Open Catalyst Experiments 2024 (OCx24): Bridging Experiments and Computational Models (2411.11783v1)

Abstract: The search for low-cost, durable, and effective catalysts is essential for green hydrogen production and carbon dioxide upcycling to help in the mitigation of climate change. Discovery of new catalysts is currently limited by the gap between what AI-accelerated computational models predict and what experimental studies produce. To make progress, large and diverse experimental datasets are needed that are reproducible and tested at industrially-relevant conditions. We address these needs by utilizing a comprehensive high-throughput characterization and experimental pipeline to create the Open Catalyst Experiments 2024 (OCX24) dataset. The dataset contains 572 samples synthesized using both wet and dry methods with X-ray fluorescence and X-ray diffraction characterization. We prepared 441 gas diffusion electrodes, including replicates, and evaluated them using zero-gap electrolysis for carbon dioxide reduction (CO$_2$RR) and hydrogen evolution reactions (HER) at current densities up to $300$ mA/cm$^2$. To find correlations with experimental outcomes and to perform computational screens, DFT-verified adsorption energies for six adsorbates were calculated on $\sim$20,000 inorganic materials requiring 685 million AI-accelerated relaxations. Remarkably from this large set of materials, a data driven Sabatier volcano independently identified Pt as being a top candidate for HER without having any experimental measurements on Pt or Pt-alloy samples. We anticipate the availability of experimental data generated specifically for AI training, such as OCX24, will significantly improve the utility of computational models in selecting materials for experimental screening.

• The paper introduces a combined experimental and computational methodology that synthesizes 572 catalyst samples and screens 19,406 materials.
• The paper demonstrates that correlating experimental data with computed descriptors enables accurate predictions of cell voltage and identification of promising HER catalysts.
• The study establishes a unified dataset that informs AI-driven catalyst discovery and highlights the need for refined models to tackle CO2RR complexities.

Insights into "Open Catalyst Experiments 2024 (OCx24): Bridging Experiments and Computational Models"

The paper "Open Catalyst Experiments 2024 (OCx24): Bridging Experiments and Computational Models" addresses two prominent chemical reactions necessary for a shift toward renewable energy: the hydrogen evolution reaction (HER) and the carbon dioxide reduction reaction (CO2RR). These reactions are crucial for developing sustainable energy technologies, yet they face significant challenges due to the inefficiencies and costs associated with catalyst discovery. The authors propose to leverage both experimental studies and computational models to ameliorate these challenges, offering a robust data set for training AI models tasked with predicting and enhancing catalyst performance under industrial conditions.

Experimental and Computational Methodology

The OCx24 initiative employs an integrated approach including high-throughput experimental techniques and extensive computational screening for catalyst discovery. For the experimental component, 572 samples were synthesized using chemical reduction and spark ablation, covering 13 elements and characterized via XRF and XRD techniques. The data were filtered to focus on single-phase materials with potential industrial applicability. These samples were then subjected to HER and CO2RR testing.

Computationally, the authors conducted a large-scale screening of 19,406 materials by calculating the adsorption energies of six intermediates (OH, CO, CHO, C, COCOH, H) using the AdsorbML pipeline. These involved upwards of 685 million structural relaxations and approximately 20 million DFT single-point calculations, marking it as the largest computational screening for catalysts to date.

Model Development and Predictive Capacity

The authors aim to predict catalyst performances by blending experimental results with computational descriptors. For HER, the correlation between experimental data and descriptors was significant. Models were capable of predicting the cell voltage for 19,406 materials, identifying promising HER catalysts, including copper and other low-cost elements.

Conversely, CO2RR posed more significant prediction challenges due to its complexity and diverse potential product outputs. Although correlations were weaker, these initial efforts underscore the potential for model refinement and necessitate the development of more nuanced computational descriptors or experimental data to enhance future models.

Implications and Future Directions

This paper comprehensively assembles a large, well-characterized dataset bridging real-world experiments and advanced computational simulations. The implications for AI-driven catalyst discovery are substantial, as a unified dataset can significantly inform models that predict practical catalyst applications without needing excessive lab resources. These insights are expected to guide the future discovery of efficient, low-cost, and scalable catalysts essential for renewable energy conversion technologies.

In future developments, expanding the dataset size and diversity, including exploring varied synthesis methods and incorporating complex reaction conditions, could further improve the robustness of predictive models, particularly for complex reactions like CO2RR. The success of these endeavors could bolster the collaboration between computational chemistry and experimental methodology, hastening the discovery of sustainable energy solutions.