Unveiling the Knowledge of a Parallel Synthesizer Generated Database of RAFT Polymerizations

Combinatorial chemistry and high-throughput testing have been indispensable tools in the materials sciences since their introduction in the 1970s. Several configurations, including automated parallel synthesizers, flow-chemistry setups, and pipetting robots, have been used to produce experimental data automatically.<br/>Since there are so many combinations of useable monomers, solvents, initiators, and other reaction conditions that can be combined with different methods of polymerization to turn a monomer into a polymer, the topic of polymer sciences has attracted a lot of attention. In the past, human intuition was usually utilized to gather the knowledge produced by campaigns to sample a particular subset of polymerizations. The conclusions of the experimental work were then published, and with any luck, the original and primary data supporting those conclusions were also made available to the public as part of the publication's supporting documentation. As the community strives to implement machine learning models, some of which are data-hungry, to further explain certain trends in the data collected, it is of utmost importance to simplify access to the data generated.<br/>Hence, in this work, a synthesis robot has been utilized to perform more than 450 polymerizations in an automated approach in batches of 15 reactions. During the reaction time of 15 hours, 14 samples per polymerization – either for size-exclusion chromatography or NMR spectroscopy – were collected automatically to generate a dataset consisting of more than 4600 data points. The polymerization parameters were varied between 15 different monomers, seven RAFT-agents and three solvents. The subset of monomers consisted of five monomers each, from the group of acrylates and methacrylates with differing substituents at the ester moiety. Additionally, five styrenic monomers with differing <i>para</i>-position substituents were examined. Dimethylsulfoxide, dimethylformamide, and toluene were the investigated solvents utilized to collect data on contrasting polarity regimes. All of the radical polymerizations utilized azobis(isobutyronitril) as initiator, and were carried out at 70 °C. The collected data points were curated into several hundred kinetics and fitted with suitable growth functions so that data points could be extrapolated and interpolated from one another. The resulting knowledge was made accessible through an online interface, enabling a rapid search for the ideal reaction conditions to synthesize a polymer with desired characteristics, like molar mass.<br/>This work cannot only be seen as a presentation of a high-throughput generated database of homopolymers. It also illustrates the possibilities and restrictions of using such a database and provides insights into data curation and augmentation.<br/>In the future, an extension of the dataset seems to be possible with, e.g., further RAFT-agents, different temperatures, additional initiators or further monomers as test points. Furthermore, the data provided could be utilized to examine trends regarding the polymerization behavior of certain monomer groups with different substituents at the relevant moiety in combination with specific RAFT-agents.