ChemXDyn: Dynamics-Informed Species and Reaction Detection Methodology from Atomistic Simulations

Area of Science:

• Computational Chemistry and Chemical Physics
• Materials Science and Engineering
• Reaction Dynamics and Kinetics

Background:

• Accurate chemical identification from molecular dynamics (MD) is crucial for kinetic modeling and mechanistic discovery.
• Current methods using distance thresholds misclassify transient interactions, leading to inaccurate reaction networks and rates.
• Need for a dynamics-aware approach to robustly identify chemical bonds and reactions.

Purpose of the Study:

• Introduce ChemXDyn, a computational methodology for dynamics-aware identification of chemical species and reaction pathways.
• Leverage time-resolved interatomic distance signatures to robustly identify bonded interactions.
• Distinguish genuine bond dynamics from non-reactive encounters using valence and coordination constraints.

Main Methods:

• Developed ChemXDyn, a novel computational methodology.
• Utilizes time-resolved interatomic distance (IAD) signatures.
• Propagates molecular connectivity while enforcing atomic valence and coordination constraints.

Main Results:

• ChemXDyn suppresses unphysical species and recovers experimentally consistent reaction pathways in MD simulations.
• Successfully applied to hydrogen, ammonia, and methane oxidation simulations using ReaxFF and neural-network potentials.
• Improves fidelity of rate constant estimation compared to static threshold-based methods.

Conclusions:

• ChemXDyn provides a robust and transferable foundation for MD-derived reaction networks and kinetics.
• Enhances mechanistic understanding in reactive systems like combustion, catalysis, and plasma chemistry.
• Enables more accurate prediction of chemical behavior at the atomistic level.