Searcharxiv⌕ Search

arXiv subjects

Victor Fung

Publications and source records attributed to Victor Fung.

At least 19 recordsLinked to original sources

Synthetic Crystal Data Generation for Diffusion Pretraining in Material Property Prediction

Diffusion-based pretraining has emerged as a powerful strategy for learning transferable representations in graph neural networks for materials property prediction without requiring quantum-mechanical labels such as energies, forces, and stresses for every pretraining structure. However, the method retains a critical dependency: it requires structures residing near local minima of a potential energy surface, which have traditionally been obtained through costly DFT relaxations of known or physically synthesized compounds. This reliance constrains both the scale and the structural diversity of available pretraining data, leaving large regions of crystal structure space unexplored. In this work, we connect diffusion-based pretraining with synthetic structure generation to reduce this bottleneck. We introduce a synthetic data generation pipeline combining three complementary strategies: perturb-and-relax, generative sampling via MatterGen, and protostructure enumeration. Where structural relaxation is required, we use fast machine-learned interatomic potentials instead of performing a new DFT relaxation for each generated structure. We further compare a coordinate-denoising control with continuous-time variance-preserving (VP) and sub-VP diffusion objectives using a randomly initialized Orb-v2 backbone. The source comparison covers seven MatBench property-prediction tasks, and an eighth task, \texttt{mp\_gap}, is included in the protostructure scaling study. Every synthetic-pretraining condition numerically outperforms random initialization on all seven source-comparison tasks. Scaling the protostructure corpus from 20k to 165k under a fixed 400-epoch schedule numerically matches or exceeds MP-20k pretraining on most tasks; because this protocol increases both data diversity and optimization updates, the result reflects joint data-and-compute scaling.

cs.LG↗

LATHE: LAnguage-driven Toolkit for Hypothesis-based crystal Editing

Manipulating crystal structures towards targeted geometric properties or symmetry constraints is a longstanding challenge. Existing computational frameworks fall short: LLMs have poor spatial awareness and operate at coarse granularity in the token space, while diffusion- and gradient-based structure generation approaches generally produce complete structures \textit{de novo} rather allowing for fine-grained editing of a given input under arbitrary constraints. We present LATHE, a geometric toolkit that closes this gap by expressing seven classes of crystallographic properties --- bond length, bond angle, dihedral angle, coordination environment, lattice parameters, cell volume, and space group, as differentiable objectives to enable direct geometric editing of crystal structures. To demonstrate its usage in materials design, we further expose LATHE through a Model Context Protocol server and embed it in a closed-loop multi-agent system which translates hypotheses intogeometric modifications in natural language towards a given design objective. Across single-property benchmarks, LATHE attains near-perfect constraint satisfaction on all seven property types while keeping optimized structures close to local energy minima. The LATHE-equipped agent translates over 87.5\% of natural-language prompts into valid executable configurations and faithfully completes them. In a band-gap inverse-design case study, the multi-agent loop reaches the target tolerance window in nine of ten independent runs at a typical cost of thirteen hypothesis-evaluation cycles. By bridging natural-language hypothesis generation and physically grounded gradient-based structural editing, this work establishes a paradigm for interpretable, closed-loop computational materials discovery that is immediately applicable to a broad class of functional material design tasks.

cond-mat.mtrl-sci↗

The convergent laboratory: when AI reasoning, autonomous experiments, high performance and quantum computing reshape chemistry

This Comment emerges from TPC26 (https://tpc26.org), a conference convening leaders from academia, national laboratories, and industry who are reshaping materials science discovery. The meeting explored how AI, autonomous agents, self-driving labs, higher performance and quantum computing converge to amplify their individual impact on materials science discovery. The perspectives here reflect the firsthand experiences of researchers at these frontiers and capture the essence of this global endeavor. As AI-driven reasoning, autonomous agentic frameworks, self-driving laboratories, and fault-tolerant quantum processors mature simultaneously, we offer this Comment as a reference at what we believe is a tipping point of transformative advances and productive disruption in the chemical sciences.

cs.AI↗

Symbolic Predicate-Guided Language Agents for Inverse Design of Perovskite Oxides

Efficient discovery of high-performance materials has been pursued through a variety of data- and AI-driven strategies, among which inverse design, generating materials from desired target properties, has emerged as an important paradigm. Large language models (LLMs) offer a complementary route for inverse materials design because their reasoning and in-context learning capability can be used not only to propose candidates but also to demonstrate interpretable design principles. In this work, we introduce a domain specific language (DSL)-guided strategy to improve the reasoning and design capability of LLM agents by translating natural language design rules into symbolic predicates encoded in a predefined chemistry DSL. These predicates allow the LLM agent to obtain statistical evidence from the accumulated materials data, enabling the agent to evaluate and refine its own reasoning during the design loop. Based on this strategy, we developed a multi-agent materials design framework, called Operational Rule-grounded CHEmical Search Through Reasoning Agents (ORCHESTRA), and applied it to the inverse design of double perovskite oxides under multiple target-property objectives. The results show that symbolic predicates help the LLM identify unsupported rules, validate newly proposed rules and improve the rule store over iterative design cycles. Compared with a strategy relying only on natural language rules, the DSL-guided framework showed the potential to improve materials design performance, particularly for challenging target properties. These findings suggest that mathematical and statistical grounding can enhance the reasoning capability of LLM agents in materials science and that LLM-based inverse design can be performed effectively without large task-specific datasets or additional model training.

cond-mat.mtrl-sci↗

Ligand-Controlled Phonon Dynamics in CsPbBr3 Nanocrystals Revealed by Machine-Learned Interatomic Potentials

Halide perovskite nanocrystals are leading candidates for next-generation optoelectronics, yet the role of surface ligands in controlling their phonon dynamics remains poorly understood. These dynamics critically govern nonradiative relaxation, energy up-conversion, and phonon-assisted anti-Stokes emission. Conventional ab initio methods, while accurate, are computationally infeasible for experimentally relevant nanocrystal sizes that require thousands of atoms to capture realistic ligand shells and dynamic disorder at finite temperatures. Here, we introduce a machine- learned interatomic potential fine-tuned on small CsPbBr3 nanocrystals with diverse ligands, enabling accurate prediction of ligand-induced phonon properties far beyond the spatial and temporal scales of ab initio methods. We find that both cationic and anionic ligands systematically redshift Pb-Br-Pb stretching modes while blueshifting the PbBr64- octahedral rotation mode, with stronger overall effects for anionic passivation. Notably, anionic ligands stiffen the rotation mode non-monotonically with respect to the ligand binding energy. Our findings reveal important roles of cationic and anionic ligands in modulating key dynamic modes of halide perovskite nanocrystals associated with detrimental nonradiative losses, offering mechanistic insights and design principles for high-performance perovskite nanocrystal optoelectronics.

cond-mat.mtrl-sci↗

Reasoning-Driven Design of Single Atom Catalysts via a Multi-Agent Large Language Model Framework

Large language models (LLMs) are becoming increasingly applied beyond natural language processing, demonstrating strong capabilities in complex scientific tasks that traditionally require human expertise. This progress has extended into materials discovery, where LLMs introduce a new paradigm by leveraging reasoning and in-context learning, capabilities absent from conventional machine learning approaches. Here, we present a Multi-Agent-based Electrocatalyst Search Through Reasoning and Optimization (MAESTRO) framework in which multiple LLMs with specialized roles collaboratively discover high-performance single atom catalysts for the oxygen reduction reaction. Within an autonomous design loop, agents iteratively reason, propose modifications, reflect on results and accumulate design history. Through in-context learning enabled by this iterative process, MAESTRO identified design principles not explicitly encoded in the LLMs' background knowledge and successfully discovered catalysts that break conventional scaling relations between reaction intermediates. These results highlight the potential of multi-agent LLM frameworks as a powerful strategy to generate chemical insight and discover promising catalysts.

cond-mat.mtrl-sci↗

Determining Atomic Structure from Spectroscopy via an Active Learning Framework

Determining atomic structure from spectroscopic data is central to materials science but remains restricted to a limited set of techniques and material classes, largely due to the computational cost and complexity of structural refinement. Here we introduce ActiveStructOpt, a general framework that integrates graph neural network surrogate models with active learning to efficiently determine candidate structures that reproduce target spectra with minimal computational expenditure. Benchmarking with X-ray pair distribution function data, and with the more computationally demanding simulations of X-ray absorption near-edge spectra (XANES) and extended X-ray absorption fine structure (EXAFS), demonstrate that ActiveStructOpt reliably determines structures that match closely in spectra across diverse materials classes. Under equivalent computational budgets, ActiveStructOpt outperforms existing structure determination methods. By enabling data-efficient, multi-objective structural refinement across a broad range of computable spectroscopic techniques, ActiveStructOpt provides a flexible and extensible approach to atomic structure determination in complex materials.

cond-mat.mtrl-sci↗

Improving Reliability of Machine Learned Interatomic Potentials With Physics-Informed Pretraining

Machine learned interatomic potentials (MLIPs) have emerged as powerful tools for molecular dynamics (MD) simulations with their competitive accuracy and computational efficiency. However, MLIPs are often observed to exhibit un-physical behavior when encountering configurations which deviate significantly from their training data distribution, leading to simulation instabilities and unreliable dynamics, thus limiting the reliability of MLIPs for materials simulations. We present a physics-informed pretraining strategy that leverages simple physical potentials which can improve the robustness and stability of graph-based MLIPs for MD simulations. We demonstrate this approach by deploying a pretraining-finetuning pipeline where MLIPs are initially pretrained on data labelled with embedded atom model potentials and subsequently finetuned on the quantum mechanical ground truth data. By evaluating across three diverse material systems (phosphorus, silica, and a subset of Materials Project) and three representative MLIP architectures (CGCNN, M3GNet, and TorchMD-NET), we find that this physics-informed pretraining consistently improves both prediction accuracy as well as stability in MD compared to the baselines.

cond-mat.mtrl-sci↗

RoFt-Mol: Benchmarking Robust Fine-Tuning with Molecular Graph Foundation Models

In the era of foundation models, fine-tuning pre-trained models for specific downstream tasks has become crucial. This drives the need for robust fine-tuning methods to address challenges such as model overfitting and sparse labeling. Molecular graph foundation models (MGFMs) face unique difficulties that complicate fine-tuning. These models are limited by smaller pre-training datasets and more severe data scarcity for downstream tasks, both of which require enhanced model generalization. Moreover, MGFMs must accommodate diverse objectives, including both regression and classification tasks. To better understand and improve fine-tuning techniques under these conditions, we classify eight fine-tuning methods into three mechanisms: weight-based, representation-based, and partial fine-tuning. We benchmark these methods on downstream regression and classification tasks across supervised and self-supervised pre-trained models in diverse labeling settings. This extensive evaluation provides valuable insights and informs the design of a refined robust fine-tuning method, ROFT-MOL. This approach combines the strengths of simple post-hoc weight interpolation with more complex weight ensemble fine-tuning methods, delivering improved performance across both task types while maintaining the ease of use inherent in post-hoc weight interpolation.

cs.LG↗

Scalable Foundation Interatomic Potentials via Message-Passing Pruning and Graph Partitioning

Atomistic foundation models (AFMs) have great promise as accurate interatomic potentials, and have enabled data-efficient molecular dynamics simulations with near quantum mechanical accuracy. However, AFMs remain markedly slower at inference and are far more memory-intensive than conventional interatomic potentials, due to the need to capture a wide range of chemical and structural motifs in pre-training datasets requiring deep, parameter-rich model architectures. These deficiencies currently limit the practical use of AFMs in molecular dynamics (MD) simulations at extended temporal and spatial scales. To address this problem, we propose a general workflow for accelerating and scaling AFMs containing message-passing architectures. We find that removing low-contribution message-passing layers from AFM backbones serves as an effective pruning method, significantly reducing the parameter count while preserving the accuracy and data-efficiency of AFMs. Once pruned, these models become more accessible for large scale simulations via a graph-partitioned, GPU-distributed strategy, which we implement and demonstrate within the AFM fine-tuning platform MatterTune. We show that this approach supports million-atom simulations on both single and multiple GPUs, and enables task-specific large-scale simulations at nanosecond timescales with AFM-level accuracy.

cond-mat.mtrl-sci↗

Facet: highly efficient E(3)-equivariant networks for interatomic potentials

Computational materials discovery is limited by the high cost of first-principles calculations. Machine learning (ML) potentials that predict energies from crystal structures are promising, but existing methods face computational bottlenecks. Steerable graph neural networks (GNNs) encode geometry with spherical harmonics, respecting atomic symmetries -- permutation, rotation, and translation -- for physically realistic predictions. Yet maintaining equivariance is difficult: activation functions must be modified, and each layer must handle multiple data types for different harmonic orders. We present Facet, a GNN architecture for efficient ML potentials, developed through systematic analysis of steerable GNNs. Our innovations include replacing expensive multi-layer perceptrons (MLPs) for interatomic distances with splines, which match performance while cutting computational and memory demands. We also introduce a general-purpose equivariant layer that mixes node information via spherical grid projection followed by standard MLPs -- faster than tensor products and more expressive than linear or gate layers. On the MPTrj dataset, Facet matches leading models with far fewer parameters and under 10% of their training compute. On a crystal relaxation task, it runs twice as fast as MACE models. We further show SevenNet-0's parameters can be reduced by over 25% with no accuracy loss. These techniques enable more than 10x faster training of large-scale foundation models for ML potentials, potentially reshaping computational materials discovery.

cond-mat.mtrl-sci↗

A Comprehensive Assessment and Benchmark Study of Large Atomistic Foundation Models for Phonons

The rapid development of universal machine learning potentials (uMLPs) has enabled efficient, accurate predictions of diverse material properties across broad chemical spaces. While their capability for modeling phonon properties is emerging, systematic benchmarking across chemically diverse systems remains limited. We evaluate six recent uMLPs (EquiformerV2, MatterSim, MACE, and CHGNet) on 2,429 crystalline materials from the Open Quantum Materials Database. Models were used to compute atomic forces in displaced supercells, derive interatomic force constants (IFCs), and predict phonon properties including lattice thermal conductivity (LTC), compared with density functional theory (DFT) and experimental data. The EquiformerV2 pretrained model trained on the OMat24 dataset exhibits strong performance in predicting atomic forces and third-order IFC, while its fine-tuned counterpart consistently outperforms other models in predicting second-order IFC, LTC, and other phonon properties. Although MACE and CHGNet demonstrated comparable force prediction accuracy to EquiformerV2, notable discrepancies in IFC fitting led to poor LTC predictions. Conversely, MatterSim, despite lower force accuracy, achieved intermediate IFC predictions, suggesting error cancellation and complex relationships between force accuracy and phonon predictions. This benchmark guides the evaluation and selection of uMLPs for high-throughput screening of materials with targeted thermal transport properties.

cond-mat.mtrl-sci↗

MatterTune: An Integrated, User-Friendly Platform for Fine-Tuning Atomistic Foundation Models to Accelerate Materials Simulation and Discovery

Geometric machine learning models such as graph neural networks have achieved remarkable success in recent years in chemical and materials science research for applications such as high-throughput virtual screening and atomistic simulations. The success of these models can be attributed to their ability to effectively learn latent representations of atomic structures directly from the training data. Conversely, this also results in high data requirements for these models, hindering their application to problems which are data sparse which are common in this domain. To address this limitation, there is a growing development in the area of pre-trained machine learning models which have learned general, fundamental, geometric relationships in atomistic data, and which can then be fine-tuned to much smaller application-specific datasets. In particular, models which are pre-trained on diverse, large-scale atomistic datasets have shown impressive generalizability and flexibility to downstream applications, and are increasingly referred to as atomistic foundation models. To leverage the untapped potential of these foundation models, we introduce MatterTune, a modular and extensible framework that provides advanced fine-tuning capabilities and seamless integration of atomistic foundation models into downstream materials informatics and simulation workflows, thereby lowering the barriers to adoption and facilitating diverse applications in materials science. In its current state, MatterTune supports a number of state-of-the-art foundation models such as ORB, MatterSim, JMP, and EquformerV2, and hosts a wide range of features including a modular and flexible design, distributed and customizable fine-tuning, broad support for downstream informatics tasks, and more.

cond-mat.mtrl-sci↗

Electronic Structure Guided Inverse Design Using Generative Models

The electronic structure of a material fundamentally determines its underlying physical, and by extension, its functional properties. Consequently, the ability to identify or generate materials with desired electronic properties would enable the design of tailored functional materials. Traditional approaches relying on human intuition or exhaustive computational screening of known materials remain inefficient and resource-prohibitive for this task. Here, we introduce DOSMatGen, the first instance of a machine learning method which generates crystal structures that match a given desired electronic density of states. DOSMatGen is an E(3)-equivariant joint diffusion framework, and utilizes classifier-free guidance to accurately condition the generated materials on the density of states. Our experiments find this approach can successfully yield materials which are both stable and match closely with the desired density of states. Furthermore, this method is highly flexible and allows for finely controlled generation which can target specific templates or even individual sites within a material. This method enables a more physics-driven approach to designing new materials for applications including catalysts, photovoltaics, and superconductors.

cond-mat.mtrl-sci↗

Pre-training Graph Neural Networks with Structural Fingerprints for Materials Discovery

In recent years, pre-trained graph neural networks (GNNs) have been developed as general models which can be effectively fine-tuned for various potential downstream tasks in materials science, and have shown significant improvements in accuracy and data efficiency. The most widely used pre-training methods currently involve either supervised training to fit a general force field or self-supervised training by denoising atomic structures equilibrium. Both methods require datasets generated from quantum mechanical calculations, which quickly become intractable when scaling to larger datasets. Here we propose a novel pre-training objective which instead uses cheaply-computed structural fingerprints as targets while maintaining comparable performance across a range of different structural descriptors. Our experiments show this approach can act as a general strategy for pre-training GNNs with application towards large scale foundational models for atomistic data.

cond-mat.mtrl-sci↗

GeSS: Benchmarking Geometric Deep Learning under Scientific Applications with Distribution Shifts

Geometric deep learning (GDL) has gained significant attention in scientific fields, for its proficiency in modeling data with intricate geometric structures. However, very few works have delved into its capability of tackling the distribution shift problem, a prevalent challenge in many applications. To bridge this gap, we propose GeSS, a comprehensive benchmark designed for evaluating the performance of GDL models in scientific scenarios with distribution shifts. Our evaluation datasets cover diverse scientific domains from particle physics, materials science to biochemistry, and encapsulate a broad spectrum of distribution shifts including conditional, covariate, and concept shifts. Furthermore, we study three levels of information access from the out-of-distribution (OOD) test data, including no OOD information, only unlabeled OOD data, and OOD data with a few labels. Overall, our benchmark results in 30 different experiment settings, and evaluates 3 GDL backbones and 11 learning algorithms in each setting. A thorough analysis of the evaluation results is provided, poised to illuminate insights for GDL researchers and domain practitioners who are to use GDL in their applications.

cs.LG↗

Representation-space diffusion models for generating periodic materials

Generative models hold the promise of significantly expediting the materials design process when compared to traditional human-guided or rule-based methodologies. However, effectively generating high-quality periodic structures of materials on limited but diverse datasets remains an ongoing challenge. Here we propose a novel approach for periodic structure generation which fully respect the intrinsic symmetries, periodicity, and invariances of the structure space. Namely, we utilize differentiable, physics-based, structural descriptors which can describe periodic systems and satisfy the necessary invariances, in conjunction with a denoising diffusion model which generates new materials within this descriptor or representation space. Reconstruction is then performed on these representations using gradient-based optimization to recover the corresponding Cartesian positions of the crystal structure. This approach differs significantly from current methods by generating materials in the representation space, rather than in the Cartesian space, which is made possible using an efficient reconstruction algorithm. Consequently, known issues with respecting periodic boundaries and translational and rotational invariances during generation can be avoided, and the model training process can be greatly simplified. We show this approach is able to provide competitive performance on established benchmarks compared to current state-of-the-art methods.

cond-mat.mtrl-sci↗

LLMatDesign: Autonomous Materials Discovery with Large Language Models

Discovering new materials can have significant scientific and technological implications but remains a challenging problem today due to the enormity of the chemical space. Recent advances in machine learning have enabled data-driven methods to rapidly screen or generate promising materials, but these methods still depend heavily on very large quantities of training data and often lack the flexibility and chemical understanding often desired in materials discovery. We introduce LLMatDesign, a novel language-based framework for interpretable materials design powered by large language models (LLMs). LLMatDesign utilizes LLM agents to translate human instructions, apply modifications to materials, and evaluate outcomes using provided tools. By incorporating self-reflection on its previous decisions, LLMatDesign adapts rapidly to new tasks and conditions in a zero-shot manner. A systematic evaluation of LLMatDesign on several materials design tasks, in silico, validates LLMatDesign's effectiveness in developing new materials with user-defined target properties in the small data regime. Our framework demonstrates the remarkable potential of autonomous LLM-guided materials discovery in the computational setting and towards self-driving laboratories in the future.

cond-mat.mtrl-sci↗