An experimental platform for semi-autonomous kinetic model refinement combining optimal experimental design, computer-controlled experiments, and optimization leads to new understanding of N[formula omitted]O + O

Abstract

Automated platforms could enable the unprecedented pace required for modeling the countless proposed alternative fuels relevant to future technologies, but existing platforms rely exclusively on automatically generated computational data and lack experimental validation. Here, we introduce an experimental platform for rapid kinetic model refinement that combines optimal experimental design, computer-controlled experiments, and optimization—each of which are tailored in novel ways to form a cohesive platform together. The optimal experimental design considers (1) realistic uncertainties in both experimental conditions and measurements and (2) diverse reactant mixtures that can include “chemical sensitizers,” which are not necessarily reactants in a specified Quantity of Interest (QoI) but sensitize kinetic information relevant to a QoI. Similarly, our High-Throughput Jet-Stirred Reactor (HT-JSR) features (1) rapid multi-species diagnostics that can measure dozens of species within minutes, (2) a flow delivery system that can prepare up to ∼10-component reactant mixtures, and (3) computer-controlled operation of all components. Finally, a post-processing code automatically retrieves data from the instruments, quantifies uncertainties in both experimental conditions and measurements, and produces self-contained files usable for optimization in our MultiScale Informatics software. This platform is demonstrated for the rate constant for N2O + O ⇌ N2 + O2 as the QoI where, unlike for N2O + O ⇌ NO + NO, proposed values at ∼1000 K span ∼5 orders of magnitude yet give equally good agreement with previous experimental data—suggesting that previous experiments fail to constrain the branching ratio of N2O + O. The results show that optimal conditions with more species as both reactants and analytes—particularly NO2 as both a reactant and analyte—enable unambiguous discrimination of the main products of N2O + O. Specifically, the data preclude N2 + O2 as the main products at ∼1000 K—contrary to most recently proposed values.Novelty and significance statementWe introduce a novel experimental platform for rapid kinetic model refinement that combines optimal design, computer-controlled experiments, and optimization—each uniquely tailored to form a cohesive platform. It notably employs high levels of automation, high-throughput multi-species diagnostics, and uniquely diverse reactant mixtures to accelerate scientific discovery. This platform is shown to achieve a goal that no previous experiment has: decipher the main products of N2O + O. This success—attributable to the platform’s unique design principles—demonstrates the effectiveness of this experimental platform as a tool for accelerating scientific discovery and kinetic model development.