A large-scale pretraining corpus for materials science, containing approximately 780M tokens of RefinedWeb data, designed for training AI models for materials discovery and molecular dynamics research.
See LICENSE filematerials sciencepretraining corpusAI materials discovery
A technical Q&A dataset covering fundamental and advanced topics in materials science, including material properties, synthesis methods, characterization techniques, and engineering applications, designed for training technical response models.
See LICENSE filematerials sciencetechnical Q&Asynthesis methods
A materials science demonstration dataset hosted on HuggingFace under the SciFy account, showcasing data organization and access patterns for materials science applications.
See LICENSE fileMaterials ScienceDatasetHuggingFace
HetCat26 provides benchmark tests for assessing the transferability of pre-trained machine learning interatomic potentials across key aspects of heterogeneous catalytic modeling.
See LICENSE fileHeterogeneous CatalysisMachine Learning Interatomic PotentialsBenchmarking
A materials science dataset hosted on HuggingFace, supporting machine learning and materials science cross-disciplinary research, containing relevant domain training or evaluation samples.
See LICENSE fileMaterials ScienceDatasetMachine Learning
A collection of domain ontologies for materials science and engineering, providing semantic organization of core concepts including structure, properties, processing, and performance for knowledge reasoning applications.
See LICENSE fileMaterials ScienceOntologyKnowledge Graph
Open-source classical molecular dynamics engine from Sandia, covering materials and soft matter with multiscale modelling and custom potentials, plus ML-IAP interfaces for machine-learned potentials.
NewGNU General Public License 2.0Molecular DynamicsMaterials SimulationPotentials
Presents a general theory for rolling arbitrary 2D crystals into nanotubes, implemented as open-source NTBuilder with desktop and web interfaces, including a catalog of 20 million nanotube structures.
See LICENSE filenanotube2D crystalmaterials discovery
The NIST-led additive manufacturing benchmark series tests melt-pool, residual-stress and microstructure models through controlled machines, standardized artifacts and blind challenges, with nine benchmark sets and 85 challenge submissions in 2025.
NewPublic data (U.S. NIST, public domain)Additive ManufacturingMetal 3D PrintingBenchmark
The open-source chemical kinetics suite originally initiated at Caltech and hosted under NumFOCUS provides thermodynamics, detailed kinetics, transport, zero-dimensional reactor networks and one-dimensional reacting flows in one object model, shipping 22 built-in mechanisms such as GRI-Mech 3.0 as a reproducible source of physical ground truth for combustion, electrochemistry and plasma modelling.
NewBSD 3-Clause style (California Institute of Technology 2001-2009, Sandia Corporation 2009, Cantera Developers 2011-2026)Canterachemical kineticscombustion
Materials and Chemistry / Atomistic Machine Learning
The consolidated repository of Meta FAIR Chemistry has grown from an OC20 electrocatalysis toolkit into a unified platform spanning molecules, materials, and catalysts, with the UMA model using a mixture-of-linear-experts architecture trained on 500 million structures and switching task names to serve seven domains from OC20 to OC25, OMat24, and OMol25.
NewRepository code is MIT (Meta, Inc.); the fairchem-lammps package declares GnuV2 on PyPI; UMA weights are licensed separately under gated model terms; datasets carry their own termsAtomic Foundation ModelMachine Learning PotentialElectrocatalysis
Google DeepMind's graph-network materials discovery system pushes the candidate pool past a billion structures through active learning, verifies them with DFT, and feeds the results back into training, adding 381,000 stable materials to the convex hull while shipping a downloadable convex-hull dataset and pretrained potential implementations.
NewCode and notebooks are Apache-2.0; the GNoME dataset is released under CC BY-NC 4.0 (non-commercial); the paper is CC BY 4.0Materials DiscoveryGraph Neural NetworkFirst Principles
An open-source equivariant graph-neural-network potential framework from the Csanyi group that unifies atomic cluster expansion with high-order equivariant message passing, reaching high accuracy with only two interaction layers, and now shipping several generations of foundation weights (MACE-MP, MH, OMOL, POLAR) for out-of-the-box molecular dynamics of crystals, liquids, gases, surfaces and reactive systems.
NewMIT (code); some weights are released under the Academic Software License (ASL) for academic use onlyMolecular DynamicsMachine-Learned PotentialEquivariant Neural Network
This roadmap reviews the evolution of polymer informatics toward on-demand inverse design, causal reasoning across chemistry, processing, and performance, and closed-loop autonomous experimentation.
See LICENSE filepolymer informaticsinverse designautonomous experimentation
This article reviews how artificial intelligence and machine learning support physical modeling in chemical engineering, including atomic-scale simulations, materials and catalyst discovery, transport and thermodynamics, separations, and process systems engineering.
See LICENSE filechemical engineeringmaterials discoveryphysical modeling
This survey reviews optimization of mathematical models combining integer decisions with ReLU neural networks for chemical and process systems engineering, particularly superstructure optimization using surrogates for complex nonlinear processes.
See LICENSE fileReLU neural networkssurrogate modelschemical process optimization
A systematic survey of operator learning covering the theory, architectures and error analysis of DeepONet, FNO and their variants - the reference entry point to learning infinite-dimensional maps.
arXiv preprint (copyright held by authors)Operator LearningSurveyDeepONet
The Python analysis library behind Materials Project - crystal structures, phase diagrams, band structures, DOS out of the box; the standard data layer between DFT and ML workflows.
ICLR 2026 paper proposing ATOM, a quasi-equivariant pretrained transformer neural operator that requires no explicit molecular graph and uses temporal attention for parallel decoding of multiple future states; curates TG80 (80 compounds, 2.5M fs trajectories) and demonstrates zero-shot generalization on MD17/RMD17/MD22.
arXiv preprint (copyright held by authors)Neural OperatorMolecular DynamicsTransformer
Investigating nonequilibrium phase-ordering kinetics in binary mixtures of uniaxial ellipsoidal particles using large-scale molecular dynamics simulations with anisotropic Gay-Berne potential, revealing two-stage ordering processes across different time scales after temperature quench.
See LICENSE filemolecular dynamicsphase-ordering kineticsanisotropic potential
Presents FRPN, a Full-Resolution Polymer Network that retains both atom-level details and monomer-level connectivity for multi-scale polymer property prediction.
See LICENSE filepolymermulti-scale modelingmaterials discovery
Presents the first complete machine learning method for accelerating plane-wave DFT in materials under the PAW formalism, formalizing seven criteria that a Complete Neural Electronic Initializer must satisfy and addressing two missing structure-dependent components: augmentation occupancies and spin initialization.
See LICENSE fileDensity Functional TheoryMachine LearningMaterials Science
A physics-informed dual-state query learning framework for GPCR ligand bioactivity prediction, addressing the neglect of dynamic conformational equilibria in existing deep learning approaches to improve functional efficacy discrimination.
See LICENSE fileGPCR ligandbioactivity predictionphysics-informed learning
Proposes finite-state automata encoding recursive lattice geometry for exact MPO construction of tight-binding Hamiltonians across fractals, quasicrystals, and hyperbolic lattices.
See LICENSE filetight-binding modelmatrix product operatorfinite-state automaton
A method to predict how force correction changes energy exchange and structural statistics in molecular dynamics by measuring leading response coefficients on shared reference trajectories.
See LICENSE filemolecular dynamicsforce correctionresponse coefficients
A digital simulation toolkit for physics-based generation of realistic STM images of material surfaces, supporting atomic-scale surface characterization while addressing the extreme noise sensitivity of scanning tunneling microscopy.
See LICENSE filescanning tunneling microscopymaterial surface imagingphysics-based image generation
Systematically compares hard-constrained and soft-constrained physics-informed molecular machine learning on experimental data, examining how fundamental thermodynamic relations such as the Clapeyron equation are embedded into architectures or training losses for materials property prediction.
See LICENSE filePhysics-informed molecular MLClapeyron equationThermodynamic constraints
AbsorbEvo is an agentic framework for autonomous inverse design of microwave absorbers, translating natural-language performance objectives into candidate designs verified through full-wave electromagnetic simulations.
See LICENSE filemicrowave absorbersinverse designfull-wave simulation
Proposes a nearly-instant latent space Bayesian optimization method that replaces expensive sequential decision-making with sampling overhead, making BO practical for generative model-driven de novo design pipelines with cheap virtual screens.
See LICENSE fileBayesian OptimizationLatent Space OptimizationGenerative Design
Presents a method using sparse differentiation to compute exact molecular Hessians from machine learning interatomic potentials, reducing computational complexity from O(N²) to O(N), overcoming limitations for large atomistic systems.
See LICENSE filemolecular dynamicsinteratomic potentialsHessian matrix
Uses explicit-ion molecular dynamics and charge-regulation Monte Carlo simulations to compute the interaction force between two spherical, charge-regulated amphoteric nanoparticles, addressing how adjustable surface ionization governs their electrostatic interaction behavior.
See LICENSE fileamphoteric nanoparticlescharge regulationmolecular dynamics
Co-PiLOT combines a generative encoder-decoder with physics-informed constraints to support inverse design of molecules, devices, and microstructures against expensive black-box simulators.
See LICENSE filephysics-informed machine learninginverse designlatent space optimization
A self-driving laboratory strategy that simultaneously and dynamically assesses composition-dependent synthetic accessibility and target properties via aggregated cost functions to accelerate optimization of multicomponent NASICON fast ionic conductors while maintaining phase purity.
See LICENSE fileNASICONfast ionic conductorself-driving laboratory
CP-Agent is an agentic harness for crystal plasticity simulation workflows in polycrystalline metals, automating heterogeneous tool configuration, multi-step data pipeline orchestration, and constitutive parameter calibration against experiments.
See LICENSE filecrystal plasticitypolycrystalline metalssimulation workflow
This materials-science study presents an energetically driven structure-matching framework for analyzing ongoing total X-ray scattering experiments, using gold nanoparticle data to compare idealized spherical, octahedral, decahedral, and icosahedral structures.
See LICENSE fileTotal X-ray scatteringNanoparticle structureAutonomous laboratories
An SE(3)-equivariant denoising diffusion model learns the configurational distribution of amorphous oxides and generates diverse structural ensembles beyond conventional DFT and machine-learned-potential simulations.
See LICENSE fileamorphous oxidesequivariant diffusion modelmaterials generation
MARC-TS introduces a two-stage framework that learns endpoint-conditioned continuous reaction paths, queries at arbitrary resolution, and refines transition-state candidates via local path context, with T1x-IRC-8K dataset of 8,209 reactions and 1.09M path-resolved geometries.
See LICENSE filereaction pathtransition statechemical reaction
Presents a 3D multiphysics FEM framework for high-frequency induction tube welding, coupling time-harmonic magnetic scalar potential with quasi-stationary heat transport for coupled electromagnetic-thermal simulation.
See LICENSE fileinduction tube weldingmultiphysics simulationfinite element method
Neural networks trained on 4DSTEM diffraction patterns enable rapid identification of LiNiO2 and NiO material phases, providing a computationally efficient approach for analyzing large sets of materials image data.
See LICENSE filePhase Prediction4DSTEM DiffractionNeural Networks
An SO(3)-equivariant neural network predicts the full 6×6 grand resistance matrix for microparticles in Stokes flow, outperforming traditional scalar drag correlations by preserving orientation dependence and rotational response.
See LICENSE fileStokes flowresistance tensorequivariant neural network
FracGen generates controllable fracture dynamics from a single image of an intact object using fracture-related physical signals. Its training data comes from FracSim, a fracture simulation framework combining material point dynamics with a continuum damage model.
See LICENSE filefracture mechanicsphysics-informed generationmaterial failure simulation
This materials-science study examines how hydrogen affects phase transformation and deformation twinning in body-centered-cubic iron using a density-functional-theory-trained deep-neural-network interatomic potential and large-scale molecular dynamics. It reports that hydrogen lowers the yield stress.
See LICENSE filehydrogen effectsdeformation twinningplasticity of iron
The paper applies knowledge-driven multi-view retrieval and explanation to identify crystalline phases from X-ray diffraction patterns, addressing reference-pattern scarcity in high-throughput materials discovery.
See LICENSE fileX-ray diffractionphase identificationcrystalline materials
A region-held-out evaluation protocol for hydrogen embrittlement detection in 316L stainless steel, preventing data leakage from same specimen regions, comparing texture features vs deep learning for SEM micrograph classification.
See LICENSE fileHydrogen EmbrittlementStainless SteelSEM Imaging
For materials discovery, this work separates a crystal structure prediction physical prior from adaptive search in composition space to support de novo crystal generation.
See LICENSE filecrystal generationcrystal structure predictioncomposition space search
This paper studies the structural, thermophysical, and dynamical properties of MgCl2 molten salt over a wide temperature range using AIMD and molecular dynamics simulations driven by machine-learning interaction potentials. The MLIP approach is reported to achieve near-AIMD accuracy at a fraction of the computational cost, enabling larger systems and longer timescales.
See LICENSE fileMgCl2 molten saltMolecular dynamicsMachine-learning interaction potentials
Using leakage-free simulations across three temperatures, the study shows that mobility-network topology adds incremental out-of-fold predictive information beyond local structure in aging glass-forming liquids and remains highly complementary to softness-based representations.
See LICENSE fileGlass-forming liquidMobility networkAging dynamics
A physics-informed computational screening framework that decomposes effective second-harmonic response to optimize nonlinear optical properties of 2D semiconductor/organic molecule interfaces.
See LICENSE filesecond-harmonic generation2D materialsinterface engineering
This paper applies physics-informed neural networks to reconstruct molten pool flow fields from high-speed video for wire-arc directed energy deposition, where extreme temperatures and intense arc illumination make direct flow sensing difficult.
See LICENSE fileMolten Pool FlowHigh-Speed Video ReconstructionPhysics-Informed Neural Networks
This paper develops reduced-order modeling for mechanical metamaterials, addressing buckling and pattern transformation while accounting for local, nonlocal, and patterning fluctuation effects.
See LICENSE filemechanical metamaterialsreduced-order modelingcomputational homogenization
This materials-science paper applies reinforcement learning to the discrete composition channel of a crystal generator for property-targeted inverse design, while examining validated gains and reward hacking.
See LICENSE fileinverse materials designcrystal generationreinforcement learning
Proposes SO(3) Equivariant Neural Kalman Networks (SENK) with response-state cascade to improve transferable vibrational spectral prediction across chemical spaces using equivariant transformer backbone.
See LICENSE filevibrational spectroscopymolecular property predictionequivariant neural networks
Wave physics-informed regression enables spatially resolved characterization of heterogeneous plates with varying thickness, damage, or material properties for structural health monitoring.
See LICENSE filestructural health monitoringplate thickness mappingwave physics-informed regression
The paper presents a structure-preserving Fourier Neural Operator for long-time Cahn-Hilliard phase-field dynamics using coarse temporal supervision, targeting efficient surrogate prediction during computationally intensive late-stage coarsening while addressing error accumulation in repeated time stepping.
See LICENSE fileCahn-Hilliard dynamicsFourier neural operatorphase-field simulation
Investigates the relationship between residual scale and learnability in Δ-learning, demonstrating through molecular graph neural network evaluation on total energy targets that smaller residual targets do not necessarily yield better learning outcomes.
See LICENSE fileDelta learningresidual learningmolecular graph neural network
The paper generalizes fixed-lattice cluster expansions and atomic cluster expansions by representing both rotation and permutation symmetries, supporting equivariant representations for atomistic materials modeling.
See LICENSE filecluster expansionatomic interactionsrotation equivariance
The paper introduces topo-spectral percolation descriptors derived from static crystal structures to identify ion diffusion pathways and bottleneck atoms, supporting mechanistic analysis of ion transport in solid materials.
See LICENSE fileion transportcrystal structuressolid electrolytes
The paper introduces U(1) gauge-equivariant representation learning for mean-field Bloch states, explicitly addressing the gauge redundancy of electronic states in first-principles condensed-matter systems.
See LICENSE fileBloch state representationGauge-equivariant learningCondensed-matter electronic structure
CrystAF studies where physics should enter all-atom molecular crystal generation by comparing training, post-training, and inference strategies with the generator and physical signal held fixed.
See LICENSE fileMolecular crystalsGenerative modelingInteratomic potential