Development of a Machine Accessible Kinetic Databank for Radical Polymerizations

Development of a Machine Accessible Kinetic Databank for Radical Polymerizations

Machine-learning is a young discipline in the chemical sciences that has nonetheless led to significant changes in research approaches in a relatively short time span. Any machine-assisted research approach requires training sets and machine-readable databases to retrieve information from. A standardization of notations allows for data exchange between computer systems and softwares. As an example, the recently introduced BigSmiles (simplified molecular-input line-entry system) notation (Lin et al. ACS Cent. Sci. 2019, 5, 9, 1523-1531; https://doi.org/10.1021/acscentsci.9b00476) allows to exchange structural data for polymer across computer systems.

The IUPAC working party on Modelling of Polymerization Kinetics and Processes has collated significant kinetic data on free radical polymerization in recent years, and published a series of benchmark papers on the topic. While generally available, still many researchers do not make full use of these data sets. A central database will increase awareness and foster better use of the data. More importantly, a machine-readable database will allow for direct and automated exchange of data. For example, kinetic models can always retrieve the latest and most updated kinetic data for specific monomers. In machine-learning approaches, algorithms can make use the data for deep learning and interconnection with other data such as molecular characteristics, physical properties or further kinetic data. This can range from prediction of materials properties to automated process control in synthesis.

The kinetic database will consist of all IUPAC benchmarked kinetic data for free radical polymerization. A further selection of reliable kinetic data will be made to also include monomers that have not yet been critically assessed. For these monomers, the database can serve as a future starting point for data collection. While not part of this project, the same database could later be extended by other parameters, such as overall time conversion relations, molecular weights, and physical properties of the resulting polymers from polymerization. The database will be designed in a fashion to allow facile extension to either direction.
First versions of the database will be hosted via Monash University. Source codes will be published open access and long term migration of the database to central servers is envisaged.