarkable level of accuracy. This demonstrates the outstanding performance of the obtained EANN-PES that is able to reproduce available density functional theory (DFT) data for an extensive range of surface temperatures (90–1000 K); a large number of degrees of freedom, those corresponding to six CO adsorbates and 24 moving surface atoms; and the varying CO coverage caused by the abundant desorption events.

Machine-learned potentials for next-generation matter simulations

The choice of simulation methods in computational materials science is driven by a fundamental trade-off: bridging large time- and length-scales with highly accurate simulations at an affordable computational cost. Venturing the investigation of complex phenomena on large scales requires fast yet accurate computational methods. We review the emerging field of machine-learned potentials, which promises to reach the accuracy of quantum mechanical computations at a substantially reduced computational cost. This Review will summarize the basic principles of the underlying machine learning methods, the data acquisition process and active learning procedures. We highlight multiple recent applications of machine-learned potentials in various fields, ranging from organic chemistry and biomolecules to inorganic crystal structure predictions and surface science. We furthermore discuss the developments required to promote a broader use of ML potentials, and the possibility of using them to help solve open questions in materials science and facilitate fully computational materials design.

Machine learning of free energies in chemical compound space using ensemble representations: Reaching experimental uncertainty for solvation

Free energies govern the behavior of soft and liquid matter, and improving their predictions could have a large impact on the development of drugs, electrolytes, or homogeneous catalysts. Unfortunately, it is challenging to devise an accurate description of effects governing solvation such as hydrogen-bonding, van der Waals interactions, or conformational sampling. We present a Free energy Machine Learning (FML) model applicable throughout chemical compound space and based on a representation that employs Boltzmann averages to account for an approximated sampling of configurational space. Using the FreeSolv database, FML's out-of-sample prediction errors of experimental hydration free energies decay systematically with training set size, and experimental uncertainty (0.6 kcal/mol) is reached after training on 490 molecules (80% of FreeSolv). Corresponding FML model errors are on par with state-of-the art physics based approaches. To generate the input representation for a new query compound, FML requires approximate and short molecular dynamics runs. We showcase its usefulness through analysis of solvation free energies for 116k organic molecules (all force-field compatible molecules in the QM9 database), identifying the most and least solvated systems and rediscovering quasi-linear structure-property relationships in terms of simple descriptors such as hydrogen-bond donors, number of NH or OH groups, number of oxygen atoms in hydrocarbons, and number of heavy atoms. FML's accuracy is maximal when the temperature used for the molecular dynamics simulation to generate averaged input representation samples in training is the same as for the query compounds. The sampling time for the representation converges rapidly with respect to the prediction error.

Comparative Study of Cold Sintering Process and Autoclave Thermo-Vapor Treatment on a ZnO Sample

Analysis of scanning electron microscopy images was used to study the changes in the crystal size distribution of ZnO, which occurred during its processing in an aqueous medium at 220–255 °C and an equilibrium vapor pressure in an autoclave. The results were compared with those of ZnO placed in a die for treatment under similar conditions supplemented with mechanical pressure application in the cold sintering process. In both cases, ZnO was treated in the presence of an activating additive: either zinc acetate or ammonium chloride. During autoclaving, a powder consisting of fine ZnO monocrystals was obtained, while the cold sintering process led to ceramics formation. Under vapor pressure and mechanical pressure, the aqueous medium affected ZnO transformation by the same mechanism of solid-phase mobility activation due to the additives’ influence. The higher the content of additives in the medium, and the higher the mechanical pressure, the more pronounced activating effect was observed. Mass transfer during the cold sintering process occurred mainly by the coalescence of crystals, while without mechanical pressure, the predominance of surface spreading was revealed. In the initial ZnO powder, the average crystal size was 0.193 μm. It grew up to 0.316–0.386 μm in a fine-crystalline powder formed in the autoclave and to an average grain size of 0.244–0.799 μm in the ceramics, which relative density reached 0.82–0.96. A scheme explaining the influence of an aqueous medium on the solid-phase mobility of ZnO structure was proposed. It was found that the addition of 7.6 mol% ammonium chloride to the reaction medium causes the processes of compaction and grain growth similar to those observed in ZnO Cold Sintering Process with the addition of 0.925 mol% zinc acetate.

Measuring transferability issues in machine-learning force fields: the example of gold–iron interactions with linearized potentials

Machine-learning force fields have been increasingly employed in order to extend the possibility of current first-principles calculations. However, the transferability of the obtained potential cannot always be guaranteed in situations that are outside the original database. To study such limitation, we examined the very difficult case of the interactions in gold–iron nanoparticles. For the machine-learning potential, we employed a linearized formulation that is parameterized using a penalizing regression scheme which allows us to control the complexity of the obtained potential. We showed that while having a more complex potential allows for a better agreement with the training database, it can also lead to overfitting issues and a lower accuracy in untrained systems.

Insights into Water Permeation through hBN Nanocapillaries by Ab Initio Machine Learning Molecular Dynamics Simulations

Water permeation between stacked layers of hBN sheets forming 2D nanochannels is investigated using large-scale ab initio-quality molecular dynamics simulations. A high-dimensional neural network potential trained on density-functional theory calculations is employed. We simulate water in van der Waals nanocapillaries and study the impact of nanometric confinement on the structure and dynamics of water using both equilibrium and nonequilibrium methods. At an interlayer distance of 10.2 Å confinement induces a first-order phase transition resulting in a well-defined AA-stacked bilayer of hexagonal ice. In contrast, for h < 9 Å,  the 2D water monolayer consists of a mixture of different locally ordered patterns of squares, pentagons, and hexagons. We found a significant change in the transport properties of confined water, particularly for monolayer water where the water-solid friction coefficient decreases to half and the diffusion coefficient increases by a factor of 4 as compared to bulk water. Accordingly, the slip-velocity is found to increase under confinement and we found that the overall permeation is dominated by monolayer water adjacent to the hBN membranes at extreme confinements. We conclude that monolayer water in addition to bilayer ice has a major contribution to water transport through 2D nanochannels.

Computing Surface Acidity Constants of Proton Hopping Groups from Density Functional Theory-Based Molecular Dynamics: Application to the SnO2(110)/H2O Interface

Proton transfer at metal oxide/water interfaces plays an important role in electrochemistry, geochemistry, and environmental science. The key thermodynamic quantity to characterize this process is the surface acidity constant. An ab initio method that combines density functional theory-based molecular dynamics (DFTMD) and free energy perturbation theory has been established for computing surface acidity constants. However, it involves a reversible proton insertion procedure in which frequent proton hopping, e.g., for strong bases and some oxide surfaces (e.g., SnO₂), can cause instability issues in electronic structure calculation. In the original implementation, harmonic restraining potentials are imposed on all O-H bonds (denoted by the V_rH scheme) to prevent proton hopping and thus may not be applicable for systems involving spontaneous proton hopping. In this work, we introduce an improved restraining scheme with a repulsive potential V_rep to compute the surface acidities of systems in which proton hopping is spontaneous and fast. In this V_rep scheme, a Buckingham-type repulsive potential V_rep is applied between the deprotonation site and all other protons in DFTMD simulations. We first verify the V_rep scheme by calculating the pK_a values of H₂O and aqueous HS^- solution (i.e., strong conjugate bases) and then apply it to the SnO₂(110)/H₂O interface. It is found that the V_rep scheme leads to a prediction of the point of zero charge (PZC) of 4.6, which agrees well with experiment. The intrinsic individual pK_a values of the terminal five-coordinated Sn site (Sn_5cOH₂) and bridge oxygen site (Sn₂O_brH⁺) are 4.4 and 4.7, respectively, both being almost the same as the PZC. The similarity of the two pK_a values indicates that dissociation of terminal water has almost zero free energy at this proton hopping interface (i.e., partial water dissociation), as expected from the acid-base equilibrium on SnO₂.

Efficient Atomic-Resolution Uncertainty Estimation for Neural Network Potentials Using a Replica Ensemble

Neural network potentials (NNPs) are gaining much attention as they enable fast molecular dynamics (MD) simulations for a wide range of systems while maintaining the accuracy of density functional theory calculations. Since NNP is constructed by machine learning on training data, its prediction uncertainty increases drastically as atomic environments deviate from training points. Therefore, it is essential to monitor the uncertainty level during MD simulations to judge the soundness of the results. In this work, we propose an uncertainty estimator based on the replica ensemble in which NNPs are trained over atomic energies of a reference NNP that drives MD simulations. The replica ensemble is trained quickly, and its standard deviation provides atomic-resolution uncertainties. We apply this method to a highly reactive silicidation process of Si(001) overlaid with Ni thin films and confirm that the replica ensemble can spatially and temporally trace simulation errors at atomic resolution, which in turn guides the augmentation of the training set. The refined NNP completes a 3.6 ns simulation without any noticeable problems. By suggesting an efficient and atomic-resolution uncertainty indicator, this work will contribute to achieving reliable MD simulations by NNPs.

Dynamical matrix propagator scheme for large-scale proton dynamics simulations

We derive a matrix formalism for the simulation of long range proton dynamics for extended systems and timescales. On the basis of an ab initio molecular dynamics simulation, we construct a Markov chain, which allows us to store the entire proton dynamics in an M × M transition matrix (where M is the number of oxygen atoms). In this article, we start from common topology features of the hydrogen bond network of good proton conductors and utilize them as constituent constraints of our dynamic model. We present a thorough mathematical derivation of our approach and verify its uniqueness and correct asymptotic behavior. We propagate the proton distribution by means of transition matrices, which contain kinetic data from both ultra-short (sub-ps) and intermediate (ps) timescales. This concept allows us to keep the most relevant features from the microscopic level while effectively reaching larger time and length scales. We demonstrate the applicability of the transition matrices for the description of proton conduction trends in proton exchange membrane materials.