Learning Chemistry of Complex Reaction Systems via a Python First-Principles Reaction Rule Stencil (pReSt) Generator

Area of Science:

• Computational Chemistry
• Catalysis
• Chemical Reaction Engineering

Background:

• Automated network generators create complex reaction networks from reactants and rules.
• Current methods rely on user-defined rules, often neglecting catalyst geometry, which impacts selectivity.
• Transferring rules between gas-phase and surface processes is common but can be imprecise.

Purpose of the Study:

• To develop a first-principles-based framework for reaction mechanism generation.
• To "learn chemistry" from published reaction mechanisms using density functional theory (DFT) data.
• To enable the discovery of novel reaction pathways and assess data quality.

Main Methods:

• Utilizing density functional theory (DFT) data from published reaction mechanisms.
• Developing a framework that learns chemical rules and reaction pathways.
• Implementing the framework with Python Reaction Stencil (pReSt) software.
• Testing the framework across multiple catalytic chemistries.

Main Results:

• The framework generates previously unstudied reaction networks.
• It identifies and flags new reactions for further computational testing (DFT convergence).
• It reconciles differences between catalysts and reactants, revealing new pathways.
• Demonstrated efficacy across various catalytic systems.

Conclusions:

• The proposed framework offers a novel approach to reaction mechanism generation.
• It serves as a diagnostic tool for assessing mechanism data quality.
• It facilitates the discovery of new molecular pathways in catalysis.