道达尔

Geological carbon sequestration (GCS) will play a critical role in decarbonization and in facilitating the transition to clean energy systems. Because CO2 is highly mobile, ensuring its safe and permanent injection into subsurface geological formations involves monitoring over larger spatial domains and longer time periods than is typical for hydrocarbon reservoirs. This can benefit from simulation tools capable of modeling key CO2 trapping mechanisms, particularly those optimized for speed and scalability on high- performance computing systems. Using isothermal versions of the SPE11B and SPE11C benchmark cases, we conduct a mesh refinement study simulating CO2 injection into kilometer- scale rock formations at centimeter resolution with the GEOS open- source simulation framework. We focus on how mesh refinement improves the accuracy of convective mixing in both 2D and 3D simulations. The computational costs associated with achieving a converged solution highlight the need for predictive upscaling techniques. A systematic performance scaling analysis- including both central processing unit (CPU) and graphics processing unit (GPU) architectures-complements the "Results" section.

Important hydrocarbon resources are preserved in deeply buried (>6500 m), intensely karstified Ordovician carbonates of the Tarim Basin, where reservoir development is closely associated with strike–slip fault systems. However, the links among faulting, karstification, reservoir formation, and subsequent diagenetic modification remain poorly understood because of the complex tectonic and burial history of the basin. In this study, we compared deeply buried Ordovician karst reservoirs from the Halahatang oil field in the Tabei Uplift, Tarim Basin, with near-surface Cretaceous karst systems at Fontaine de Vaucluse, southeastern France. Both of these systems were influenced by multiphase strike–slip faults. Despite their contrasting deposition timing and burial histories, the two systems exhibited comparable linear reservoir distributions controlled by a sequential evolution process that formed under compressional stress. In both cases, diagenetic fluids were characterized by mixing between two end-members, meteoric water and formation water, although they may have been circulated to different depths. Moreover, epigenic karstification was the principal mechanism of reservoir formation in both cases, although hypogenic products occur locally. These similarities indicated that the deeply buried (>8000 m) Ordovician paleokarst reservoirs of the Halahatang area formed primarily because of fault-controlled epigenic karstification and associated meteoric-water diagenesis. This comparative framework increases our understanding of the evolution of ultradeep carbonate reservoirs and has implications for hydrocarbon exploration in the Ordovician carbonates of the Tabei Uplift.

Synthetic data has gained attention for training large language models, but poor-quality data can harm performance (see, e.g., Shumailov et al. (2023); Seddik et al. (2024)). A potential solution is data pruning, which retains only high-quality data based on a score function (human or machine feedback). Previous work Feng et al. (2024) analyzed models trained on synthetic data as sample size increases. We extend this by using random matrix theory to derive the performance of a binary classifier trained on a mix of real and pruned synthetic data in a high dimensional setting. Our findings identify conditions where synthetic data could improve performance, focusing on the quality of the generative model and verification strategy. We also show a smooth phase transition in synthetic label noise, contrasting with prior sharp behavior in infinite sample limits. Experiments with toy models and large language models validate our theoretical results.

The Comoros archipelago located in the western Indian Ocean has been a topic of debate for ca. 50 years regarding its origin. Various mechanisms have been proposed to explain its formation, ranging from the impact of a mantle plume to the development of a plate boundary between the Somalia and Lwandle plates. Determining the timing of the volcanic activity is crucial to understanding the archipelago's origin. Despite recent geochronological studies, the age of the initial volcanic eruptions on the islands remains uncertain due to the difficulty of accessing the earliest lavas. This study uses high- quality seismic reflection profiles and regional stratigraphy to identify the first volcanic series that marked the onset of volcanism on each island of the Comoros Archipelago. Our findings reveal that localized volcanic activity began ca. 32 Ma at Z & eacute;l & eacute;e and Geyser banks and Mayotte edifice, the eastern portions of the archipelago, much earlier than previously believed. Volcanism spread across the Comoros archipelago ca. 9-8 Ma, from the northern part of the Mayotte edifice to Moh & eacute;li Island in the west. Ca. 4 Ma, volcanic activity occurred on Anjouan Island and the Jumelles seamounts, followed by Grande Comore Island ca. 2 Ma. This progression of ages from east to west indicates a chronological sequence over time. The timing of volcanic activity in the Comoros archipelago is similar to the magmatism evolution documented at Madagascar and along the East African Rift System (EARS). Magmatic activity began in the late Oligocene, followed by quiescence during the middle Miocene, and resumed in the late Miocene, coinciding with widespread deformation along the EARS, including its offshore branches and Madagascar. Our study shows that the regional tectonic control of volcanic activity in the Comoros archipelago began during the early Miocene, thereby suggesting that the Comoros archipelago developed as an offshore branch of the EARS south of the Somalian plate at that time.

This paper examines the strategic adoption of artificial intelligence (AI) by authoritarian leaning regimes in the Middle East, focusing on Iran, Saudi Arabia, and the United Arab Emirates (UAE). By analyzing the commonalities and divergences in AI-driven governance practices across these states, the study explores the broader implications of AI on authoritarian control. Employing an interdisciplinary theoretical approach, it traces how AI-powered technologies such as facial recognition, predictive policing, and digital censorship have transformed surveillance into an adaptive and pervasive mechanism of state control. The study highlights the dual nature of these technologies, demonstrating their capacity to enhance governance efficiency while simultaneously reinforcing political repression. To conceptualize the evolving landscape of digital governance, the paper, drawing on comparative case studies from the Middle East, introduces the TRIAD Framework (Technology, Regulation, Influence, and Adaptive Dynamics). This novel analytical model elucidates the complex interplay between technological advancements, regulatory changes, and geopolitical influence, thereby contributing to the understanding of how these factors sustain authoritarian resilience within the region. The framework illustrates how AI not only fortifies existing mechanisms of control but also fosters a feedback loop wherein technological innovations drive regulatory evolution and reinforce autocratic adaptability.