Catlas: an automated framework for catalyst discovery demonstrated for direct syngas conversion

Abstract

Catalyst discovery is paramount to support access to energy and key chemical feedstocks in a post fossil fuel era. Exhaustive computational searches of large material design spaces using ab-initio methods...

Vision 2050: Reaction Engineering Roadmap

“…Transient kinetics, using techniques such as the temporal analysis of products (TAP) and the Steady-State Isotopic Transient Kinetic Analysis (SSITKA), have been advanced but remain singular expertise . Machine learning techniques for the design of experiments, the discovery of descriptors, and improving the accuracy of models are becoming popular. , Still, their use for catalyst discovery and reactor design is not routine yet. The widespread application of these tools is aided by easier access to many open-source libraries of machine learning tools, the vast improvements to the machine learning API (Application Programming Interface), and the high expectations of the community.…”

Catalyst Energy Prediction with CatBERTa: Unveiling Feature Exploration Strategies through Large Language Models

“…The search for optimal catalyst materials for specific reactions poses a significant challenge in the development of sustainable chemical processes. Traditional avenues of exploration have involved laborious experiments or computationally intensive quantum chemistry calculations, exemplified by density-functional theory (DFT) simulations. , Nevertheless, the requirement to assess a vast array of systems makes the catalyst screening for optimality more challenging. , This is attributed to the fact that a singular bulk catalyst can exhibit a range of surface orientations. ,, Additionally, adsorbates have the potential to bind to numerous distinct adsorption sites on these surfaces, with varying orientations . As such, relying solely on DFT calculations proves inadequate for swiftly assessing the vast array of potential adsorbate–catalyst combinations because of their time and resource demands.…”

“…1,2 Nevertheless, the requirement to assess a vast array of systems makes the catalyst screening for optimality more challenging. 3,4 This is attributed to the fact that a singular bulk catalyst can exhibit a range of surface orientations. 3,5,6 Additionally, adsorbates have the potential to bind to numerous distinct adsorption sites on these surfaces, with varying orientations.…”

“…3,4 This is attributed to the fact that a singular bulk catalyst can exhibit a range of surface orientations. 3,5,6 Additionally, adsorbates have the potential to bind to numerous distinct adsorption sites on these surfaces, with varying orientations. 7 As such, relying solely on DFT calculations proves inadequate for swiftly assessing the vast array of potential adsorbate−catalyst combinations because of their time and resource demands.…”

Clarifying trust of materials property predictions using neural networks with distribution-specific uncertainty quantification

“…Effective uncertainty estimates of predictions are an important design element to enhance several advanced ML strategies for materials discovery, such as high-throughput screening [17,18], transfer learning [19], and active learning [20,21], as illustrated in figure 1. For catalyst discovery, one can ascertain if an ML model is making accurate predictions of catalyst behavior by confirming the predicted properties experimentally or through first-principles calculations.…”