(513fa) Bayesian Uncertainty Quantification for Tafel Slope Estimation

The Tafel slope is an important metric for catalyst performance that is nearly ubiquitously reported in papers that develop a new electrocatalyst. Established literature practices for estimating the Tafel slope from experimental current-voltage data rely heavily on subjective human intervention and fail to leverage all available data, resulting in flawed uncertainty estimates. We have developed a Bayesian data analysis approach that obviates human intervention and makes use of the data in its entirety to produce a distributional Tafel uncertainty estimate. We have validated our approach by re-analyzing over 100 Tafel datasets from the CO2 reduction literature. Re-analysis of the data using our method reveals no systematic tendency towards “cardinal values” of the Tafel slope predicted by simple electrochemical rate theories. This suggests that microkinetic mechanisms in CO2 reduction routinely violate the ideality assumptions employed in deriving these theories. Additionally, we use synthetic data to demonstrate that current literature practices for Tafel estimation are susceptible to serious latent data insufficiency issues. Our method flags these issues, and can be employed in an iterative fashion with data collection to address removable experimental sources of uncertainty. Finally, we show that our approach can be generalized to fitting more complicated kinetic models encountered regularly across several sub-fields in catalysis. We anticipate that the distributional error quantification afforded by our method will be indispensable for accurately proposing and discriminating between microkinetic mechanisms on the basis of experimental data.