Adaptive mixed variable Bayesian self-optimisation of catalytic reactions

Catalytic reactions play a central role in many industrial processes, owing to their ability to enhance efficiency and sustainability. However, complex interactions between the categorical and continuous variables leads to non-smooth response surfaces, which traditional optimisation methods struggle...

Full description

Saved in:
Bibliographic Details
Published inReaction chemistry & engineering Vol. 9; no. 2; pp. 38 - 316
Main Authors Aldulaijan, Naser, Marsden, Joe A, Manson, Jamie A, Clayton, Adam D
Format Journal Article
LanguageEnglish
Published Cambridge Royal Society of Chemistry 30.01.2024
Subjects
Algorithms
Bayesian analysis
Chemical reactions
Continuous flow
Cross coupling
Optimization
Online AccessGet full text

Cover

More Information
Summary:Catalytic reactions play a central role in many industrial processes, owing to their ability to enhance efficiency and sustainability. However, complex interactions between the categorical and continuous variables leads to non-smooth response surfaces, which traditional optimisation methods struggle to navigate. Herein, we report the development and benchmarking of a new adaptive latent Bayesian optimiser (ALaBO) algorithm for mixed variable chemical reactions. ALaBO was found to outperform other open-source Bayesian optimisation toolboxes, when applied to a series of test problems based on simulated kinetic data of catalytic reactions. Furthermore, through integration of ALaBO with a continuous flow reactor, we achieved the rapid self-optimisation of an exemplar Suzuki-Miyaura cross-coupling reaction involving six distinct ligands, identifying a 93% yield within a budget of just 25 experiments. A novel adaptive latent Bayesian optimisation (ALaBO) algorithm accelerates the development of mixed variable catalytic reactions.
Bibliography:Electronic supplementary information (ESI) available. See DOI
https://doi.org/10.1039/d3re00476g
ISSN:2058-9883
2058-9883
DOI:10.1039/d3re00476g