Search for: All records

Abstract Physics-informed machine learning (PIML), the combination of prior physics knowledge with data-driven machine learning models, has emerged as an effective means of mitigating a shortage of training data, increasing model generalizability, and ensuring physical plausibility of results. In this paper, we survey a wide variety of recent works in PIML and summarize them from three key aspects: 1) motivations of PIML, 2) physics knowledge in PIML, and 3) methods of physics knowledge integration in PIML. We additionally discuss current challenges and corresponding research opportunities in PIML. 

Abstract. Groundwater serves as a crucial freshwater resource for people and ecosystems, playing a vital role in adapting to climate change. Yet, its availability and dynamics are affected by climate variations, changes in land use, and abstraction. Despite its importance, our understanding of how global change will influence groundwater in the future remains limited. Multi-model ensembles are powerful tools for impact assessments; compared to single-model studies, they provide a more comprehensive understanding of uncertainties and enhance the robustness of projections by capturing a range of possible outcomes. However, to date, no ensemble of groundwater models has been available to assess the impacts of global change. Here, we present the new Groundwater sector within ISIMIP, which combines multiple global, continental, and regional-scale groundwater models. We describe the rationale for the sector, the sectoral output variables that underpinned the modeling protocol, and showcase current model differences and possible future analysis. Currently, eight models are participating in this sector, ranging from gradient-based groundwater models to specialized karst recharge models, each producing up to 19 out of 23 modeling protocol-defined output variables. To showcase the benefits of a joint sector, we utilize available model outputs of the participating models to show the substantial differences in estimating water table depth (global arithmetic mean 6–127 m) and groundwater recharge (global arithmetic mean 78–228 mm yr−1), which is consistent with recent studies on the uncertainty of groundwater models, but with distinct spatial patterns. We further outline synergies with 13 of the 17 existing ISIMIP sectors and specifically discuss those with the global water and water quality sectors. Finally, this paper outlines a vision for ensemble-based groundwater studies that can contribute to a better understanding of the impacts of climate change, land use change, environmental change, and socio-economic change on the world's largest accessible freshwater store – groundwater. 

Artificial Intelligence (AI) is poised to revolutionize numerous aspects of human life, with healthcare among the most critical fields set to benefit from this transformation. Medicine remains one of the most challenging, expensive, and impactful sectors, with challenges such as information retrieval, data organization, diagnostic accuracy, and cost reduction. AI is uniquely suited to address these challenges, ultimately improving the quality of life and reducing healthcare costs for patients worldwide. Despite its potential, the adoption of AI in healthcare has been slower compared to other industries, highlighting the need to understand the specific obstacles hindering its progress. This review identifies the current shortcomings of AI in healthcare and explores its possibilities, realities, and frontiers to provide a roadmap for future advancements. 

Metal-organic frameworks (MOFs) with tunable structures and unique host-guest chemistry have emerged as promising candidates for conductive materials. However, the tunability of conductivity and porosity in conductive MOFs and their interrelationship still lack a systematic study. Herein, we report the synthesis of a series of 3D copper MOFs (NU-4000 to NU-4003) using a triphenylene-based hexatopic carboxylate linker. By modulating the ratio of mixed solvents, distinct structural topologies and π-π stacking arrangements were achieved, resulting in electrical conductivity ranging from insulators (˂ 10-6 S/cm) to semiconductors (10-8 ~ 102 S/cm). Among them, NU-4003 features continuous π-π stacking and exhibits a conductivity of 1.7 × 10-6 S/cm. To further enhance conductivity, we encapsulated C60, a strong electron acceptor, within the circular channels of NU-4003, resulting in a remarkable conductivity increase to 140 S/cm with approximately 100% pore occupancy. Even at lower C60 loadings that leave 54% of the pore volume remaining accessible, the conductivity remains exceptionally high at 104 S/cm. This represents an eight-order magnitude enhancement and positions NU-4003-C60 as one of the most conductive 3D MOFs reported to date. This work integrates two charge transport pathways (through-space and electron donor and acceptor) into a single MOF host-guest material, achieving a significant enhancement in conductivity. This study demonstrates the potential of combining host-guest chemistry and π-π stacking to design conductive MOFs with permanent porosity maintained, providing a blueprint for the development of next-generation materials for electronic and energy-related applications.