Fast and Reliable Metamodeling of Complex Reaction Spaces Using Universal Kriging

Abstract

We report the application of metamodeling algorithms based on Universal Kriging for the controlled synthesis of compound semiconductor nanoparticles. Application of such a metamodel allows the prediction of reaction outcomes at arbitrary points within sparsely sampled parameter spaces as a function of reaction conditions. To demonstrate the applicability of Universal Kriging to chemical reaction screening within microfluidic reaction systems CdSe and CdSeTe quantum dots were synthesized by using a segmented flow capillary reactor. Variation of input reagent flows (to control reagent concentrations and reaction residence times) and online spectroscopic monitoring of product characteristics was achieved in a fully automated manner. The resulting fluorescence spectra are analyzed to extract the fwhm, wavelength maximum, and intensity of the band-edge emission. These values are subsequently used as inputs for the Universal Kriging metamodeling algorithm to predict the reactor output at arbitrary points within accessible parameter space. Results demonstrate that the algorithm can predict reaction outcomes with high accuracy and reliability.