Predicting Long-Time-Scale Kinetics under Variable Experimental Conditions with Kinetica.jl

Area of Science:

• Computational Chemistry
• Materials Science
• Chemical Engineering

Background:

• Predicting molecular degradation over long timescales is crucial for industrial materials design.
• Constructing accurate chemical reaction networks with kinetic data is computationally intensive.
• Existing methods struggle with the scale and complexity of realistic reaction systems.

Purpose of the Study:

• To introduce Kinetica.jl, a novel software package for automated chemical reaction network generation and kinetic modeling.
• To enable the efficient simulation of complex chemical systems over extended time scales.
• To bridge the gap between theoretical reaction networks and experimental observations.

Main Methods:

• A kinetics-driven algorithm explores chemical reaction space to construct large-scale networks.
• Machine learning models predict activation energies for elementary reactions.
• Symbolic-numeric modeling and discrete kinetic approximation enable efficient long-time-scale simulations.
• Hydrocarbon pyrolysis was simulated using transient temperature profiles.

Main Results:

• Kinetica.jl can generate and characterize networks with approximately 10^3 chemical species and 10^4-10^5 reactions.
• The software allows flexible and efficient computation of kinetic profiles under variable temperature conditions.
• Accurate long-time-scale kinetic profiles were propagated for automated reaction network refinement.
• Successful demonstration of hydrocarbon pyrolysis simulation over second timescales.

Conclusions:

• Kinetica.jl provides an automated solution for generating, characterizing, and modeling complex chemical reaction systems.
• The package facilitates direct connections between computational models and experimental data.
• This approach significantly advances the capability for predicting molecular degradation and designing new materials.