Effect of errors in linear scaling relations and Brønsted–Evans–Polanyi relations on activity and selectivity maps

[Display omitted] •Uncertainty quantification and global sensitivity analysis of high throughput microkinetic models.•Errors in linear scaling relations dominate uncertainty in identifying most active and selective catalysts.•Uncertainty in predicted materials is low but in rates and selectivities is high.•Key paths identified via global sensitivity and stochastic reaction path analyses. We introduce for the first time uncertainty quantification and global sensitivity analysis to assess the effect of error in linear scaling relations (LSRs) and Brønsted–Evans–Polanyi (BEP) relations on activity and selectivity maps of microkinetic models, which have recently been used for in silico prediction of new materials. The method is applied to ethanol hydrodeoxygenation (HDO). Selectivity trends are driven by adsorbate thermochemistry rather than kinetics. Uncertainty quantification calculations show that the most likely location of the maximum conversion can be estimated to be within about 10kcal/mol in the C and O binding energies. Broad selectivity trends are even more robust. Model rates show uncertainties of 2–3 orders of magnitude about the median. Uncertainty in the activity predictions is dominated by that of the LSRs. Our calculations demonstrate that there is a common initial mechanism of ethanol HDO and decomposition, and the stability of decomposition products is crucial to determining the selectivity.