哈尔滨师范大学

Magnetic field-driven spin polarization modulation has emerged as an effective way to boost the electrocatalytic oxygen evolution reaction (OER). However, the correlation among catalyst structure, magnetic property, and magnetic field enhanced-electrochemical activity remains to be fully elucidated. Herein, single-domain CoFe2O4 catalysts with tunable oxygen vacancies (CFO-VO) were synthesized to probe how VO mediates magnetism and OER activity under magnetic field. The introduction of VO can simultaneously modulate saturation magnetization ( Ms ) and coercivity (Hc), where the increased Ms dominates the magnetic field-enhanced OER activity. Under a 14,0 0 0 G magnetic field, the optimized CFO-VO exhibits up to 16.1 % reduction in overpotential and 365 % enhancement in magnetocurrent (MC). Electrochemical analyses and post-OER characterization reveal that the magnetic field synergistically improves OER kinetics through lattice distortion induction, magnetohydrodynamic effect, and spin charge transfer effect. Importantly, the magnetic field promotes additional Co3+ generation to compensate for charge imbalance caused by VO filling, maintaining dynamic equilibrium of VO and effective reactant adsorption-conversion processes. This work unveils the synergistic mechanism of VO and magnetic parameters for enhancing OER performance under the magnetic field, providing new insights into the design of high-efficiency spin-regulated OER catalysts. (c) 2025 Published by Elsevier B.V. on behalf of Chinese Chemical Society and Institute of Materia Medica, Chinese Academy of Medical Sciences.

The development of multifunctional electromagnetic wave-absorbing materials is essential for next-generation flexible electronics and intelligent protection systems. Herein, a novel three-dimensional porous MXene-based film integrated with metallic nickel nanoparticles (Ni-PMF) is designed and synthesized with the potential to address the urgent need for multifunctional electromagnetic wave-absorbing materials in next-generation intelligent systems. By using polystyrene spheres as sacrificial templates, a hierarchical porous architecture is constructed to prevent MXene nanosheet restacking, extend electromagnetic wave propagation paths, and optimize impedance matching. Simultaneously, uniformly distributed Ni nanoparticles introduce abundant heterogeneous interfaces, enhancing interfacial polarization and magnetic loss, which significantly improve electromagnetic wave attenuation. The Ni-PMF film achieves a minimum reflection loss of –64.7 dB and a broad effective absorption bandwidth of 7.2 GHz, covering the full Ku-band and outperforming most reported MXene thin film absorbers. In addition to superior electromagnetic wave absorption, the film demonstrates excellent electrothermal conversion and flexible strain-sensing capabilities, enabling integrated protection and real-time sensing functions. This multifunctional material offers promising potential for next-generation smart flexible electronic systems.

Hyperspectral Unmixing (HU) is a pivotal technology for separating mixed pixels, extracting endmember spectral signatures, and quantifying their spatial distribution from remote sensing images. In recent years, the state space model mamba has brought breakthrough progress to hyperspectral unmixing due to its long-sequence modeling capability and linear computational complexity. However, the complex coupling effect among spectral variability, noise interference, and model ill-posedness remains a critical bottleneck constraining the accuracy of hyperspectral unmixing. To address this challenge, we innovatively propose a multimodal uncertain state space fusion network based on generative endmember modeling (MUSF-GEM). This network employs a dual-branch architecture capable of synergistically extracting feature representations from complementary input modalities. Specifically, this network introduces an uncertainty modeling mechanism, which dynamically generates adaptive weights by explicitly modeling the intrinsic noise characteristics and their uncertainties of hyper-spectral image (HSI) and light detection and ranging (LiDAR). These weights dynamically adjust the contribution of each modality's features during the fusion process. Meanwhile, to effectively capture long-range dependencies in pixel-level spectra, we propose an uncertainty-guided Mamba module. This module employs a noise-aware mechanism to adaptively suppress noise interference during feature extraction, reducing potential noise disturbances when acquiring global features. Furthermore, to more effectively extract features with clearer spatial structures, we enhance the spatial features of LiDAR by constructing morphological profile (MP). The outcomes on two real datasets further validate the efficacy of the proposed approach.

Deep learning-based fault diagnosis methods often assume that training and testing labels are consistent, which limits their ability to detect unknown fault types. Moreover, another problem is the scarcity of labeled samples in actual engineering, which hinders the effective training of deep learning models. Therefore, a new semi-supervised adversarial discrimination and deep evidential fusion (SAD-DEF) approach is proposed. Firstly, a semi-supervised deep neural network is designed through adversarial learning, which enables the model to capture more general features. Subsequently, a new classifier based on deep evidence fusion method is proposed to achieve known class diagnosis and uncertainty estimation. This uncertainty can effectively detect unknown fault class. Finally, a modified uncertainty threshold is designed. Four case studies illustrate that the SAD-DEF effectively diagnoses known class faults and detect unknown class faults under limited labeled samples.

The development of efficient, low-cost, and stable electrocatalysts remains a central challenge for sustainable hydrogen production. In this work, we systematically designed a series of single transition metal (TM) atoms (including Ti, V, Cr, Mn, Fe, Co, Ni, Cu, Zr, Nb, Mo, Ru, Rh, Pd, Ag, Hf, Ta, W, Os, Ir, Pt, and Au) anchored on a WS2 monolayer, and investigated their catalytic activity for the hydrogen evolution reaction (HER) using density functional theory (DFT) calculations. The comprehensive results reveal that TM atoms can effectively tune the electronic structure, electrical conductivity, and hydrogen adsorption behavior of WS2. In particular, Cr@WS2, Fe@WS2, Mo@WS2, and Ru@WS2 catalysts exhibit Gibbs free energies of hydrogen adsorption close to zero, indicating their promising HER activity. Further crystal orbital Hamilton population (COHP) analyses and Bader charge calculations demonstrate that the chemical bonding characteristics and charge transfer between the TM atoms and the WS2 substrate play a pivotal role in tuning adsorption strength and catalytic performance. This work provides a theoretical foundation for optimizing WS2-based single-atom catalysts and opens new avenues for the design of highly efficient HER electrocatalysts based on two-dimensional materials.