Search

Abstract The advent of artificial intelligence (AI) has enabled a comprehensive exploration of materials for various applications. However, AI models often prioritize frequently encountered material examples in the scientific literature, limiting the selection of suitable candidates based on inherent physical and chemical attributes. To address this imbalance, we generated a dataset consisting of 1,453,493 natural language-material narratives from OQMD, Materials Project, JARVIS, and AFLOW2 databases based on ab initio calculation results that are more evenly distributed across the periodic table. The generated text narratives were then scored by both human experts and GPT-4, based on three rubrics: technical accuracy, language and structure, and relevance and depth of content, showing similar scores but with human-scored depth of content being the most lagging. The integration of multimodal data sources and large language models holds immense potential for AI frameworks to aid the exploration and discovery of solid-state materials for specific applications of interest.

Metal-organic frameworks (MOFs) are attractive materials with easily tunable porous structures. Their selective carbon dioxide (CO2) capture ability can be varied by altering the functionality of the organic ligands. However, rule-based approaches to tuning and developing MOFs with high CO2 capture and conversion abilities are hindered by the numerous possible combinations of metal ions and organic linkers. Recently, machine learning (ML) has been applied to unravel key descriptors in predicting the performance of MOFs. This review summarizes recent advancements in ML models for MOFs in CO2 capture and utilization, including high-throughput screening, neural network interatomic potential, and generative models. The development of sophisticated ML models for designing high-performance MOFs will play a critical role in addressing climate change in the future. Finally, the main challenges and limitations of current approaches in designing high-performance MOFs are discussed.

Coupled electrochemical systems for the direct capture and conversion of CO2 have garnered significant attention owing to their potential to enhance energy- and cost-efficiency by circumventing the amine regeneration step. However, optimizing the coupled system is more challenging than handling separated systems because of its complexity, caused by the incorporation of solvent and heterogeneous catalysts. Nevertheless, the deployment of machine learning can be immensely beneficial, reducing both time and cost owing to its ability to simulate and describe complex systems with numerous parameters involved. In this review, we summarized the machine learning techniques employed in the development of CO2 capture solvents such as amine and ionic liquids, as well as electrochemical CO2 conversion catalysts. To optimize a coupled electrochemical system, these two separately developed systems will need to be combined via machine learning techniques in the future.

Summary: CO2 capture and its electrochemical conversion have historically developed as two distinct technologies and scientific fields. Each process possesses unique energy penalties, inefficiencies, and costs, which accrue along the mitigation pathway from emissions to product. Recently, the concept of integrating CO2 capture and electrochemical conversion, or “electrochemically reactive capture,” has aroused attention following early laboratory-scale proofs-of-concept. However, the integration of the two processes introduces new complexities at a basic science and engineering level, many of which have yet to be clearly defined. The key parameters to guide reaction, electrolyte, electrode, and system design would, therefore, benefit from delineation. To begin this effort, this perspective outlines several crucial physicochemical and electrochemical considerations, where we argue that the absence of basic knowledge leaves the field of designing metaphorically in the dark. The considerations make clear that there is ample need for fundamental science that can better inform design, following which the potential impacts of integration can be rigorously assessed beyond what is possible at present.