CSwinUKAN, a crack-segmentation model for tunnel linings that combines Swin Transformer and Tokenized KAN within an enhanced U-Net, is introduced in the study. The model outperformed multiple baselines and delivered notably higher recall, reducing missed detections in safety-critical inspections.

Researchers have developed a programmable macrophage engineering strategy that rapidly equips immune cells with pathogen-specific recognition and antibacterial functions through membrane interface modification. The engineered macrophages efficiently capture and eliminate bacteria while simultaneously reshaping the immune microenvironment, providing a promising therapeutic approach for severe bacterial infections occurring under immunosuppressed conditions.

Researchers from Beihang University, Truth Memory Corporation, Aix-Marseille University and Université Paris-Saclay has successfully transitioned the emerging physics concept of "orbitronics" from theoretical study to practical application in next-generation magnetic random-access memory (MRAM). By designing a highly efficient ruthenium-tungsten write electrode, the researchers demonstrated ultrafast writing of MRAM devices with a pulse of 28.7 ps, and solved critical industrial bottlenecks in MRAM yield and area penalty, paving the way for ultra-low-power memories.

Landau quantization is a cornerstone of quantum Hall effects. However, controlling Landau-level spacing and the spatial profiles of Landau modes through complex-valued gauge fields has remained largely unexplored. Now, a collaborative team from Beijing Institute of Technology, The Hong Kong University of Science and Technology, University of Exeter, and The University of Hong Kong, writing in Science Bulletin, demonstrates a mechanism for tuning Landau quantization. By constructing a real pseudo-magnetic field and an imaginary pseudo-electric field, the researchers enlarge the Landau-level spacing and enhance mode localization. These findings establish a versatile framework for tailoring non-Hermitian Landau levels and may advance applications in mode-selective filtering, field concentration, and wave-based sensing.

The release of marine isoprene is an important component of material exchange between the ocean and the atmosphere, and its long-term variability and climatic drivers have attracted widespread attention. A research team led by Professor Jialei Zhu at Tianjin University constructed a global marine isoprene emission flux dataset spanning 1950–2020 using machine learning and an improved parameterization scheme. Basing on this dataset, they have revealed mechanisms through which the El Niño–Southern Oscillation (ENSO) modulates the interannual variability of marine isoprene emissions by influencing sea surface temperature, solar radiation, surface winds, phytoplankton activity, and other environmental factors. This study has been published in 2026 Chinese and English editions of SCIENCE CHINA Earth Sciences.

A multi-institutional research team reviews progress in CO₂-involved organic synthesis over the past five years. By organizing reactions according to C-C, C-N, and C-O bond formation, the review highlights the growing role of photochemistry and electrochemistry in converting CO₂ into value-added organic products.

A recent study by Lanzhou University's Center for Pan-Third Pole Environment has revealed a climate-sensitive, multi-sphere bacterial metacommunity on the Tibetan Plateau, reshaping our understanding of microbial connectivity across alpine ecosystems. Analyzing 3,373 samples from five habitats, the team identified precipitation as a key driver with dual effects—homogenizing most ecosystems while enhancing microbial uniqueness in glaciers. The breakthrough calls for integrating microbial monitoring into plateau ecological security assessments.