Ab initio calculations have greatly advanced our understanding of homogeneous catalysts and crystalline heterogeneous catalysts. In contrast, amorphous heterogeneous catalysts remain poorly understood. The principal difficulties include (i) the nature of the disorder is quenched and unknown; (ii) each active site has a different local environment and activity; (iii) active sites are rare, often less than ∼20% of potential sites, depending on the catalyst and its preparation method. Few (if any) studies of amorphous heterogeneous catalysts have ever attempted to compute site-averaged kinetics, because the exponential dependence on variable activation energy requires an intractable number of ab initio calculations to converge. We present a new algorithm using machine learning techniques (metric learning kernel regression) and importance sampling to efficiently learn the distribution of activation energies. We demonstrate the algorithm by computing the site-averaged activity for a model amorphous catalyst with quenched disorder.
31 October 2019 in Reaction Chemistry & Engineering
Ab initio computational studies have made tremendous progress in describing the behavior of molecular (homogeneous) catalysts and crystalline versions of heterogeneous catalysts, but not for amorphous heterogeneous catalysts. Even widely used industrial amorphous catalysts like atomically dispersed Cr on silica remain poorly understood and largely intractable to computational investigation. The central problems are that (i) the amorphous support presents an unknown quenched disordered structure, (ii) metal atoms attach to various surface grafting sites with different rates, and (iii) the resulting grafted sites have different activation and catalytic reaction kinetics. This study combines kernel regression and importance sampling techniques to efficiently model grafting of metal ions onto a non-uniform ensemble of support environments. Our analysis uses a simple model of the quenched disordered support environment, grafting chemistry, and catalytic activity of the resulting grafted sites.
In an infrequent metadynamics (iMetaD) simulation, a well-tempered metadynamics bias accumulates in the reactant basin, accelerating escapes to the product state. Like the earlier hyperdynamics strategy, iMetaD enables estimates of the unbiased escape rates. However, iMetaD applies the bias to visited locations in a collective variable (CV) space, not to the more specific visited locations in a full configuration space as done in hyperdynamics. This difference makes rate estimates from iMetaD sensitive to the choice of CVs, to parameters that control the bias deposition rate, and to the preparation of the initial state within the reactant basin. This paper uses an extremely simple discrete state model to illustrate complications that can arise in systems that exhibit fluxional transitions between sub-basins of the reactant state. Specifically, we show how the reactant-to-product escape time and relaxation times within the reactant basin(s) impose bounds on the admissible parameter choices for an iMetaD calculation. Predictions from the discrete state model are validated by iMetaD simulations on a corresponding two-dimensional potential energy surface.
For disordered catalysts such as atomically dispersed “single-atom” metals on amorphous silica, the active sites inherit different properties from their quenched-disordered local environments. The observed kinetics are site-averages, typically dominated by a small fraction of highly active sites. Standard sampling methods require expensive ab initio calculations at an intractable number of sites to converge on the site-averaged kinetics. We present a new method that efficiently estimates the site-averaged turnover frequency (TOF). The new estimator uses the same importance learning algorithm [Vandervelden et al., React. Chem. Eng. 5, 77 (2020)] that we previously used to compute the site-averaged activation energy. We demonstrate the method by computing the site-averaged TOF for a simple disordered lattice model of an amorphous catalyst. The results show that with the importance learning algorithm, the site-averaged TOF and activation energy can now be obtained concurrently with orders of magnitude reduction in required ab initio calculations.
An efficient data-driven literature-based discovery approach is utilized, that is capable of extracting knowledge from materials science-related text, related to Pulsed Laser Deposition (PLD), aiming to extract various processing parameters that are critical during the PLD process.