DATA-DRIVEN CAUSAL DISCOVERY IN CATALYSIS

ChemRxiv

Abstract

Catalysts are essential for the transition towards more sustainable chemical processes. However, their design still relies largely on empirical strategies, limiting the systematic understanding of how synthesis variables and reaction conditions influence catalytic performance and final material properties. This work applies causal discovery methods to two catalytic case studies: an OCM reaction dataset and an in-house dataset on metal-supported zeolite synthesis. The aim is to investigate the intrinsic causal structures underlying catalytic systems and to move beyond purely correlational analysis. The generated graphs revealed only two causal relationships for the first dataset and showed a frequent identification of latent variables. These results highlight the complexity of catalytic datasets and demonstrate the potential of causal discovery as an exploratory tool for studying synthesis–property–performance relationships. Take-home message. Causal discovery can support catalysis research as an exploratory tool, helping identify robust causal signals while exposing hidden confounding and uncertainty. In the studied datasets, only a few directed relations were consistently recovered, showing that causal interpretation remains strongly limited by data size, variability and latent factors.

Authors 5

1. Bárbara Alves Sequeira

   University of Lisbon · Instituto Superior Técnico · Centro de Química Estrutural

   Affiliation as printed

   Centro de Química Estrutural-Institute of Molecular Sciences Chemical Engineering Department, Instituto Superior Técnico

   Universidade de Lisboa

2. Adèle H. Ribeiro Aachen

   RWTH Aachen University

   Affiliation as printed

   Faculty of Mathematics and Natural Sciences & Center for Computational Life Sciences

   RWTH Aachen University

3. Catarina Barata

   University of Lisbon · Instituto Superior Técnico

   Affiliation as printed

   Instituto Superior Técnico

   Universidade de Lisboa

4. Carmen Bacariza

   University of Lisbon · Instituto Superior Técnico · Centro de Química Estrutural

   Affiliation as printed

   Centro de Química Estrutural-Institute of Molecular Sciences Chemical Engineering Department, Instituto Superior Técnico

   Universidade de Lisboa

5. Pedro S. F. Mendes

   University of Lisbon · Instituto Superior Técnico · Centro de Química Estrutural

   Affiliation as printed

   Centro de Química Estrutural-Institute of Molecular Sciences Chemical Engineering Department, Instituto Superior Técnico

   Universidade de Lisboa