arXiv subjects

Philipp Benner

Publications and source records attributed to Philipp Benner.

10 recordsLinked to original sources

arXiv2026

Representations from Pretrained Machine-Learning Interatomic Potentials as Coarse Coordinates for Material Generation and Evaluation

Generative machine learning is increasingly used for inorganic crystal structure generation. Most models and the corresponding evaluation approaches rely on simple forms of crystal structure representation. In this paper, we showcase the power of atom-averaged features from pretrained Machine-Learning Interatomic Potentials (MLIPs), such as MACE, for such tasks. We first introduce a distance measure that assesses the output of material generative models by capturing both quality and novelty in a single distribution-based evaluation framework. In particular, we introduce the Coarse-Fine Transport Distance (CFTD) using two different featurizers, where the quality component is based on coarse MACE features. We showcase CFTD's versatility in capturing crystal-structure quality while also detecting memorization, and compare it with the recently introduced continuous SUN metrics. We further show that coarse MACE features can be used as guidance for a material generative model.

cs.LG↗

arXiv2026

Parameter-Efficient Fine-Tuning of Machine-Learning Interatomic Potentials for Phonon and Thermal Properties

Machine-learning interatomic potentials are widely used as computationally efficient surrogates for density functional theory in atomistic simulations, enabling large-scale, long-time modeling of materials systems. We investigate how different fine-tuning strategies influence the prediction of harmonic phonon band structures, thermal properties, and the potential energy surface along imaginary phonon modes. We achieve substantial accuracy improvements with minimal additional data, with as few as 10 additional training structures already yielding significant gains. In addition to existing approaches, we introduce Equitrain, a finetuning framework that implements LoRA-based adaptation. Across 53 materials systems, we show that fine-tuned models consistently outperform both the underlying pretrained model and models trained from scratch. Equitrain achieves the best overall performance, and our results demonstrate that fine-tuning enables accurate phonon predictions.

cond-mat.mtrl-sci↗

arXiv2026

A critical assessment of bonding descriptors for predicting materials properties

Most machine learning models for materials science rely on descriptors based on materials compositions and structures, even though the chemical bond has been proven to be a valuable concept for predicting materials properties. Over the years, various theoretical frameworks have been developed to characterize bonding in solid-state materials. However, integrating bonding information from these frameworks into machine learning pipelines at scale has been limited by the lack of a systematically generated and validated database. Recent advances in high-throughput bonding analysis workflows have addressed this issue, and our previously computed Quantum-Chemical Bonding Database for Solid-State Materials was extended to include approximately 13,000 materials. This database is then used to derive a new set of quantum-chemical bonding descriptors. A systematic assessment is performed using statistical significance tests to evaluate how the inclusion of these descriptors influences the performance of machine-learning models that otherwise rely solely on structure- and composition-derived features. Models are built to predict elastic, vibrational, and thermodynamic properties typically associated with chemical bonding in materials. The results demonstrate that incorporating quantum-chemical bonding descriptors not only improves predictive performance but also helps identify intuitive expressions for properties such as the projected force constant and lattice thermal conductivity via symbolic regression.

cond-mat.mtrl-sci↗

arXiv2025

Transport Novelty Distance: A Distributional Metric for Evaluating Material Generative Models

Recent advances in generative machine learning have opened new possibilities for the discovery and design of novel materials. However, as these models become more sophisticated, the need for rigorous and meaningful evaluation metrics has grown. Existing evaluation approaches often fail to capture both the quality and novelty of generated structures, limiting our ability to assess true generative performance. In this paper, we introduce the Transport Novelty Distance (TNovD) to judge generative models used for materials discovery jointly by the quality and novelty of the generated materials. Based on ideas from Optimal Transport theory, TNovD uses a coupling between the features of the training and generated sets, which is refined into a quality and memorization regime by a threshold. The features are generated from crystal structures using a graph neural network that is trained to distinguish between materials, their augmented counterparts, and differently sized supercells using contrastive learning. We evaluate our proposed metric on typical toy experiments relevant for crystal structure prediction, including memorization, noise injection and lattice deformations. Additionally, we validate the TNovD on the MP20 validation set and the WBM substitution dataset, demonstrating that it is capable of detecting both memorized and low-quality material data. We also benchmark the performance of several popular material generative models. While introduced for materials, our TNovD framework is domain-agnostic and can be adapted for other areas, such as images and molecules.

cond-mat.mtrl-sci↗

arXiv2024

SynCoTrain: A Dual Classifier PU-learning Framework for Synthesizability Prediction

Material discovery is a cornerstone of modern science, driving advancements in diverse disciplines from biomedical technology to climate solutions. Predicting synthesizability, a critical factor in realizing novel materials, remains a complex challenge due to the limitations of traditional heuristics and thermodynamic proxies. While stability metrics such as formation energy offer partial insights, they fail to account for kinetic factors and technological constraints that influence synthesis outcomes. These challenges are further compounded by the scarcity of negative data, as failed synthesis attempts are often unpublished or context-specific. We present SynCoTrain, a semi-supervised machine learning model designed to predict the synthesizability of materials. SynCoTrain employs a co-training framework leveraging two complementary graph convolutional neural networks: SchNet and ALIGNN. By iteratively exchanging predictions between classifiers, SynCoTrain mitigates model bias and enhances generalizability. Our approach uses Positive and Unlabeled (PU) Learning to address the absence of explicit negative data, iteratively refining predictions through collaborative learning. The model demonstrates robust performance, achieving high recall on internal and leave-out test sets. By focusing on oxide crystals, a well-characterized material family with extensive experimental data, we establish SynCoTrain as a reliable tool for predicting synthesizability while balancing dataset variability and computational efficiency. This work highlights the potential of co-training to advance high-throughput materials discovery and generative research, offering a scalable solution to the challenge of synthesizability prediction.

cond-mat.mtrl-sci↗

arXiv2023

A foundation model for atomistic materials chemistry

Atomistic simulations of matter, especially those that leverage first-principles (ab initio) electronic structure theory, provide a microscopic view of the world, underpinning much of our understanding of chemistry and materials science. Over the last decade or so, machine-learned force fields have transformed atomistic modeling by enabling simulations of ab initio quality over unprecedented time and length scales. However, early ML force fields have largely been limited by: (i) the substantial computational and human effort of developing and validating potentials for each particular system of interest; and (ii) a general lack of transferability from one chemical system to the next. Here we show that it is possible to create a general-purpose atomistic ML model, trained on a public dataset of moderate size, that is capable of running stable molecular dynamics for a wide range of molecules and materials. We demonstrate the power of the MACE-MP-0 model - and its qualitative and at times quantitative accuracy - on a diverse set of problems in the physical sciences, including properties of solids, liquids, gases, chemical reactions, interfaces and even the dynamics of a small protein. The model can be applied out of the box as a starting or "foundation" model for any atomistic system of interest and, when desired, can be fine-tuned on just a handful of application-specific data points to reach ab initio accuracy. Establishing that a stable force-field model can cover almost all materials changes atomistic modeling in a fundamental way: experienced users get reliable results much faster, and beginners face a lower barrier to entry. Foundation models thus represent a step towards democratising the revolution in atomic-scale modeling that has been brought about by ML force fields.

physics.chem-ph↗

arXiv2023

Matbench Discovery -- A framework to evaluate machine learning crystal stability predictions

The rapid adoption of machine learning (ML) in domain sciences necessitates best practices and standardized benchmarking for performance evaluation. We present Matbench Discovery, an evaluation framework for ML energy models, applied as pre-filters for high-throughput searches of stable inorganic crystals. This framework addresses the disconnect between thermodynamic stability and formation energy, as well as retrospective vs. prospective benchmarking in materials discovery. We release a Python package to support model submissions and maintain an online leaderboard, offering insights into performance trade-offs. To identify the best-performing ML methodologies for materials discovery, we benchmarked various approaches, including random forests, graph neural networks (GNNs), one-shot predictors, iterative Bayesian optimizers, and universal interatomic potentials (UIP). Our initial results rank models by test set F1 scores for thermodynamic stability prediction: EquiformerV2 + DeNS > Orb > SevenNet > MACE > CHGNet > M3GNet > ALIGNN > MEGNet > CGCNN > CGCNN+P > Wrenformer > BOWSR > Voronoi fingerprint random forest. UIPs emerge as the top performers, achieving F1 scores of 0.57-0.82 and discovery acceleration factors (DAF) of up to 6x on the first 10k stable predictions compared to random selection. We also identify a misalignment between regression metrics and task-relevant classification metrics. Accurate regressors can yield high false-positive rates near the decision boundary at 0 eV/atom above the convex hull. Our results demonstrate UIPs' ability to optimize computational budget allocation for expanding materials databases. However, their limitations remain underexplored in traditional benchmarks. We advocate for task-based evaluation frameworks, as implemented here, to address these limitations and advance ML-guided materials discovery.

cond-mat.mtrl-sci↗

arXiv2023

A Quantum-Chemical Bonding Database for Solid-State Materials

An in-depth insight into the chemistry and nature of the individual chemical bonds is essential for understanding materials. Bonding analysis is thus expected to provide important features for large-scale data analysis and machine learning of material properties. Such chemical bonding information can be computed using the LOBSTER software package, which post-processes modern density functional theory data by projecting the plane wave-based wave functions onto a local, atomic orbital basis. With the help of a fully automatic workflow, the VASP and LOBSTER software packages are used to generate the data. We then perform bonding analyses on 1520 compounds (insulators and semiconductors) and provide the results as a database. The database structure of the bonding analysis database, which allows easy data retrieval, is also explained. The projected densities of states and bonding indicators are benchmarked on standard density-functional theory computations and available heuristics, respectively. Lastly, we illustrate the predictive power of bonding descriptors by constructing a machine-learning model for phononic properties, which shows an increase in prediction accuracies by 27 % (mean absolute errors) compared to a benchmark model differing only by not relying on any quantum-chemical bonding features.

cond-mat.mtrl-sci↗

arXiv2023

"Ultima Ratio": Simulating wide-range X-ray scattering and diffraction

We demonstrate a strategy for simulating wide-range X-ray scattering patterns, which spans the small- and wide scattering angles as well as the scattering angles typically used for Pair Distribution Function (PDF) analysis. Such simulated patterns can be used to test holistic analysis models, and, since the diffraction intensity is on the same scale as the scattering intensity, may offer a novel pathway for determining the degree of crystallinity. The "Ultima Ratio" strategy is demonstrated on a 64-nm Metal Organic Framework (MOF) particle, calculated from Q < 0.01 1/nm up to Q < 150 1/nm, with a resolution of 0.16 Angstrom. The computations exploit a modified 3D Fast Fourier Transform (3D-FFT), whose modifications enable the transformations of matrices at least up to 8000^3 voxels in size. Multiple of these modified 3D-FFTs are combined to improve the low-Q behaviour. The resulting curve is compared to a wide-range scattering pattern measured on a polydisperse MOF powder. While computationally intensive, the approach is expected to be useful for simulating scattering from a wide range of realistic, complex structures, from (poly-)crystalline particles to hierarchical, multicomponent structures such as viruses and catalysts.

cond-mat.mtrl-sci↗

arXiv2014

Point estimates in phylogenetic reconstructions

Motivation: The construction of statistics for summarizing posterior samples returned by a Bayesian phylogenetic study has so far been hindered by the poor geometric insights available into the space of phylogenetic trees, and ad hoc methods such as the derivation of a consensus tree makeup for the ill-definition of the usual concepts of posterior mean, while bootstrap methods mitigate the absence of a sound concept of variance. Yielding satisfactory results with sufficiently concentrated posterior distributions, such methods fall short of providing a faithful summary of posterior distributions if the data do not offer compelling evidence for a single topology. Results: Building upon previous work of Billera et al., summary statistics such as sample mean, median and variance are defined as the geometric median, Fréchet mean and variance, respectively. Their computation is enabled by recently published works, and embeds an algorithm for computing shortest paths in the space of trees. Studying the phylogeny of a set of plants, where several tree topologies occur in the posterior sample, the posterior mean balances correctly the contributions from the different topologies, where a consensus tree would be biased. Comparisons of the posterior mean, median and consensus trees with the ground truth using simulated data also reveals the benefits of a sound averaging method when reconstructing phylogenetic trees.

stat.AP↗