The first work package of SELECTFLOW focused on the rational design of additives (specifically, Lewis acids) for the development of selective photochemical C4-functionalization of N-heterocycles (such as pyridines), which are scaffolds commonly found in pharmaceutically-active molecules, using ubiquitous inert alkanes as coupling partners. In order to identify the most suitable additives, a combined computational and experimental mechanistic study was conducted. First, the interaction between pyridine and different additives was studied computationally. By comparing the stabilization of the different pyridine-additive adducts using frontier molecular orbital theory (FMO), we could determine, potentially, which additives would form the more reactive adduct. We observed that most of the typically-used Lewis acids, did not stabilize the adducts enough to undergo reactivity. Alongside the FMO study, we also analysed the electrophilicity of the adducts by conducting natural population analysis (NPO). By doing so, we could study the regioselectivity of the reaction. Among all the computed structures, only a few additives were found to be promising. We then studied the pyridine-additive adducts experimentally by 1H NMR spectroscopy. Despite testing a wide variety of additives, we could not find establish relevant correlation. In parallel, we conducted experimental validation of the most promising results from the mechanistic studies. For this, we performed various photochemical reactions and studied the results, both in terms of yield and regioselectivity. After testing several different Lewis acids under various conditions, we did not observe any striking difference in terms of selectivity towards the targeted C4-functionalized pyridine. Zinc-based additives displayed selectivity, but at the expense of very slow reactivity. When we attempted to translate these results to microfluidic conditions, we could not observe any product formation, most likely due to technical issues, such as clogging of the photoreactor.
For the second work package, we decided to turn our attention toward the use of gaseous reagents for the modification of drug-like compounds. To achieve this, we studied the regioselective installation of carbon monoxide gas and light and heavy alkanes into organic compounds. Due to the advantages of working under microfluidic conditions, the safe handling of hazardous CO and gaseous alkanes could be achieved. By employing a photocatalytic methodology, we developed an unprecedented method for the upgrading of light hydrocarbons at ambient temperature, thereby opening the door to the use of these readily available feedstocks as coupling partners.
Finally, we targeted the third and last work package of SELECTFLOW: the development of environmentally-friendly photochemical processes that minimize waste and resource consumption. Key challenges developing efficient photochemical reactions include optimization, replication and scale-up. These challenges arise from non-reproducible light absorption and experimental variability and are responsible for large use of resources and produce waste. In response to the need for efficient optimization of complex photocatalytic reaction conditions, we developed a robotic platform that iteratively determines optimal, substrate-specific reaction conditions for photochemical processes in a scalable, flow-based architecture. A machine learning algorithm enabled the development of a closed-loop optimization process, describing tailored reaction conditions for each substrate across different reactivities without human intervention. Operating autonomously, the platform eliminates the need for extensive expertise in photocatalysis or scaling processes to achieve optimal results. This makes this platform a valuable collaborative robotic setup suitable for any synthetic organic chemistry laboratory, irrespective of users’ specific familiarity with photocatalysis.
