Real-time inline-IR-analysis via linear-combination strategy and machineś learning for automated reaction optimization.

Automation has revolutionized many fields by improving efficiency, accuracy, and reproducibility. However, in organic chemistry, automating key tasks such as reaction optimization and analysis remains a significant challenge. To accelerate advancements in organic chemistry research and development, we propose a fully automated system based on real-time inline analysis performed by Fourier-transform infrared spectroscopy and assisted by a neural network model. To rapidly collect data, a linear combination of spectral intensities was used as training data for a yield prediction model. Using this model, we demonstrated real-time yield prediction of Suzuki-Miyaura cross-coupling with remarkable accuracy. By combining this yield prediction model with real-time inline analysis and a flow chemistry setup, we have developed a fully automated system for the rapid and efficient optimization of reaction conditions and process analysis.