Comparing machine learning potentials for water: Kernel-based regression and Behler-Parrinello neural networks.

In this paper, we investigate the performance of different machine learning potentials (MLPs) in predicting key thermodynamic properties of water using RPBE + D3. Specifically, we scrutinize kernel-based regression and high-dimensional neural networks trained on a highly accurate dataset consisting of about 1500 structures, as well as a smaller dataset, about half the size, obtained using only on-the-fly learning. This study reveals that despite minor differences between the MLPs, their agreement on observables such as the diffusion constant and pair-correlation functions is excellent, especially for the large training dataset. Variations in the predicted density isobars, albeit somewhat larger, are also acceptable, particularly given the errors inherent to approximate density functional theory. Overall, this study emphasizes the relevance of the database over the fitting method. Finally, this study underscores the limitations of root mean square errors and the need for comprehensive testing, advocating the use of multiple MLPs for enhanced certainty, particularly when simulating complex thermodynamic properties that may not be fully captured by simpler tests.

Mechanism of Ionomer Film Formation via Solution Drying.

Film formation via the drying of ionomer solutions is a crucial process that has a strong influence on the morphology and transport properties of polymer electrolyte membranes and thin films. However, the microscopic mechanism of this process remains unclear. Here, we elucidate this mechanism using a coarse-grained model based on all-atom molecular dynamics that accurately reproduces small-angle X-ray scattering spectra. In dilute ionomer solutions, ionomers form rod-like bundles with diameters of 1.5-2 nm. As the water solvent evaporates, these bundles gradually aggregate and connect to each other, while maintaining their diameter. Finally, the remaining water forms nanosized clusters surrounded by the surfaces of the bundles with hydrophilic sulfonate groups.

Proton Transport in Perfluorinated Ionomer Simulated by Machine-Learned Interatomic Potential.

Polymers are a class of materials that are highly challenging to deal with using first-principles methods. Here, we present an application of machine-learned interatomic potentials to predict structural and dynamical properties of dry and hydrated perfluorinated ionomers. An improved active-learning algorithm using a small number of descriptors allows to efficiently construct an accurate and transferable model for this multielemental amorphous polymer. Molecular dynamics simulations accelerated by the machine-learned potentials accurately reproduce the heterogeneous hydrophilic and hydrophobic domains formed in this material as well as proton and water diffusion coefficients under a variety of humidity conditions. Our results reveal pronounced contributions of Grotthuss chains consisting of two to three water molecules to the high proton mobility under strongly humidified conditions.

Discovering surface reaction pathways using accelerated molecular dynamics and network analysis tools.

We present an automated method that maps surface reaction pathways with no experimental data and with minimal human interventions. In this method, bias potentials promoting surface reactions are applied to enable statistical samplings of the surface reaction within the timescale of ab initio molecular dynamics (MD) simulations, and elementary reactions are extracted automatically using an extension of the method constructed for gas- or liquid-phase reactions. By converting the extracted elementary reaction data to directed graph data, MD trajectories can be efficiently mapped onto reaction pathways using a network analysis tool. To demonstrate the power of the method, it was applied to the steam reforming of methane on the Rh(111) surface and to propane reforming on the Pt(111) and Pt3Sn(111) surfaces. We discover new energetically favorable pathways for both reactions and reproduce the experimentally-observed materials-dependence of the surface reaction activity and the selectivity for the propane reforming reactions.

Mechanisms of temperature-dependent oxygen absorption/release and appearance of intermediate phase in κ-Ce2Zr2O8: study based on oxygen vacancy formation energy computations.

This study clarified the mechanisms of the temperature-dependent oxygen absorption/release properties and appearance of the intermediate phase for κ-Ce2Zr2O8, which is known to have a high oxygen storage/release capacity (OSC). First-principle computations revealed that the vacancy formation energies depend on the number of vacancies and can be categorized into two groups: low-energy and high-energy. The intermediate phase observed experimentally was assigned to the state after all the oxygen vacancies in the low-energy group were formed. We also found that the mechanism of the improved OSC performance by Ti substitution could be explained in terms of the vacancy formation energies.

Molecular dynamics simulations of proton conducting media containing phosphoric acid.

An anhydrous proton conducting electrolyte is a key material in realizing medium-temperature fuel cells that can drastically simplify heat radiation systems in transportation applications. However, practical applications are limited by the low proton conductivity. To clarify the rate limiting process, molecular dynamics simulations using machine-learned inter-atomic potentials were conducted on three materials: liquid phosphoric acid, solid acid CsH2PO4, and coordination polymer [Zn(HPO4)(H2PO4)2](ImH2)2. The simulations showed proton hopping between phosphoric acid anions in the 100-300 fs time scale in all the three materials. However, the calculated diffusion coefficient of protons in three materials spans over orders of magnitude as observed experimentally. The rotational rates of the anions showed a remarkable difference; in the proton conducting liquid phosphoric acid and cubic CsH2PO4, the anions rotate at the 100 ps time scale whereas they rarely rotate in other media. This result clearly indicates that proton conduction is limited by the reorientations of the anions.

The role of oxygen-permeable ionomer for polymer electrolyte fuel cells.

In recent years, considerable research and development efforts are devoted to improving the performance of polymer electrolyte fuel cells. However, the power density and catalytic activities of these energy conversion devices are still far from being satisfactory for large-scale operation. Here we report performance enhancement via incorporation, in the cathode catalyst layers, of a ring-structured backbone matrix into ionomers. Electrochemical characterizations of single cells and microelectrodes reveal that high power density is obtained using an ionomer with high oxygen solubility. The high solubility allows oxygen to permeate the ionomer/catalyst interface and react with protons and electrons on the catalyst surfaces. Furthermore, characterizations of single cells and single-crystal surfaces reveal that the oxygen reduction reaction activity is enhanced owing to the mitigation of catalyst poisoning by sulfonate anion groups. Molecular dynamics simulations indicate that both the high permeation and poisoning mitigation are due to the suppression of densely layered folding of polymer backbones near the catalyst surfaces by the incorporated ring-structured matrix. These experimental and theoretical observations demonstrate that ionomer's tailored molecular design promotes local oxygen transport and catalytic reactions.

First-principles hydration free energies of oxygenated species at water-platinum interfaces.

The hydration free energy of atoms and molecules adsorbed at liquid-solid interfaces strongly influences the stability and reactivity of solid surfaces. However, its evaluation is challenging in both experiments and theories. In this work, a machine learning aided molecular dynamics method is proposed and applied to oxygen atoms and hydroxyl groups adsorbed on Pt(111) and Pt(100) surfaces in water. The proposed method adopts thermodynamic integration with respect to a coupling parameter specifying a path from well-defined non-interacting species to the fully interacting ones. The atomistic interactions are described by a machine-learned inter-atomic potential trained on first-principles data. The free energy calculated by the machine-learned potential is further corrected by using thermodynamic perturbation theory to provide the first-principles free energy. The calculated hydration free energies indicate that only the hydroxyl group adsorbed on the Pt(111) surface attains a hydration stabilization. The observed trend is attributed to differences in the adsorption site and surface morphology.

Challenges in applying highly active Pt-based nanostructured catalysts for oxygen reduction reactions to fuel cell vehicles.

The past 30 years have seen progress in the development of Pt-based nanocatalysts for the oxygen reduction reaction, and some are now in production on a commercial basis and used for polymer electrolyte fuel cells (PEFCs) for automotives and other applications. Further improvements in catalytic activity are required for wider uptake of PEFCs, however. In laboratories, researchers have developed various catalysts that have much higher activities than commercial ones, and these state-of-the-art catalysts have potential to improve energy conversion efficiencies and reduce the usage of platinum in PEFCs. There are several technical issues that must be solved before they can be applied in fuel cell vehicles, which require a high power density and practical durability, as well as high efficiency. In this Review, the development history of Pt-based nanocatalysts and recent analytical studies are summarized to identify the origin of these technical issues. Promising strategies for overcoming those issues are also discussed.

On-the-Fly Active Learning of Interatomic Potentials for Large-Scale Atomistic Simulations.

The on-the-fly generation of machine-learning force fields by active-learning schemes attracts a great deal of attention in the community of atomistic simulations. The algorithms allow the machine to self-learn an interatomic potential and construct machine-learned models on the fly during simulations. State-of-the-art query strategies allow the machine to judge whether new structures are out of the training data set or not. Only when the machine judges the necessity of updating the data set with the new structures are first-principles calculations carried out. Otherwise, the yet available machine-learned model is used to update the atomic positions. In this manner, most of the first-principles calculations are bypassed during training, and overall, simulations are accelerated by several orders of magnitude while retaining almost first-principles accuracy. In this Perspective, after describing essential components of the active-learning algorithms, we demonstrate the power of the schemes by presenting recent applications.

Descriptors representing two- and three-body atomic distributions and their effects on the accuracy of machine-learned inter-atomic potentials.

When determining machine-learning models for inter-atomic potentials, the potential energy surface is often described as a non-linear function of descriptors representing two- and three-body atomic distribution functions. It is not obvious how the choice of the descriptors affects the efficiency of the training and the accuracy of the final machine-learned model. In this work, we formulate an efficient method to calculate descriptors that can separately represent two- and three-body atomic distribution functions, and we examine the effects of including only two- or three-body descriptors, as well as including both, in the regression model. Our study indicates that non-linear mixing of two- and three-body descriptors is essential for an efficient training and a high accuracy of the final machine-learned model. The efficiency can be further improved by weighting the two-body descriptors more strongly. We furthermore examine a sparsification of the three-body descriptors. The three-body descriptors usually provide redundant representations of the atomistic structure, and the number of descriptors can be significantly reduced without loss of accuracy by applying an automatic sparsification using a principal component analysis. Visualization of the reduced descriptors using three-body distribution functions in real-space indicates that the sparsification automatically removes the components that are less significant for describing the distribution function.

CO oxidation activity of non-reducible oxide-supported mass-selected few-atom Pt single-clusters.

Platinum nanocatalysts play critical roles in CO oxidation, an important catalytic conversion process. As the catalyst size decreases, the influence of the support material on catalysis increases which can alter the chemical states of Pt atoms in contact with the support. Herein, we demonstrate that under-coordinated Pt atoms at the edges of the first cluster layer are rendered cationic by direct contact with the Al2O3 support, which affects the overall CO oxidation activity. The ratio of neutral to cationic Pt atoms in the Pt nanocluster is strongly correlated with the CO oxidation activity, but no correlation exists with the total surface area of surface-exposed Pt atoms. The low oxygen affinity of cationic Pt atoms explains this counterintuitive result. Using this relationship and our modified bond-additivity method, which only requires the catalyst-support bond energy as input, we successfully predict the CO oxidation activities of various sized Pt clusters on TiO2.

Phase Transitions of Hybrid Perovskites Simulated by Machine-Learning Force Fields Trained on the Fly with Bayesian Inference.

Realistic finite temperature simulations of matter are a formidable challenge for first principles methods. Long simulation times and large length scales are required, demanding years of computing time. Here we present an on-the-fly machine learning scheme that generates force fields automatically during molecular dynamics simulations. This opens up the required time and length scales, while retaining the distinctive chemical precision of first principles methods and minimizing the need for human intervention. The method is widely applicable to multielement complex systems. We demonstrate its predictive power on the entropy driven phase transitions of hybrid perovskites, which have never been accurately described in simulations. Using machine learned potentials, isothermal-isobaric simulations give direct insight into the underlying microscopic mechanisms. Finally, we relate the phase transition temperatures of different perovskites to the radii of the involved species, and we determine the order of the transitions in Landau theory.

Band bending and dipole effect at interface of metal-nanoparticles and TiO2 directly observed by angular-resolved hard X-ray photoemission spectroscopy.

This paper describes the observation of band bending and band edge shifts at the interfaces between nanoscale metals and TiO2 film over a wide depth range by angular-resolved hard X-ray photoemission spectroscopy (HAXPES). The HAXPES results indicate strong electrostatic interactions between the TiO2 semiconductor and metal nanoparticles, while density functional theory (DFT) calculations suggest that these interactions are primarily associated with charge transfer leading to electric dipole moments at the interface in the ground state. The effects of these dipole moments are not limited to the surface but also occur deep in the bulk of the semiconductor, and are highly dependent on the coverage of the metal nanoparticles on the semiconductor species.

A Universal 3D Voxel Descriptor for Solid-State Material Informatics with Deep Convolutional Neural Networks.

Material informatics (MI) is a promising approach to liberate us from the time-consuming Edisonian (trial and error) process for material discoveries, driven by machine-learning algorithms. Several descriptors, which are encoded material features to feed computers, were proposed in the last few decades. Especially to solid systems, however, their insufficient representations of three dimensionality of field quantities such as electron distributions and local potentials have critically hindered broad and practical successes of the solid-state MI. We develop a simple, generic 3D voxel descriptor that compacts any field quantities, in such a suitable way to implement convolutional neural networks (CNNs). We examine the 3D voxel descriptor encoded from the electron distribution by a regression test with 680 oxides data. The present scheme outperforms other existing descriptors in the prediction of Hartree energies that are significantly relevant to the long-wavelength distribution of the valence electrons. The results indicate that this scheme can forecast any functionals of field quantities just by learning sufficient amount of data, if there is an explicit correlation between the target properties and field quantities. This 3D descriptor opens a way to import prominent CNNs-based algorithms of supervised, semi-supervised and reinforcement learnings into the solid-state MI.

Predicting Catalytic Activity of Nanoparticles by a DFT-Aided Machine-Learning Algorithm.

Catalytic activities are often dominated by a few specific surface sites, and designing active sites is the key to realize high-performance heterogeneous catalysts. The great triumphs of modern surface science lead to reproduce catalytic reaction rates by modeling the arrangement of surface atoms with well-defined single-crystal surfaces. However, this method has limitations in the case for highly inhomogeneous atomic configurations such as on alloy nanoparticles with atomic-scale defects, where the arrangement cannot be decomposed into single crystals. Here, we propose a universal machine-learning scheme using a local similarity kernel, which allows interrogation of catalytic activities based on local atomic configurations. We then apply it to direct NO decomposition on RhAu alloy nanoparticles. The proposed method can efficiently predict energetics of catalytic reactions on nanoparticles using DFT data on single crystals, and its combination with kinetic analysis can provide detailed information on structures of active sites and size- and composition-dependent catalytic activities.

Activities and Stabilities of Au-Modified Stepped-Pt Single-Crystal Electrodes as Model Cathode Catalysts in Polymer Electrolyte Fuel Cells.

The purpose of this study is to test the concept of protecting vulnerable sites on cathode catalysts in polymer electrolyte fuel cells. Pt single-crystal surfaces were modified by depositing Au atoms selectively on (100) step sites and their electrocatalytic activities for oxygen reduction reaction (ORR) and stabilities against potential cycles were examined. The ORR activities were raised by 70% by the Au modifications, and this rise in the activity was ascribed to enhanced local ORR activities on Pt(111) terraces by the surface Au atoms. The Au modifications also stabilized the Pt surfaces against potential cycles by protecting the low-coordinated (100) step sites from surface reorganizations. Thus, the surface modification by selective Au depositions on vulnerable sites is a promising method to enhance both the ORR activity and durability of the catalysts.

What Makes the Photocatalytic CO2 Reduction on N-Doped Ta2O5 Efficient: Insights from Nonadiabatic Molecular Dynamics.

Recent experimental studies demonstrated that photocatalytic CO2 reduction by Ru catalysts assembled on N-doped Ta2O5 surface is strongly dependent on the nature of the anchor group with which the Ru complexes are attached to the substrate. We report a comprehensive atomistic analysis of electron transfer dynamics in electroneutral Ru(di-X-bpy) (CO)2Cl2 complexes with X = COOH and PO3H2 attached to the N-Ta2O5 substrate. Nonadiabatic molecular dynamics simulations indicate that the electron transfer is faster in complexes with COOH anchors than in complexes with PO3H2 groups, due to larger nonadiabatic coupling. Quantum coherence counteracts this effect, however, to a small extent. The COOH anchor promotes the transfer with significantly higher frequency modes than PO3H2, due to both lighter atoms (C vs P) and stronger bonds (double vs single). The acceptor state delocalizes onto COOH, but not PO3H2, further favoring electron transfer in the COOH system. At the same time, the COOH anchor is prone to decomposition, in contrast to PO3H2, making the former show smaller turnover numbers in some cases. These theoretical predictions are consistent with recent experimental results, legitimating the proposed mechanism of the electron transfer. We emphasize the role of anchor stability, nonadiabatic coupling, and quantum coherence in determining the overall efficiency of artificial photocatalytic systems.

Kinetically induced irreversibility in electro-oxidation and reduction of Pt surface.

A mean field kinetic model was developed for electrochemical oxidations and reductions of Pt(111) on the basis of density functional theory calculations, and the reaction mechanisms were analyzed. The model reasonably describes asymmetric shapes of cyclic voltammograms and small Tafel slopes of relevant redox reactions observed in experiments without assuming any unphysical forms of rate equations. Simulations using the model indicate that the oxidation of Pt(111) proceeds via an electrochemical oxidation from Pt to PtOH and a disproportionation reaction from PtOH to PtO and Pt, while its reduction proceeds via two electrochemical reductions from PtO to PtOH and from PtOH to Pt.