The fabrication of self-biased microwave devices demands ferrite materials that integrate high magnetic performance with low-loss characteristics at operational frequencies. M-type barium hexaferrites are promising candidates for millimeter-wave self-biased circulators, yet simultaneously optimizing their saturation magnetization, ferromagnetic resonance (FMR) linewidth, dielectric loss, and thermal conductivity remains a key challenge. To address this issue, Cu-Zn-Sn co-substituted BaFe11.5CuxZn0.25-xSn0.25O19 (x = 0.05-0.20) were synthesized. The optimal composition (x = 0.10) achieves a high saturation magnetization of 65.95 emu g-1, a narrow FMR linewidth of similar to 500 Oe at 9 GHz, a low dielectric loss (epsilon '' approximate to 0.970 at 1 MHz), and a thermal conductivity of 1.944 W m(-1) K-1. Magnetic-field-assisted texturing applied to the x = 0.10 composition induces strong c-axis orientation, as evidenced by significantly enhanced (00l) diffraction peaks and {000l} pole figure analysis, yielding a high remanence ratio (Mr/Ms). The improved remanence ratio arises from preferential grain alignment along the easy magnetization axis, which is critical for self-biased circulator operation. These findings demonstrate that multi-element co-doping combined with texture control provides an effective strategy for balancing the key magnetic, dielectric, and thermal properties of BaM hexaferrites, establishing a viable material platform for millimeter-wave self-biased circulators. (c) 2026 Published by Elsevier Ltd on behalf of The editorial office of Journal of Materials Science & Technology.
Federated Learning (FL) enables collaborative model training without centralizing raw data and has become an important paradigm for privacy-sensitive intelligent systems. Nevertheless, its distributed architecture, heterogeneous data, and unreliable participants introduce substantial robustness and security challenges, ranging from communication noise and system failures to poisoning, privacy inference, and Byzantine attacks. In this paper, we present a structured survey of robust federated learning. We first organize the literature through a threat-oriented taxonomy that covers poisoning attacks, privacy leakage risks, Byzantine behavior, and scenario-specific robustness challenges. We explicitly analyze stronger adversaries that collude, coordinate poisoning campaigns, replace models behind secure aggregation, and adapt their optimization to defense-specific weaknesses. We then review representative defense mechanisms, including robust aggregation, backdoor defense, privacy-preserving training, secure aggregation, and personalized robustness enhancement. Particular attention is given to three recurring tensions in the literature: robustness under Non-IID data, robustness-efficiency trade-offs, and robustness-privacy trade-offs. We further discuss robust federated learning in wireless and IoT environments, where communication uncertainty and resource constraints become first-order concerns. We examine how existing robustness mechanisms translate, or fail to translate, to foundation-model-based FL, federated fine-tuning of large language models, federated reinforcement learning, agentic federated systems, and federated retrieval-augmented generation. Finally, we summarize open challenges and outline future research directions for building more reliable, efficient, and deployable federated learning systems.
Radio-frequency (RF) and microwave dielectric ceramics are key materials for high-frequency electronic devices. Among them, perovskites outperform conventional dielectrics due to their high quality factor ( Q = 1/dielectric loss) and exceptional reliability. However, their practical application is limited by the trade-off between Q & times; f and the temperature coefficient of resonant frequency ( rf ), as well as by their high densification temperature ( >= 1500 degrees C). Here, a synergistic strategy combining B-site complex-ion modification with composite sintering-aid engineering is proposed. The optimized 95CZT05CMN composition exhibits excellent microwave dielectric properties of dielectric constant sr = 32.4, Q & times; f = 34,240 GHz, and rf = + 16.8 ppm/ degrees C at 1540 degrees C. More importantly, the 5G1L0.5C-modified ceramics can be densified at 950-1050 degrees C while retaining favorable dielectric properties ( sr = 27.3-31.9, Q & times; f = 12,920-13,400 GHz, rf = -6.1 to -11.6 ppm/ degrees C), together with a high flexural strength of 212 MPa. Furthermore, a C-band dielectric resonator antenna delivers radiation efficiency above 85%, while multilayer ceramic capacitors (MLCCs) fabricated show excellent C0G-type stability ( AC / C25 degrees C within +/- 0.3%) and a dielectric loss of similar to 0.06% at 1 MHz. These results establish a viable route toward low sintering temperature, low-loss and thermally stable CaZrO3 -based perovskites for advanced RF and integrated electronic applications. (c) 2026 Published by Elsevier Ltd on behalf of The editorial office of Journal of Materials Science & Technology.
Integrating artificial intelligence (AI) into Internet of Things (IoT) is bringing unprecedented conveniences to daily life, but it also introduces severe cybersecurity challenges due to the widespread deployment of IoT devices. Federated learning, with its distributed nature and privacy-preserving characteristics, has emerged as a powerful framework for building Intrusion Detection Systems (IDSs) in IoT environments. It enables collaborative model training across distributed devices without centralizing sensitive data. In this paper, we systematically present typical federated learning-based IDS architectures in IoT environments, along with optimization algorithms that enhance their performance and robustness. We explicitly define a threat model that distinguishes two fundamentally different security layers. The first layer consists of external network intrusions that federated learning-based IDSs aim to detect. The second layer comprises internal model-layer attacks, such as poisoning, backdoors, and gradient leakage, which target the federated learning training process itself. Our review methodology involves retrieving literature from major academic databases (Web of Science, IEEE Xplore, ACM Digital Library, Springer Link, and ScienceDirect) and Google Scholar, with selection based on strict criteria. This survey contributes a structured synthesis and taxonomy of existing federated learning-based IDS literature, a comparative experimental analysis of datasets, and a detailed discussion of optimization strategies organized around the defined threat model. We conclude by discussing current challenges and outlining future research directions, including the critical need for lightweight model design and robust defenses against adversarial attacks on the federated learning.
Vat photopolymerization (VPP) of ceramics has garnered considerable attention owing to its exceptional capability to fabricate components with intricate and complex geometries. Nevertheless, the pronounced scale-dependent effects in ceramic VPP give rise to a widespread yet poorly understood phenomenon: as-printed green bodies frequently fail to withstand subsequent thermal treatment, i.e., “printable-but-not-sinterable.” This dilemma has severely impeded the advancement of ceramic VPP. These issues stem fundamentally from the difficulty in fully understanding and precisely regulating the complex multiphysics coupling phenomena induced by structural features. In single-step processing scenarios, forming quality and precision are governed by the intricate interplay of multiple factors. In multi-stage integrated manufacturing workflows, imperfections originating from upstream processes tend to accumulate and be amplified in subsequent steps. Furthermore, certain geometrical features are inherently incompatible with both shaping and thermal post-processing, rendering the resultant components highly susceptible to warpage, cracking, and loss of dimensional accuracy. This paper reviews multiscale structural forming in ceramic VPP additive manufacturing, focusing on multiphysics coupling mechanisms and their effects on forming quality. To address challenges arising from the long process chain and multiphysics coupling, we propose a holistic process-chain coordination strategy with four-level optimization, guiding the controllable fabrication of high-quality multiscale ceramic components.
