Machine Learning-Assisted Development of High-Performance Ethanol Synthesis Catalysts via CO2 Hydrogenation

Area of Science:

• Catalysis
• Materials Science
• Chemical Engineering

Background:

• High-performance catalyst discovery is vital for chemical processes.
• Machine learning (ML) shows potential for accelerating catalyst development but has limitations in discovering novel catalysts.
• Current ML approaches often rely on traditional element pools, restricting innovation.

Purpose of the Study:

• To develop highly efficient catalysts for ethanol synthesis via CO2 hydrogenation using an ML approach.
• To expand catalyst discovery beyond traditional elements by incorporating previously unstudied ones.
• To create a robust, data-driven discovery system combining ML predictions and experimental validation.

Main Methods:

• Utilized a closed-loop discovery system involving 24 iterations of ML predictions and experimental validation.
• Started with an initial dataset of 58 catalysts and expanded it to 555 tested catalysts.
• Employed advanced characterization techniques, including in situ/operando X-ray absorption spectroscopy (XAS), ambient-pressure X-ray photoelectron spectroscopy (AP-XPS), and diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS).

Main Results:

• Discovered over 50 catalysts with superior activity for ethanol synthesis.
• Identified a highly effective multielemental catalyst: Pd(0.8)-Au(0.3)/K(2.5)-Sr(1)-Fe(20)-Zn(4)-Cd(2)-Yb(1)-Re(1)/CeO2(25%)-ZrO2.
• Achieved an ethanol space-time yield of 8.2 mmol gcat-1 h-1 with 57.6% CO2 conversion and 23.2% ethanol selectivity.

Conclusions:

• The ML approach successfully identified novel, high-performance catalysts by exploring an expanded element pool.
• The study highlights the critical role of individual elements in the multielemental catalyst for enhancing ethanol synthesis efficiency.
• This data-driven methodology provides a powerful framework for accelerating the discovery of advanced catalytic materials.