Optimal on-line sampling of parallel reactions: general concept and a specific spectroscopic example

Abstract

The ‘gap’ between the discovery and the optimisation phases in high-throughput systems is described and the problems that stem from the centralised nature of parallel reactor set-ups are outlined. It is proposed that on-line analysis methods (e.g. fast spectroscopic analysis) can be adapted to perform kinetic studies in high-throughput environments. Strategies for performing time-resolved analysis are examined using a (pseudo)first-order reaction that is monitored by on-line FT-NIR as an example. Stochastic modelling is used to compare the efficiency of the various strategies in the determination of accurate first-order kinetic constants. The effects of the sampling time, the ‘sample window’ size, and the number of samples on the error in k obs are studied. A new sampling strategy is introduced that is based on equidistant sampling along the concentration axis. This strategy is adapted to realistic conditions using a simple iterative algorithm. The application of this iterative approach to automatic intelligent on-line analysis of chemical reaction kinetics is discussed.