A5084028723- Scientific Computing and Data Management
- Machine Learning in Materials Science
- Advanced Database Systems and Queries
- Computational Physics and Python Applications
- X-ray Diffraction in Crystallography
- Electron and X-Ray Spectroscopy Techniques
- Computational Drug Discovery Methods
- Inorganic Chemistry and Materials
- Cosmology and Gravitation Theories
- Surface Chemistry and Catalysis
- Noncommutative and Quantum Gravity Theories
- Fuel Cells and Related Materials
- BIM and Construction Integration
- Advanced Materials Characterization Techniques
- Delphi Technique in Research
- Reservoir Engineering and Simulation Methods
- Advanced Battery Technologies Research
- Ferroelectric and Negative Capacitance Devices
- Electronic and Structural Properties of Oxides
- Black Holes and Theoretical Physics
- Catalytic Processes in Materials Science
- Neural Networks and Applications
Lawrence Berkeley National Laboratory
2023-2025
University of Cambridge
2024-2025
Abstract Large-scale simulations with complex electron interactions remain one of the greatest challenges for atomistic modelling. Although classical force fields often fail to describe coupling between electronic states and ionic rearrangements, more accurate ab initio molecular dynamics suffers from computational complexity that prevents long-time large-scale simulations, which are essential study technologically relevant phenomena. Here we present Crystal Hamiltonian Graph Neural Network...
Machine-learned force fields have transformed the atomistic modelling of materials by enabling simulations ab initio quality on unprecedented time and length scales. However, they are currently limited by: (i) significant computational human effort that must go into development validation potentials for each particular system interest; (ii) a general lack transferability from one chemical to next. Here, using state-of-the-art MACE architecture we introduce single general-purpose ML model,...
Abstract Machine learning interatomic potentials (MLIPs) have introduced a new paradigm for atomic simulations. Recent advancements led to universal MLIPs (uMLIPs) that are pre-trained on diverse datasets, providing opportunities force fields and foundational machine models. However, their performance in extrapolating out-of-distribution complex environments remains unclear. In this study, we highlight consistent potential energy surface (PES) softening effect three uMLIPs: M3GNet, CHGNet,...
High-throughput density functional theory (DFT) calculations have become a vital element of computational materials science, enabling screening, property database generation, and training “universal” machine learning models. While several software frameworks emerged to support these efforts, new developments such as learned force fields increased demands for more flexible programmable workflow solutions. This manuscript introduces atomate2, comprehensive evolution our original atomate...
We present Jobflow, a domain-agnostic Python package for writing computational workflows tailored high-throughput computing applications.With its simple decorator-based approach, functions and class methods can be transformed into compute jobs that stitched together complex workflows.Jobflow fully supports dynamic where the full acyclic graph of is not known until runtime, such as launch other based on results previous steps in workflow.The all Jobflow easily stored variety filesystem-and...
Reducing hallucination of Large Language Models (LLMs) is imperative for use in the sciences where reproducibility crucial. However, LLMs inherently lack long-term memory, making it a nontrivial, ad hoc, and inevitably biased task to fine-tune them on domain-specific literature data. Here we introduce LLaMP, multimodal retrieval-augmented generation (RAG) framework multiple data-aware reasoning-and-acting (ReAct) agents that dynamically interact with computational experimental data Materials...
The Open Databases Integration for Materials Design (OPTIMADE) application programming interface (API) empowers users with holistic access to a growing federation of databases, enhancing the accessibility and discoverability materials chemical data. Since first release OPTIMADE specification (v1.0), API has undergone significant development, leading v1.2 release, underpinned multiple scientific studies. In this work, we highlight latest features format, accompanying software tools, provide...
Machine learning interatomic potentials (MLIPs) have introduced a new paradigm for atomic simulations. Recent advancements seen the emergence of universal MLIPs (uMLIPs) that are pre-trained on diverse materials datasets, providing opportunities both ready-to-use force fields and robust foundations downstream machine refinements. However, their performance in extrapolating to out-of-distribution complex environments remains unclear. In this study, we highlight consistent potential energy...
Matbench Discovery simulates the deployment of machine learning (ML) energy models in a high-throughput search for stable inorganic crystals. We address disconnect between (i) thermodynamic stability and formation (ii) in-domain vs out-of-distribution performance. Alongside this paper, we publish Python package to aid with future model submissions growing online leaderboard further insights into trade-offs various performance metrics. To answer question which ML methodology performs best at...
Materials with high-dielectric constant easily polarize under external electric fields, allowing them to perform essential functions in many modern electronic devices. Their practical utility is determined by two conflicting properties: high dielectric constants tend occur materials narrow band gaps, limiting the operating voltage before breakdown. We present a high-throughput workflow that combines element substitution, ML pre-screening, ab initio simulation and human expert intuition...
The Open Databases Integration for Materials Design (OPTIMADE) application programming interface (API) empowers users with holistic access to a growing federation of databases, enhancing the accessibility and discoverability materials chemical data. Since first release OPTIMADE specification (v1.0), API has undergone significant development, leading upcoming v1.2 release, underpinned multiple scientific studies. In this work, we highlight latest features format, accompanying software tools,...
Highlights•High-throughput workflow to explore materials with high dielectric constants•Combining machine learning and density functional perturbation theory for screening•A promising material, Bi2Zr2O7, is successfully discovered synthesizedSummaryMaterials constants polarize easily under external electric fields, making them essential in modern electronics. However, their utility often limited by narrow band gaps, which reduce the operating voltage before breakdown. We present a...
Large-scale simulations with complex electron interactions remain one of the greatest challenges for atomistic modeling electrochemical systems. Our recent work, Crystal Hamiltonian Graph Neural Network (CHGNet), presents foundational graph-neural-network-based machine-learning interatomic potential (MLIP). The inclusion charge information by magnetic moment prediction enables CHGNet to better describe both atomic and electronic degrees freedom, which opens possibility a understanding...