(560co) Enhancing Organic Electrosynthesis through Artificial Intelligence: The Case of Adiponitrile Electrohydrodimerization

The increase of fossil fuel-derived carbon dioxide (CO₂) emissions has made their impact in the environment self-evident. The chemical sector is a major contributor to this increase, accounting for more than 5.5% of the global CO₂ emissions. The electrification of the chemical industry can represent a major step in the integration of renewable electricity in the industrial sector. Inorganic electrosynthetic processes have been largely implemented in industry (e.g., chloro-alkali and Aluminum production), however electrosynthetic methods for the production of organic chemicals is severely limited. Adiponitrile (ADN) is a precursor to Nylon 6,6 and it can be obtained via the electrohydrodimerization of acrylonitrile (AN). Although the electrochemical production of adiponitrile (ADN) is the largest organic electrochemical process in industry, it faces many challenges in terms of selectivity control. Optimizing the composition of the electrolyte has resulted in significant performance improvements, however mass transport limitations have proven to be a challenge at high current densities, where propionitrile (PN) is the main reaction by-product. Here, we provide insights of the effects of pulsed potential techniques on the mitigation of mass transport limitations and the effective control of product distribution. Moreover, an approach that combines experimental insights and artificial intelligence is proposed for the rapid optimization of operation conditions and the maximization of reaction performance.

The regulation of electron-transfer rates with the use of complex pulse potential waveforms is implemented to control the electrochemical environment surrounding the reaction surface, ultimately favoring ADN production. Electrochemical pulses balance reactant diffusive fluxes to the electrode and the generation/consumption of species in the electrical double layer (EDL), helping mitigate mass transport limitations at high current densities. An improvement of over 250% in the production ratio of ADN:PN (desired to undesired product) was achieved optimizing pulse duration and amplitude. Furthermore, the careful control of overall reaction time and composition of the EDL led to a 20% increase in ADN production with respect to DC operation.

The integrated use of machine learning (ML) algorithms enabled the prediction of ADN formation in a complete landscape of pulse durations. Using ML, a new set of optimal conditions was identified, leading to a 30% increase in ADN production with respect to DC operation for periodic pulses varying from -60 mA cm^-2 (120 ms) to 0 mA cm^-2 (5 ms). A new paradigm in organic electrosynthesis research is proposed, where a combination of electrochemical design principles, artificial neural networks, and judiciously designed experimental campaigns are used to uncover optimal electrosynthetic processes.