(653d) Leveraging Disparate Experimental and Ab Initio Data to Model the Reaction Network of CO2 Upgrading to C1 Products

CO₂ can be upgraded to a number of C1 molecules, viz., carbon monoxide, methanol, formic acid, methane, and formaldehyde on transition metals. In this research, we aim to develop a unified and accurate predictive model of this overall reaction network to reliably infer the mechanism on a specific metal facet and thereby identify globally sensitive parameters for catalyst design. In our experience, microkinetic models parameterized using density functional theory (DFT) alone (which is the state-of-the-art) often do not provide the requisite predictive accuracy. In such circumstances, the sensitive kinetic and thermodynamic parameters can be reparameterized to related experimental data. However, such closely related and reliable kinetic data may not always be available. For instance, no experimental dataset is available that simultaneously analyzes the entire C1 product spectrum on a set of transition metal facets.

In this work, we develop a new data-driven method to systematically fuse the disparate datasets available to us, viz., ab initio (DFT) and experimental (singly crystal thermochemistry and kinetics) data, to build a model for calculating the energetics of all species and reactions in the aforementioned reaction network on any specified transition metal surface. We then use these energies to build an improved microkinetic model of the reaction network.