Atomic layer deposition (ALD) is a promising approach to stabilizing supported metal nanoparticles; however, conventional oxide overcoating often forms continuous layers that obstruct active sites and compromise catalytic performance. Here, we present a synthesis-integrated strategy for inherent area-selective ALD, in which the interfacial reactivity established during nanoparticle synthesis directs subsequent oxide nucleation. Using Ag/ZnO as a model system, we demonstrate that synthesis-programmed surface chemistry governs the nucleation and spatial distribution of ALD-grown ZnO. Microwave-assisted polyol synthesis generates oxygen-containing Ag surface species that serve as favorable nucleation sites, creating an intrinsic reactivity contrast between the Ag nanoparticles and carbon support. This promotes preferential ZnO nucleation on Ag without additional surface pretreatment or blocking agents. The resulting nucleation-dominated ZnO nanodomains help stabilize the electrochemical response of Ag and mitigate surface changes during repeated cycling in alkaline media while preserving catalytic accessibility. In H2/O2 anion exchange membrane fuel cells, the selectively modified Ag/C cathode delivers improved performance and stability compared with its conventionally prepared counterpart. More broadly, this work demonstrates that synthesis-programmed surface chemistry provides a practical and versatile approach for controlling ALD nucleation and engineering metal–oxide interfaces, with broader applicability to other supported catalyst systems for electrochemical energy conversion.
In this paper, we study piezoelectric materials with multiscale porous microstructure. Piezoelectric materials have wide applications in modern electronic devices, while the porous microstructure can be beneficial for specific applications, such as ultrasonic transducers. For the numerical simulation of such materials, homogenization methods are widely employed to determine the effective material properties at each macroscopic point. However, their applicability becomes limited for materials with complex multiscale porous microstructures, where intricate microscale interactions cannot be adequately represented by standard homogenization frameworks. In this paper, we propose a multicontinuum modeling approach for porous piezoelectric materials based on multicontinuum homogenization. We introduce several interacting continua representing channels of different sizes and derive the corresponding multicontinuum model. To this end, we formulate multicontinuum expansions of the fine-scale mechanical displacement and electric potential based on macroscopic variables. Coupled constrained cell problems are then introduced to capture the effective homogenized behavior. Utilizing the corresponding cell solutions, we rigorously derive a generalized multicontinuum piezoelectricity model valid for an arbitrary number of continua. Numerical experiments involving model problems with two continua are presented to validate the proposed model.
Plant disease detection and severity estimation are critical for agricultural sustainability, yet existing approaches struggle with limited severity annotations, fail to incorporate disease progression dynamics, and lack biological grounding. This paper presents a novel epidemiologically-constrained multimodal framework that addresses these challenges through four integrated stages. First, we introduce Epidemiologically-Constrained Differential GANs (Delta-GAN) that generate biologically plausible diseased images by incorporating epidemiological growth models and environmental modulation factors Phi(T, H, L), eliminating the need for manual severity labels. Second, Disease-Aware Progressive Super-Resolution Network (DAPSR-Net) enhances image resolution while preserving pathological features through attention-guided upsampling, achieving 94.1% lesion preservation accuracy. Third, Severity-Aware Multimodal Knowledge Fusion (SA-MKF-KG) integrates visual, semantic, and severity embeddings through Graph Convolutional Networks, creating hierarchical disease-severity representations. Finally, Dynamic Multi-Task Learning employs attention-based fusion of five CNN architectures with Wasserstein distance for optimal transport-based severity matching. The framework achieves state-of-the-art performance with 99.78% classification accuracy and 99.52% severity estimation accuracy on a multi-crop dataset spanning seven plant species and 12 disease types. Cross-dataset evaluation on PlantDoc field images demonstrates 91.3% accuracy without retraining, reducing the laboratory-to-field gap from over 34% to under 9%. SHAP-CAM analysis and counterfactual explanations provide interpretable insights, revealing task-specific feature utilization and minimal intervention strategies. By integrating epidemiological modeling, adaptive fusion, and explainable AI, this framework offers a scientifically grounded, transparent solution for precision agriculture, enabling early disease detection and informed intervention strategies across diverse environmental conditions.
Abstract Tubulin glutamylation is an essential post-translational modification that expands the functional diversity of microtubules in many cellular structures, including flagella, motile cilia, primary cilia, centrosomes, and neurons. This modification adds variable lengths of glutamate side chains to the C-terminal tails of tubulin, creating a finely tuned biochemical signal that regulates microtubule stability, motor protein movement, the activity of severing enzymes, and the recruitment of key signaling molecules. Growing evidence shows that glutamylation is not uniformly distributed but instead forms distinct spatial patterns along microtubule arrays, particularly within the axonemes of flagella and cilia, centriolar triplets, and long-lived neuronal microtubules. These patterns are established by tubulin ligase–like enzymes that add glutamates and by carboxypeptidases that remove them, together shaping a dynamic “tubulin code.” In motile cilia and flagella, glutamylation fine-tunes dynein-driven force generation and the coordination of axonemal bending. Disruption of this modification impairs ciliary beating and sperm flagellar motility, leading to disorders such as primary ciliary dyskinesia, which manifests as chronic respiratory infections and laterality defects, and can also disrupt cerebrospinal fluid flow, causing hydrocephalus and male infertility such as asthenozoospermia. In primary cilia, reduced glutamylation perturbs intraflagellar transport and ciliary signaling and contributes to ciliopathies including Joubert syndrome and retinal degeneration. In dividing cells, altered glutamylation on centrosomes leads to errors in chromosome segregation and is associated with cancer progression. This review summarizes current knowledge of the enzymes, structural principles, and cellular mechanisms governing tubulin glutamylation, highlights its emerging roles in human diseases, and discusses new technological advances—including biochemical reconstitution, super-resolution imaging, and live-cell manipulation tools—that are beginning to reveal how this modification dynamically controls microtubule properties and the functions of flagella, cilia, and centrosomes in health and disease.
Grid resilience is crucial in light of power interruptions caused by increasingly frequent extreme weather events. Well-designed energy management systems (EMS) have made progress in improving microgrid resilience through the coordination of distributed energy resources (DERs), but still face significant challenges in addressing the uncertainty of load demand caused by extreme weather. The integration of deep reinforcement learning (DRL) into EMS design enables optimized microgrid control strategies for coordinating DERs. Building on this, we proposed a cooperative multi-agent deep reinforcement learning (MADRL)-based EMS framework to provide flexible scalability for microgrids, enhance resilience and reduce operational costs during power outages. Specifically, the gated recurrent unit with a gating mechanism was introduced to extract features from temporal data, which enables the EMS to coordinate DERs more efficiently. Next, the proposed MADRL method incorporating action masking techniques was evaluated in the IEEE 33-Bus system using real-world data on renewable generation and power load. Finally, the numerical results demonstrated the superiority of the proposed method in reducing operating costs as well as the effectiveness in enhancing microgrid resilience during power interruptions.
