Shu Hu

Abstract Ruthenium dioxide (RuO 2 ) shows excellent activity toward the acidic oxygen evolution reaction (OER); however, its practical application is limited by poor long‐term stability. Herein, a single‐phase Mo 0.5 Ru 0.5 O 2 nanoparticle catalyst is reported with a high Mo content, synthesized via high‐temperature thermal shock treatment under an oxygen atmosphere (HTSO), exhibiting high activity and stability in OER. The HTSO technique involves rapidly heating the precursor to ≈1200 °C for ≈0.05 s in oxygen, followed by immediate quenching at a rate of ≈10 4 °C s −1 . The resulting nanoparticles exhibit a uniform size of ≈10 nm and homogeneous elemental mixing, overcoming the thermodynamic barriers that typically lead to phase separation in conventional synthesis methods. The Mo 0.5 Ru 0.5 O 2 catalyst achieves an overpotential of 210 mV at 10 mA cm −2 and maintains stable performance over 300 h at 50 mA cm −2 in OER, significantly surpassing the stability of RuO 2 and other reported high‐metal‐content doped RuO 2 catalysts. High‐valence Mo, with its multiple accessible oxidation states and compatible ionic radius, serves as an ideal dopant for RuO 2 , enabling stable lattice substitution, effective electron donation, and ultimately suppressing Ru over‐oxidation while enhancing stability. This approach enhances catalyst stability and Ru utilization, providing a versatile platform for synthesizing other metal‐doped RuO 2 systems toward cost‐effective and stable OER catalysts.

The laser-induced ionization of Kr He mixture was used to prepare metastable 5s[3/2] 2 level Kr atoms. By analyzing the temporal evolution of the Kr 760.15 nm (5p[3/2] 2 → 5s[3/2] 2 ) spectral line intensity, three stages of metastable production were confirmed: the “photon excitation + radiation” process, the “electron-impact excitation + radiation” process, and the “ion-electron recombination” process. In the “electron-impact excitation + radiation” stage, the total collisional decay rate constant for Kr ⁎ atoms at the 5p[3/2] 2 level associated with the buffer gas He atoms was (0.0179 ± 0.0012) × 10 −11 cm 3 s −1 , which confirmed that under laser-induced ionization, the collision contributions from other levels to the 5p[3/2] 2 level Kr atoms were significant. This revealed that He atoms inhibit the reionization of Kr atoms due to their interference and slowing effect on the energy pooling process of Kr atomic levels. In the “ion-electron recombination” stage, the total collisional decay rate constant for Kr ⁎ atoms at the 5p[3/2] 2 level associated with He atoms was (0.0008 ± 0.0001) × 10 −11 cm 3 s −1 , which was essentially the same as the total collisional decay rate constant for pure Kr gas in this stage. This again indicated that in the continuous recombination process of ions and electrons, the collisional replenishment rate from the Rydberg states to the 5p[3/2] 2 level dominates, masking the difference in collisional decay rates between the Kr He mixture and pure Kr gas. • 5p[3/2] 2 level Kr atoms are produced by photon excitation, electron-impact excitation, and recombination processes. • Total collisional decay rate constant for He on 5p[3/2] 2 level Kr is small in electron-impact excitation stage. • Reionization issue is mitigated because the energy pooling process of Kr atoms was interfered with by He atoms. • Collisional replenishment rate constant dominates in recombination stage.

Visual anomaly detection is a highly challenging task, often categorized as a one-class classification and segmentation problem. Recent studies have demonstrated that the student-teacher (S-T) framework effectively addresses this challenge. However, most S-T frameworks rely solely on pre-trained teacher networks to guide student networks in learning multi-scale similar features, overlooking the potential of the student networks to enhance learning through multi-scale feature fusion. In this study, we propose a novel model named PFADSeg, which integrates a pre-trained teacher network, a denoising student network with multi-scale feature fusion, and a guided anomaly segmentation network into a unified framework. By adopting a unique teacher-encoder and student-decoder denoising mode, the model improves the student network's ability to learn from teacher network features. Furthermore, an adaptive feature fusion mechanism is introduced to train a self-supervised segmentation network that synthesizes anomaly masks autonomously, significantly increasing detection performance. Rigorous evaluations on the widely-used MVTec AD dataset demonstrate that PFADSeg exhibits excellent performance, achieving an image-level AUC of 98.9%, a pixel-level mean precision of 76.4%, and an instance-level mean precision of 78.7%.

Artificial Intelligence (AI) has made it possible for anyone to create images, audio, and video with unprecedented ease, enriching education, communication, and creative expression. At the same time, the rapid rise of AI-generated media has introduced serious risks, including misinformation, identity misuse, and the erosion of public trust as synthetic content becomes increasingly indistinguishable from real media. Although deepfake detection has advanced, many existing tools remain closed-source, limited in modality, or lacking transparency and educational value, making it difficult for users to understand how detection decisions are made. To address these gaps, we introduce SynthGuard, an open, user-friendly platform for detecting and analyzing AI-generated multimedia using both traditional detectors and multimodal large language models (MLLMs). SynthGuard provides explainable inference, unified image and audio support, and an interactive interface designed to make forensic analysis accessible to researchers, educators, and the public. The SynthGuard platform is available at: https://in-engr-nova.it.purdue.edu/

Longer wavelengths of light contain less energy but comprise more of the solar spectrum, making them important to incorporate into any process aiming for high efficiency. Here, we developed a novel redox-mediated synthetic mechanism to construct a heterojunction with strongly coupled interfaces. Specifically, an α-Fe 2 O 3 /Cu 2 O/CuO nanosheet composite was synthesized, forming an S-scheme α-Fe 2 O 3 /Cu 2 O electronic interface, a burgeoning class of materials designed to upconvert longer wavelengths of light and utilize solar energy more effectively. Through a series of experiments including X-ray photoelectron spectroscopy (XPS), ultraviolet–visible (UV–Vis) diffuse reflectance spectroscopy (UV–Vis-DRS), electrochemical impedance spectroscopy (EIS), and photocatalytic measurements, we were able to fully confirm the electronic structure of the α-Fe 2 O 3 /Cu 2 O interfacial heterojunction. These characterizations demonstrate the S-scheme flow of electrons, which is further supported by COMSOL numerical simulations. The successful formation of the S-scheme heterojunction is made possible through the direct Fe–O–Cu covalent bonding at the interface. These bonds provide ultrafast interfacial charge transfer pathways on picosecond time scales followed by long-lived charge-separated states, as quantified by our transient optical experiments. The proposed redox-mediated synthetic strategy provides a valuable guideline for constructing effective solid heterojunctions with strongly coupled interfaces, which are desirable for various applications in catalysis, energy storage, electronics, photovoltaics, and beyond.

This paper continues the work of our previous study [J. Quant. Spectrosc. Radiat. Transf. 330 109233], further exploring the dynamics of the metastable 5s[3/2] 2 level of Kr atoms prepared by laser-induced ionization. Focusing on the “photon excitation + radiation” process, “electron-impact excitation + radiation” process, and “ion-electron recombination” process, we obtained the corresponding total collisional decay rate constants k col for the 5p[3/2] 2 level of (1.11±0.09)×10 −11 cm 3 s −1 , (0.09±0.03)×10 −11 cm 3 s −1 , and (0.0010±0.0001)×10 −11 cm 3 s −1 , respectively, based on time-resolved spectra at 760.15 nm. While the corresponding total radiative decay rates k rad were (2.58±0.75)×10 7 s −1 , (0.88±0.29)×10 7 s −1 , and (0.004±0.001)×10 7 s −1 . Through the analysis and comparison of the total collisional decay rate constants, and radiative decay rates obtained from the three stages, this study provides an in-depth understanding of the dynamic processes of the interactions between the relevant physical mechanisms during laser-induced ionization to produce metastable atoms. It clearly identifies that the changes in the population replenishment rate of the 5p[3/2] 2 level, due to the collision and radiation channels at different stages, are key factors influencing the values of k col and k rad .

Liquid-state optical gain media are desired for high-power lasing applications due to their easier heat management than solid-state media, as well as for the emerging field of optofluidics. Colloidal quantum dots (QDs) are solution-processed materials that carry many attractive properties suitable for liquid lasing. To date, however, their lasing action is often achieved in close-packed films, as a high volume fraction of QDs is required for stimulated emission to outpace the ultrafast Auger decay of gain-active multiexciton states. Here we report liquid lasing from color-tunable (red, orange, and green) alloyed core/shell QDs with impeded Auger recombination. Lasing action is achieved by loading the QD-solutions into cavities under quasi-continuous-wave excitation. The light amplification behaviors of QD-solutions under ambient conditions are much more stable compared to those of both solid-state QD-films and dye solutions. Compatibility with aqueous solvents and ease of scalability are also demonstrated. An optimized optical power efficiency of 17.2% has been achieved. These results indicate that liquid lasing from colloidal QDs holds strong promise for real-world implementation.

Hydrogen (H2) is a pivotal chemical in modern industry, serving as a versatile chemical feedstock and an energy carrier. Photoelectrocatalytic (PEC) water-splitting cells that utilize only sunlight can significantly reduce the total energy input of green H2 production. I will introduce and discuss the Energy Materials Network (EMN) approach to accelerating the design, manufacturing, and scale-up of PEC water-splitting devices. The current perovskite-based PEC cells are size-limited due to pinhole defects and long-term stability issues in large-area perovskite-based tandem light absorbers. To address this challenge, we have developed a hybrid molecular-inorganic coating approach that reduces pinhole defects and passivates defective interfaces or grain boundaries. I will also discuss the scaling up of the coating process, PEC system designs, and advanced manufacturing techniques supporting the upcoming meter-scale PEC device demonstration. I will discuss the use of a 1.7/1.2 eV tandem PSC-PSC design to increase solar-to-hydrogen (STH) efficiency to 18%. Besides, a 1.7/1.1 eV PSC/Si tandem configuration promises kilogram-scale solar-driven H2 production per day. A numerically controlled fabrication process has been developed to ensure quality control and volume manufacturing, including completely separated H2 and O2 evolution chambers which ensure high purity (> 99%) of the produced H2. I will also cover benchmarking protocols and best practices for outdoor testing of perovskite-based PEC water-splitting devices. Multi-physics modeling, simulations, and iterative on-sun testing, supported by standardized data collection and machine intelligence, can expedite the optimization process for solar-to-H2 efficiency and production costs. Additionally, we will continuously revisit technoeconomic analysis (TEA) through collaboration with the Technoeconomic Analysis node experts within the Energy Materials Network. Figure 1

Particulate photocatalysts, usually in a powder suspension or immobilized on a panel, host multiple concurring redox processes such as coevolving H 2 and O 2 . While co-evolving H 2 and O 2 is unsafe, instead, one can develop schemes of redox-mediated water splitting: H 2 -evolving photocatalysts will produce hydrogen while selectively oxidizing, e.g., I - to IO 3 - in solutions; a dichroic mirror splits the solar spectrum to allow O 2 -evolving photocatalysts to absorb the solar light unused by the H 2 -evolving photocatalysts; and the O 2 -evolving catalysts produce oxygen while selectively reducing, e.g., IO 3 - back to I - in a second solution. The challenges for photocatalysis, though, have been the lack of understanding and operando probing of photocatalyst/liquid interfaces. Photocatalysis features nanoscale electrochemical cells, where those reductive and oxidative sites are in nanoscale proximity, both supported on the photocatalyst surface. In all these cases, the conversion efficiency remains low, and trial-and-error approaches remain obscure to accelerate photocatalyst discovery. We need tools to probe the photocatalyst/liquid interfaces. In this talk, I discuss the probing of front and back potentials of thin-film model photocatalysts using nanoscale scanning electrochemical potentiometry. First, we developed new methods to probe the deep hole charge potentials of O2-evolving photocatalysts having O 2p or N 2p levels at the valence band maxima. Through collaborations, we further probed the spatial distribution of H 2 -evolving and O 2 -evolving rates and their operando electrochemical potentials, using scanning electrochemical microscopy (SECM). We will show the variation of reductive reaction rates as SECM probes move away from the IrO x or CoO x O 2 -evolving sites. Finally, I will discuss a digital multi-physics model to quantify and visualize the QFLs, band edge positions, and catalyst potentials. We show a series of experimental parameter sets to validate our digital model.

Group III-V semiconductors, such as InP, GaInP 2 , and GaAs, has been found as excellent photoabsorbers when performing light-to-chemical conversion due to their high quantum efficiency 1 . These efficient semiconductors are being manufactured at scale for optoelectronic devices and solar cells. However, they are not only very easily photocorroded, but they form oxide on the interface serving as recombination sites. Although “Leaky” TiO 2 coatings can effectively protect III-V semiconductors for anodic reactions 1, 2 , these coatings are intrinsically unstable under cathodic conditions, and their oxide can form massive recombination sites on the interface 3, 4 . To address these problems, we herein developed a novel gallium nitride (GaN) coating by atomic layer deposition (ALD) that can passivate and protect III-V semiconductors by forming an oxygen-free interface. Using InP as the photoabsorber, we grow GaN coating with sputtered Pt as cocatalyst and achieved a record-breaking performance for photoelectrochemistry (PEC) hydrogen evolution reaction (HER). During photo-electrocatalytic H 2 evolution, the photocathode has a photovoltage of 0.8 V, photocurrent of 34 mA/cm 2 under 1 sun illumination, and stability of at least 150 hours in pH 0 acid. In this work, we will discuss about the various characterizations on the interface, which shows a oxide-free interface and lower interfacial recombination rates. Beyond stabilization, this GaN serves as a multifunctional coating that can reduce surface recombination of oxide terminated III-V surfaces, provide a high driving force for charge separation, and tune the band edge position for desired photocatalytic reactions 5 . X. Shen, T.S. Zhao, H.Q. Su, M.Q. Yang, J.Y. Chen, Y.L. Liu, R. Yanagi, D. Solanki and S. Hu: Tuning Intermediate Bands of Protective Coatings to Reach the Bulk-Recombination Limit of Stable Water-Oxidation GaP Photoanodes. Adv Energy Mater 12 (2022). S. Hu, M.R. Shaner, J.A. Beardslee, M. Lichterman, B.S. Brunschwig and N.S. Lewis: Amorphous TiO 2 coatings stabilize Si, GaAs, and GaP photoanodes for efficient water oxidation. Science 344 , 1005 (2014). W. Yu, J.L. Young, T.G. Deutsch and N.S. Lewis: Understanding the Stability of Etched or Platinized p-GaInP Photocathodes for Solar-Driven H2 Evolution. Acs Appl Mater Inter 13 , 57350 (2021). W.L. Yu, M.H. Richter, P. Buabthong, I.A. Moreno-Hernandez, C.G. Read, E. Simonoff, B.S. Brunschwig and N.S. Lewis: Investigations of the stability of etched or platinized p-InP(100) photocathodes for solar-driven hydrogen evolution in acidic or alkaline aqueous electrolytes. Energy & Environmental Science 14 , 6007 (2021). R. Yanagi, T.S. Zhao, M. Cheng, B. Liu, H.Q. Su, C.X. He, J. Heinlein, S. Mukhopadhyay, H.Y. Tan, D. Solanki and S. Hu: Photocatalytic CO2 Reduction with Dissolved Carbonates and Near-Zero CO2(aq) by Employing Long-Range Proton Transport. Journal of the American Chemical Society 145 , 15381 (2023). Figure 1

Accurate and efficient question-answering systems are essential for high-quality patient care in the medical field. While Large Language Models (LLMs) have made remarkable strides across various domains, they still face challenges in medical question answering, particularly in understanding domain-specific terminology and performing complex reasoning, limiting their effectiveness in critical applications. To address this, we propose a multi-agent medical question-answering (MedQA) system incorporating similar case generation. We leverage the Llama3.1:70B model in a multi-agent architecture to enhance enhance zero-shot classification on the MedQA dataset, utilizing the model’s inherent medical knowledge and reasoning capabilities without additional training data. Experimental results show substantial gains over existing benchmark models, with improvements of 7% in both accuracy and F1-score across various medical QA tasks. Furthermore, we examine the model’s interpretability and reliability in addressing complex medical queries. This research not only offers a robust solution for medical question answering but also establishes a foundation for broader applications of LLMs in the medical domain.

Deformable registration is a fundamental task in medical image processing, aiming to achieve precise alignment by establishing nonlinear correspondences between images. Traditional methods offer good adaptability and interpretability but are limited by computational efficiency. Although deep learning approaches have significantly improved registration speed and accuracy, they often lack flexibility and generalizability across different datasets and tasks. In recent years, foundation models have emerged as a promising direction, leveraging large and diverse datasets to learn universal features and transformation patterns for image registration, thus demonstrating strong cross-task transferability. However, these models still face challenges in generalization and robustness when encountering novel anatomical structures, varying imaging conditions, or unseen modalities. To address these limitations, this paper incorporates Sharpness-Aware Minimization (SAM) into foundation models to enhance their generalization and robustness in medical image registration. By optimizing the flatness of the loss landscape, SAM improves model stability across diverse data distributions and strengthens its ability to handle complex clinical scenarios. Experimental results show that foundation models integrated with SAM achieve significant improvements in cross-dataset registration performance, offering new insights for the advancement of medical image registration technology. Our code is available at https://github.com/Promise13/fm_sam.

Deformable registration is a fundamental task in medical image processing, aiming to achieve precise alignment by establishing nonlinear correspondences between images. Traditional methods offer good adaptability and interpretability but are limited by computational efficiency. Although deep learning approaches have significantly improved registration speed and accuracy, they often lack flexibility and generalizability across different datasets and tasks. In recent years, foundation models have emerged as a promising direction, leveraging large and diverse datasets to learn universal features and transformation patterns for image registration, thus demonstrating strong cross-task transferability. However, these models still face challenges in generalization and robustness when encountering novel anatomical structures, varying imaging conditions, or unseen modalities. To address these limitations, this paper incorporates Sharpness-Aware Minimization (SAM) into foundation models to enhance their generalization and robustness in medical image registration. By optimizing the flatness of the loss landscape, SAM improves model stability across diverse data distributions and strengthens its ability to handle complex clinical scenarios. Experimental results show that foundation models integrated with SAM achieve significant improvements in cross-dataset registration performance, offering new insights for the advancement of medical image registration technology. Our code is available at https://github.com/Promise13/fm_sam}{https://github.com/Promise13/fm\_sam.

The four-wave mixing (FWM) process is an important method for generating collimated blue light from alkali vapor. In this paper, we introduced high-pressure He gas to broaden the absorption linewidth of Rb atoms for the pump laser. By using an optical parametric oscillator (OPO) that outputs tunable wavelength pump laser, we performed two-photon excitation of the Rb-He mixture and achieved redshifted and asymmetric broadened collimated light (CL) around 420 nm. We demonstrated the participation of Rb-He excimers in the FWM process for the first time, to the best of our knowledge, from both spectral and temporal dimensions, and revealed the underlying mechanism. Experiments showed that the peak of the ∼420 nm CL intensity appeared at higher heating temperatures, influenced by changes in the optical depth of the Rb-He mixture. This study provides both theoretical and experimental support for the promotion of a collimated blue light scheme in alkali metal -buffer gas systems, balancing the pump laser absorption and FWM efficiency.

The advent of Vision-Language Models (VLMs) in medical image analysis has the potential to help process multimodal inputs and increase performance over traditional inference methods. However, when considering the domain in which these models will be implemented, fairness and robustness are important to ensure the model stays true for any patient. In this paper, we introduce a framework for ensuring robustness and fairness of VLM models. This framework modifies the loss function at training by identifying and adjusting faulty image-text pairs through a Dynamic Bad Pair Mining algorithm and also utilizing Sinkhorn distance to ensure the loss distributions of protected groups do not deviate from the total loss. Experimental testing of our framework shows up to a 8.6\% improvement when looking at equity-scaled AUC.

Nanowire arrays present many unique advantages for solar-to-chemical energy conversion. One possible advantage is that photon recycling between neighboring nanowires has the potential to increase solar energy conversion efficiencies. Here, we explore three underlying mechanisms of optical and electronic coupling between neighboring nanowires─incident photon scattering, photon recycling, and charge-carrier transport from the photoexcited nanowire to the neighboring nanowire via the underlying substrate─using single nanowire-level microscopy and spectroscopy measurements. We present a comprehensive analysis of light absorption and emission of a single nanowire at open circuit, and subsequent re-absorption and re-emission by a neighboring nanowire. We developed a novel correlated single nanowire microspectroscopy and widefield imaging methodology to spatially resolve photon communication pathways between neighboring nanowires and selectively image re-emitted and reflected photons. We developed unique multiphysics models to couple wave optics and semiconductor photophysics to especially isolate contributions from photon recycling and electronic transport to photon emission from neighboring nanowires. By systematically varying the morphologies of the nanowires modeled, we identified pathways to maximize photon recycling between neighboring nanowires. We concluded that the measured photoluminescence is more strongly influenced by the diffusion of charge carriers as compared to photon recycling in materials with moderate-to-large charge-carrier mobilities (>10 cm 2 V –1 s –1 ), and that photon recycling dictates photoluminescence intensity only when the charge-carrier mobility is low (<1 cm 2 V –1 s –1 ). The experimental and simulation platforms developed herein for photon management strategies can be leveraged by the semiconductor photocatalysis community to enhance solar-to-chemical conversion efficiencies in semiconductor nanowire arrays.

Abstract The direct utilization of dissolved inorganic carbon in seawater for CO 2 conversion promises chemical production on-demand and with zero carbon footprint. Photoelectrochemical (PEC) CO 2 reduction (CO 2 R) devices promise the sustainable conversion of dissolved carbon in seawater to carbon products using sunlight as the only energy input. However, the diffusion-dominant transport mechanism and the near-zero concentration of CO 2 (aq) (CO 2 dissolved in aqueous solution) in static seawater has made it extremely challenging to achieve high solar-to-fuel (STF) efficiency and high carbon-product selectivity. Here, where CO 2 (aq) as a reactant generated in situ by acidification of HCO 3 - flows continuously from BiVO 4 photoanodes to Si photocathodes, enabling a single-step conversion of dissolved carbon into products. Our PEC device significantly increases the CO selectivity from 3% to 21%, which approaches the 30% theoretical limit according to multi-physics modeling. Meanwhile, the Si/BiVO 4 PEC CO 2 R device achieved a STF efficiency of 0.71%. Such flow engineering achieves flow-dependent selectivity, rate, and stability in simulated seawater, thus promising practical solar fuel production at scale.

MedSAM, a medical foundation model derived from the SAM architecture, has demonstrated notable success across diverse medical domains. However, its clinical application faces two major challenges: the dependency on labor-intensive manual prompt generation, which imposes a significant burden on clinicians, and the absence of semantic labeling in the generated segmentation masks for organs or lesions, limiting its practicality for non-expert users. To address these limitations, we propose AutoMedSAM, an end-to-end framework derived from SAM, designed to enhance usability and segmentation performance. AutoMedSAM retains MedSAM's image encoder and mask decoder structure while introducing a novel diffusion-based class prompt encoder. The diffusion-based encoder employs a dual-decoder structure to collaboratively generate prompt embeddings guided by sparse and dense prompt definitions. These embeddings enhance the model's ability to understand and process clinical imagery autonomously. With this encoder, AutoMedSAM leverages class prompts to embed semantic information into the model's predictions, transforming MedSAM's semi-automated pipeline into a fully automated workflow. Furthermore, AutoMedSAM employs an uncertainty-aware joint optimization strategy during training to effectively inherit MedSAM's pre-trained knowledge while improving generalization by integrating multiple loss functions. Experimental results across diverse datasets demonstrate that AutoMedSAM achieves superior performance while broadening its applicability to both clinical settings and non-expert users. Code is available at https://github.com/HP-ML/AutoPromptMedSAM.git.

Graph generative models have become increasingly prevalent across various domains due to their superior performance in diverse applications. However, as their application rises, particularly in high-risk decision-making scenarios, concerns about their fairness are intensifying within the community. Existing graph-based generation models mainly focus on synthesizing minority nodes to enhance the node classification performance. However, by overlooking the node generation process, this strategy may intensify representational disparities among different subgroups, thereby further compromising the fairness of the model. Moreover, existing oversampling methods generate samples by selecting instances from corresponding subgroups, risking overfitting in those subgroups owing to their underrepresentation. Furthermore, they fail to account for the inherent imbalance in edge distributions among subgroups, consequently introducing structural bias when generating graph structure information. To address these challenges, this paper elucidates how existing graph-based sampling techniques can amplify real-world bias and proposes a novel framework, Fair Graph Synthetic Minority Oversampling Technique (FG-SMOTE), aimed at achieving a fair balance in representing different subgroups. Specifically, FG-SMOTE starts by removing the identifiability of subgroup information from node representations. Subsequently, the embeddings for simulated nodes are generated by sampling from these subgroup-information-desensitized node representations. Lastly, a fair link predictor is employed to generate the graph structure information. Extensive experimental evaluations on three real graph datasets show that FG-SMOTE outperforms the state-of-the-art baselines in fairness while also maintaining competitive predictive performance.

To date, III-V semiconductor-based tandem devices with GaInP top photoabsorbers show the highest solar-to-electricity or solar-to-fuel conversion efficiencies. In photoelectrochemical (PEC) cells, however, III-V semiconductors are sensitive, in terms of photochemical stability and, therefore, require suitable functional layers for electronic and chemical passivation. GaN films are discussed as promising options for this purpose. The band alignment between such a protection layer and the III-V semiconductor should be aligned to minimize corrosion and nonradiative interfacial recombination and to promote selective charge carrier transport. Here, we investigate the band alignment between GaN passivation layers and n-type doped GaInP(100) photoabsorbers and grew n-type GaInP(100) epitaxially by metalorganic vapor phase epitaxy on oxidized GaAs(100) substrates to mimic a realistic preparation sequence. We prepared 1-20 nm GaN films on top employing atomic layer deposition and studied the band alignment at the GaN/GaInP(100) heterointerface by X-ray and ultraviolet photoelectron spectroscopy. Due to the limited emission depth of photoelectrons, we determined the band alignment by a series of measurements, in which we increased the thickness of the GaN films successively. The n-GaInP(100) surfaces, prepared with a well-known phosphorus-terminated p(2 × 2)/c(4 × 2) reconstruction, show an upward surface band bending (BB) of 0.38 eV and a Fermi level pinning due to the present surface states. Upon oxidation, the surface states are partially passivated, resulting in a reduction of the BB to 0.16 eV and a valence band offset (VBO) between the GaInP(100) and the thin oxide layer of 2.01 eV. Applying Kraut's approach, we identified a VBO of 1.90 eV and a conduction band offset of 0.44 eV between GaInP(100) with a thin oxide layer and the GaN passivation layer. We conclude that the GaN is a well-suited passivation layer for PEC cells and facilitates selective transport of photogenerated electrons.

Nanowire arrays present many unique advantages for solar-to-chemical energy conversion and are good model systems to investigate how the performance of one nanowire can influence others in an array. Spatially resolved photoluminescence is a powerful experimental characterization tool to quantify optical and electronic coupling between nanowires in an array. However, three underlying mechanisms of incident photon scattering, photon recycling, and charge-carrier diffusion dictate this coupling. In this study, we present a comprehensive analysis of light absorption and emission of a single nanowire at open circuit, and subsequent re-absorption and re-emission by a neighboring nanowire. We developed a novel correlated single nanowire micro-spectroscopy and widefield imaging methodology to spatially resolve photon communication pathways between neighboring nanowires and selectively image re-emitted and reflected photons. Unique multiphysics models have been developed to couple wave optics and semiconductor photophysics to especially isolate contributions from photon recycling and electronic transport to photon emission from neighboring nanowires. By systematically varying the morphologies of the nanowires modeled, we identify pathways to maximize photon recycling between neighboring nanowires. We conclude that the measured photoluminescence is more strongly influenced by the diffusion of charge-carriers as compared to photon recycling in materials with moderate-to-large charge-carrier mobilities (&gt; 10 cm2 V-1 s-1), and that photon recycling dictates photoluminescence intensity only when the charge-carrier mobility is low (&lt;1 cm2 V-1 s-1). The experimental and simulation platforms developed herein for photon management strategies can be leveraged by the semiconductor photocatalysis community to enhance solar-to-chemical conversion efficiencies in semiconductor nanowire arrays.

Electrothermal synthesis of commodity chemicals has received notable interest in recent decades as renewable electricity becomes more available and environmental challenges are increasingly recognized. Representative electrothermal approaches, such as Joule heating, microwaves, induction heating and plasma, have rapidly evolved from operating in millimeter-sized micro-reactors toward modular and even industrial-scale systems. Meanwhile, new chemical engineering concepts, such as dynamic and programmable operation for non-equilibrium chemical reactions using nanosecond- to millisecond-long energy pulsing, spatial and temporal heating by electrifying various reactor components (for example, the reactor walls, catalyst bed or reactant in porous media), and field-enhanced reactions and catalysis, have been discovered to improve synthesis outcomes. Despite the rapid progress of this field, there remain many knowledge gaps and technical hurdles. Here we review the critical engineering advances, analyze the unaddressed challenges and discuss the potential directions for the electrothermal synthesis of commodity chemicals toward its broader implementation for future chemical manufacturing. Electrothermal synthesis of commodity chemicals has received notable interest as renewable electricity becomes more available and environmental challenges are increasingly recognized. This Perspective discusses critical engineering advances, unaddressed challenges and potential directions for the electrothermal synthesis of commodity chemicals toward its broader implementation for future chemical manufacturing.

Multilayered microstructures can accelerate the proton conduction process in proton exchange membranes (PEMs). Herein, we design and construct PEMs with microstructures based on bifunctional nanofibers and sulfonated poly(ether ether ketone) (SPEEK) nanofibers. Specifically, the bifunctional nanofibers composed of poly(vinyl alcohol) and chitosan are prepared and then combined with the electrospun SPEEK nanofibers. The stable microstructure is derived from the compatible interfacial property of nanofibers and the formed hydrogen bonds. The multilayered microstructure consisting of nanofibers accelerates the proton conduction even at subzero temperature because of regulating the proton conduction pathways. Specifically, the (SKNF/CPNF/SKNF)/PA membrane exhibits the proton conductivities of (0.951 ± 0.138) × 10 –2 S/cm at −30 °C and (7.32 ± 0.37) × 10 –2 S/cm at 160 °C. Additionally, the fine proton conductivity stability is demonstrated by the proton conductivity in the long-term test and the cooling/heating cycle test, such as 1.67 × 10 –2 S/cm at −30 °C (after 1000 h), 4.52 × 10 –2 S/cm at 30 °C (after 810 h), 1.12 × 10 –2 S/cm at −30 °C, and 1.01 × 10 –1 S/cm at 30 °C in the cooling/heating process (5 cycles). The single fuel cell possesses an open-circuit voltage of 0.886 V and a peak power density of 0.508 W/cm 2 at 130 °C.

The robustness of unmanned aerial vehicle (UAV) tracking is crucial in many tasks like surveillance and robotics. Despite its importance, little attention is paid to the performance of UAV trackers under common corruptions due to lack of a dedicated platform. Addressing this, we propose UAV-C, a large-scale benchmark for assessing robustness of UAV trackers under common corruptions. Specifically, UAV-C is built upon two popular UAV datasets by introducing 17 common corruptions from 4 representative categories including adversarial, sensor, blur, and composite corruptions in different levels. Finally, UAV-C contains more than 10K sequences. To understand the robustness of existing UAV trackers against corruptions, we extensively evaluate 12 representative algorithms on UAV-C. Our study reveals several key findings: 1) Current trackers are vulnerable to corruptions, indicating more attention needed in enhancing the robustness of UAV trackers; 2) When accompanying together, composite corruptions result in more severe degradation to trackers; and 3) While each tracker has its unique performance profile, some trackers may be more sensitive to specific corruptions. By releasing UAV-C, we hope it, along with comprehensive analysis, serves as a valuable resource for advancing the robustness of UAV tracking against corruption. Our UAV-C will be available at https://github.com/Xiaoqiong-Liu/UAV-C.

In recent years, efficient super-resolution research has focused on reducing model complexity and improving efficiency by leveraging deep small-kernel convolution, but it has the problem of a small receptive field, which leads to a limited ability of the network to reconstruct details. Large kernel convolution can provide a large receptive field and lead to a substantial enhancement in the quality of image reconstruction, but its computational cost is too high. To minimize the model’s parameter count and achieve efficient super-resolution reconstruction, this study introduces a symmetric visual attention network. The network decomposes the large kernel convolution into three different lightweight and efficient convolutions. It then forms a bottleneck structure by leveraging the varied receptive field sizes of these convolutions in combination. The attention mechanism is integrated to create a bottleneck attention module, enhancing the network’s feature awareness. Furthermore, the bottleneck attention modules are symmetrically arranged to construct a symmetric large kernel attention block, thereby further enhancing the network’s capability to extract deep features. The experimental results demonstrate that the proposed model achieves competitive quantitative metrics when compared to other lightweight super-resolution methods, and the details of the reconstructed images are enhanced. With only 183K parameters, the model achieves a lightweight yet high-quality super-resolution model, offering a novel solution approach for efficient super-resolution.

Besides the scattering structures, the energy transfer (ET) process in the gain medium plays a significant role in the competition between coherent (comprising strongly coherent components) and incoherent (consisting of weakly coherent or "hidden" coherent components) modes of random lasers. In this study, bichromatic emission random lasers were successfully created using polydimethylsiloxane (PDMS) replicas with grooved structures that imitate the inner surface of abalone shells as scattering substrates. The influence mechanism of the ET process from the monomer to dimer in the Rhodamine 640 dye on the competition of random laser modes was thoroughly investigated from both spectral and temporal dimensions. It was confirmed that the ET process can reduce the gain of monomers while amplifying the gain of dimers. By considering the dominant high-efficiency ET processes, an energy transfer factor associated with the pump energy density was determined. Notably, for the first time, it was validated that the statistical distribution characteristics of the time sequence variations in the coherent random laser generated by dimers closely resemble a normal distribution. This finding demonstrates the feasibility of producing high-quality random number sequences.

The robustness of unmanned aerial vehicle (UAV) tracking is crucial in many tasks like surveillance and robotics. Despite its importance, little attention is paid to the performance of UAV trackers under common corruptions due to lack of a dedicated platform. Addressing this, we propose UAV-C, a large-scale benchmark for assessing robustness of UAV trackers under common corruptions. Specifically, UAV-C is built upon two popular UAV datasets by introducing 18 common corruptions from 4 representative categories including adversarial, sensor, blur, and composite corruptions in different levels. Finally, UAV-C contains more than 10K sequences. To understand the robustness of existing UAV trackers against corruptions, we extensively evaluate 12 representative algorithms on UAV-C. Our study reveals several key findings: 1) Current trackers are vulnerable to corruptions, indicating more attention needed in enhancing the robustness of UAV trackers; 2) When accompanying together, composite corruptions result in more severe degradation to trackers; and 3) While each tracker has its unique performance profile, some trackers may be more sensitive to specific corruptions. By releasing UAV-C, we hope it, along with comprehensive analysis, serves as a valuable resource for advancing the robustness of UAV tracking against corruption. Our UAV-C will be available at https://github.com/Xiaoqiong-Liu/UAV-C.