INVESTIGATION OF CHEMISTRY IN MATERIALS USING FIRST-PRINCIPLES METHODS AND MACHINE LEARNING FORCE FIELDS

In computational catalysis, density-functional theory (DFT) calculations are usually utilized, although they suffer from high computational costs. Thus, it would be challenging to explicitly predict the catalytic properties of nanoparticles (NPs) at the nanoscale under solvents. Using molecular dynamics (MD) simulations with a reactive force field (ReaxFF), we investigated the catalytic performance of Ni-Pt NPs for the direct synthesis of hydrogen peroxide (H2O2), in which water solvents were explicitly considered along with the effects of the sizes (1.5, 2.0, 3.0, and 3.5 nm) and compositions (Ni90Pt10, Ni80Pt20, and Ni50Pt50) of the NPs. Among the Ni-Pt NPs, 3.0 nm NPs show the highest activity and selectivity for the direct synthesis of H2O2, revealing that the catalytic performance is not well correlated with the surface areas of NPs. The superior catalytic performance results from the high H2 dissociation and low O2 dissociation properties, which are correlated with the numbers of NiNiPt-fcc and NiNi-bridge sites on the surface of Ni-Pt NPs, respectively. The ReaxFF-MD simulations propose the optimum composition (Ni80Pt20) of 3.0 nm Ni-Pt NPs, which is also explained by the numbers of NiNiPt-fcc and NiNi-bridge sites. Furthermore, from the ReaxFF-MD simulations, the direct synthesis of H2O2 for the Ni-Pt NPs can be achieved not only with the Langmuir-Hinshelwood mechanism, which has been conventionally considered, but also with the water-induced mechanism, which is unlikely to occur on pure Pd and Pd-based alloy catalysts; these results are supported by DFT calculations. These results reveal that the ReaxFF-MD method provides significant information for predicting the catalytic properties of NPs, which could be difficult to provide with DFT calculations; thus, it can be a useful framework for the design of nanocatalysts through complementation with a DFT method.

To make a practical molecular dynamics (MD) simulation of the large-scale reactive chemical systems of Li-H and Li-C, we have optimized parameters of the reactive force field (ReaxFF) for these systems. The parameters for this force field were obtained from fitting to the results of density functional theory (DFT) calculations on the structures and energy barriers for a number of Li-H and Li-C molecules, including Li(2), LiH, Li(2)H(2), H(3)C-Li, H(3)C-H(2)C-Li, H(2)C=C-LiH, HCCLi, H(6)C(5)-Li, and Li(2)C(2), and to the equations of state and lattice parameters for condensed phases of Li. The accuracy of the developed ReaxFF was also tested by comparison to the dissociation energies of lithium-benzene sandwich compounds and the collision behavior of lithium atoms with a C(60) buckyball.

Stable molecular dynamics simulations of halide perovskites from a temperature-ensemble gradient-domain machine learning approach

In this work, we present the parametrization of Ca-O/H interactions within the reactive force field ReaxFF, and its application to study the hydration of calcium oxide surface. The force field has been fitted using density functional theory calculations on gas phase calcium-water clusters, calcium oxide bulk and surface properties, calcium hydroxide, bcc and fcc Ca, and proton transfer reactions in the presence of calcium. Then, the reactive force field has been used to study the hydration of the calcium oxide {001} surface with different water contents. Calcium oxide is used as a catalyzer in many applications such as CO(2) sequestration and biodiesel production, and the degree of surface hydroxylation is a key factor in its catalytic performance. The results show that the water dissociates very fast on CaO {001} bare surfaces without any defect or vacancy. The surface structure is maintained up to a certain amount of water, after which the surface undergoes a structural rearrangement, becoming a disordered calcium hydroxyl layer. This transformation is the most probable reason for the CaO catalytic activity decrease.

Deformation of metal–organic frameworks (MOFs)induced byadsorbate molecules can affect adsorption properties such as capacityand selectivity, but most computational studies of MOFs assume frameworkrigidity to simplify calculations. Although flexible force fields(FFs) for MOFs have been parametrized for specific materials, thegenerality of FFs for reliably modeling adsorbate-induced deformationto accuracy nearing that of density functional theory (DFT) has notbeen established. This work confirms using DFT calculations that adsorbate-induceddeformation can affect CO2 and H2O adsorptionenergies in a considerable fraction of MOFs promising for direct aircapture (DAC). We then benchmark the efficacy of several general-purposeFFs in describing adsorbate-induced deformation for DAC against DFT.Our results show that current classical FFs are insufficient for describingMOF deformation, especially in cases of interest for DAC where stronginteractions exist between adsorbed molecules and MOF frameworks.Some emerging machine learning force fields (MLFFs) we tested, particularlyCHGNet, MACE-MP-0, and Equiformer V2, appear to be more promisingthan the classical FF for emulating the deformation behavior describedby DFT. The best performing FF (CHGNet), however, fails to achievethe accuracy required for practical predictions with a mean absoluteadsorption energy error of 0.124 eV.

Theoretical study of the CO formation mechanism in the CO2 gasification of lignite

We developed a framework for predicting the energies and ground state configurations of native point defects, extrinsic dopants and impurities, and defect complexes across zinc blende-phase Cd/Zn-Te/Se/S compounds, important for CdTe-based solar cells. This framework, named DeFecT-FF, is powered by high-throughput density functional theory (DFT) computations and crystal graph-based machine learning force field (MLFF) models trained on the DFT data. The Cd/Zn-Te/Se/S chemical space is chosen because alloying at Cd or Te sites is a promising avenue to tailor the electronic and defect properties of the CdTe absorber layer to potentially improve solar cell performance. The sheer number of defect configurations achievable when considering all possible singular defects and their combinations, symmetry-breaking operations, and defect charge states, as well as the expense of running large supercell calculations, makes this an ideal problem for developing accurate and widely-applicable force field models. Here, we introduce our dataset of structures and energies from HSE06 geometry optimization, including bulk and alloyed supercells with and without defects. Data were gradually expanded using active learning and accurate MLFF models were trained to predict energies and atomic forces across different charge states. Via accelerated prediction and screening, we identified many new low energy defect configurations and obtained high-fidelity defect formation energy diagrams using HSE06 calculations with spin-orbit coupling. The DeFecT-FF framework has been released publicly as an online tool on the nanoHUB platform, allowing users to upload any crystallographic information file, generate defects of interest, and compute defect formation energies as a function of Fermi level and chemical potential conditions, thus bypassing expensive DFT calculations.

Exploring carbon dioxide generation in combustion of long-flame coal in Huating mining area by using ReaxFF MD

Coal tar, a by-product of the pyrolysis of coal, is rich in aromatic compounds that have the potential to facilitate the synthesis of graphene, a high-quality carbon material, via low-temperature chemical vapor deposition (CVD). This approach offers a promising avenue for the cost-effective and large-scale industrial production of graphene while minimizing energy consumption. Nevertheless, there is a paucity of research focused on the low-temperature synthesis mechanisms of graphene derived from aromatic compounds in the context of graphene growth. To achieve high-quality graphene synthesis from coal tar and its aromatic constituents at reduced temperatures, a comprehensive investigation into the reaction pathways of these aromatic compounds is essential. In this study, we meticulously simulate the pyrolysis of benzene, a key aromatic component of coal tar, across various temperature settings utilizing reactive force field (ReaxFF) methodology. Furthermore, we apply density functional theory (DFT) calculations, executed through the Vienna Ab initio Simulation Package (VASP), to assess the dehydrogenation energy associated with the adsorption of benzene on vapor-deposited copper foils. Our molecular dynamics simulations, enhanced by a mixed force field approach, revealed that the dehydrogenated benzene ring (C6 intermediate) acts as a critical precursor for graphene synthesis. This research significantly elucidates the reaction pathways of aromatic benzene in coal tar through molecular simulations conducted at different temperatures, both in the gas phase and on solid copper foil substrates.

In the pursuit of higher energy densities crucial for electrifying the transportation sector, this study employs a machine learning (ML) force field to explore the conversion reaction mechanisms of Li-FeF2 battery cathodes. By leveraging ML, we accelerate simulation times significantly while maintaining the fidelity of traditional ab-initio methods. Our analysis focuses on the cation diffusion pathways and charge transfer processes during battery operation. We specifically investigate how the coherence of the LiF/Fe interface impacts the diffusion activation barriers of Fe. The ML-enhanced simulations uncover fundamental mechanisms governing cathode cycling behavior, previously unachievable with compute-intensive density functional theory. Additionally, we explore the intrinsic diffusion limitations of Fe at the LiF/Fe interface, providing insights crucial for optimizing cathode particle sizes to enhance battery performance. Our findings not only deepen the understanding of battery chemistry but also guide the design of next-generation batteries, underscoring the broad applicability of AI and ML in characterizing battery materials.

Combustion of an Illinois No. 6 coal char simulated using an atomistic char representation and the ReaxFF reactive force field

Halide perovskite optoelectronic devices suffer from chemical degradation and current-voltage hysteresis induced by migration of highly mobile charged defects. Atomic scale molecular dynamics simulations can capture the motion of these ionic defects, but classical force fields are too inflexible to describe their dynamical charge states. Using CsPbI3 as a case study, we train machine learned force fields from density functional theory calculations and study the diffusion of charged halide interstitial and vacancy defects in bulk CsPbI3. We find that negative iodide interstitials and positive iodide vacancies, the most stable charge states for their respective defect type, migrate at similar rates at room temperature. Neutral interstitials are faster, but neutral vacancies are 1 order of magnitude slower. Oppositely charged interstitials and vacancies, as they can occur in device operation or reverse bias conditions, are significantly slower and can be considered relatively immobile.

A force field as accurate as quantum mechanics (QMs) and as fast as molecular mechanics (MMs), with which one can simulate a biomolecular system efficiently enough and meaningfully enough to get quantitative insights, is among the most ardent dreams of biophysicists—a dream, nevertheless, not to be fulfilled any time soon. Machine learning force fields (MLFFs) represent a meaningful endeavor in this direction, where differentiable neural functions are parametrized to fit ab initio energies and forces through automatic differentiation. We argue that, as of now, the utility of the MLFF models is no longer bottlenecked by accuracy but primarily by their speed, as well as stability and generalizability—many recent variants, on limited chemical spaces, have long surpassed the chemical accuracy of 1 kcal/mol—the empirical threshold beyond which realistic chemical predictions are possible—though still magnitudes slower than MM. Hoping to kindle exploration and design of faster, albeit perhaps slightly less accurate MLFFs, in this review, we focus our attention on the technical design space (the speed-accuracy trade-off) between MM and ML force fields. After a brief review of the building blocks (from a machine learning-centric point of view) of force fields of either kind, we discuss the desired properties and challenges now faced by the force field development community, survey the efforts to make MM force fields more accurate and ML force fields faster, and envision what the next generation of MLFF might look like.

In this study, wedevelop a reactive force field (ReaxFF) for aSi/O/H/F system to perform etching simulations of SiO2 withan HF etchant. Quantum mechanical (QM) training sets from densityfunctional theory calculations, which contain structures of reactant/productand energies with bond dissociation, valence angle distortions, andreactions between SiO2 clusters and SiO2 slabwith HF gases, are used to optimize the ReaxFF parameters. Structuresand energies calculated using the ReaxFF match well with the QM trainingsets. Using the optimized ReaxFF, we conduct molecular dynamics simulationsof the etching process of SiO2 substrates with active HFmolecules. The etching yield and number of reaction products withdifferent incident energies of the HF etchant are investigated. Thesesimulations show that the developed ReaxFF offers insights into theatomistic surface reaction of the SiO2 etching process.

Development of a Ni–Al reactive force field for Ni-based superalloy: revealing electrostatic effects on mechanical deformation