Exploration of Chemical Space Through Automated Reasoning

Clymo, Judith, Collins, Christopher M ORCID: 0000-0002-0101-4426, Atkinson, Katie ORCID: 0000-0002-5683-4106, Dyer, Matthew S ORCID: 0000-0002-4923-3003, Gaultois, Michael W ORCID: 0000-0003-2172-2507, Gusev, Vladimir V ORCID: 0000-0002-2815-607X, Rosseinsky, Matthew J ORCID: 0000-0002-1910-2483 and Schewe, Sven ORCID: 0000-0002-9093-9518 (2025) Exploration of Chemical Space Through Automated Reasoning Angewandte Chemie, 137 (6). ISSN 0044-8249, 1521-3757

Access the full-text of this item by clicking on the Open Access link.

Abstract

Abstract The vast size of composition space poses a significant challenge for materials chemistry: exhaustive enumeration of potentially interesting compositions is typically infeasible, hindering assessment of important criteria ranging from novelty and stability to cost and performance. We report a tool, Comgen, for the efficient exploration of composition space, which makes use of logical methods from computer science used for proving theorems. We demonstrate how these techniques, which have not previously been applied to materials discovery, can enable reasoning about scientific domain knowledge provided by human experts. Comgen accepts a variety of user‐specified criteria, converts these into an abstract form, and utilises a powerful automated reasoning algorithm to identify compositions that satisfy these user requirements, or prove that the requirements cannot be simultaneously satisfied. In contrast to machine learning techniques, explicitly reasoning about domain knowledge, rather than making inferences from data, ensures that Comgen's outputs are fully interpretable and provably correct. Users interact with Comgen through a high‐level Python interface. We illustrate use of the tool with several case studies focused on the search for new ionic conductors. Further, we demonstrate the integration of Comgen into an end‐to‐end automated workflow to propose and evaluate candidate compositions quantitatively, prior to experimental investigation. This highlights the potential of automated formal reasoning in materials chemistry.

Item Type:	Article
Uncontrolled Keywords:	46 Information and Computing Sciences, 4602 Artificial Intelligence, Networking and Information Technology R&D (NITRD), Machine Learning and Artificial Intelligence
Divisions:	Faculty of Science & Engineering Faculty of Science & Engineering > School of Electrical Engineering, Electronics and Computer Science Faculty of Science & Engineering > School of Physical Sciences
Depositing User:	Symplectic Admin
Date Deposited:	07 Feb 2025 17:02
Last Modified:	23 May 2026 09:27
DOI:	10.1002/ange.202417657
Open Access URL:	https://doi.org/10.1002/anie.202417657
Related Websites:	Publisher
URI:	https://livrepository.liverpool.ac.uk/id/eprint/3190207
Disclaimer:	The University of Liverpool is not responsible for content contained on other websites from links within repository metadata. Please contact us if you notice anything that appears incorrect or inappropriate.