Reaction Network and Kinetic Modeling of Free Radical Polymerization Reactions Via First-Principles and Machine Learning Approach

Research Interests: reaction network, kinetic modeling, computational chemistry, first-principles, machine-learning.

Polymerization modeling remains less understood due to its complexity involving enormous amount of intermediates despite of its wide application in chemical industry. Generation of reaction network and kinetic modeling of polymerization can shed light to dominant reaction pathway under specific process environment. Using predetermined SMARTS reaction templates of typical free radical polymerization reaction of styrene, acrylic acid, and acrolein, possible elementary reactions and intermediates are exhaustively enumerated and the stable ensemble of conformers of each molecule has been obtained. Among vast intermediate chemical space, fraction of them were optimized with density function theory and their energies were trained with machine learning (ML) model with kernel ridge regression to predict atomization energy of the remaining part. Kinetic parameters are determined using automated fashion in conjunction with ML, which is fed as input for subsequent kinetic modeling. The computed molecular weight distribution and thermochemistry data is compared with experimental result and the sources of error are discussed.