The development of strategic materials such as spinels and perovskites is hampered by the long cycles and low efficiency of traditional trial-and-error methods, for which machine learning (ML) offers a disruptive data-driven paradigm. This review dissects the field's evolution from traditional machine learning (TML), reliant on manual feature engineering, to deep learning (DL) models capable of end-to-end autonomous feature extraction, examining their applications within forward screening and inverse design frameworks. We critically assess how DL overcomes TML's limitations by mapping atomic-level structures to macroscopic properties with greater precision, while also elucidating cutting-edge approaches to address the core challenge of high-quality data scarcity. However, we argue that fully realizing ML's potential requires shifting focus beyond predictive accuracy toward model reliability and robustness. In materials discovery and inverse design, the credibility of predictions depends on accurate uncertainty calibration. Integrating Bayesian learning and confidence-aware modeling can yield statistically reliable guidance for experiments and design decisions, enhancing the stability and trust of AI-assisted research. Future breakthroughs must pivot toward building standardized multi-modal databases, developing interpretable models that can unveil underlying physical mechanisms to overcome the "black-box" problem, and deeply integrating AI with automated experimental platforms to establish a closed-loop research ecosystem that truly accelerates scientific discovery.

Ectopic pregnancy (EP) is a leading cause of maternal morbidity and mortality in early pregnancy. Current diagnostic approaches, which rely on ultrasound scanning and blood measurements, are limited by low detection sensitivity. Herein, we propose a one-step system that uses EP-associated serum metabolic fingerprints (ESF) for efficient EP diagnosis and risk prediction. The system employs nanoparticle-assisted laser desorption/ionization mass spectrometry to rapidly record ESF. A machine learning-based diagnostic model was then used to analyze ESF from 722 participants, achieving an area under the curve (AUC) of 0.913. Simultaneously, potential metabolic biomarkers from ESF were annotated, enabling accurate diagnosis of EP across various clinical profiles (AUC 0.922). Moreover, a rupture risk prediction model was constructed, yielding an AUC of 0.885, significantly surpassing conventional clinical indicators (AUC 0.702, p < 0.05). Our work offers a rapid, effective tool for early EP diagnosis and risk stratification, marking a pivotal advancement toward precision diagnostics.

ABSTRACTWe present particle‐resolved methods for determining the porosity and surface area of microparticles, based on single‐particle trajectory analysis conducted via optical video microscopy integrated with microfluidics and LED illumination. This technique introduces a unique combination of analytical advantages that address key limitations of conventional methods, such as BET nitrogen adsorption. Notably, (1) the method operates with extremely low analyte quantities on the order of micrograms or less and in the form of dilute aqueous dispersions, eliminating the need for drying or bulk sample preparation; (2) surface area quantification is performed on a per‐particle basis, with total surface area determined by summing up all individual particle measurements, enabling high‐resolution analysis of particle‐to‐particle heterogeneity; and (3) the entire workflow from sample preparation to data acquisition and surface area calculation is rapid, straightforward, and relies only on a pre‐defined, self‐generated reference data library. We validate the method using both plain and mesoporous SiO2 microparticles, demonstrating a relative precision of approximately 9%, in line with benchmark techniques such as nitrogen sorption. This approach thus offers a robust, accessible alternative for surface area analysis, utilizing standard optical microscopy equipment available in most laboratory settings, and is particularly well suited for low‐sample‐volume applications in material characterization.

Synthesis of nanomaterials via plasma plume ablation offers a rapid and surfactant-free route for nanoparticles production. However, controlling nucleation under highly non-equilibrium conditions remains a major challenge, often leading to low yield and poor size uniformity. Here we demonstrate that introducing heterogeneous metal ions into the ablation medium provides an effective strategy to regulate plasma plume. Using pulsed laser ablation in liquids as a model system, we show that dissolved metal ions undergo heterogeneous nucleation, thereby simultaneously suppressing plume redeposition. Using silver ions （Ag+） as a representative ion, we identify an optimal concentration window (∼1mm) that maximized nanoparticle yield while producing uniformly small IrAg nanoparticles. A systematic ion library further reveals that the promotion efficiency across different ions correlates strongly with the standard reduction potential (E°), establishing a quantitative link between ion reducibility and plume behavior. These results demonstrate that thermodynamic driving forces bias the relative rates of reduction and nucleation, thereby selecting the dominant condensation pathway under non-equilibrium conditions. This ion-regulated strategy extends the classical LaMer framework into plasma plume synthesis and provides predictive guidelines for scalable, size-controlled nanomaterial production.

Aqueous zinc ion batteries (AZIBs) are emerging as safe, low-cost, and high-capacity devices for sustainable energy storage. Herein, Cobalt-based Prussian Blue Analogues (PBAs: a subclass of porous metal-organic frameworks-MOFs) have been strategically synthesized with different secondary transition metal ions: manganese, iron, and copper, namely MnCoPBA, FeCoPBA, and CuCoPBA for systematic performance evaluation as cathode materials for AZIBs. Among these, MnCoPBA demonstrates the most promising performance, attributed to its highest specific surface area, uniform cube-shaped particle morphology, and dual redox centers. It delivers a specific discharge capacity of approximately 117 mAh g-1 at a current density of 0.2 A g-1 within the voltage window of 0.3-2.0V (vs. Zn2+/Zn), significantly outperforming FeCoPBA (60 mAh g-1) and CuCoPBA (45 mAh g-1). Moreover, the assembled Zn//MnCoPBA cell demonstrates excellent cyclic stability, with capacity retention of ∼89% after 1000 cycles, a coulombic efficiency of ∼99.7%, and superior rate capability, indicating its potential as a viable cathode material for AZIBs. Ex situ PXRD, FE-SEM, HR-TEM, and XPS measurements, along with device-level demonstrations, further confirm the structural robustness and reversibility of Zn2+ storage in the MnCoPBA electrode. This work represents the first methodical assessment of Co-based PBAs in AZIBs, providing valuable mechanistic insights into structure-property relationships.

Flexible and wearable triboelectric nanogenerators (TENGs) are considered promising candidates for mechanical energy harvesting and self-powered sensing, yet simultaneously achieving high output performance together with environmental adaptability remains challenging. In this study, a hydrophobic nanofiber-reinforced paper-based TENG (HF-PTENG) was developed by integrating (3-aminopropyl)triethoxysilane (APTES)-modified polyacrylonitrile/aramid nanofiber (PANA) membranes and a flexible conductive paper electrode (FCPE). The HF-PTENG delivered an open-circuit voltage of 384.3V, a short-circuit current of 15.1µA, and a transferred charge of 149.5 nC, achieving a maximum instantaneous power destiny of 2.1mW/cm2. Stable electrical output was maintained over 11000 operation cycles, and efficient biomechanical energy harvesting was demonstrated from finger tapping, wrist bending, and knee motions to power capacitors, commercial LEDs, and sensors. A smart glove integrated with the HF-PTENG was designed to achieve wireless control of a miniature car via WiFi communication. In addition, a 3×3 TENG sensor array was constructed to visualize planar pressure distribution. Separately, convolutional neural network (CNN) analysis enabled the recognition of six finger-drawn patterns with an accuracy of 98.32%. This study offers a practical and scalable strategy for constructing high-performance, flexible TENGs with board potential for applications in wearable electronics, human-machine interfaces, and intelligent motion sensing.

Programmable patterning of nematic liquid crystals (LCs) enables spatially encoded optical functionality, but existing approaches often face trade-offs between patterned area, feature fidelity, dimensionality, and device integration. Many methods rely on surface alignment, electrode patterning, or global illumination, limiting local addressability, depth control, or scalability. Field-assisted polymerization of reactive mesogens offers a means to capture designed LC director configurations as permanent, structurally encoded profiles. Here, we report a hybrid strategy combining wide-field one-photon polymerization (1PP) for rapid, large-area templating with two-photon polymerization direct laser writing (2PP-DLW) for localized, maskless microstructuring with depth control. This decoupling of patterned area from feature fidelity allows multiple, spatially co-located director profiles to be encoded within a single glass cell, enabling voltage-selective visibility and reconfigurable optical responses. Local 2PP-DLW features define confinement boundaries and deterministic defect nucleation and guidance, while 1PP establishes the global architecture with high throughput. Because patterning is encoded in the polymer rather than electrode geometry, complex profiles can be realized using uniform electrodes. In a nematic Pi-cell, this approach enables controlled defect channeling, programmable topological transitions, and multistate optical patterns, offering a scalable route to high-fidelity reconfigurable LC micro-optics.

Proper physicochemical characterization of advanced materials and complex industrial composites remains a significant challenge, particularly for nanomaterials, whose nanoscale dimensions and mostly complex chemistry challenge the analysis. In this work, we employed a correlative analytical approach that integrates atomic force microscopy (AFM), scanning electron microscopy (SEM) coupled with energy-dispersive X-ray spectroscopy (EDS), time-of-flight secondary ion mass spectrometry (ToF-SIMS), Auger electron spectroscopy (AES), and Raman spectroscopy. This combination enables detailed chemical and structural characterization with sub-micrometer spatial resolution. Three commercial graphene-based materials of varying complexity were selected and investigated to test the analytical performance of this approach. Furthermore, one of the commercial graphene oxide samples was chemically functionalized via amination and fluorination. This allowed us to assess how surface modifications influence both the material properties and the limits of the applied analytical techniques.

Triboelectrification has been known for millennia and remains of great practical importance, with applications spanning pharmaceuticals, electronics, industrial processes, and especially energy harvesting. Its inherent complexity, however, has long hindered a definitive understanding. In recent years, the advent of triboelectric nanogenerators (TENGs) has stimulated extensive experimental investigations and, in parallel, the development of promising first-principles-based theoretical models. Together, these advances have begun to clarify fundamental aspects of triboelectric charge transfer. This review traces the progression from early electron-transfer concepts to recent frameworks, including the backflow-stuck charges, thermoelectric, flexoelectric, and mechanochemical models. We discuss their evolution, the problems they address, the discoveries and insights they have enabled, as well as their limitations. At the beginning of the manuscript, special emphasis is given to the development of TENGs, which have both advanced practical applications and uncovered new triboelectric phenomena. Finally, we identify key open challenges and provide an outlook on future directions. The integration of complementary models through multiscale approaches, supported by systematic experimental validation, offers the most promising pathway toward a comprehensive understanding of this long-standing problem and the rational design of triboelectric materials and devices.