We introduce a simple, easy to implement, and computationally efficient tropical convolutional neural network architecture that is robust against adversarial attacks. We exploit the tropical nature of piece-wise linear neural networks by embedding the data in the tropical projective torus. This can be accomplished with a single additional hidden layer called a tropical embedding layer, and can in principle be added to any neural network architecture. We study the geometry of the resulting decision boundary, and find that like adversarial training and various regularization techniques that have been proposed, adding the tropical embedding layer tends to increase the number of linear regions associated with the decision boundaries. Our numerical experiments show that our approach achieves state-of-the-art levels of adversarial robustness, while requiring much less computational time than adversarial training.

Publisher Copyright: © 2026 The Authors.

The current work presents an optimized neural network for predicting depth of cut (DoC) in the abrasive water jet machining (AWJM) of Titanium Ti6Al4V Grade 5 material. Five process parameters were used for experimentation i.e., water pressure (Wp), traverse speed (Ts), nozzle to orifice diameter (N/Odia), abrasive mass flow rate (Amf), and abrasive orifice size (Aos). Experiments were carried out using the Taguchi based L27 Orthogonal array, resulting in the development of a regression equation describing the process behaviour. The DoC was modelled using two neural network architectures: a single hidden layer network (NN1), where neurons were varied from 1 to 10, and a deep neural network (NND), where both neurons and hidden layers were varied to determine the optimum network configuration. Four different activation functions, namely, Sigmoidal, Gaussian, Tanh, and Linear functions were employed to perform the optimization. The dataset for training and testing the neural network models was generated using the regression equation based on experimental data, rather than direct experimental measurements. The optimization of both the neural network architectures was evaluated using two performance metrics, i.e., root mean squared error (RMSE) and co-efficient of determination ([Formula: see text]). The results concluded that both the neural network (NN) models were able to accurately predict DoC, however, the NND was found to be better in terms of accuracy with a 12.5% reduction in testing RMSE and 4.2% improvement in testing ([Formula: see text]). Amongst the activation functions used, the Gaussian functions outperformed the other functions across both neural network architectures, as its symmetric response reflects the jet energy distribution peaking at the centre and decaying radially, thereby effectively capturing the non-linear penetration behavior. SEM analysis supported the computational results, showing micro-cutting with stable grooves for machined surfaces with higher DoC, whereas lower DoC experienced micro-ploughing and ridge formation. Contour plots further revealed that DoC in AWJM of Ti6Al4V is primarily governed by water pressure and traverse speed, thereby confirming the capability of the optimised NND architecture to capture the underlying physical mechanisms of AWJM.

Ion-gated transistors inherently exhibit time-dependent behavior governed by ionic motion associated with electric double-layer formation; however, their practical implementation has been limited by insufficient control over ionic dynamics and poor compatibility with scalable thin-film integration. Here, we present carbon nanotube (CNT) solid-ion-gated transistors (sIGTs) that allow the wide-range engineering of ionic dynamics while remaining fully compatible with wafer-scale thin-film processing. Tunable ionic conductance is achieved by ionic content engineering in the film and thickness scaling into the sub-micron regime, enabling ionic time constants from microseconds to milliseconds. CNT sIGTs demonstrate robust DC operation at low ionic content with an optimized polymer matrix and wafer-scale fabrication on flexible substrates. Frequency-dependent gate modulation governed by ionic conductance is systematically investigated through electrical impedance spectroscopy and small-signal analysis, including a comparison of the -3dB cutoff frequency and the transit frequency. This analysis provides direct insight into the relationship between ionic conductance and frequency-dependent device response, exhibiting consistent trends across both two-terminal and three-terminal device configurations. Monolithic three-dimensional integration of two-tier CNT sIGTs with engineered dynamic responses is demonstrated as a compact dual-timescale physical reservoir for neuromorphic computing that enables classification of time-varying inputs using a single readout layer.

Dipolar excitons typically emerge in weakly coupled van der Waals heterostructures (vdWHs), where electrons and holes are confined in different layers. However, the tunability of these extrinsic interlayer dipolar excitons under external out-of-plane electric fields is constrained by built-in interfacial electric fields and significant nonradiative processes. Here, we propose a dipolar exciton in monolayer Hf2SiCO2, where vertically separated electrons and holes reside in a single layer of several atoms' thickness. The dipolar excitons in the two X valleys, connected by rotoreflection symmetry, possess alternating antiparallel out-of-plane electric dipole moments, which are termed an alterexciton. These dipolar excitons exhibit electrically tunable polarization in a single valley, which further leads to a single-valley excitonic insulator under an increasing electric field. Because of the optical selection rules, the layer-locked valley excitons exhibit linear dichroism and valley-dependent electrical tunability. Furthermore, under linearly polarized light, the Coulomb-bound electrons and holes of the excitons are simultaneously deflected by the Berry curvature in each layer-locked valley, giving rise to the exciton Hall effect. These results not only contribute to the valley-polarized manipulation of dipolar excitons but also facilitate the exploration of single-valley single-photon emitters.

Natural fibre-reinforced polymer composites (NFRPCs) are sustainable alternatives to synthetic laminates; conversely, their structural reliability is influenced by type of fibre, fibre orientation and stacking architecture. In this work, ten sample laminate configurations (S1-S10) with Flax (F), Hemp (H), and Kenaf (K) fibres were fabricated using a hand lay-up method with single-layer (SL), double-layer (DL), and alternating-layer (AL) sequences under 0/90° and ± 45° orientations. Mechanical properties, namely tensile, flexural, and compressive strengths, were experimentally evaluated. The double-layer 0/90° configuration (S7) demonstrated optimal mechanical performance, achieving Tensile, Flexural, and Compressive strengths of 57.13MPa, 60.50MPa, and 638.63MPa, respectively. Shifts in the laminate orientation from 0/90° to ± 45° resulted in tensile reductions of up to 54%, thereby enhancing shear-dominated stress redistribution. A mechanics-informed Structural Performance Index (SPI) framework was developed, incorporating normalised tensile, flexural, and compressive responses into application-specific structural indicators. The SPI-based evaluation recognised S7 as the best arrangement for structural and crash-dominant scenarios (SPI = 0.99), whereas kenaf-dominant laminates (S5) were recommended for the stiffness-oriented applications. The proposed SPI methodology delivers a decision tool that associates experimental characterisation and reliability-driven laminate selection for multi-axial engineering applications.

We present an experimentally feasible implementation of a secure multiparty computation application enabled by quantum oblivious transfer (QOT) on an entanglement-based physical layer. The QOT protocol uses polarization-encoded entangled states to share oblivious keys between two parties with quantum key distribution (QKD) providing authentication. Our system integrates the post-processing for QOT and QKD, both sharing a single physical layer, ensuring efficient key generation and authentication, respectively. Authentication involves hashing messages into a cryptographic context, verifying tags, and replenishing keys. This process uses a parallel QKD pipeline specifically for authentication, not for secure key generation. Oblivious keys are generated over a distance up to 25.8 km with a channel loss of 8.47 dB. In a back-to-back setup, a QOT rate of [Formula: see text] OTs/second is achieved, corresponding to 1 minute and 53 seconds per OT, primarily limited by the entanglement source. Using pre-distributed oblivious keys improved the rate to 0.11 OTs/second, or 9.1 seconds per OT. The considered QOT protocol is statistically correct, computationally secure for an honest receiver, and statistically secure for an honest sender, assuming a computationally hiding, statistically binding commitment. An experimentally feasible use case is demonstrated for privacy-preserving fingerprint matching against no-fly lists for border control. The fingerprint is secret-shared across two sites, ensuring security, while the matching is performed using the MASCOT protocol, supported by QOT. The application required 128 1-out-of-2 OTs, each with message length of 128 bits, with the highest security achieved in 20 minutes and 39 seconds. This work demonstrates the feasibility of QOT in secure quantum communication applications.

The durability of platinum (Pt) electrocatalysts in electrochemical energy conversion is fundamentally challenged by surface oxidation and dissolution during electrochemical operation. Although Pt surface oxidation has commonly been discussed in terms of a "place exchange" mechanism between Pt and oxygen, atomic-scale insight into its potential-dependent progression has remained limited. Herein, we directly visualize atomic electrooxidation and dissolution of {111}-terminated surfaces of octahedral Pt nanoparticles by employing ex situ differential-phase-contrast scanning transmission electron microscopy combined with online inductively coupled-plasma mass spectroscopy and density functional theory calculations. We reveal the atomistic structural evolution of the {111} nanoparticle surface with progressively increasing electrode potentials (0.8-1.5 V), from the initial lattice expansion induced by adsorbed oxygen, to vacancy-induced formation of two-dimensional, lattice-contracted PtOx monolayers, and finally to a dimensional transition to three-dimensional PtO2 growth. Furthermore, we demonstrate how the potential cycling protocols (triangular versus square wave cycling) decisively control the final oxide's dimensionality (multilayer versus single layer) and stability. These atomic-scale insights establish how electrochemical conditions dictate Pt oxidation pathways and atomistic structural evolution, providing a mechanistic basis for understanding and improving the durability of Pt-based electrocatalysts.

Accurate measurement of ammonium (NH4+) and nitrate (NO3-) concentrations in soils is essential for understanding factors that determine nitrogen bioavailability and potential losses to the wider environment. Diffusive gradients in thin-films (DGT) can be used to examine NH4+ and NO3- dynamics in soils, while overcoming some of the limitations faced by conventionally used extraction methods. This study evaluated the effects of temperature, soil type, and moisture content on NH4+ and NO3- dynamics using DGT across two experiments. In the first experiment, NH4+ and NO3- were applied to three South Australian soils and incubated at three moisture contents. Single and mixed binding layer DGT probes were deployed for 12, 24, and 48 h to examine short-term changes in N species concentrations. In the second experiment, NH4+ and NO3- concentrations and transformation rates were monitored in one soil over 42 days following urea application, at three soil moisture contents and two temperatures (15 and 24 °C). Soil type, moisture content, and DGT deployment time affected measured NH4+ and NO3- concentrations in the first experiment. However, the mixed binding layer DGT underestimated concentrations of both species by up to 66%. In the second experiment, DGT-measured NH4-N increased to up to 3044 ± 379 µg L-1 within 3 days under the warmest, wettest treatment, with transformation rates of both N species further influenced by temperature, moisture content, and time after urea application. These dynamics were clearly resolved by DGT, thus demonstrating the technique's potential to provide valuable new insight into soil nitrogen cycling.

Emphysematous cystitis, also known as 'cystitis emphysematosa,' is a rare complicated urinary tract infection characterized by gas within the bladder wall. Most cases are mild and detected incidentally; however, delayed management can lead to significant morbidity and mortality, particularly when infection extends to the upper urinary tract. With increasing use of cross-sectional imaging, radiologists are encountering this condition more frequently. This pictorial review presents 20 cases of emphysematous cystitis from our institution, illustrating the range of CT imaging patterns and clinical features from representative cases. We identified three distinct patterns of gas distribution: loculated or curvilinear focal collections (40% of cases), single layer of cobblestone/beaded necklace pattern (30% of cases), and circumferential dissection of the bladder wall with multilayered gas (30% of cases). Diabetes is the most common predisposing factor. Other risk factors included neurogenic bladder, long-term urinary catheters, immunosuppression, chemotherapy, recurrent urinary stones, and frequent urologic procedures. The clinical presentation ranges from incidental detection on imaging to symptomatic patients presenting with acute abdomen and dysuria in the emergency department (ED). Treatment primarily consists of antibiotic therapy, with surgical intervention required only in select severe cases. Understanding these CT patterns along with the associated clinical features and risk factors allows radiologists to accurately diagnose emphysematous cystitis and guide timely clinical management.

To compare the sectoral inner macular layers thicknesses among eyes with pseudoexfoliation syndrome (PXS), pseudoexfoliation glaucoma (PXG), and healthy controls. This cross-sectional study included 27 eyes from 14 patients with early pseudoexfoliation glaucoma, 39 eyes from 29 patients with pseudoexfoliation syndrome, and 37 eyes from 19 healthy controls. All eyes underwent spectral-domain optical coherence tomography (SD-OCT) of the macula using the Glaucoma Module Premium Edition of Spectralis SD-OCT (Heidelberg Engineering, Heidelberg, Germany). The sectoral thicknesses of the retinal nerve fiber layer (RNFL), ganglion cell layer (GCL), and inner plexiform layer (INL) were measured. Differences in sectoral thicknesses of the RNFL, GCL, and IPL between groups were evaluated using generalized estimating equations. The diagnostic performance of single and composite macular layers in detecting PXG was assessed using the area under the receiver operating characteristic curve (AUC). There were no significant differences in age or sex between groups (p > 0.05). We observed significant differences in ganglion cell and inner plexiform layer thicknesses in sectors 2-5 between the groups (p < 0.05). Both PXG and PXS eyes showed significant sectoral thinning of the GCL and IPL compared to controls. The ganglion cell complex (GCC) provided the highest AUC for discriminating between PXG eyes, particularly in sector 4. Early PXG preferentially affects the inner macular layers, particularly the IPL, with sectoral vulnerability. Composite macular layers, such as the GCC, demonstrated modest discriminatory ability and may have adjunctive value in the detection of early glaucoma.

Hydroxyapatite (HA) on a magnesium (Mg) surface is studied using density functional theory, to help understand the effect of HA coating and alloying in the surfaces of Mg-based biodegradable implants. We determine the adsorption energies and structural changes of a single layer of HA on pure Mg(0001) and on sparsely calcium (Ca) or zinc (Zn) doped Mg(0001) and find that both Zn and Ca doping improves the adsorption, except in a few positions of HA relative to the dopant position. All adsorption configurations, whether with pure or doped Mg surfaces, show deformation of the surface and HA layer. For Ca doping, we found that for a certain adsorption configuration, the dopant Ca atom moves out of the Mg surface and into the HA layer, leaving behind a Mg vacancy in the top layer of the Mg surface. Plots of electron density changes show that electrons accumulate around the Ca dopant and the neighboring Mg atoms, while in Zn doping this is less pronounced. Overall, our results demonstrate that the dopant choice and relative position of HA influence the interaction between HA and Mg-surfaces, and affect both adsorption energies and atomic and electronic structures.

Two-dimensional hole gases (2DHGs) offer enhanced spin-orbit coupling and correlations, making them attractive for spintronic and quantum devices. Yet, most realizations require complex epitaxial structures and protective encapsulation, limiting their applicability. Here, we show that a stable 2DHG emerges in Ge(111) upon deposition of a single monolayer of Pt. Angle-resolved photoemission spectroscopy reveals multiple Ge-derived hole subbands, including light-hole-, heavy-hole-, and split-off-like branches with two-dimensional dispersion, effective masses enhanced by a factor of about 3.5 compared to bulk Ge, and a spin-orbit splitting of about 0.3 eV. Remarkably, the subsurface hole bands persist after air exposure. Transport measurements confirm conductive behavior and hole-type carriers down to cryogenic temperatures. The coexistence of strong spin-orbit interaction and exceptional ambient robustness, combined with CMOS compatibility and without requiring complex heterostructures, gating, or encapsulation, establish Pt/Ge(111) as a versatile platform for correlated phenomena and spintronic and quantum devices.

To investigate microstructural alterations in the retina, choroid, and optic nerve head in children with spina bifida using optical coherence tomography. Thirty-eight children with spina bifida and 32 children without spina bifida were included in this cross-sectional study. Both groups were matched for age and sex. Comprehensive ophthalmological examinations included optical coherence tomography imaging of the macula, choroid, and peripapillary retinal nerve fiber layer. The mean age was 10.2 ± 3.5 years. Axial length was significantly shorter in patients with spina bifida (P = .025), and hyperopia was more frequent (P < .001). Central macular thickness was similar between the two groups (P = .553). However, single retinal layer analysis revealed significant differences. In patients with spina bifida, the outer nuclear layer was thicker (P = .019) and the inner nuclear and outer plexiform layers were thinner (P = .038, P = .035). Subfoveal and nasal choroidal thicknesses were greater in patients with spina bifida (P = .012, P = .038). However, there was no significant difference in choroidal thickness after comparing the two groups by controlling the axial length variable. Children with spina bifida may have retinal and choroidal alterations detectable with optical coherence tomography. Choroidal changes may be related to shorter axial length, whereas retinal alterations could reflect impaired macular development in spina bifida.

Metal halide perovskites combine mixed ionic-electronic conduction with low-temperature processing, yet most perovskite memristors operate in a single mode, either nonvolatile bipolar resistive switching (RS) or volatile threshold switching (TS). This limits in-device multi-role reuse and drives peripheral overhead. We introduce a single asymmetric transport layer at one contact that makes bias polarity the mode selector. The built-in field adds to or opposes the applied field, yielding nonvolatile RS under positive sweeps and volatile TS with self-reset under negative sweeps in the same device. The approach is process-compatible across C60, NiOx, and Spiro-OMeTAD. In C60-based devices, we further observe improved retention (∼2.0 × 104 s), endurance (∼6 × 103 cycles), and storage stability (∼1300 h) vs. symmetric control. Mechanistically, Mott-Schottky analysis reveals a finite built-in voltage that stabilizes or dissolves halide-vacancy filaments depending on polarity, explaining the reconfigurability. Functionally, negative pulses realize short-term, self-erasing updates, while positive pulses consolidate long-term weights under a unified read bias-providing an intrinsic stability-plasticity balance. This minimal asymmetry streamlines compute-in-memory and neuromorphic systems.

EWOD fabrication is typically costly, requiring specialist microfabrication facilities, equipment and techniques like physical vapour deposition. Printed electronics methods like inkjet printing offer cheaper and faster alternatives to typical fabrication methods, and conductive polymers like PEDOT:PSS offer advantages in flexibility, cost, and transparency. We report a low-cost fabrication approach and characterise its performance against standard ITO-glass electrodes. Ink-on-paper electrodes and patterned-on-glass electrodes are restricted to a single plane, limiting the number of electrically independent electrodes on a single conductive layer. Our design and fabrication approach, employing inkjet printing and double-sided patterning, overcomes topological constraints, addressing both connectivity and cost. PEDOT:PSS is highly compatible with inkjet printing technology, and a single pass of PEDOT:PSS printing was sufficient to achieve the desired conductivity, offering a cost-effective solution for fabricating large, multilevel, flexible EWOD electrode arrays.

ABSTRACTA comprehensive understanding of age‐related changes in the human immune system is critical for deciphering the mechanisms of immunosenescence. Although transcriptomic studies have described immune alterations during aging, investigations into the proteomic layer and integrated multi‐omics approaches remain scarce. Here, we performed multi‐omics sequencing on peripheral blood samples from 69 individuals aged 23–75 years. We identified widespread dysregulation of RNA splicing in aging immune cells, with exon skipping being the most prevalent splicing event, which was mainly enriched in protein regulation‐related processes. Notably, lncRNAs such as SNHG1 and RP1‐3J17.3 modulated the expression of splicing‐regulatory mRNAs and proteins in aging immune cells, thereby influencing the RNA splicing of downstream genes. Splicing dysregulation of genes including EIF4G1, a translation initiation factor, led to alterations in protein translation, modification, and degradation, ultimately reshaping protein profiles in aging leukocytes. These proteomic alterations in our data reflect a gradual weakening of T‐cell function during aging. Together, unlike previous studies that focused only on single molecular layers, our study elucidates the interconnected nature of molecular regulation within leukocytes during aging, uncovers cross‐level interactions and regulatory networks, and provides a comprehensive multi‐omics resource for understanding immune aging and identifying potential strategies to delay immunosenescence.

Chiral liquid crystal elastomers (CLCEs) have been widely explored for structural color modulation via tunable photonic bandgaps, yet their functionality has remained largely limited to optical responses. Here, we introduce an opto-acoustic chiral liquid crystal elastomer actuator (OA-CLCEA) that redefines CLCEs as intrinsically multimodal transducers, enabling simultaneous color modulation and sound generation within a single soft material layer. Under a static DC electric field, Maxwell stress induces thickness compression and reduces the chiral pitch, producing continuous, wavelength-resolved structural color tuning, while an AC electric field drives membrane vibration to generate audible sound across 20 Hz-20kHz. Notably, the intrinsic DC-AC cross-term in the Maxwell stress enables simultaneous yet decoupled control of optical wavelength and acoustic frequency, allowing genuine orthogonal modulation within a single material system. Unlike conventional multimodal platforms based on heterogeneous architectures, the OA-CLCEA functions as both a photonic modulator and an acoustic emitter through a unified electromechanical mechanism. This monolithic design simplifies device architecture while enabling compactness and electrical programmability, and establishes a material-level opto-acoustic transduction strategy. These results further position CLCEs as an unexplored platform for active acoustic emission and provide a scalable route toward reconfigurable multimodal devices and synesthetic human-machine interfaces.

Interface passivation is crucial to reduce nonradiative recombination losses at the perovskite-electron transport layer interface and enhance the power conversion efficiency (PCE) of perovskite solar cells. Various molecules are known to result in a gain in open-circuit voltage (VOC). However, this gain is often associated with increased instability. Here, we investigate the interface passivation of a Cs0.1FA0.6MA0.3Pb0.5Sn0.5I3 narrow-bandgap (1.26 eV) perovskite by ammonium iodide derivatives with multiple, primary, secondary, or tertiary ammonium iodide groups, connected via alkane linkers of different lengths to establish structure-property relationships. The impact of these passivators on interfacial recombination, charge extraction, and device stability is elucidated by tracking the quasi-Fermi level splitting (QFLS) of perovskite layers and perovskite/C60 bilayers, together with the VOC of complete solar cell devices over one month. All tested passivators reduce the nonradiative recombination losses at the perovskite/C60 interface, but the extent to which this translates into an improved photovoltaic performance strongly depends on the molecular structure. Short-chain primary diammonium passivators provide the most favorable balance between effective passivation and charge extraction, yielding QFLS values that closely match the device VOC. In contrast, extended and branched multiammonium passivators improve photovoltage stability but impede charge carrier extraction, leading to reduced fill factors (FF) and short-circuit current densities (JSC). Secondary ammonium terminal groups promote crystallite formation on the perovskite surface, which reduces the efficacy of passivation. Importantly, combining small with larger-sized passivators in a single layer enables simultaneous enhancement of QFLS, VOC, FF, and operational stability.

This study aimed to redefine the precise morphology and relationships of the temporal fasciae using high resolution sectioned images and three dimensional (3D) models. In the temporal region, fascial layers are difficult to distinguish by cadaveric dissection, which has contributed to inconsistent anatomical terminology. Therefore, we undertook this study to investigate and clarify these fasciae. Using the sectioned images with true color and high resolution, fascial, muscular, and vascular structures in the temporal region were traced, segmented, and reconstructed into surface models using MATLAB and Slicer. Maya scripts integrated these models with the sectioned images for 3D visualization. Three fascial layers (the temporoparietal fascia, the superficial layer of the deep temporal fascia, and the deep layer of the deep temporal fascia) were clearly distinguished. The deep temporal fascia presented as a single layer posteriorly but divided into two layers anteriorly. Distinct fat pads were consistently identified between each fascial layer and between the deep fascia and temporal muscle. The courses of the superficial temporal artery and middle temporal vein were precisely visualized. This study provides a clarified, 3D anatomical framework of the temporal fascia, resolving discrepancies in traditional descriptions. High fidelity, artifact free image data offer a reliable basis for safer surgical approaches, flap elevation, filler injections, and regional anesthesia in the temporal region.