New tricks by very old dogs: predicting the catalytic hydrogenation of HMF derivatives using Slater-type orbitals

Abstract

We report new experimental results on the hydrogenation of 5-ethoxymethylfurfural, an important intermediate in the conversion of sugars to industrial chemicals, using eight different M/Al 2 O 3 catalysts (M = Au, Cu, Ni, Ir, Pd, Pt, Rh, and Ru) under various conditions. These data are then compared with the results for 48 bimetallic supported catalysts. The results are explained using a simple and effective model, applying catalyst descriptors based on Slater type orbitals (STOs). Each metal is described using four parameters: the height of the orbital peak, the distance of the peak from the metal atom centre, the peak width at half height, and the peak skewness. Importantly, all these parameters are derived from one simple equation, so the calculation is fast and robust. We then apply these descriptors for modeling the hydrogenation data using multivariate methods. Despite the inherent complexity of the reaction network, these simple models describe the catalysts' performance well. The general application of such descriptor models to in silico design and performance prediction of solid catalysts is discussed.

Designing bifunctional alkene isomerization catalysts using predictive modelling

“…This workflow is an iterative data-driven approach that enhances synthetic efforts by using descriptor models to highlight "good regions" in the catalyst space. 9 Such a data-driven approach has proven highly effective for predicting catalyst performance in a variety of reactions, including hydrocyanation, 10,11 fluoroarylation, 12 hydrogenation, 13,14 oxidation, 15,16 C-H functionalization, 17 oxygen reduction, 18 C-N coupling, 19 and C-C coupling. [20][21][22][23][24] Here, in a collaboration between the two groups, we applied this workflow to predicting the activity and selectivity of ruthenium zipper catalysts for isomerising 1-hexene or 1-heptene to the respective internal 2-E and 3-alkenes (Scheme 2).…”

Machine-learning prediction of the d-band center for metals and bimetals

“…[20][21][22][23][24] Considering that first-principles calculations are too time-consuming to explore the full spectrum of possibilities, and on the other hand, a great amount of data is being generated and accumulated in the field, ML methods can give a fast and high-precision alternative to the first-principles models. However, ML methods in catalysis [25][26][27][28][29][30][31][32][33][34] are still in their infancy.…”

Machine learning for heterogeneous catalyst design and discovery

“…Data mining methods are powerful ML tools to find nontrivial insights in big data and to help build predictive models. Efforts have been made to integrate data mining methods with heterogeneous or homogeneous catalysis data to promote catalyst characterization and to build quantitative structure‐property relationship models . An early study used data mining to help make predictive models of cyclohexene epoxidation yield by mesoporous titanium‐silicate catalysts .…”