This work employed the attapulgite (ATP) clay as a potentially cost-effective mineral support and constructed a series of high-performance Ni-based catalysts for CO2 methanation through a multi-step synergistic strategy, including acid washing pretreatment, calcination temperature optimization, and promoter incorporation (La2O3, CeO2, and Co2O3). The effects of different preparation parameters on the structure and catalytic performance of the catalysts were systematically investigated. The results demonstrated that acid washing pretreatment effectively removed surface impurities from ATP, optimized the pore structure, and enhanced the accessibility of active sites. The calcination temperature at 400 degrees C enabled high NiO dispersion while preserving the structural stability of the support. Notably, the introduction of La2O3 as a promoter further refined the NiO crystallite size, strengthened metal-support interactions, and enriched surface basic sites and active oxygen species. As a result, the optimal catalyst, Ni-La2O3/A-ATP-400, achieved a CO2 conversion exceeding 50% at 300 degrees C, accompanied by a low apparent activation energy of 54.23 kJ/mol. In addition, it exhibited excellent CH4 selectivity (>90%) and good catalytic stability, maintaining high conversion during a 20 h time-on-stream test. Comprehensive characterization using N-2 physisorption, XRD, TEM, H-2-TPR, CO2-TPD, XPS, and in-situ DRIFTS revealed the preservation of a stable mesoporous structure, highly dispersed active species, regulated surface chemical properties, and a dominant formate-mediated reaction pathway. This study provided systematic experimental insights and theoretical guidance for the rational design of efficient CO2 methanation catalysts based on natural mineral supports.
Photocatalytic nitrate reduction to ammonia has attracted increasing attention as a sustainable strategy for simultaneous ammonia synthesis and nitrate pollutant remediation under mild conditions. Compared with conventional nitrogen fixation routes, nitrate reduction possesses higher reactant solubility and lower activation barriers, making it more favorable for solar-driven nitrogen upcycling. Nevertheless, the reaction involves complicated multistep proton-coupled electron transfer processes and suffers from sluggish kinetics, charge recombination, and competing side reactions, which severely limit ammonia selectivity and efficiency. This review systematically summarizes recent advances in photocatalytic nitrate-to-ammonia conversion from both mechanistic and catalyst-design perspectives. Fundamental thermodynamics, reaction pathways, intermediate evolution, and rate-determining steps are first discussed, followed by recent progress in electronic structure regulation, defect and active-site engineering, heterostructure construction, interfacial polarization modulation, reaction pathway control, and microenvironment engineering. Particular attention is given to the intrinsic correlation among charge dynamics, intermediate stabilization, and reaction selectivity. Emerging insights from operando characterization and theoretical calculations are also highlighted to bridge experimental observations with mechanistic understanding. Finally, current challenges involving ammonia quantification, selectivity regulation, mechanistic ambiguity, and practical scalability are critically discussed, together with future opportunities in mechanism-oriented catalyst design and data-driven catalyst discovery.
Deep hashing has potential application prospects in large-scale image retrieval due to its low-cost and efficient retrieval advantages. However, traditional instance-based hashing methods match images by directly comparing their hash codes, but this process is easily disturbed by appearance factors such as color, angle, and background, resulting in semantic mismatches. Approaches based on centroid/proxy representations have been proposed to alleviate the above problems. Existing centroid/proxy-based approaches are limited to training and do not optimize retrieval efficiency, which goes against the core benefits of hashing. In this paper, by using the mean centroid representation in both training and retrieval, a novel Mean Centroid Representation for Deep Hashing framework (MCRDH) is proposed to learn a highly discriminative embedding space, realizing intra-class aggregation and inter-class separation. Specifically, by calculating the mean centroid of each class of samples, a novel mean centroid triplet loss is designed to alleviate the negative impact of outliers, thereby generating compact and discriminative hash codes. In addition, since the samples of each class are represented by a mean centroid, the number of target vectors in the database set is significantly reduced, thereby greatly narrowing the search space, reducing retrieval time and storage requirements, which is suitable for production deployment. Comprehensive experiments conducted on three public datasets demonstrate that our proposed MCRDH framework surpasses various state-of-the-art methods. The source code of our proposed MCRDH framework is available at: https://github.com/QinLab-WFU/MCRDH.
With the dual advantages of low storage cost and high retrieval efficiency, deep cross-modal hashing has shown great application prospects in multimedia retrieval scenarios. Deep cross-modal hashing typically projects heterogeneous samples into the Hamming space and generates binary codes by capturing class-defined semantic information. However, sample variability stems not only from class differences but also from structured features (e.g., the angle or orientation of an image, the sentence pattern of a text). Therefore, existing methods that only focus on class-defined semantic features and regard such structured features as unknown random noise may weaken the discriminative ability of standard hash algorithms in generating binary codes. To address the above challenges, a novel Deep Augmented Inter-class Hashing (DAIH) framework is proposed to generate highly discriminative discrete hash codes by learning latent features shared across categories. Specifically, a multi-modal dual-encoder architecture is designed to learn class-discriminative information while explicitly separating and learning cross-class shared information. In this architecture, the class encoder focuses on learning discriminative features that distinguish different categories, while the auxiliary encoder specializes in learning latent features shared across categories. Secondly, a clustering label strategy is proposed to capture class-irrelevant information without additional attribute annotations, enabling the auxiliary encoder to effectively model the features shared across categories. Thirdly, to further enhance the independence between the feature spaces of the dual-encoders, an adversarial mutual information loss is proposed, which reduces the overlap between the two feature spaces by utilizing the adversarial setup of the gradient reversal layer, thereby improving the discriminative ability of the class encoder. Finally, extensive experiments are conducted on three benchmark datasets, verifying the advantage of the DAIH model over various state-of-the-art cross-modal baseline methods. The source code of the DAIH model is available for download at https://github.com/QinLab-WFU/DAIH.
Clinopyroxene (Cpx), a common mineral in basalt, has become increasingly important in elucidating mantle source and magmatic processes. Basalts collected from Pohnpei Island, a site of hotspot activity linked to the Caroline mantle plume, are rich in Cpx phenocrysts, of which some exhibit zoning textures. This study presents the in situ major and trace element compositions, as well as the Sr isotope ratios, of Cpx phenocrysts in Pohnpei basalts. The Cpx crystals in samples DS1 and DS2 exhibit core-rim zoning, characterized by a large and homogeneous core and a slightly brighter, very thin rim. The Cpx crystals in samples S7 and S8 display complex core-mantle-rim zoning. The dark domains of these Cpx crystals have higher Mg# values but lower FeO and TiO2 contents than the bright domains. The zoned Cpx crystals in S7 and S8 basalts display larger compositional variations than those in DS1 and DS2. Notably, the Cpx zones in four samples (S7, S8, DS1, and DS2) exhibit similar rare earth element (REE) distribution patterns and 87Sr/86Sr values (0.703 426–0.704 622), suggesting a common or similar mantle source. We suggest that the Cpx crystals in the Pohnpei basalts crystallized from multiple batches of heterogeneous mixing magma that were derived from an EM II-like end-member and mixed with Pacific-type depleted mantle components. The zoning textures and compositional variations recorded in these Cpx crystals provide insights into complex magma chamber processes, including fractional crystallization (involving olivine, Cpx, and Fe-Ti oxide) and magma mixing. Pressure-temperature calculations using the Cpx-only thermometer and barometer indicated that the Cpx crystals in DS1, DS2, S7, and S8 crystallized within the lower crust under similar temperatures and pressures. This study contributes to a comprehensive understanding of mantle source compositions, magmatic processes, and the complex mixing within magma chambers beneath Pohnpei Island.
