Mechanistically Interpretable Artificial Intelligence for Designing Oxygen Electrocatalysts

Area of Science:

• Electrochemistry
• Materials Science
• Artificial Intelligence

Background:

• Oxygen reduction and evolution reactions (ORR/OER) are crucial for energy conversion but designing effective electrocatalysts is challenging due to their complex nature.
• Structure-property relationships for ORR/OER electrocatalysts are difficult to establish across diverse operating conditions.

Purpose of the Study:

• To develop an AI-driven framework for rapid discovery and evaluation of novel electrocatalysts.
• To overcome the limitations in rational design of electrocatalysts for ORR/OER.
• To accelerate materials discovery for energy conversion applications.

Main Methods:

• Developed Two-Stage Material Screening (TSMS), an AI framework integrating DFT computations and active-learning.
• Incorporated an experimental feedback loop and mechanistic interpretation into the screening process.
• Applied TSMS to screen over 6.9 million compositions for protonic solid oxide cells (P-SOCs).

Main Results:

• Identified and experimentally validated top-performing electrocatalyst candidates.
• Achieved a peak power density of 2.68 W cm^-2 in fuel cell mode and 3.51 A cm^-2 at 1.3 V in electrolysis mode.
• Demonstrated stable performance over 500 hours at 600°C, with electron affinity, d-p hybridization, and densification resistance identified as key descriptors.

Conclusions:

• TSMS framework enables rapid and systematic discovery of high-performance electrocatalysts.
• The study provides key descriptors for understanding and designing ORR/OER electrocatalysts.
• TSMS offers a versatile and generalizable paradigm for accelerating materials discovery in electrochemistry.