News Picture Generic

Closed-Loop Bayesian Optimization for the Synthesis of Amine-Grafted Resins for Direct Air Capture

March 24, 2026

ChemRxiv

Direct air capture (DAC) of CO2 remains limited by adsorbent performance. Adsorption kinetics are a particularly underexplored lever to improve adsorbents, despite their significance in realistic processes. Today, the profile of 'the' optimal DAC adsorbent displaying high CO2 uptake and selectivity as well as fast sorption kinetics remains unknown. Here, we present a closed-loop approach combining high-throughput synthesis and characterisation with Bayesian optimization to understand how the material synthesis space impacts adsorbent properties and how it can ultimately be optimized to enhance adsorption-based DAC performance. To carry out this study, we chose amine-functionalized porous adsorbents as our model system. By jointly optimizing adsorption kinetics and equilibrium CO2 uptake at 400 ppm, the framework identifies high-performing materials while experimentally mapping a Pareto front. Sorbents with the highest CO2 uptakes exhibit the slowest kinetics, and vice versa. We link this trade-off to the pore structure of the materials and how the pore geometry imposes diffusion constraints. The optimization study uncovers adsorbents with order-of-magnitude higher adsorption kinetics than benchmark adsorbents at half the uptakes. Beyond this model system, our Bayesian optimization framework is a blueprint for multi-objective materials optimization in fields beyond DAC adsorbents.

For details: 

Closed-Loop Bayesian Optimization for the Synthesis of Amine-Grafted Resins for Direct Air Capture

Tristan L Spreng 1, Paul Schweng 2, David Danaci 1,3,4, Ronny Pini 1,3, and Camille Petit 1

1) Department of Chemical Engineering, Imperial College London, London, United Kingdom
2) Institute of Materials Chemistry and Research, University of Vienna, Vienna, Austria,
3) The Sargent Centre for Process Systems Engineering, Imperial College London, London, United Kingdom,
4) I-X Centre for AI in Science, Imperial College London, London, United Kingdom
5) National University of Singapore (Suzhou) Research Institute, Suzhou, Jiangsu, China

ChemRxiv
https://chemrxiv.org/doi/full/10.26434/chemrxiv.15000640/v1

Contact us to learn more about this exciting article:https://www.chemspeed.com/contact/

Other Recent News

Discover more news articles you might be interested in

Read more about High-throughput solid microsampling through stochastic robotic automation
News Picture 1 1 V2
Featured
Jul
14

High-throughput solid microsampling through stochastic robotic automation

Solid sampling at the sub-milligram and milligram scale remains a major bottleneck for high-throughput chemistry or material sciences, as existing approaches rely on manual handling or slow deterministic microsampling that do not readily scale.

Here we present STORMS, an automated STOchastic Robotic MicroSampling system that enables fast, reliable and parallel sampling of solid materials at sub-milligram and milligram masses. 

Read more about Parameter efficient multi-model vision assistant for polymer solvation behaviour inference
News Picture 1 1 V2
Jun
30

Parameter efficient multi-model vision assistant for polymer solvation behaviour inference

Polymer–solvent systems exhibit complex solvation behaviours encompassing a diverse range of phenomena, including swelling, gelation, and dispersion. Accurate interpretation is often hindered by subjectivity, particularly in manual rapid screening assessments. While computer vision models hold significant promise to replace the reliance on human evaluation for inference, their adoption is limited by the lack of domain-specific datasets tailored, in our case, to polymer–solvent systems.

© Chemspeed (part of Bruker) 2026