Active learning streamlines development of high performance catalysts for higher alcohol synthesis

Area of Science:

• Catalysis
• Materials Science
• Chemical Engineering

Background:

• Higher alcohol synthesis (HAS) from syngas is crucial but challenging due to complex catalyst requirements.
• Designing multicomponent catalysts for HAS is hindered by vast chemical spaces and intricate reaction dynamics.

Purpose of the Study:

• To develop an efficient catalyst for syngas-based higher alcohol synthesis.
• To integrate active learning into experimental workflows for streamlined catalyst design.
• To optimize catalyst composition and reaction conditions for enhanced productivity and selectivity.

Main Methods:

• Employed an active learning framework to guide experimental design for the FeCoCuZr catalyst family.
• Conducted 86 experiments to navigate composition and reaction condition space.
• Utilized multi-objective optimization to balance higher alcohol productivity with minimized CO2 and CH4 selectivity.
• Performed feature-importance analysis for data-informed guideline formulation.

Main Results:

• Identified Fe65Co19Cu5Zr11 catalyst achieving 1.1 gHA h-1 gcat-1 higher alcohol productivity, a 5-fold improvement.
• Demonstrated stable operation for 150 hours.
• Uncovered an intrinsic trade-off between higher alcohol productivity and undesired byproduct selectivity.
• Developed Pareto-optimal catalysts beyond human expert discernment.

Conclusions:

• Active learning significantly reduces experimental costs and environmental footprint in catalyst development.
• The developed FeCoCuZr catalyst and optimization strategy offer a new paradigm for HAS.
• The data-driven approach is adaptable to other catalytic transformations, promoting laboratory sustainability.