About

Paulette Clancy is the Edward J. Schaefer Professor in Engineering at the Department of Chemical and Biomolecular Engineering at Johns Hopkins University. She is known for her work in computational materials processing, focusing on developing new machine learning approaches to advance materials discovery and processing, including Bayesian optimization and active learning. Her research encompasses a broad scope of materials, from semiconductors to image processing and biomimetic situations, with a specialization in complex solution processing scenarios. Clancy is the director of research for the JHU Data Science and AI initiative, associate director of the Johns Hopkins Center for Integrated Structure-Mechanical Modeling and Simulation (CISMMS), and a fellow of the Hopkins Extreme Materials Institute (HEMI). She leads a prominent research group studying atomic- and molecular-scale modeling of semiconductor materials, including traditional silicon-based compounds and all-organic materials. Her group's research areas include advanced organic materials such as covalent organic frameworks and organic electronics, algorithm development involving force field development and Bayesian optimization, electronic materials like III-IV semiconducting materials, and nucleation and crystal growth of hybrid organic/inorganic perovskites and quantum dot nanocrystals. Her lab focuses on understanding the links between processing, structure, and function in advanced materials, with current projects involving the development of Bayesian optimization methods, modeling woven materials, creating near-perfect quantum dots, and discovering polymorphs of electronic materials for shape memory applications. Clancy has a distinguished background, having earned her bachelor’s degree in chemistry from Queen Elizabeth College (London University) in 1974 and a DPhil in physical chemistry from Oxford University in 1977. She conducted postdoctoral research at Cornell University and London University before joining the faculty at Cornell in 1987. During her tenure at Cornell, she served as the inaugural director of the Institute for Computational Science and Engineering, the Samuel W. and Diane M. Bodman Chair of Chemical Engineering, and director of the School of Chemical & Biomolecular Engineering. She also held roles as associate director of the College of Engineering’s Energy Institute. In 2018, she joined the faculty of the Whiting School of Engineering at Johns Hopkins University.

Research topics

• Computer Science
• Artificial Intelligence
• Information Retrieval
• Nanotechnology
• Library science
• Organic chemistry
• Chemistry
• Computational biology
• Programming language
• Inorganic chemistry
• Biology
• Materials science

Selected publications

• Can Coding Agents Reproduce Findings in Computational Materials Science?

  arXiv (Cornell University) · 2026-05-01

  articleOpen access

  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 54.1%. 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.

  PublisherOA PDF

• Can Coding Agents Reproduce Findings in Computational Materials Science?

  arXiv (Cornell University) · 2026-05-01

  preprintOpen access

  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 54.1%. 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.

  PublisherDOI

• Uncovering the Subtleties of Solvation and Speciation of Organic Halide Salts in Organic Solvents

  The Journal of Physical Chemistry B · 2026-04-27

  articleCorresponding

  Despite the importance of solution chemistry in the design of synthesis and processing routes for lead halide perovskite thin films and crystals, a detailed molecular-level understanding of key solution species prior to the addition of lead salts remains unclear. Here we study solutions comprised of an organic cation (CH3NH3+ (MA+) or CHN2H4+ (FA+)) and a halide (Cl–, Br–, or I–) in one of four organic solvents: γ-butyrolactone (GBL), N,N-dimethylformamide (DMF), dimethyl sulfoxide (DMSO), or tetrahydrothiophene 1-oxide (THTO). We interrogate speciation and solvation dynamics through a combination of 1H nuclear magnetic resonance (NMR) spectroscopy and room temperature ab initio molecular dynamics (AIMD) simulations. We elucidate the roles of each species in directing cation-halide ion pair formation, cation–solvent hydrogen bonding, and the intramolecular conformational freedom of the organic cation. In doing so, we identify trends in the propensity of different solvents to favor the formation of contact ion pairs (DMSO and THTO) versus solvent-separated ion pairs (DMF and GBL as well as THTO). Using insights from AIMD simulations, we propose factors that govern the solubility of the organic halide salt in organic solvent that are consistent with experimental observations. AIMD simulations additionally reveal a “halide-hopping” phenomenon in certain systems. This experimental-computational investigation has defined the key solution interactions that exist in perovskite precursor solutions prior to the addition of lead salts, thus laying the foundation for future experiments investigating how solvation and solution speciation direct the crystallization pathway of lead halide perovskites. The nuance and subtlety of the effects of changing the cation, halide, or solvent is surprisingly rich, and not easily categorized in simplistic correlations.

  PublisherDOI

• Synthesis and Characterization of Ultrasonically Atomized Al-Based Alloy Powders for Tunable Thermal Reactivity

  ˜The œminerals, metals & materials series · 2026-01-01

  book-chapter
  PublisherDOI

• Author response for "A Bayesian approach providing design choices and chemical insight for solution-processed thermoelectric polymers"

  2026-01-15

  peer-review
  PublisherDOI

• Perovskites Dataset

  Research Data Repository, Duke University · 2026-01-01

  datasetOpen access1st authorCorresponding

  Description and instructions for accessing the data can be found at <a href=https://www.sciserver.org/datasets/materials/perovskites/>https://www.sciserver.org/datasets/materials/perovskites/</a>

  PublisherDOI

• Author response for "A Bayesian approach providing design choices and chemical insight for solution-processed thermoelectric polymers"

  2026-02-09

  peer-review
  PublisherDOI

• Low-energy pathways lead to self-healing defects in CsPbBr₃

  Physical Chemistry Chemical Physics · 2025-01-01 · 8 citations

  articleOpen accessSenior author

  . Our work reveals the existence of a low-energy diffusion pathway involving a concerted "domino effect" of interstitials, with the net result that interstitials can diffuse more readily over longer distances than expected. This observation suggests that defect self-healing can be promoted if the "domino effect" strategy can be engaged.

  PublisherOA PDFDOI

• Bayesian optimization of solution-processed thermoelectric polymers using a database of ab initio electronic structure data

  Research Square · 2025-02-07

  preprintOpen accessSenior author
  PublisherOA PDFDOI

• Author Correction: A call to elevate the role of processing in AI-driven materials design

  Nature Reviews Materials · 2025-10-31

  articleOpen access
  PublisherOA PDFDOI

Recent grants

• New, GK-12 Grass Roots: Advancing Education in Renewable Energy and Cleaner Fuels through Collaborative Graduate Fellow/Teacher/Grade-School Student Interactions

  NSF · $1.4M · 2011–2015

• NEB: Ultimate Electronic Device Scaling Using Structurally Precise Graphene Nanoribbons

  NSF · $1.2M · 2011–2016

Frequent coauthors

• Kentaro Okano

  Kobe University

  1600 shared

• Raymond Woon Sing Wong

  King's College London

  1600 shared

• I Ibrahim

  1600 shared

• Jodie L. Lutkenhaus

  Texas A&M University

  1600 shared

• Jeremy J. Baumberg

  University of Cambridge

  1600 shared

• Paul F. Scott

  University of Cambridge

  1600 shared

• Bernhardt L. Trout

  1600 shared

• Allison Holloway

  Nanjing University

  1600 shared

Labs

• Computational Materials LabPI

Education

• D.Phil., Chemistry

  University of Oxford

  1977

Awards & honors

• American Institute of Chemical Engineers (AIChE) National Wo…
• Alice Cook Award for services promoting women in science at…
• Zellman Warhaft award for the promotion of diversity in Corn…
• Edward J. Schaefer Professor in Engineering