arXiv subjects

Wyatt Bunstine

Publications and source records attributed to Wyatt Bunstine.

5 recordsLinked to original sources

arXiv2026

Human and LLM Collaboration for Accelerated Materials Synthesis and Discovery

Although Large Language Models (LLM) and Artificial Intelligence (AI) tools have enabled a rapid increase in the generation rate of predicted materials, the rate of new materials discovery has lagged behind. This is due to the challenges associated with designing a sequence of chemical reactions to predictably produce new materials, especially in new structure types. Here, we report a study of human and LLM generated recipes for the synthesis of known and new materials. The success of the recipes is determined through in-lab experimentation, and the results are passed back to the humans and LLMs in a closed-loop process to study the effects of their collaboration. The Ruddlesden-Popper homologous series was selected for all material candidates to provide a materials phase space that is simultaneously well studied and likely to host undiscovered materials. We find that humans (H) and LLM (L) have similar success rates: 83(8)% (H) and 75(9)% (L) [known materials, round one], 17(9)% (H) and 22(10)% (L) [unknown materials, round one], 79(8)% (H) and 71(9)% (L) [known materials, round two], and 22(7)% (H) and 14(6)% (L) [unknown materials, round two]. Through this collaborative human-LLM effort, we discovered Ba3PtO5, a material with a new structural prototype that constitutes the missing 1D member of the herein reported dimensionally tunable Rock-Salt Perovskite (RSP) homologous series of the form (AX)m(ABX3)p, of which the Ruddlesden-Popper series is a subset.

cond-mat.mtrl-sci↗

arXiv2026

Coupling Language Models with Physics-based Simulation for Synthesis of Inorganic Materials

Modern generative machine learning (ML) models can propose novel inorganic crystalline materials with targeted properties; however, synthesis planning of these materials remains difficult due to the complexity of the associated physical processes and limited availability of computational tools. We introduce a novel hybrid framework to evaluate Large Language Models (LLMs) in inorganic synthesis planning by combining thermodynamic databases with simplified kinetics models to approximate realistic synthesis conditions. As a case study, we focus on the niobium-oxygen system, which features multiple industrially relevant oxide phases with well-characterized data. In computational simulations, we compare LLM-generated synthesis routes with classical path-planning algorithms, showing that the implicit priors in LLMs can yield more viable strategies. In our evaluation setting, classical search methods serve primarily as a foil rather than a direct competitor. This illustrates the relative complexity of the problem and highlights where the LLM's implicit priors add value.

cs.AI↗

arXiv2026

Can Coding Agents Reproduce Findings in Computational Materials Science?

Large language models are increasingly deployed as autonomous coding agents and have achieved remarkably strong performance on software engineering benchmarks. However, it is unclear whether such success transfers to computational scientific workflows, where tasks require not only strong coding ability, but also the ability to navigate complex, domain-specific procedures and to interpret results in the context of scientific claims. To address this question, we present AutoMat, a benchmark for evaluating LLM-based agents' ability to reproduce claims from computational materials science. AutoMat poses three interrelated challenges: recovering underspecified computational procedures, navigating specialized toolchains, and determining whether the resulting evidence supports a claim. By working closely with subject matter experts, we curate a set of claims from real materials science papers to test whether coding agents can recover and execute the end-to-end workflow needed to support (or undermine) such claims. We then evaluate multiple representative coding agent settings across several foundation models. Our results show that current LLM-based agents obtain low overall success rates on AutoMat, with the best-performing setting achieving a success rate of only 53%. Error analysis further reveals that agents perform worst when workflows must be reconstructed from paper text alone and that they fail primarily due to incomplete procedures, methodological deviations, and execution fragility. Taken together, these findings position AutoMat as both a benchmark for computational scientific reproducibility and a tool for diagnosing the current limitations of agentic systems in AI-for-science settings.

cs.CL↗

arXiv2025

Real-space orbital tiling approach for the design of novel superconductors

Despite substantial advances in the field, we still lack a predictive framework capable of guiding the discovery of new families of superconductors. While momentum-space approaches have advanced the microscopic understanding of superconductivity, they offer limited guidance for materials design based on atomic building blocks. Here, we propose a real-space framework which conceptualizes Cooper pairs as confined standing waves resulting from coherent tilings of atomic orbitals. We call this model the Real-space Orbital Superconducting Pathway (ROSP). Using a tight-binding toy model, we show that the energetics of electron pairing depend on the configuration and overlap of real-space orbitals, which motivates \textit{a priori} design of superconducting families from orbital tiling. We connect the ROSP model to Roald Hoffmann's isolobal analogy to classify families of superconductors based on shared orbital tilings, rather than structure or electron count. As an example, we suggest that superconductivity in La$_{3}$Ni$_{2}$O$_{7}$ and LaNiO$_{2}$, despite differing structures and electron counts, may arise from a common ROSP. We introduce a new notation to classify two-dimensional square-net ROSPs and further propose several new families of superconductors on the anti-cuprate lattice. This framework provides a new model for predicting and designing families of high-T$_c$ superconductors from real-space orbital architecture, even without microscopic knowledge of the attractive pairing interaction.

cond-mat.supr-con↗

arXiv2024

Hydroflux-Controlled Growth of Magnetic K-Cu-Te-O(H) Phases

Innovative synthetic approaches can yield new phases containing novel structural and magnetic motifs. In this work, we show the synthesis and magnetic characterization of three new and one previously reported layered phase in the K-Cu-Te-O(H) phase space using a tunable hydroflux technique. The hydroflux, with a roughly equal molar ratio of water and alkali hydroxide, is a highly oxidizing, low melting solvent which can be used to isolate metastable phases unattainable through traditional solid state or flux techniques. The newly synthesized phases, K$_{2}$Cu$_{2}$TeO$_{6}$, K$_{2}$Cu$_{2}$TeO$_{6}$ $\cdot$ H$_{2}$O, and K$_{6}$Cu$_{9}$Te$_{4}$O$_{24}$ $\cdot$ 2 H$_{2}$O, contain Cu$^{2+}$ within CuO$_{4}$ square planar plaquettes and TeO$_{6}$ octahedra ordering to form structural honeycomb layers isolated by interlayer K$^{+}$ ions and H$_{2}$O molecules. We find the synthesized structures display varying tilt sequences of the CuO$_{4}$ plaquettes, leading to distinct Cu$^{2+}$ magnetic motifs on the structural honeycomb lattice and a range of effective magnetic dimensionalities. We find that K$_{2}$Cu$_{2}$TeO$_{6}$ $\cdot$ H$_{2}$O does not order and displays alternating chain Heisenberg antiferromagnetic (AFM) behavior, while K$_{2}$Cu$_{2}$TeO$_{6}$ and K$_{6}$Cu$_{9}$Te$_{4}$O$_{24}$ $\cdot$ 2 H$_{2}$O order antiferromagnetically (T$_{N}$ = 100 K and T$_{N}$ = 6.5 K respectively). The previously known phase, K$_{2}$CuTeO$_{4}$(OH)$_{2}$ $\cdot$ H$_{2}$O, we find contains structurally and magnetically one-dimensional CuO$_{4}$ plaquettes leading to uniform chain Heisenberg AFM behavior and shows no magnetic order down to T = 0.4 K. We discuss and highlight the usefulness of the hydroflux technique in novel syntheses and the interesting magnetic motifs that arise in these particular phases.

cond-mat.str-el↗