Towards Universal Neural Network Potential for Material Discovery Applicable to Arbitrary Combination of 45 Elements

Abstract

Computational material discovery is under intense study owing to its ability to explore the vast space of chemical systems. Neural network potentials (NNPs) have been shown to be particularly effective in conducting atomistic simulations for such purposes. However, existing NNPs are generally designed for narrow target materials, making them unsuitable for broader applications in material discovery. To overcome this issue, we have developed a universal NNP called PreFerred Potential (PFP), which is able to handle any combination of 45 elements. Particular emphasis is placed on the datasets, which include a diverse set of virtual structures used to attain the universality. We demonstrated the applicability of PFP in selected domains: lithium diffusion in LiFeSO${}_4$F, molecular adsorption in metal-organic frameworks, an order-disorder transition of Cu-Au alloys, and material discovery for a Fischer-Tropsch catalyst. They showcase the power of PFP, and this technology provides a highly useful tool for material discovery.

Previous title: "PFP: Universal Neural Network Potential for Material Discovery"

References in corpus (5)

• PyTorch: An Imperative Style, High-Performance Deep Learning Library
• Accurate and efficient algorithm for Bader charge integration
• ANI-1: An extensible neural network potential with DFT accuracy at force field computational cost
• Scaling Laws for Neural Language Models
• The Open Catalyst 2020 (OC20) Dataset and Community Challenges

Cited by in corpus (42)

• DeePMD-kit v2: A software package for Deep Potential models
• Scalable Parallel Algorithm for Graph Neural Network Interatomic Potentials in Molecular Dynamics Simulations
• A practical guide to machine learning interatomic potentials -- Status and future
• General-purpose machine-learned potential for 16 elemental metals and their alloys
• Evaluation of the MACE Force Field Architecture: from Medicinal Chemistry to Materials Science
• DPA-2: a large atomic model as a multi-task learner
• On the redundancy in large material datasets: efficient and robust learning with less data
• The Evolution of Machine Learning Potentials for Molecules, Reactions and Materials
• Artificial Intelligence in Material Engineering: A review on applications of AI in Material Engineering
• A Comprehensive and Versatile Multimodal Deep Learning Approach for Predicting Diverse Properties of Advanced Materials
• GPTFF: A high-accuracy out-of-the-box universal AI force field for arbitrary inorganic materials
• A reactive neural network framework for water-loaded acidic zeolites
• Universal Machine Learning Kohn-Sham Hamiltonian for Materials
• Tensor-reduced atomic density representations
• BOTAN: BOnd TArgeting Network for prediction of slow glassy dynamics by machine learning relative motion
• An equivariant graph neural network for the elasticity tensors of all seven crystal systems
• Navigating chemical reaction space with a steering wheel
• MGNN: Moment Graph Neural Network for Universal Molecular Potentials
• Segregation and Ordering of Light Interstitials (B, C, H, and N) in Cr-Ni Alloys: Implications for Grain Boundary Stability in Superalloy Design
• Artificial Intelligence for Science in Quantum, Atomistic, and Continuum Systems
• How Graph Neural Network Interatomic Potentials Extrapolate: Role of the Message-Passing Algorithm
• Facilitating {\it ab initio} configurational sampling of multicomponent solids using an on-lattice neural network model and active learning
• Modeling intercalation chemistry with multi-redox reactions by sparse lattice models in disordered rocksalt cathodes
• Accurate and efficient machine learning interatomic potentials for finite temperature modeling of molecular crystals
• Global optimization in the discrete and variable-dimension conformational space: The case of crystal with the strongest atomic cohesion
• Pre-training, fine-tuning, and distillation (PFD): Automatically generating machine learning force fields from universal models
• Equivariant Neural Networks for Spin Dynamics Simulations of Itinerant Magnets
• Generative Model for Constructing Reaction Path from Initial to Final States
• A statistical understanding of oxygen vacancies in distorted high-entropy oxides
• Atomistic Simulations of Oxide-Water Interfaces using Machine Learning Potentials
• Atomistic mechanisms of oxidation and chlorine corrosion in Ni-based superalloys: The role of boron and light interstitial segregation
• Neural network potential molecular dynamics simulations of (La,Ce,Pr,Nd)0.95(Mg,Zn,Pb,Cd,Ca,Sr,Ba)0.05F2.95
• Moiré superlattices of antimonene on a Bi(111) substrate with van Hove singularity and Rashba-type spin polarization
• Atomistic Simulations of Short-range Ordering with Light Interstitials in Inconel Superalloys
• Combining graph deep learning and London dispersion interatomic potentials: A case study on pnictogen chalcohalides
• Massive Discovery of Low-Dimensional Materials from Universal Computational Strategy
• FastTrack: a fast method to evaluate mass transport in solid leveraging universal machine learning interatomic potential
• Managing Expectations and Imbalanced Training Data in Reactive Force Field Development: an Application to Water Adsorption on Alumina
• A flexible kinetic Monte Carlo framework for GaN molecular beam epitaxy with adaptive on-the-fly barrier evaluation
• Spectral Analysis of Light Interstitial Segregation Energies in Ni: The Role of Local Cr Coordination for Boron and Carbon
• Revealing trends in catalytic activity of adatoms for hydrogen adsorption on carbon: a case study of graphene and carbon nanotube
• Spectral Sampling of Boron Diffusion in Ni Alloys: Cr and Mo Effects on Bulk and Grain Boundary Transport