Related Experiment Video
Updated: May 12, 2026

Automated Robotic Liquid Handling Assembly of Modular DNA Devices
Published on: December 1, 2017
Rethinking catalysis: interpretable AI and description of real-world conditions via materials genes.
Lucas Foppa1,2, Matthias Scheffler1
1The NOMAD Laboratory at the Fritz Haber Institute of the Max Planck Society, Berlin, Germany. foppa@fhi-berlin.mpg.de.
Researchers developed an interpretable artificial intelligence (AI) method to identify key "materials genes" that predict catalytic performance. This approach links material properties to catalytic outcomes without complex atomistic details.
Area of Science:
- Catalysis
- Materials Science
- Artificial Intelligence
Background:
- Traditional catalysis descriptors rely on mechanistic understanding under vacuum conditions.
- Real-world catalysis involves complex interplay of high pressures, temperatures, material restructuring, and transport phenomena.
- Bridging the gap between fundamental understanding and industrial conditions is crucial for catalyst design.
Purpose of the Study:
- To introduce an interpretable artificial intelligence (AI) approach for identifying key physicochemical parameters correlating with catalytic performance.
- To develop statistically derived descriptors ('materials genes') that bypass explicit atomistic descriptions.
- To determine influential parameters for supported palladium-based metal alloy nanoparticle selectivity in acetylene hydrogenation.
Main Methods:
- Utilized the sure-independence-screening-and-sparsifying-operator (SISSO) symbolic-regression AI approach.
- Combined SISSO with sensitivity analysis based on partial derivatives.
- Applied the method to supported palladium-based metal alloy nanoparticles for concentrated acetylene stream hydrogenation.
Main Results:
- Identified key 'materials genes' correlating with catalytic selectivity.
- The most influential genes included the calculated average d-band center and measured average particle diameter.
- These findings highlight the importance of adsorption properties and structure sensitivity in ethylene formation.
Conclusions:
- The interpretable AI approach effectively identifies critical descriptors for catalysis under realistic conditions.
- Average d-band center and particle diameter are key factors influencing selectivity in palladium-catalyzed acetylene hydrogenation.
- This 'materials gene' concept offers a powerful statistical framework for catalyst discovery and optimization.
Related Concept Videos
Introduction to Mechanisms of Enzyme Catalysis
Introduction to Mechanisms of Enzyme Catalysis
Catalysis
Catalysis
Catalytically Perfect Enzymes
Heterogeneous Catalysis
