arXiv subjects

Wesley F. Reinhart

Publications and source records attributed to Wesley F. Reinhart.

14 recordsLinked to original sources

arXiv2026

Lifetime Sample Tracking (LiST): A Data Platform for Materials Science

The 2D Crystal Consortium Materials Innovation Platform (2DCC-MIP) is an NSF supported national user facility focused on advancing the synthesis of 2D materials, monolayers, surfaces, and interfaces. The need for the facility to organize and share data with users led to the development of an internal data management and analysis engine, the Lifetime Sample Tracking platform (LiST). This infrastructure allows the automated capture, curation, analysis and dissemination of data ranging from experimental materials synthesis parameters and characterization, to theoretical first-principles and ReaxFF molecular dynamics modeling1. The system currently hosts synthesis and property data (accessible via a REST API) on approximately twenty thousand samples produced by the 2DCC grown using a variety of techniques from bulk crystal growth to metal-organic chemical vapor deposition (MOCVD) and molecular beam epitaxy (MBE), among others. Data used in publications can easily be grouped by the system into data packages that are given digital object identifiers (DOIs) for inclusion with each publication. The LiST platform is now being used by groups outside of the 2DCC as a solution for data curation in materials science. Data management tools such as LiST support the materials development process by allowing a closed loop iteration between synthesis, characterization, theory, and targeted materials design. This also enables machine learning (ML) research, artificial intelligence (AI) analysis, and the potential for autonomous synthesis in the future.

cond-mat.mtrl-sci↗

arXiv2026

Multi-modal machine learning analysis of GaSe molecular beam epitaxy growth conditions

Autonomous synthesis platforms integrating machine learning with in situ diagnostics have the potential to revolutionize thin-film growth by enabling real-time process optimization and reducing the need for manual tuning. However, their application to molecular beam epitaxy (MBE) remains underdeveloped. Here, we present a machine learning-guided framework for MBE growth of GaSe films, leveraging reflection high-energy electron diffraction (RHEED) as an in situ diagnostic alongside ex situ characterization via X-ray diffraction and atomic force microscopy. Unsupervised learning on RHEED patterns reveals a well-defined boundary between high- and low-quality samples, capturing physically meaningful features. Mutual information analysis shows a strong correlation between RHEED embeddings and rocking curve full-width at half-maximum (fwhm), while the correlation with AFM root-mean-square (RMS) roughness is weak. Among key growth conditions, growth rate most strongly influences fwhm, whereas the Se/Ga flux ratio primarily affects RMS roughness and the RHEED embeddings. Supervised learning models trained to predict fwhm and RMS roughness demonstrate moderate accuracy, with significant improvement achieved by incorporating RHEED embeddings. Furthermore, anomaly detection via residual analysis in supervised learning aligns well with unsupervised classification from RHEED, reinforcing the reliability of the predictive models. This study establishes a data-driven framework for machine learning-assisted MBE, paving the way for real-time process control and accelerated optimization of thin-film synthesis.

cond-mat.mtrl-sci↗

arXiv2026

A Constrained Natural-Language Interface for Variational Multi-Physics Finite Element Simulations in FEniCS

Large language models can reduce the manual effort required to set up finite element simulations, but they introduce reliability risks when generated solver code lies on the critical path. We present a constrained natural-language interface for multi-physics finite element analysis in which the LLM is limited to front-end tasks: parsing prompts into structured JSON, generating Gmsh code only for non-catalog geometries, and using retry feedback for those stages. It never writes FEniCS solver templates, derives weak forms, or writes the numerical solver core. A deterministic dispatcher maps the validated specification to five human-written FEniCS/UFL templates: linear elasticity, hyperelasticity, elastoplasticity, thermo-mechanical coupling, and phase-field fracture. We validate this deterministic template layer against analytical solutions and published 2D/3D benchmarks. Smooth cases reach sub-percent agreement on adequate meshes, while harder nonlinear cases reach the 2-5 percent range. We also evaluate the LLM-facing front end directly. In a 15-prompt parser benchmark, first-pass valid parses were obtained for 9 cases, and all remaining cases were repaired after retry, giving a final valid parse rate of 100.0 percent, 100.0 percent problem-class accuracy, and 97.1 percent field-extraction accuracy. In a 10-case custom-geometry benchmark routed through the real LLM-to-Gmsh path, first-pass and final success were both 90.0 percent, with one unrecovered invalid-geometry failure. These results show that the parser and constrained prompt/validation design are effective on these benchmarks. As an end-to-end demonstration, the system generates and analyzes a 3D elastoplastic L-bracket with a fillet and bolt hole from one natural-language prompt. The contribution is a measured architecture for natural-language-driven variational simulation, not open-ended autonomous code generation.

cs.CE↗

arXiv2025

Cross-Modal Characterization of Thin Film MoS$_2$ Using Generative Models

The growth and characterization of materials using empirical optimization typically requires a significant amount of expert time, experience, and resources. Several complementary characterization methods are routinely performed to determine the quality and properties of a grown sample. Machine learning (ML) can support the conventional approaches by using historical data to guide and provide speed and efficiency to the growth and characterization of materials. Specifically, ML can provide quantitative information from characterization data that is typically obtained from a different modality. In this study, we have investigated the feasibility of projecting the quantitative metric from microscopy measurements, such as atomic force microscopy (AFM), using data obtained from spectroscopy measurements, like Raman spectroscopy. Generative models were also trained to generate the full and specific features of the Raman and photoluminescence spectra from each other and the AFM images of the thin film MoS$_2$. The results are promising and have provided a foundational guide for the use of ML for the cross-modal characterization of materials for their accelerated, efficient, and cost-effective discovery.

cond-mat.mtrl-sci↗

arXiv2024

Transfer Learning for Multi-material Classification of Transition Metal Dichalcogenides with Atomic Force Microscopy

Deep learning models are widely used for the data-driven design of materials based on atomic force microscopy (AFM) and other scanning probe microscopy. These tools enhance efficiency in inverse design and characterization of materials. However, limited and imbalanced experimental materials data typically available is a major challenge. Also important is the need to interpret trained models, which have typically been complex enough to be uninterpretable by humans. Here, we present a systemic evaluation of transfer learning strategies to accommodate low-data scenarios in materials synthesis and a model latent feature analysis to draw connections to the human-interpretable characteristics of the samples. Our models show accurate predictions in five classes of transition metal dichalcogenides (TMDs) (MoS$_2$, WS$_2$, WSe$_2$, MoSe$_2$, and Mo-WSe$_2$) with up to 89$\%$ accuracy on held-out test samples. Analysis of the latent features reveals a correlation with physical characteristics such as grain density, DoG blob, and local variation. The transfer learning optimization modality and the exploration of the correlation between the latent and physical features provide important frameworks that can be applied to other classes of materials beyond TMDs to enhance the models' performance and explainability which can accelerate the inverse design of materials for technological applications.

cond-mat.mtrl-sci↗

arXiv2024

Efficient Generation of Grids and Traversal Graphs in Compositional Spaces towards Exploration and Path Planning

Many disciplines of science and engineering deal with problems related to compositions, ranging from chemical compositions in materials science to portfolio compositions in economics. They exist in non-Euclidean simplex spaces, causing many standard tools to be incorrect or inefficient, which is significant in combinatorically or structurally challenging spaces exemplified by Compositionally Complex Materials (CCMs) and Functionally Graded Materials (FGMs). Here, we explore them conceptually in terms of problem spaces and quantitatively in terms of computational feasibility. This work implements several essential methods specific to the compositional (simplex) spaces through a high-performance open-source library nimplex. Most significantly, we derive and implement an algorithm for constructing a novel n-dimensional simplex graph data structure, which contains all discretized compositions and all possible neighbor-to-neighbor transitions as pointer arrays. Critically, no distance or neighborhood calculations are performed, instead leveraging pure combinatorics and the ordering in procedurally generated simplex grids, keeping the algorithm $\mathcal{O}(N)$, so that graphs with billions of transitions take seconds to construct on a laptop. Furthermore, we demonstrate how such graph representations can be combined to express path-planning problem spaces and to incorporate prior knowledge while keeping the problem space homogeneous. This allows for efficient deployment of existing high-performance gradient descent, graph traversal search, and other path optimization algorithms.

cond-mat.mtrl-sci↗

arXiv2024

Crystal Growth Characterization of WSe$_2$ Thin Film Using Machine Learning

Materials characterization remains a labor-intensive process, with a large amount of expert time required to post-process and analyze micrographs. As a result, machine learning has become an essential tool in materials science, including for materials characterization. In this study, we perform an in-depth analysis of the prediction of crystal coverage in WSe$_2$ thin film atomic force microscopy (AFM) height maps with supervised regression and segmentation models. Regression models were trained from scratch and through transfer learning from a ResNet pretrained on ImageNet and MicroNet to predict monolayer crystal coverage. Models trained from scratch outperformed those using features extracted from pretrained models, but fine-tuning yielded the best performance, with an impressive 0.99 $R^2$ value on a diverse set of held-out test micrographs. Notably, features extracted from MicroNet showed significantly better performance than those from ImageNet, but fine-tuning on ImageNet demonstrated the reverse. As the problem is natively a segmentation task, the segmentation models excelled in determining crystal coverage on image patches. However, when applied to full images rather than patches, the performance of segmentation models degraded considerably, while the regressors did not, suggesting that regression models may be more robust to scale and dimension changes compared to segmentation models. Our results demonstrate the efficacy of computer vision models for automating sample characterization in 2D materials while providing important practical considerations for their use in the development of chalcogenide thin films.

cond-mat.mtrl-sci↗

arXiv2024

nimCSO: A Nim package for Compositional Space Optimization

nimCSO is a high-performance tool implementing several methods for selecting components (data dimensions) in compositional datasets, which optimize the data availability and density for applications such as machine learning. Making said choice is a combinatorically hard problem for complex compositions existing in highly dimensional spaces due to the interdependency of components being present. Such spaces are encountered, for instance, in materials science, where datasets on Compositionally Complex Materials (CCMs) often span 20-45 chemical elements, 5-10 processing types, and several temperature regimes, for up to 60 total data dimensions. At its core, nimCSO leverages the metaprogramming ability of the Nim language (nim-lang.org) to optimize itself at the compile time, both in terms of speed and memory handling, to the specific problem statement and dataset at hand based on a human-readable configuration file. As demonstrated in this paper, nimCSO reaches the physical limits of the hardware (L1 cache latency) and can outperform an efficient native Python implementation over 400 times in terms of speed and 50 times in terms of memory usage (not counting interpreter), while also outperforming NumPy implementation 35 and 17 times, respectively, when checking a candidate solution. It is designed to be both (1) a user-ready tool, implementing two efficient brute-force approaches (for handling up to 25 dimensions), a custom search algorithm (for up to 40 dimensions), and a genetic algorithm (for any dimensionality), and (2) a scaffold for building even more elaborate methods in the future, including heuristics going beyond data availability. All configuration is done with a simple human-readable YAML config file and plain text data files, making it easy to modify the search method and its parameters with no knowledge of programming and only basic command line skills.

cond-mat.mtrl-sci↗

arXiv2024

Quantitative analysis of MoS$_2$ thin film micrographs with machine learning

Isolating the features associated with different materials growth conditions is important to facilitate the tuning of these conditions for effective materials growth and characterization. This study presents machine learning models for classifying atomic force microscopy (AFM) images of thin film MoS$_2$ based on their growth temperatures. By employing nine different algorithms and leveraging transfer learning through a pretrained ResNet model, we identify an effective approach for accurately discerning the characteristics related to growth temperature within the AFM micrographs. Robust models with up to 70% test accuracies were obtained, with the best performing algorithm being an end-to-end ResNet fine-tuned on our image domain. Class activation maps and occlusion attribution reveal that crystal quality and domain boundaries play crucial roles in classification, with models exhibiting the ability to identify latent features beyond human visual perception. Overall, the models demonstrated high accuracy in identifying thin films grown at different temperatures despite limited and imbalanced training data as well as variation in growth parameters besides temperature, showing that our models and training protocols are suitable for this and similar predictive tasks for accelerated 2D materials characterization.

cond-mat.mtrl-sci↗

arXiv2023

Comparing Forward and Inverse Design Paradigms: A Case Study on Refractory High-Entropy Alloys

The rapid design of advanced materials is a topic of great scientific interest. The conventional, ``forward'' paradigm of materials design involves evaluating multiple candidates to determine the best candidate that matches the target properties. However, recent advances in the field of deep learning have given rise to the possibility of an ``inverse'' design paradigm for advanced materials, wherein a model provided with the target properties is able to find the best candidate. Being a relatively new concept, there remains a need to systematically evaluate how these two paradigms perform in practical applications. Therefore, the objective of this study is to directly, quantitatively compare the forward and inverse design modeling paradigms. We do so by considering two case studies of refractory high-entropy alloy design with different objectives and constraints and comparing the inverse design method to other forward schemes like localized forward search, high throughput screening, and multi objective optimization.

cond-mat.mtrl-sci↗

arXiv2022

Predicting aggregate morphology of sequence-defined macromolecules with Recurrent Neural Networks

Self-assembly of dilute sequence-defined macromolecules is a complex phenomenon in which the local arrangement of chemical moieties can lead to the formation of long-range structure. The dependence of this structure on the sequence necessarily implies that a mapping between the two exists, yet it has been difficult to model so far. Predicting the aggregation behavior of these macromolecules is challenging due to the lack of effective order parameters, a vast design space, inherent variability, and high computational costs associated with currently available simulation techniques. Here, we accurately predict the morphology of aggregates self-assembled from sequence-defined macromolecules using supervised machine learning. We find that regression models with implicit representation learning perform significantly better than those based on engineered features such as $k$-mer counting, and a Recurrent-Neural-Network-based regressor performs the best out of nine model architectures we tested. Furthermore, we demonstrate the high-throughput screening of monomer sequences using the regression model to identify candidates for self-assembly into selected morphologies. Our strategy is shown to successfully identify multiple suitable sequences in every test we performed, so we hope the insights gained here can be extended to other increasingly complex design scenarios in the future, such as the design of sequences under polydispersity and at varying environmental conditions.

cond-mat.soft↗

arXiv2021

Generative deep learning as a tool for inverse design of high-entropy refractory alloys

Generative deep learning is powering a wave of new innovations in materials design. In this article, we discuss the basic operating principles of these methods and their advantages over rational design through the lens of a case study on refractory high-entropy alloys for ultra-high-temperature applications. We present our computational infrastructure and workflow for the inverse design of new alloys powered by these methods. Our preliminary results show that generative models can learn complex relationships in order to generate novelty on demand, making them a valuable tool for materials informatics.

cond-mat.mtrl-sci↗

arXiv2021

Unsupervised learning of sequence-specific aggregation behavior for a model copolymer

We apply a recently developed unsupervised machine learning scheme for local atomic environments to characterize large-scale, disordered aggregates formed by sequence-defined macromolecules. This method provides new insight into the structure of these disordered, dilute aggregates, which has proven difficult to understand using collective variables manually derived from expert knowledge. In contrast to such conventional order parameters, we are able to classify the global aggregate structure directly using descriptions of the local environments. The resulting characterization provides a deeper understanding of the range of possible self-assembled structures and their relationships to each other. We also provide a detailed analysis of the effects of finite system size, stochasticity, and kinetics of these aggregates based on the learned collective variables. Interestingly, we find that the spatiotemporal evolution of systems in the learned latent space is smooth and continuous, despite being derived from only a single snapshot from each of about 1000 monomer sequences. These results demonstrate the insight which can be gained by applying unsupervised machine learning to soft matter systems, especially when suitable order parameters are not known.

cond-mat.soft↗

arXiv2021

Unsupervised learning of atomic environments from simple features

I present a strategy for unsupervised manifold learning on local atomic environments in molecular simulations based on simple rotation- and permutation-invariant three-body features. These features are highly descriptive, generalize to multiple chemical species, and are human-interpretable. The low-dimensional embeddings of each atomic environment can be used to understand and quantify messy crystal structures such as those near interfaces and defects or well-ordered crystal lattices such as in bulk materials without modification. The same method can also yield collective variables describing collections of particles such as for an entire simulation domain. I demonstrate the method on colloidal crystallization, ice crystals, and binary mesophases to illustrate its broad applicability. In each case, the learned latent space yields insights into the details of the observed microstructures. For ices and mesophases, supervised classifiers are trained based on the learned manifolds and directly compared against a recent neural-network-based approach. Notably, while this method provides comparable classification performance, it can also be deployed on even a handful of observed environments without labels or \textit{a priori} knowledge. Thus, the current approach provides an incredibly versatile strategy to characterize and classify local atomic environments, and may unlock insights in a wide variety of molecular simulation contexts.

cond-mat.mtrl-sci↗