Research

The Wexler Group develops computational materials chemistry methods for energy conversion and environmental applications. The projects described here concern catalyst surface reconstruction, solar thermochemical hydrogen production, and nanocrystal synthesis. Related work on CO2 conversion, ferroelectric energy harvesting, and solar energy conversion is included on the Papers page.

Across these projects, we use statistical thermodynamics, first-principles quantum-mechanical calculations, Monte Carlo simulations, data science, and machine learning in collaboration with experimental groups. We use these methods to connect atomic-scale structures and energetics to thermodynamic observables and materials behavior during synthesis or under operating conditions. The current research questions are:

  • How can catalyst surface structures be predicted as functions of temperature and chemical environment?
  • How does perovskite composition affect redox thermodynamics and stability during solar thermochemical water splitting?
  • How do precursors and ligands affect the crystal structure and phase of chalcogenide nanocrystals during synthesis?

Surface Phase Diagrams

Surface phase diagram with ordered, disordered, and gas-phase adsorbates

Catalyst activity and selectivity depend on surface structure and composition under reaction conditions. Temperature, pressure, and chemical environment can drive surface reconstruction and degradation, while measurements under reactive conditions remain challenging. We develop computational methods to predict equilibrium catalyst surface structures and use them as reference states for studying catalytic turnover.

Our initial demonstration applied nested sampling to Lennard-Jones gas particles adsorbed on flat and stepped Lennard-Jones surfaces. From the sampled energies, we constructed a canonical partition function and calculated heat capacities and structural order parameters to identify adsorbate phases and transitions.

  • Yang, M.; Pártay, L. B.; Wexler, R. B. Surface Phase Diagrams from Nested Sampling. Phys. Chem. Chem. Phys. 2024, 26 (18), 13862–13874. PDF | DOI
  • Yang, R.; Chen, J.; Thibodeaux, D.; Wexler, R. B. FreeBird.jl: An Extensible Toolbox for Simulating Interfacial Phase Equilibria. J. Chem. Theory Comput. 2025, 21 (21), 10765–10779. PDF | DOI
  • Chatbipho, T.; Yang, R.; Wexler, R. B.; Pártay, L. B. Adsorbate Phase Transitions on Nanoclusters from Nested Sampling. J. Chem. Phys. 2025, 163 (17), 174701. PDF | DOI
  • Tipping, A. D.; Huggins, B.; Yang, R.; Wexler, R. B.; Garnett, R. Accelerating Materials Discovery with Active Thermodynamic Surface Shape Mapping. AI for Accelerated Materials Design (AI4Mat) Workshop, NeurIPS 2026. PDF | OpenReview

Solar Thermochemical Hydrogen Production

Solar thermochemical hydrogen-production cycle and perovskite redox process

Two-step solar thermochemical hydrogen production cycles use redox-active metal oxides to split water. Concentrated solar heat removes oxygen from the oxide at high temperature and low oxygen partial pressure. Steam then restores the oxygen and releases hydrogen. We study how perovskite composition and oxygen-vacancy thermodynamics affect this cycle's kinetics, stability, and durability.

Experiments on our (Ca, Ce)(Ti, Mn)O3−δ perovskite showed reversible oxygen-vacancy formation and filling without a reported bulk phase transition under the conditions studied. Building on HydroGEN results, we combine modeling, synthesis, characterization, and thermodynamic measurements with reactor design, system analysis, and techno-economic analysis to evaluate performance, cost, and scalability.

  • Choudhary, K.; Wines, D.; Li, K.; Garrity, K. F.; Gupta, V.; Romero, A. H.; Krogel, J. T.; Saritas, K.; Fuhr, A.; Ganesh, P.; Kent, P. R. C.; Yan, K.; Lin, Y.; Ji, S.; Blaiszik, B.; Reiser, P.; Friederich, P.; Agrawal, A.; Tiwary, P.; Beyerle, E.; Minch, P.; Rhone, T. D.; Takeuchi, I.; Wexler, R. B.; Mannodi-Kanakkithodi, A.; Ertekin, E.; Mishra, A.; Mathew, N.; Wood, M.; Rohskopf, A. D.; Hattrick-Simpers, J.; Wang, S.-H.; Achenie, L. E. K.; Xin, H.; Williams, M.; Biacchi, A. J.; Tavazza, F. JARVIS-Leaderboard: A Large-Scale Benchmark of Materials Design Methods. npj Comput. Mater. 2024, 10, 93. PDF | DOI
  • Way, L.; Spataru, C. D.; Jones, R. E.; Trinkle, D. R.; Rowberg, A. J. E.; Varley, J. B.; Wexler, R. B.; Smyth, C. M.; Douglas, T. C.; Bishop, S. R.; Fuller, E. J.; McDaniel, A. H.; Lany, S.; Witman, M. D. Defect Diffusion Graph Neural Networks for Materials Discovery in High-Temperature Energy Applications. Chem. Mater. 2025, 37 (17), 6473–6484. PDF | DOI
  • Douglas, T. C.; Dzara, M. J.; Rowberg, A. J. E.; King, K. A.; Syrigou, M.; Strange, N. A.; Bell, R. T.; Goyal, A.; Guan, P.-W.; Wexler, R. B.; Varley, J. B.; Ogitsu, T.; Lany, S.; McDaniel, A. H.; Bishop, S. R.; Witman, M. D. Large-Scale Experimental Validation of Thermochemical Water-Splitting Oxides Discovered by Defect Graph Neural Networks. Mater. Horiz. 2026, 13 (2), 829–839. PDF | DOI
  • Witman, M. D.; Pujet, S.; Rowberg, A. J. E.; Sutton, C.; Varley, J. B.; Lany, S.; Wexler, R. B. Transfer Learning on Universal Interatomic Potential Embeddings Improves Generalization in Structure-Property Defect Models. ChemRxiv 2026. Submitted for publication. PDF | DOI
  • Kumar, M.; Ali, N.; Witman, M. D.; Zhai, S.; Miller, J. E.; Ermanoski, I.; Stechel, E. B.; Wexler, R. B. Local B-Site Chemistry Controls Oxygen-Vacancy Energetics in Ca–Ce–Ti–Mn Perovskites for Thermochemical Hydrogen Production. J. Mater. Chem. A 2026, in press. PDF | DOI
  • Kumar, M.; Ferris, B. S.; Isakov, A.; Wexler, R. B. Periodic Trends in Oxygen Vacancy Formation Energies Across Binary Main-Group Metal Oxides. arXiv 2026, arXiv:2611.15297. Submitted for publication. PDF | DOI

Nanocrystal Synthesis

Halide-dependent synthesis pathways to wurtzite and rock-salt manganese sulfide nanocrystals

We combine experimental and computational methods to determine how halides affect the crystal structure and phase of manganese chalcogenide nanocrystals during synthesis. We identify prenucleation species, measure the thermochemistry of reactions and surface-ligand interactions, and monitor nucleation and growth kinetics using in situ techniques. We use first-principles calculations to characterize atomic-scale interactions and mechanisms that affect crystal structure and phase. We use these results as inputs for kinetic and thermodynamic models of nanocrystal nucleation and growth. We also examine lanthanide chalcogenide nanocrystals, which have been studied less extensively than manganese chalcogenide nanocrystals. We seek chemical principles for synthesizing Mn and Ln chalcogenide nanocrystals and test whether those principles apply to other material classes.

  • Chen, J.; Subramani, T.; Mekan, D.; Gendler, D.; Yang, R.; Kumar, M.; Householder, M.; Rosado Ortiz, A.; Hernandez-Pagan, E. A.; Lilova, K.; Wexler, R. B. Equilibrium Thermochemistry and Crystallographic Morphology of Manganese Sulfide Nanocrystals. npj Comput. Mater. 2026, in press. PDF | DOI

Computing Resources

Our group uses its bear, dragon, and wapiti systems and national allocations such as Expanse for computational research. Expand each entry for hardware details.

bear — Wexler Group
  • Dell PowerEdge T550
  • Intel Xeon Gold 6338 processors, 2.00 GHz
  • 64 cores
dragon — Wexler Group
  • Dell PowerEdge C6520
  • Intel Xeon Gold 6338 processors, 2.00 GHz
  • 256 cores across four nodes
wapiti — Wexler Group

A group-owned, single-node counterpart to dragon.

Expanse — San Diego Supercomputer Center

Through a National Artificial Intelligence Research Resource (NAIRR) Pilot Start-Up award, our group has 2,000 GPU hours on the Expanse AI resource for nested sampling selection of fine-tuning data for foundation interatomic potentials.

  • 5 petaflops peak; 93,184 CPU cores and 208 NVIDIA GPUs
  • GPU nodes with four NVIDIA V100 SXM2 GPUs (32 GB each) connected by NVLink
Theta — Past Computing Resource

Our group previously used Theta at the Argonne Leadership Computing Facility. Theta retired at the end of 2023.

  • Intel-Cray XC40; 11.7 petaflops
  • 4,392 nodes and 281,088 cores
  • Intel Xeon Phi 7230 processors