The efficacy of mathematical models heavily depends on the quality of the training data, yet collecting sufficient data is often expensive and challenging. Many modeling applications require inferring parameters only as a means to predict other quantities of interest (QoI). Because models often contain many unidentifiable (sloppy) parameters, QoIs often depend on a relatively small number of parameter combinations. Therefore, we introduce an information-matching criterion based on the Fisher information matrix to select the most informative training data from a candidate pool. This method ensures that the selected data contain sufficient information to learn only those parameters that are needed to constrain downstream QoIs. It is formulated as a convex optimization problem, making it scalable to large models and datasets. We demonstrate the effectiveness of this approach across various modeling problems in diverse scientific fields, including power systems and underwater acoustics. Finally, we use information-matching as a query function within an active learning (AL) loop for materials science applications. In all these applications, we find that a relatively small set of optimal training data can provide the necessary information for achieving precise predictions. These results are encouraging for diverse future applications, particularly AL in large machine-learning models.

Molecule generation is advancing rapidly in chemical discovery and drug design. Flow-matching methods have recently set the state of the art (SOTA) in unconditional molecule generation, surpassing score-based diffusion models. However, diffusion models still lead in property-guided generation. In this work, we introduce PropMolFlow, an approach for property-guided molecule generation based on geometry-complete SE(3)-equivariant flow matching. Integrating five different property embedding methods with a Gaussian expansion of scalar properties, PropMolFlow achieves competitive performance against previous SOTA diffusion models in conditional molecule generation while maintaining high structural stability and validity. Additionally, it enables higher sampling speed with fewer time steps compared with baseline models. We highlight the importance of validating the properties of generated molecules through density functional theory calculations. Furthermore, we introduce a task to assess the model’s ability to propose molecules with under-represented property values, assessing its capacity for out-of-distribution generalization. PropMolFlow is a flow-matching method for property-guided molecule generation that matches diffusion model performance while generating stable, valid structures more quickly and enabling the discovery of molecules with under-represented property values.

Abstract This paper describes a project to improve the ability to predict SMA device failure by using an experimentally validated high-performance 3D implementation of the multiscale quasicontinuum (QC) method. The QC method leverages state-of-the-art machine-learning interatomic potentials to accurately represent failure mechanisms in atomistic regions.

The computation of atomistic stress fields plays a central role in linking atomistic simulations to continuum models. However, commonly used stress measures, such as the per-atom virial stress, do not constitute true spatial stress fields consistent with the continuum balance laws. Moreover, when constructing continuum-consistent atomistic stress fields, fundamental non-uniqueness arises from the decomposition of atomic forces into interatomic contributions. While physically motivated force decompositions are often available for conventional physics-based interatomic potentials, these approaches become problematic for modern machine-learning potentials that are formulated in terms of high-dimensional, data-driven descriptors rather than explicit interatomic distances. In this work, we develop a computationally efficient projection-based force decomposition framework that enables the construction of smooth and continuous atomistic stress fields for both conventional and machine-learning interatomic potentials. We show that the proposed method avoids reliance on distance-based chain-rule formulations and remains robust to model representation choices. Validation across a variety of interatomic potentials for diamond-cubic silicon demonstrates that the resulting stress fields exhibit significantly reduced noise. Our method establishes a consistent and reliable basis for stress field computation and interpretation in simulations employing machine-learning interatomic potentials.

Interatomic potentials (IPs) enable large-scale atomistic simulations beyond the reach of first-principles methods, but their predictive reliability depends critically on the selection of training data, quantified uncertainty, and model expressiveness. Active learning (AL) provides a principled framework for constructing efficient and accurate IPs, yet most strategies reduce parameter uncertainty without explicitly accounting for the specific material properties being predicted. The information-matching (IM) approach addresses this limitation by requiring that the selected training data provide at least as much parameter space information as needed to achieve prescribed uncertainty targets for selected quantities of interest (QoIs). Here, we apply IM to develop bespoke IPs specifically tailored for predicting plastic strength in metals. Due to the high computational cost of simulating plastic strength, we employ an indirect IM strategy that targets inexpensive intermediate QoIs that correlate with strength. The IM method enables precise parameter constraints with minimal training data, yielding precise predictions for both the intermediate QoIs and plastic strength. Yet, model error remains a key limitation, and a post hoc uncertainty inflation correction provides a viable means to mitigate this limitation. These findings illustrate both the promise and limits of uncertainty-aware AL for predicting complex material properties.

We derive and validate a partition function for low-dimensional systems interacting with a heat bath, addressing the general issue of thermodynamic modeling of nanoscale systems. In contrast to bulk systems in the canonical (NVT) ensemble, where the partition function is solely determined by the Hamiltonian of the system and the temperature of the heat bath, our formulation demonstrates that accounting for the interactions with the heat bath is essential for describing the statistical mechanics of low-dimensional materials. To validate our theoretical findings, we develop a molecular dynamics (MD) algorithm for directly modeling the heat bath as a gas reservoir. We first validate our approach using a 1D harmonic oscillator, calculating its length distribution through explicit numerical integration and confirming these results with MD simulations. We then apply the method to an Au nanoparticle, showing that the reservoir produces pressure-dependent changes in the equilibrium effective radius relative to conventional NVT simulations. Finally, we extend our method to investigate the out-of-plane fluctuations of a 2D graphene monolayer immersed in a gas at finite temperature and pressure. Comparisons with conventional NVT ensemble simulations controlled by a thermostat reveal that environmental interactions significantly influence the properties of the 2D material system.

The inverse design of materials with specific desired properties, such as high-temperature superconductivity, represents a formidable challenge in materials science due to the vastness of chemical and structural space. We present a guided diffusion framework to accelerate the discovery of novel superconductors. A DiffCSP foundation model is pretrained on the Alexandria Database and fine-tuned on 7183 superconductors with first-principles-derived labels. Employing classifier-free guidance, we sample 200,000 structures, which lead to 34,027 unique candidates. A multistage screening process that combines machine learning and density functional theory (DFT) calculations to assess stability and electronic properties, identifies 773 candidates with DFT-calculated Tc > 5 K. Notably, our generative model demonstrates effective property-driven design. Our computational findings were validated against experimental synthesis and characterization performed as part of this work, which highlighted challenges in sparsely charted chemistries and disorder. This end-to-end workflow accelerates superconductor discovery while underscoring the challenge of predicting and synthesizing experimentally realizable materials.

The J-integral is a fundamental concept in fracture mechanics, quantifying the energy release rate that drives crack propagation. While extensively implemented in finite element (FE) codes and adapted for atomistic calculations, its application within multiscale frameworks bridging atomistic and continuum formulations remains unexplored. This work presents a rigorous implementation and validation of the J-integral within the three-dimensional quasicontinuum (QC3D) method, computed under plane strain assumptions, using continuum fields (stress, strain energy density) derived from the interatomic potential via the Cauchy-Born rule. The implementation is validated against linear elastic fracture mechanics (LEFM) theory and the virtual crack extension (VCE) method across three regimes: (1) small-strain linear elasticity, with a prescribed anisotropic K-field displacement applied throughout; (2) the same field evaluated through the nonlinear Cauchy-Born constitutive relation, without atomic relaxation; and (3) the same relation with atomic relaxation enabled, allowing the crack-tip region to equilibrate. Excellent agreement is shown throughout. We further introduce a Markov chain Monte Carlo framework to statistically quantify the uncertainty of J-integral results for a given mesh and integration domain, applicable to conventional FE methods as well. Predictive capability is demonstrated via a QC3D simulation of a three-point bending test of silicon, where the computed critical energy release rate agrees closely with the Griffith criterion. This work establishes a reliable framework for evaluating crack driving forces in multiscale fracture simulations with quantified uncertainty, enabling large-scale fracture simulations while resolving atomistic mechanisms at the crack tip.

Molecular crystal structure prediction represents a grand challenge in computational chemistry due to large sizes of constituent molecules and complex intra- and intermolecular interactions. While generative modeling has revolutionized structure discovery for molecules, inorganic solids, and metal-organic frameworks, extending such approaches to fully periodic molecular crystals is still elusive. Here, we present MolCrystalFlow, a flow-based generative model for molecular crystal structure prediction. The framework disentangles intramolecular complexity from intermolecular packing by embedding molecules as rigid bodies and jointly learning the lattice matrix, molecular orientations, and centroid positions. Centroids and orientations are represented on their native Riemannian manifolds, allowing geodesic flow construction and graph neural network operations that respects geometric symmetries. We benchmark our model against state-of-the-art generative models for large-size periodic crystals and rule-based structure generation methods on two open-source molecular crystal datasets. We demonstrate an integration of MolCrystalFlow model with universal machine learning potential to accelerate molecular crystal structure prediction, paving the way for data-driven generative discovery of molecular crystals.

Key objectives in conditional molecular generation include ensuring chemical validity, aligning generated molecules with target properties, promoting structural diversity, and enabling efficient sampling for discovery. Recent advances in computer vision introduced a range of new guidance strategies for generative models, many of which can be adapted to support these goals. In this work, we integrate state-of-the-art guidance methods─including classifier-free guidance, autoguidance, and model guidance─in a leading molecule generation framework built on an SE(3)-equivariant flow matching process. We propose a hybrid guidance strategy that separately guides continuous and discrete molecular modalities─operating on velocity fields and predicted denoising probabilities, respectively─while jointly optimizing their guidance scales via Bayesian optimization. Our implementation, benchmarked on the QM9 and QMe14S data sets, achieves new state-of-the-art performance in property alignment for de novo molecular generation. The generated molecules also exhibit high structural validity. Furthermore, we systematically compare the strengths and limitations of various guidance methods, offering insights into their broader applicability.

The materials science community seeks to support the FAIR principles for computational simulation research. The MatCore Project was recently launched to address this need, with the goal of developing an overall metadata framework and accompanying guidelines. This paper reports on the MatCore goals and overall progress. Historical background context is provided, including a review of the principles underlying successful core metadata standards. The paper also presents selected MatCore examples and discusses future plans.

Machine learning interatomic potentials (MLIPs) enable atomistic simulations with near first-principles accuracy at substantially reduced computational cost, making them powerful tools for large-scale materials modeling. The accuracy of MLIPs is typically validated on a held-out dataset of ab initio energies and atomic forces. However, accuracy on these small-scale properties does not guarantee reliability for emergent, system-level behavior – precisely the regime where atomistic simulations are most needed, but for which direct validation is often computationally prohibitive. As a practical heuristic, predictive precision – quantified as inverse uncertainty – is commonly used as a proxy for accuracy, but its reliability remains poorly understood, particularly for system-level predictions. In this work, we systematically assess the relationship between predictive precision and accuracy in both in-distribution (ID) and out-of-distribution (OOD) regimes, focusing on ensemble-based uncertainty quantification methods for neural network potentials, including bootstrap, dropout, random initialization, and snapshot ensembles. We use held-out cross-validation for ID assessment and calculate cold curve energies and phonon dispersion relations for OOD testing. These evaluations are performed across various carbon allotropes as representative test systems. We find that uncertainty estimates can behave counterintuitively in OOD settings, often plateauing or even decreasing as predictive errors grow. These results highlight fundamental limitations of current uncertainty quantification approaches and underscore the need for caution when using predictive precision as a stand-in for accuracy in large-scale, extrapolative applications.

A simulation can stand its ground against an experiment only if its prediction uncertainty is known. The unknown accuracy of interatomic potentials (IPs) is a major source of prediction uncertainty, severely limiting the use of large-scale classical atomistic simulations in a wide range of scientific and engineering applications. Here we explore covariance between predictions of metal plasticity, from 178 large-scale (~108 atoms) molecular dynamics (MD) simulations, and a variety of indicator properties computed at small-scales (≤102 atoms). All simulations use the same 178 IPs. In a manner similar to statistical studies in public health, we analyze correlations of strength with indicators, identify the best predictor properties, and build a cross-scale “strength-on-predictors” regression model. This model is then used to estimate regression error over the statistical pool of IPs. Small-scale predictors found to be highly covariant with strength are computed using expensive quantum-accurate calculations and used to predict flow strength, within the statistical error bounds established in our study.

An extension to the effective average-atom (AA) interatomic potential (IP) that accounts for local ordering effects is derived. While the AA approach is only valid for random alloys, the new local-order average-atom (LOAA) IP accounts for short-range order within a material by utilizing information from partial radial distribution functions. Simulations with a LOAA potential require smaller system sizes to achieve statistically converged results and therefore can be used to model complex materials where short-range order effects are important, such as high-entropy alloys, at a fraction of the computational cost of standard IPs. The method is validated for a two-dimensional (2D) binary hexagonal crystal with Lennard-Jones (LJ) interactions, and for three-dimensional (3D) Fe$_{(1-x)/2}$Ni$_{(1-x)/2}$Cr$_x$ and Ni$_{0.67}$Al$_{0.33}$ alloys modeled via embedded-atom method (EAM) potentials. For the 2D crystal we obtain a local ordering phase diagram in terms of the LJ parameters and demonstrate that in all cases the LOAA formulation obtains elastic properties that match the true-species case using standard IPs. The 3D alloy examples further demonstrate the ability of this method to accurately capture other material properties and phase transformations.

Correlations between fundamental microscopic properties computable from first principles, which we term canonical properties, and complex large-scale quantities of interest (QoIs) provide an avenue to predictive materials discovery. We propose that such correlations can be efficiently discovered through simulations utilizing approximate interatomic potentials (IPs), which serve as an ensemble of “synthetic materials.” As a proof of principle we build a regression model relating canonical properties to the symmetric tilt grain boundary (GB) energy curves in face-centered cubic crystals, characterized by the scaling factor in the universal lattice matching model of Runnels et al. (2016), which we take to be our QoI. Our analysis recovers known correlations of GB energy to other properties and discovers new ones. We also demonstrate, using available density functional theory (DFT) GB energy data, that the regression model constructed from IP data is consistent with DFT results, confirming the assumption that the IPs and DFT belong to same statistical pool and thereby validating the approach. Regression models constructed in this fashion can be used to predict large-scale QoIs based on first-principles data and provide a general method for training IPs for QoIs beyond the scope of first-principles calculations.

Generative models for materials, especially inorganic crystals, hold potential to transform the theoretical prediction of novel compounds and structures. Advancement in this field depends critically on robust benchmarks and minimal, information-rich datasets that enable meaningful model evaluation. This paper critically examines common datasets and reported metrics for a crystal structure prediction taskx2014generating the most likely structures given the chemical composition of a material. We focus on three key issues: First, materials datasets should contain unique crystal structures; for example, we show that the widely-utilized carbon-24 dataset only contains ≈40

The discovery of new materials is essential for enabling technological advancements. Computational approaches for predicting novel materials must effectively learn the manifold of stable crystal structures within an infinite design space. We introduce Open Materials Generation (OMatG), a unifying framework for the generative design and discovery of inorganic crystalline materials. OMatG employs stochastic interpolants (SI) to bridge an arbitrary base distribution to the target distribution of inorganic crystals via a broad class of tunable stochastic processes, encompassing both diffusion models and flow matching as special cases. In this work, we adapt the SI framework by integrating an equivariant graph representation of crystal structures and extending it to account for periodic boundary conditions in unit cell representations. Additionally, we couple the SI flow over spatial coordinates and lattice vectors with discrete flow matching for atomic species. We benchmark OMatG's performance on two tasks: Crystal Structure Prediction (CSP) for specified compositions, and de novo generation (DNG) aimed at discovering stable, novel, and unique structures. In our ground-up implementation of OMatG, we refine and extend both CSP and DNG metrics compared to previous works. OMatG establishes a new state of the art in generative modeling for materials discovery, outperforming purely flow-based and diffusion-based implementations. These results underscore the importance of designing flexible deep learning frameworks to accelerate progress in materials science. The OMatG code is available at https://github.com/FERMat-ML/OMatG.

The design of structural and functional materials for specialized applications is experiencing significant growth fueled by rapid advancements in materials synthesis, characterization, and manufacturing, as well as by sophisticated computational materials modeling frameworks that span a wide spectrum of length and time scales in the mesoscale between atomistic and homogenized continuum approaches. This is leading towards a systems-based design methodology that will replace traditional empirical approaches, embracing the principles of the Materials Genome Initiative. However, there are several gaps in this framework as it relates to advanced structural materials development: (1) limited availability and access to high-fidelity experimental and computational datasets, (2) lack of co-design of experiments and simulation aimed at computational model validation, (3) lack of on-demand access to verified and validated codes for simulation and for experimental analyses, and (4) limited opportunities for workforce training and educational outreach. These shortcomings stifle major innovations in structural materials design. This paper describes plans for a community-driven research initiative that addresses current gaps based on best-practice recommendations of leaders in mesoscale modeling, experimentation, and cyberinfrastructure obtained at an NSF-sponsored workshop dedicated to this topic and subsequent discussions. The proposal is to create a hub for Mesoscale Experimentation and Simulation co-Operation (h-MESO)-that will (I) provide curation and sharing of models, data, and codes, (II) foster co-design of experiments for model validation with systematic uncertainty quantification, and (III) provide a platform for education and workforce development. h-MESO will engage experimental and computational experts in mesoscale mechanics and plasticity, along with mathematicians and computer scientists with expertise in algorithms, data science, machine learning, and large-scale cyberinfrastructure initiatives.

Citation intention Classification (CIC) tools classify citations by their intention (e.g., background, motivation) and assist readers in evaluating the contribution of scientific literature. Prior research has shown that pretrained language models (PLMs) such as SciBERT can achieve state-of-the-art performance on CIC benchmarks. PLMs are trained via self-supervision tasks on a large corpus of general text and can quickly adapt to CIC tasks via moderate fine-tuning on the corresponding dataset. Despite their advantages, PLMs can easily overfit small datasets during fine-tuning. In this paper, we propose a multi-task learning (MTL) framework that jointly fine-tunes PLMs on a dataset of primary interest together with multiple auxiliary CIC datasets to take advantage of additional supervision signals. We develop a data-driven task relation learning (TRL) method that controls the contribution of auxiliary datasets to avoid negative transfer and expensive hyper-parameter tuning. We conduct experiments on three CIC datasets and show that fine-tuning with additional datasets can improve the PLMs' generalization performance on the primary dataset. PLMs fine-tuned with our proposed framework outperform the current state-of-the-art models by 7% to 11% on small datasets while aligning with the best-performing model on a large dataset.

A basis is derived from elastic plate theory that quantifies equilibrium and dynamic deformation and electron diffraction patterns of twisted bilayer graphene (TBG). The basis is derived by solving in-plane and out-of-plane normal modes of an unforced parallelogram elastic plate. We show that a combination of only a few basis terms successfully captures the relaxed TBG structure with and without an underlying substrate computed using atomistic simulations. The results are validated by comparison with electron diffraction experiments. A code for extracting the elastic plate basis coefficients from an experimental electron diffraction image accompanies this paper. TBG dynamics is also studied by computing the phonon band structure from atomistic simulations. Low- energy phonons at the P point are examined in terms of the mode shape and frequency. These modes are captured by simple elastic plate models with uniformly distributed springs for interlayer and substrate interactions.