Clinical evaluations of large language models (LLMs) have rapidly expanded since 2022, yet their evidence base remains opaque. The overwhelming volume of studies creates challenges for manual curation and review. However, LLMs themselves offer the scalability and capability to evaluate the ever-growing evidence base. This LLM-assisted review identified 4,609 peer-reviewed studies in clinical medicine between January 2022 and September 2025, equating to roughly 3.2 papers per day. Only 1,048 studies used real-world patient data and of these only 19 were prospective randomized trials; most addressed simulated scenarios (n = 1,857) or exam-style tasks (n = 1,704). ChatGPT and related OpenAI models constitute 65.7% of evaluated models, with Gemini/Bard a distant second constituting 13.1% of evaluated models. Patient-facing communication and education comprised 17% of tasks, followed by knowledge retrieval, and education and assessment simulation. Across 1,046 head-to-head comparisons, LLMs outperformed humans in 33% of comparisons, with a strong dependency on task realism and level of training. At least 25% of studies had sample sizes less than 30. Despite the growth of LLMs in medicine, rigorous, patient-centered evidence remains scarce, underscoring the need for larger prospective trials before clinical adoption.
Background/Objectives: Antipsychotic dosing for behavioral and psychological symptoms of dementia (BPSD) in Alzheimer's disease remains empirical and variable. This study develops a deep learning model to predict individualized antipsychotic doses from structural MRI. Methods: A transfer learning approach with a cascaded ResNet (Cas-ResNet) was used. The model was first pre-trained on a large healthy aging dataset (CBMFM, n = 646) for brain age prediction, then fine-tuned on a BPSD dataset (SMHC, n = 86) to predict the defined daily dose (DDD) of antipsychotics. Model interpretability was performed using Grad CAM to identify predictive brain regions. Results: The proposed model achieved a mean absolute error of 0.19 and a Pearson correlation of 0.66 between predicted and actual doses, outperforming baseline 3DCNN, VGG, and DenseNet. Key contributing regions included the left inferior temporal gyrus, right parahippocampal gyrus, right putamen, left middle temporal gyrus, and left caudate. Conclusions: This proof-of-concept study demonstrates that deep learning can predict personalized antipsychotic doses from structural MRI, offering an objective tool to standardize BPSD pharmacotherapy and reduce empirical prescribing. The identified brain regions provide neurobiological insights into treatment response.
Operational decisions governing patient flow, cost, and quality of care demand specialized predictive models, yet most clinical NLP efforts focus on medical knowledge benchmarks. We introduce Lang1, a family of language models (100M-7B parameters) pretrained on 80 billion clinical tokens from NYU Langone Health electronic health records blended with 627 billion internet tokens. We evaluate Lang1 on the REalistic Medical Evaluation (ReMedE), an evaluation suite derived from 668,331 Electronic Health Records (EHR) notes spanning five tasks: readmission, mortality prediction, length of stay, comorbidity coding, and insurance denial. In zero-shot settings, both general-purpose and biomedical models underperform on four of five tasks. After finetuning, Lang1-1B outperforms finetuned generalist models up to 70 × larger and zero-shot models up to 671× larger. Joint multi-task finetuning yields cross-task transfer, and Lang1-1B transfers effectively to unseen tasks and an external health system. These results demonstrate that effective healthcare AI requires in-domain pretraining, supervised finetuning, and evaluation beyond proxy benchmarks.
Hazardous chemical elements in aquatic ecosystems pose a major threat to habitats and the environment. This study is based on two parallel analyses at different scales and innovates methodologically within this approach, with continental-scale analysis using satellite imagery. At the macroscale, it aims to spectrally detect chlorophyll-a (CHL), water turbidity (TSM), and potential for suspended pollution (ADG443), through 472 satellite sampling points distributed in the Solimões River, Amazon River, and the estuarine region, during the dry (February) and rainy (August) seasons between 2019 and 2024. At the microscale, it seeks to quantify the main chemical elements present in nanoparticles and ultrafine particles incorporated into sediments collected in the Amazon River. Sediment sampling was conducted at 18 sites, nine upstream and nine downstream of the city of Manaus, Brazil, during the dry season (February) and the wet season (August) of 2024. Sediment analyses were conducted using focused ion beam scanning electron microscopy (FIB-SEM) and high-resolution transmission electron microscopy (HR-TEM), both of which were coupled with energy-dispersive X-ray spectroscopy (EDS). The results revealed the presence of toxic elements such as chromium (Cr) and vanadium (V) in nanoparticles smaller than 15 nm. The findings demonstrate that the convergence of high sediment contamination and increasing trophic instability pose a systemic risk to the world's largest freshwater system, necessitating urgent global environmental monitoring and the implementation of advanced conservation frameworks to safeguard both biodiversity and local riverside population.
Countries have been facing environmental problems in recent years, especially due to increasing energy demand. Therefore, resolving this challenge of carbon dioxide (CO2) emissions and moving to eco-friendly, sustainable, and cleaner energy sources has emerged as a major global concern. Accordingly, this research examines the nexus hip between clean electricity generation (EG) subcomponents and power sector CO2 (PCO2) emissions under the moderating role of critical minerals (CMs) prices and geopolitical risk (GPR) in China, which is the biggest economy in the world. In this vein, the study novel quantile-based approaches on data from 2nd January 2019 to 30th June 2025 uncover the daily varying effect. The empirical outcomes demonstrate that (i) EG from clean sources (i.e., hydro, solar, and wind) generally do not reduce PCO2 emissions on a daily basis in bivariate cases; (ii) price changes of CMs for each EG subtype present moderating effects, where there are reducing effects on PCO2 emissions across various quantiles; (iii) there are causal effects from EG subtypes and price changes of CMs to PCO2 emissions across almost all levels, with a few exceptions at lower and middle quantiles; (iv) the moderating role of price changes of CMs causes a weakening effect on the nexus between each EG subtype and PCO2 emissions; (v) These outcomes are robust based on an alternative approach. Thus, the study reveals that the nexus varies across EG subtypes and quantiles under the moderating effect of CM price changes. Accordingly, the study discusses policy options (e.g., ensuring a displacement in favor of renewable energy sources, focusing on hydro EG to decarbonize PCO2, providing various financial and fiscal incentives to stimulate further installation of renewable energy capacity, trying to ensure a price stability in CMs' market, prioritizing domestic market first) to decarbonize the Chinese power sector.
Specialized clinical artificial intelligence (AI) tools are entering medical practice despite scarce independent evaluation. We quantitatively evaluate two clinical AI tools, OpenEvidence and UpToDate Expert AI, built on large language models (LLMs) against three frontier LLMs: GPT-5.2, Gemini 3.1 Pro and Claude Opus 4.6. Our evaluation has three stages: (1) 500 MedQA questions testing medical knowledge, (2) 500 HealthBench items measuring alignment with clinicians and (3) the real clinical queries (RCQ) benchmark, built from 100 de-identified queries from physicians to a general-purpose language model in a live clinical environment. For the RCQ benchmark, 12 US clinicians performed randomized, blinded review of model outputs, producing 1,800 model-question annotations. Frontier LLMs outperformed clinical AI tools in all three evaluations. Clinical AI tools performed comparably to auto-enabled Google Search AI Overview on the RCQ. These findings highlight the need for independent, real-world evaluation of AI tools before they enter clinical settings.
Accurate and equitable prediction of trauma-related in-hospital mortality is critical for guiding clinical decisions and optimising trauma care resources. Traditional severity scoring systems like the Injury Severity Score (ISS) do not account for demographic factors, potentially limiting their fairness and generalisability across diverse populations. We developed and externally validated an artificial intelligence (AI) model based on ISS and integrated demographic features (age and sex) to predict in-hospital mortality after trauma. Data from the Korean Trauma Data Bank were used for model development and internal validation, comprising 121,418 patients with trauma aged ≥15 years treated at 19 trauma centres in South Korea (2017-2022). External validation was performed on an independent cohort of 7458 patients from five trauma centres (four in South Korea and one in Australia, 2022-2024). The primary outcome was trauma-related in-hospital mortality. Predictive performance was assessed using area under the receiver operating characteristic curve (AUROC), sensitivity, specificity, accuracy, and balanced accuracy. Fairness was evaluated by comparing AUROC differences across age (<65 vs ≥65 years) and sex (female vs male) subgroups. The ISS-based AI model incorporating age and sex achieved high predictive performance (internal validation AUROC, 0.934; external validation AUROC range, 0.901-0.920), outperforming conventional ISS-based methods. The model also demonstrated improved fairness, showing reduced AUROC differences across subgroups (age: 0.068 vs 0.091; sex: 0.021 vs 0.046 for AI model vs ISS, respectively). Scaling an ISS-based AI model through demographic integration yielded accurate, fair, and generalisable predictions of trauma-related in-hospital mortality. This approach may enhance trauma care decision-making and enable more equitable resource allocation across diverse clinical settings. This research was supported by the MSIT (Ministry of Science and ICT), Korea, under the ITRC (Information Technology Research Center) support program (IITP-2025-RS-2024-00438239) and the Institute of Information & Communications Technology Planning & Evaluation (IITP) grant funded by the Korea government (MSIT) (RS-2024-00509257, Global AI Frontier Lab). In addition, this research was supported by the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health & Welfare, Republic of Korea (RS-2025-02220492).
The journal retracts the article titled "Measuring Liquid Droplet Size in Two-Phase Nozzle Flow Employing Numerical and Experimental Analyses" [...].
General purpose vision-language models (VLMs) demonstrate impressive capabilities, but their opaque training on uncurated internet data poses critical limitations for high-stakes decision making, such as in neurosurgery. We present CNS-Obsidian, a neurosurgical VLM trained on peer-reviewed neurosurgical literature, and demonstrate its clinical utility compared with GPT-4o in a real-world setting. We compiled 23 984 articles from Neurosurgery Publications journals, yielding 78 853 figures and captions. Using GPT-4o and Claude Sonnet-3.5, we converted these image-text pairs into 263 064 training samples across 3 formats: instruction fine-tuning, multiple-choice questions, and differential diagnosis. We trained CNS-Obsidian, a fine-tune of the 34-billion parameter Large Language and Visual Assistant-Next model. In a blinded, randomized deployment trial at NYU Langone Health (August 30-November 30, 2024), neurosurgeons were assigned to use either CNS-Obsidian or a Health Insurance Portability and Accountability Act-compliant GPT-4o end point as a diagnostic copilot after patient consultations. Primary outcomes were diagnostic helpfulness and accuracy, assessed through user ratings and presence of the correct diagnosis within the VLM-provided differential, respectively. CNS-Obsidian matched GPT-4o on synthetic questions (76.13% vs 77.54%, P = .235), but only achieved 46.81% accuracy on human-generated questions vs GPT-4o's 65.70% (P < 10-15). In the randomized trial, 70 consultations were evaluated (32 CNS-Obsidian, 38 GPT-4o) from 959 total consults (7.3% utilization). CNS-Obsidian received positive ratings in 40.62% of cases vs 57.89% for GPT-4o (P = .230). Both models included correct diagnosis in approximately 60% of cases (59.38% vs 65.79%, P = .626). Domain-specific VLMs trained on curated scientific literature can approach frontier model performance in specialized medical domains despite being orders of magnitude smaller and less expensive to train. This establishes a transparent framework for scientific communities to build specialized artificial intelligence models. However, low clinical utilization suggests chatbot interfaces may not align with specialist workflows, indicating need for alternative artificial intelligence integration strategies.
Next-generation biomedical devices increasingly require integrated energy harvesting and miniaturized electronic components to achieve self-sustaining operation without external power sources. However, most currently available systems remain rigid and bulky, impeding effective mechanical and biological interfacing with soft tissues. Hydrogels, with their high water content, tunable elasticity, and excellent biocompatibility, have emerged as ideal candidates for developing conformal and tissue-compliant electronics. To overcome the limitations of battery-powered implants, recent studies have focused on harvesting biomechanical energy - particularly via piezoelectric nanogenerators (PENGs) - as a sustainable alternative. The integration of PENGs into hydrogel matrices has led to the emergence of piezoelectric hydrogels (PHs), a novel class of soft, biocompatible, and self-powered biomaterials. These hybrid materials combine the mechanical adaptability of hydrogels with the energy conversion capabilities of piezoelectric materials, enabling new paradigms in implantable and wearable electronics. This review presents a comprehensive overview of PHs, including piezoelectric material classifications (ceramics, polymers, nanocomposites), fabrication strategies of natural and synthetic PHs, and their device-level performance under physiological conditions. Furthermore, we discuss the application of PHs in skin wound healing, peripheral nerve regeneration, and the repair of bone and cartilage tissues. Finally, we provide a future outlook emphasizing the need for mechanistic insights, improved output performance, and standardized in vivo evaluations to accelerate the clinical translation of PH-based self-powered bioelectronics.
BACKGROUND: Mental disorders are a major and evolving contributor to morbidity among women of child-bearing age (WCBA). However, long-term trends, geographic inequalities, and the future burden in this population remain incompletely characterised. We quantified the burden of mental disorders among WCBA (15–49 years) from 1990 to 2021 and projected trends to 2050. METHODS: We assessed the burden of 10 mental disorder categories among WCBA aged 15–49 years in 204 countries and territories using GBD 2021, including 9 specified mental disorders and a residual category (“other mental disorders”). We analysed prevalence and disability-adjusted life-years (DALYs) as numbers and age-standardised rates (ASPR and ASDR). A Bayesian age–period–cohort model was used to generate projections through 2050. RESULTS: Overall, ASPR and ASDR declined from 1990 to 2019; compared with the pre-pandemic baseline in 2019, both indicators were higher in 2020–2021. Age-specific prevalence generally increased with age and peaked at 40–44 years, although patterns differed by disorder category. At the national level, Greenland and Portugal had the highest ASPR and ASDR, whereas Vietnam had the lowest prevalence and North Korea had the lowest DALYs. Across the 21 GBD regions, higher SDI was associated with higher ASPR and ASDR, with marked regional heterogeneity. Projections suggested that the burden of mental disorders among WCBA will increase through 2050, largely driven by anxiety and depressive disorders. CONCLUSIONS: The burden of mental disorders among WCBA remains substantial and is projected to rise through 2050, with increases during 2020–2021 relative to the 2019 baseline. These findings underscore the need for context-specific strategies that strengthen surveillance and expand accessible prevention and care—particularly for anxiety and depressive disorders—across diverse sociodemographic settings.
Deep convolutional neural networks (CNNs) have seen significant growth in medical image classification applications due to their ability to automate feature extraction, leverage hierarchical learning, and deliver high classification accuracy. However, Deep CNNs require substantial computational power and memory, particularly for large datasets and complex architectures. Additionally, optimising the hyperparameters of deep CNNs, although critical for enhancing model performance, is challenging due to the high computational costs involved, making it difficult without access to high-performance computing resources. To address these limitations, this study presents a fast and efficient model that aims to achieve superior classification performance compared to popular Deep CNNs by developing lightweight CNNs combined with the Nonlinear Lévy chaotic moth flame optimiser (NLCMFO) for automatic hyperparameter optimisation. NLCMFO integrates the Lévy flight, chaotic parameters, and nonlinear control mechanisms to enhance the exploration capabilities of the Moth Flame Optimiser during the search phase while also leveraging the Lévy flight theorem to improve the exploitation phase. To assess the efficiency of the proposed model, empirical analyses were performed using a dataset of 2314 brain tumour detection images (1245 images of brain tumours and 1069 normal brain images). The evaluation results indicate that the CNN_NLCMFO outperformed a non-optimised CNN by 5% (92.40% accuracy) and surpassed established models such as DarkNet19 (96.41%), EfficientNetB0 (96.32%), Xception (96.41%), ResNet101 (92.15%), and InceptionResNetV2 (95.63%) by margins ranging from 1 to 5.25%. The findings demonstrate that the lightweight CNN combined with NLCMFO provides a computationally efficient yet highly accurate solution for medical image classification, addressing the challenges associated with traditional deep CNNs.
Understanding whether individuals with mental illness, who face challenges related to healthcare barriers, are more vulnerable to postacute sequelae of COVID-19 is limited. Here, we investigated the potential association between pre-existing mental illness and postacute sequelae of COVID-19 across 12 major health domains and 141 specific diseases in COVID-19 survivors. The large-scale, population-based cohorts from South Korea (K-COV-N cohort) used in the study included 8 632 221 individuals aged 20 years or older who were infected with SARS-CoV-2 between January 1, 2020, and December 31, 2022. The risk of postacute sequelae of COVID-19 was assessed in the 1:2 propensity score-matched cohorts, comprising 12 major health domains and 141 diseases based on the ICD-10 code, following mental illness among patients with COVID-19. We assessed the time attenuation effect of major health outcomes after 30 days following SARS-CoV-2 infection. Multiple subgroup analyses were conducted by severity of mental illness, COVID-19 severity, vaccination, and SARS-CoV-2 strain. After 1:2 exposure-driven propensity score matching, we identified 1 341 320 participants with mental illness (mean age, 49.51 [SD, 13.82] years; 62.27% female) and 2 653 597 controls (mean age, 48.78 [SD, 13.75] years; 62.03% female). Individuals with mental illness exhibited significantly higher risks across all 12 major health domains, including: infectious and parasitic events (adjusted hazard ratio [aHR], 1.36 [95% CI, 1.33-1.38]), blood and immune-related events (1.21 [1.17-1.26]), endocrine, nutritional, and metabolic events (1.21 [1.18-1.24]), nerve-related events (2.13 [2.07-2.19]), eye-related events (1.29 [1.25-1.34]), ear and mastoid events (1.52 [1.50-1.54]), circulatory events (1.25 [1.17-1.35]), respiratory events (1.26 [1.24-1.29]), digestive events (1.41 [1.40-1.41]), skin-related events (1.34 [1.30-1.38]), musculoskeletal events (1.42 [1.41-1.43]), and genitourinary events (1.54 [1.18-2.01]). Of the 141 postacute sequelae of COVID-19, 133 showed significantly increased risks. The association was strongest within the first 6-12 months after SARS-CoV-2 infection, with risks progressively attenuating beyond 12 months and nearly disappearing after 18 months. Subgroup analysis revealed that individuals with mild mental illness exhibited higher aHRs for 11 of the 12 health outcome domains compared with those with severe mental illness. Altogether, our findings show the increased risk of postacute sequelae of COVID-19 across 12 major health domains in individuals with mental illness among COVID-19 survivors. These findings highlight the need for targeted monitoring and intervention strategies to address the vulnerabilities of this population, particularly during the post-COVID-19 period.
Inflammatory cytokine disturbance is a prominent outcome of immune dysregulation, extensively documented in bipolar disorder (BD). However, observational studies have exhibited inconsistent findings, and the causal relationships between inflammatory factors and BD remain unclear. Hence, this study aimed to uncover the causality between circulating inflammatory cytokines and BD. In the bidirectional Mendelian randomization (MR) analysis, two genetic instruments derived from a publicly available genomic dataset were utilized. Genetic variant data for 41 inflammatory cytokines were obtained from a meta-analysis of genome-wide association studies involving 8293 Finnish individuals. BD data included 41,917 cases and 371,549 controls from the Psychiatric Genomics Consortium Database. To estimate causal connections between inflammation cytokines and BD, we performed five methods: inverse variance weighting (IVW), MR-Egger, weighted median, simple mode, and weighted mode. Sensitivity analyses were conducted to evaluate robustness, including tests for heterogeneity, pleiotropy, and leave-one-out validation. In the IVW approach, we identified significant associations between genetic liability to elevated levels of interleukin-17 (IL-17), macrophage inflammatory protein-1α (MIP-1α), and monocyte chemotactic protein 3 (MCP-3) with increased risk of developing BD (odds ratio [OR] = 1.119, 95% confidence interval [CI] = 1.021-1.226, p = 0.016; OR = 1.084, 95% CI = 1.002-1.174, p = 0.044; OR = 1.060, 95% CI = 1.001-1.122, p = 0.046, respectively). Across the various MR methods employed, the directions of causal inferences remained consistent. However, reverse MR analysis revealed no significant evidence for causal effects of genetic predisposition to BD on inflammatory cytokines. Our findings provide robust genetic evidence supporting causal effects of elevated circulating inflammatory cytokines on BD, conferring a high susceptibility to BD. Our study emphasizes that interventions aimed at reducing peripheral inflammatory cytokine levels should be prioritized in BD management.
YidC, a prominent member of the Oxa1 superfamily, is essential for the biogenesis of the bacterial inner membrane, significantly influencing its protein composition and lipid organization. It interacts with the Sec translocon, aiding the proper folding of multi-pass membrane proteins. It also functions independently, serving as an insertase and lipid scramblase, augmenting the insertion of smaller membrane proteins while contributing to the organization of the bilayer. Despite the wealth of structural and biochemical data available, how YidC operates remains unclear. To investigate this, we employed proximity-dependent biotin labeling (BioID) in Escherichia coli, leading to the identification of YibN as a crucial component within the YidC protein environment. We then demonstrated the association between YidC and YibN by affinity purification-mass spectrometry assays conducted on native membranes, with further confirmation using on-gel binding assays with purified proteins. Co-expression studies and in vitro assays indicated that YibN enhances the production and membrane insertion of YidC substrates, such as M13 and Pf3 phage coat proteins, ATP synthase subunit c, and various small membrane proteins like SecG. Additionally, the overproduction of YibN was found to stimulate membrane lipid production and promote inner membrane proliferation, perhaps by interfering with YidC lipid scramblase activity. Consequently, YibN emerges as a significant physical and functional interactor of YidC, influencing membrane protein insertion and lipid organization.
Patients suffering from coronary artery disease (CAD) or peripheral arterial disease (PAD) can benefit from bypass graft surgery. For this surgery, arterial vascular grafts have become promising alternatives when autologous grafts are inaccessible but suffer from numerous postimplantation challenges, particularly delayed endothelialization, intimal hyperplasia, high risk of thrombogenicity and restenosis, and difficulty in timely detection of these subtle pathological changes. We present an electronic vascular conduit that integrates flexible electronics into bionic vascular grafts for in situ, real-time and long-term monitoring for hemadostenosis and thrombosis concurrent with postoperative vascular repair. Following bypass surgery, the integrated bioelectronic sensor based on the triboelectric effect enables monitoring of the blood flow in the vascular graft and identification of lesions in real time for up to three months. In male nonhuman primate cynomolgus monkeys, the electronic vascular conduit, with an integrated wireless signal transmission module, enables wireless and real-time hemodynamic monitoring and timely identification of thrombi. This electronic vascular conduit demonstrates potential as a treatment-monitoring platform, providing a sensitive and intuitive monitoring technique during the critical period after bypass surgery in patients with CAD and PAD.
Atmospheric contaminants from natural processes and anthropogenic activities pose a major problem to the environment. Here we analyze the dynamics of atmospheric and terrestrial contaminant concentrations in sediments containing chemical elements, such as nanoparticles (NPs) and ultrafine particles in hydrological sources of the Caribbean region of Colombia. Terrestrial sediments were collected from 2022 to 2024, and quantified for major chemical elements in the form of NPs and ultrafine particles in runoff receiving areas along the banks of Colombia's Ciénaga Grande in Santa Marta Bay, on the Isla de Salamanca. Additionally, atmospheric carbon monoxide (CO) and nitrogen dioxide (NO2) levels were detected using TROPOMI, coupled with the Sentinel-5P satellite, from 2019 to 2024. Sampling was performed during both summer and winter, focusing on the Sierra Nevada de Santa Marta National Park, Isla de Salamanca, and the cities of Santa Marta and Barranquilla. Sediment samples were collected from 25 fixed sites on Isla de Salamanca and analyzed in the laboratory. The CO and NO2 concentrations were detected at 122 "collection points," with a spatial resolution of 7 km × 3.5 km, an average normalization of 0.83 μg/mg, and a maximum error of 6.62 %. The results revealed abundant Fe particles containing Cr, Ti, Zn, and Zr (1-2.5 μm in size), with nanohematite (iron oxide) particles containing C, As, and S detected via EDS. The CO concentrations peaked in Barranquilla and Santa Marta (up to 0.042 mol/m2), while NO2 concentrations reached 2.4 × 10-5 mol/m2, similar to the levels in Sierra Nevada de Santa Marta National Park and Ciénaga Grande de Santa Marta. These findings highlight the impact of forest fires on air quality in the region and support the development of new public policies to mitigate pollution and to protect ecosystems in this ecologically relevant area.
暂无摘要（点击查看详情）
Triboelectric nanogenerators (TENGs), among the most simple and efficient means to harvest mechanical energy, have great potential in renewable energy utilization. While the output performance of TENGs is still not high enough, which limits its practical application. Here, a poly(vinylidene fluoride) (PVDF)/fluorinated ethylene propylene nanoparticles (FEP NPs) porous nanofiber (PFPN) membrane with waterproof, breathable, surface superhydrophobic and high tribo-negative properties is proposed for achieving high-performance of TENGs. The PFPN-based solid-solid contact PFPN-TENG achieved an optimal electric output with a net tribo-charge density of 294 µC m-2, which is 42.2% more than that of polytetrafluoroethylene (PTFE) film. The PFNP membrane can maintain the best performance with almost no attenuation after 142680 working cycles. Based on its excellent triboelectric characteristics, the PFPN membrane shows its excellent performance for self-powered body motion sensing and mechanical energy harvesting. A flexible solid-liquid contact PFPN-TENG can achieve high electrical output with an average volume power density of ≈544.1 W m-3 by harvesting water wave energy. Such excellent performance of the PFPN membrane makes it a potential candidate to promote the power density of TENGs in harvesting blue water wave energy.
The increasing energy demands of Internet of Things devices necessitate high-performance triboelectric nanogenerators (TENGs). However, the performance of TENGs is severely constrained by air and dielectric breakdown, which not only limits the energy output but also accelerates dielectric degradation and device failure. Here, a strategy is proposed to synergistically enhance the breakdown resistance and facilitate self-extinguishing of air breakdown by introducing an aluminum nitride (AlN) coating on dielectrics. The coating improves thermal dissipation capability and modulates the electric field and energy level distribution, enhancing dielectric breakdown resistance. Additionally, a spontaneously induced reversed electric field effectively suppresses air breakdown, preventing excessive discharge damage. Consequently, the modified polyimide (PI) and polytetrafluoroethylene (PTFE) exhibit significant improvements in both air and dielectric breakdown resistance. Corresponding TENGs achieve ≈ 50% increase in energy density and prominently extend operational stability by over 360 folds. This work provides insights into the thermal and interfacial processes of air breakdown, promoting the development of high-performance and durable TENGs.