Abstract
<jats:p>Electrocatalyst performance is often treated as an intrinsic property of the catalyst, although the measured activity depends strongly on the catalyst ink formulation and how the electrode is constructed. This suggests that the catalyst ink design rules established for one catalyst may not translate to other materials. Yet, the design rules for benchmark electrocatalysts are often adapted for structurally distinct materials. Here we use Bayesian optimization over a five-dimensional formulation space (water/ethanol/isopropyl alcohol ratio, ionomer loading, catalyst content and sonication time) to map the ink formulation landscape in alkaline media for two contrasting oxygen evolution reaction catalysts: crystalline IrO2 nanoparticles, the benchmark anode catalyst, and a porous NiFe metal-organic framework. The optimal IrO2 ink exhibits a smooth response surface near the optimum, whereas the metal-organic framework shows a highly non-convex response surface, with performance highly sensitive to the ionomer content. The optimal solvent composition also markedly differs between the two catalysts, with the IrO2 preferring an ethanol-rich medium while the metal-organic framework prefers isopropyl alcohol. Turnover frequency analysis of the metal-organic framework-based catalyst reveals that the ink formulation can alter the quality of the active sites, not merely their number. These results demonstrate that ink rules developed for oxide catalysts do not directly transfer to porous framework-based electrocatalysts. Bayesian optimization therefore provides a data-efficient route to catalyst-specific electrode formulation. In addition, our results highlight ink preparation as a decisive variable in electrocatalyst benchmarking.</jats:p>