搜索结果：Nature electronics

Transition metal doping in two-dimensional (2D) semiconductor materials is an efficient way to develop new spintronic materials. In this work, the effects of Fe doping on the PbI2 monolayer electronic and magnetic properties are investigated using first-principles calculations. The pristine PbI2 monolayer is an indirect-gap semiconductor with a band gap of 2.50 eV. Fe doping induces effective magnetism engineering in this 2D material with a total magnetic moment of 4.00 µ B and in-plane magnetic anisotropy (IMA). Herein, the magnetism is produced mainly by Fe impurities. Parallel spin coupling with perpendicular magnetic anisotropy (PMA) is found in the monolayer doped with two Fe atoms (2Fe@PbI2), meanwhile the triangular doping configuration (3Fe@PbI2) induces antiparallel spin coupling with the IMA. Codoping with Br atoms can enhance the thermodynamic stability and tune the system magnetism. Specifically, Br codoping induces the ferromagnetic-to-antiferromagnetic state transition in the 2Fe@PbI2 system and enhances the antiparallel spin coupling in the 3Fe@PbI2 system. Moreover, the PMA and IMA in these systems, respectively, are also affected, becoming weaker. In all cases of doping and codoping, the magnetic semiconductor nature with relatively large spin-dependent energy gaps is obtained. Our findings introduce the doped and codoped PbI2 monolayer systems as new promising 2D spintronic materials with a magnetic semiconductor nature and tunable magnetic anisotropy, which can be selectively employed for magnetic field sensing and Magnetoresistive Random Access Memories (MRAMs) fabrication.

Flexible electronic textiles (e-textiles) have emerged as a key class of wearable systems by integrating electronic functionality with the intrinsic advantages of textile materials, such as breathability, flexibility, lightweight nature, and conformability to complex surfaces. However, the effective integration of diverse textile manufacturing technologies with advanced electronic components remains a critical challenge. This review systematically discusses in detail the formation methods of fibers, yarns, and fabrics in relation to their specific structures, followed by a comprehensive overview of the advancements of e-textile fabrication techniques, emphasizing on representative examples across different scales of textile-based electronic architectures. Furthermore, key applications of e-textiles in sensing, energy harvesting and storage, actuation, and human-machine interfaces are reviewed. Finally, current challenges and future perspectives for fabricating flexible e-textiles are discussed, aiming to provide insights and guidance for the rational design of next-generation wearable electronic systems.

The separation of minor actinides, especially americium, from lanthanides in spent nuclear fuel remains a critical challenge in nuclear waste management, primarily due to their nearly identical chemical behavior in the trivalent state. To address this, we target the linear dioxo configuration of pentavalent americium (AmO2+), which offers distinct steric and electronic features compared to spherical trivalent lanthanides. This work investigates Am(V) adsorption using a phenanthroline-based covalent organic framework (DAPhen-COF), where the pre-organized N,O-donor environment from the phenanthroline-amidine motif is designed for strong actinide coordination. Multiscale computations show that DAPhen-COF forms a highly stable complex with AmO2+. Density of states analysis reveals strong orbital hybridization between Am-5f and ligand N/O-2p states, underscoring substantial covalent interaction. Topological analysis of electron density confirms the existence of bonds with pronounced covalent character within highly polarized coordination environments, while electrostatic potential analysis verifies complementary electrostatic contributions. Energy decomposition analysis further quantifies the binding as a cooperative interplay between orbital and electrostatic forces. Quantitative adsorption energy calculations further corroborate these findings, revealing that DAPhen-COF exhibits a strong affinity for Am(V) (-151.9 kcal/mol) and clear selectivity over Eu(III), with the most stable configuration arising from cooperative actinide-actinide interactions within the confined COF interlayer space. This study not only sheds light on the unique coordination chemistry of pentavalent americium but also provides a robust theoretical foundation for designing ligand architectures capable of distinguishing actinides from lanthanides based on oxidation-state-specific motifs.

Soft, conductive materials that simultaneously offer mechanical compliance, electrical conductivity, recyclability, and scalable manufacturing remain difficult to achieve. Here, we introduce recyclable, 3D-printable PEDOT:PSS syntactic foams based on polymer microspheres coated in situ, enabling precise control over porosity, mechanical properties, and electrical conductivity. Microsphere-templated conductive shells provide a unique materials architecture that unites porosity control, tailorable flexibility, electrical functionality, recyclability, and additive manufacturability, which are capabilities that are challenging to achieve in conventional conductive foams. The resulting foams exhibit low elastic moduli comparable to soft biological tissues, stable electrical conductivity over a wide void-fraction range, and excellent compatibility with fused filament fabrication. Importantly, the foams can be thermally reshaped, and the conductive microsphere component can be recovered from the matrix by selective dissolution, while preserving structural integrity and electrical performance. To demonstrate functional relevance, the conductive foams are integrated into soft wearable electrodes, enabling reliable acquisition of electrocardiogram (ECG), electromyogram (EMG), and in-ear electroencephalogram (EEG) signals, including clear alpha-band modulation. Overall, this work establishes conductive syntactic foams as a recyclable and manufacturable materials platform for soft and wearable electronic systems.

We report the fabrication and physical characterization of a self-driven heterojunction photodetector based on an all-inorganic CsPbBr3thin film interfaced with a lightweight, freestanding p-type single-crystal silicon membrane (p-Si). Different from conventional rigid substrates, this membrane-based architecture leverages the superior carrier mobility and mechanical resilience of monocrystalline silicon for flexible, high-sensitivity optoelectronics. Beyond standard characterization, an Elliott-type quantum well model was employed to resolve the excitonic nature of the CsPbBr3nanocrystals, revealing a significant exciton binding energy of 340 meV and a quasi-particle bandgap of 2.91 eV. Under zero-bias conditions, the device exhibits a robust broadband response with a peak responsivity of 0.1 A/W and a detectivity of 1.75 × 1011Jones. Crucially, the strong built-in potential (Vbi= 0.91 V) facilitates efficient exciton dissociation, overcoming Coulombic attraction even at zero bias to enable high-performance self-powered operation. To elucidate the underlying transport physics, we conduct an in-depth study of the frequency-dispersive capacitance-voltage (C-V) characteristics. Our study reveals how interface states and deep-level defects dictate the steady-state photocurrent and transient kinetics (∼130 ms).C-Vanalysis further enabled the quantification of the interface trap density (Dit∼ 1011eV-1cm-2), confirming significant interfacial defect contributions to the non-ideal device behavior. By employing a modified depletion approximation incorporating the diode ideality factor (η), we decouple parasitic trap-assisted processes from the intrinsic junction capacitance to resolve a preciseVbiof 0.91 V. This rigorous analytical approach provides critical insights into the non-ideal junction physics governing the performance of self-powered, membrane-based perovskite optoelectronics.

The advancement of flexible electronics demands conductive polymers that combine high conductivity, optical transparency, and mechanical resilience. Poly(3,4-ethylenedioxythiophene): polystyrene sulfonate (PEDOT:PSS) has emerged as a promising candidate for capacitive touchscreens and flexible energy storage devices, yet its widespread adoption has been hindered by the intrinsic trade-off between conductivity enhancement and material compatibility. While strong acids can significantly improve PEDOT:PSS conductivity, their corrosive nature causes irreversible damage to device components. Here, we develop a benign methanol-benzoic acid co-modification strategy that achieves unprecedented performance metrics. The optimized PEDOT:PSS films demonstrate a record conductivity of 3760 S/cm among organic-modified systems while maintaining optical transparency and mechanical flexibility. Comprehensive spectroscopic characterization reveals that the conductivity enhancement originates from partial PSS removal and structural reorganization into quinoid-dominated configurations. When implemented in flexible supercapacitors, these modified electrodes deliver exceptional electrochemical performance, including an areal capacitance of 852 mF/cm2 85.2 F/cm3 volumetric) at 0.5 mA/cm2 with outstanding rate capability (71% retention at 10 mA/cm2) and cycling stability (90% capacity retention after 10 000 cycles). The devices achieve remarkable energy densities of 32.8 µWh/cm2 (areal) and 7.14 mWh/cm3 (volumetric), surpassing most reported PEDOT-based supercapacitors. This work establishes a material-friendly approach to engineer high-performance conductive polymers for next-generation flexible electronics.

The increasing demand for sustainable materials in flexible printed electronics has driven interest in biobased substrates as alternatives to petroleum-derived polymers such as polyethylene terephthalate (PET). This study evaluates a poly(lactic acid-co-oxacyclohexadecenlactone) (PLH) polymer as a biobased substrate for flexible electronics. Mechanically, PLH exhibited a lower Young's modulus (26 ± 3 MPa) than PET (110 ± 10 MPa), indicating greater flexibility, and demonstrates excellent elastic recovery, with no permanent deformation observed after a 0.5 mm extension (2 cm × 0.8 cm samples). Dielectric strength tests confirmed both substrates withstand voltages exceeding 5 kV, with no variation after curing at 60 °C. Thermogravimetric analysis (TGA) of the PLH reveals a temperature at 5% mass loss (T5%) of approximately 273 °C, indicating the onset of thermal degradation. The corresponding DTG curve shows a main degradation peak at around 295 °C, associated with the primary decomposition process, confirming sufficient thermal stability for printing processes using organic and aqueous inks. Contact angle measurements indicate moderate wettability (PLH: 80° ± 4; PET: 75° ± 5), while confocal microscopy reveals a slightly rougher topside for PLH compared to the smoother PET surface. Nevertheless, PLH supports reliable screen printing of conductive tracks (0.6 cm × 100 μm) with adhesion meeting ASTM D3359 standards. The functional performance of PLH, combined with its biobased origin and validated environmental assessment, highlights its potential as a substrate for flexible electronics with moderate thermal requirements. Future work should address bending durability and surface homogeneity to further enhance PLH applicability.

The increasing complexity and density of electronic systems have amplified the need for effective electromagnetic interference (EMI) shielding materials to ensure functionality, signal integrity, and compliance with electromagnetic compatibility regulations. Conventional shielding materials such as metals suffer from inherent drawbacks including high density, rigidity, corrosion susceptibility, and processing difficulties. In this review, conductive polymeric nanocomposites have been evaluated as a highly promising group of materials for use as EMI shielding due to their distinctive blending of lightweight nature, flexibility, processability, and tuneful electrical characteristics. These nanocomposites are typically composed of insulating matrix of polymer embedded with electrical-conductive nano-fillers like carbon nanotubes (CNTs), carbon black, graphene or metallic nanoparticles. The incorporation of these fillers enables the formation of percolative conductive networks inside the polymer matrix, which significantly enhances the electrical-conductivity and EMI shielding effectiveness (SE) of the obtained material. Shielding in these materials arises from reflection and absorption of incident electromagnetic waves, with multiple internal reflections contributing under conditions of relatively low absorption. The scalability and environmental resistance of conductive polymeric nanocomposites make them attractive for use in industries such as aerospace, automotive, consumer electronics, and healthcare, where multifunctional, lightweight, and durable materials are essential. In conclusion, conductive polymeric nanocomposites show a versatile and future-ready solution for EMI shielding, bridging the gap between performance, functionality, and manufacturability. Their adaptability and technological advancement make them central to next-generation electromagnetic shielding materials. Surface functionalization of fillers enhances compatibility with the polymer matrix, promoting uniform dispersion and improved interfacial polarization, contributing to absorption-dominant shielding.

In this paper, we propose a multivariate fast iterative filtering (MvFIF) decomposition algorithm, entropy-based features, and a nature-inspired feature selection approach for Alzheimer's disease (AD) detection using electroencephalogram (EEG) signals. Where, the MvFIF decomposes the multichannel EEG signals into multichannel intrinsic mode functions (MIMFs). The entropy features: dispersion entropy (DispEn) and distribution entropy (DistEn) are extracted from the MIMFs. Afterward, five nature-inspired feature selection algorithms are applied to reduce the feature space by selecting the relevant features for the AD. The selected features are finally used for the binary classification to distinguish AD patients from healthy control (HC) subjects using different classifiers. In addition, lobe-wise analysis is performed to understand the neural activity, diagnose AD, and guide targeted treatments. We show that the binary differential evolution optimization (BDEO) feature selection method with the support vector machine (SVM) classifier achieves the highest accuracy of 90.78% with standard deviation (SD) of 1.96% using 10-fold cross-validation (CV) and 75% with SD of 18.20% using leave-one-subject-out CV (LOSO-CV). In lobe-wise analysis, XGBoost classifier with temporal lobe gives the highest accuracy of 80.06% with SD of 1.53% using 10-fold CV and 70.78% with SD of 23.93% using LOSO-CV. The proposed approach surpasses the current leading techniques in AD detection utilizing EEG signals.

Controlling molecular spin states is essential for organic electronics, yet achieving high-spin (triplet or higher) ground states in Kekulé-type π-conjugated systems remains a challenge. Achieving high-spin ground states in π-conjugated systems has traditionally relied on two established strategies: implementing specific molecular topologies to create nondisjoint frontier orbitals, or taking advantage of high molecular symmetry (Cn, 3 ≤ n) to induce orbital degeneracy. Consequently, Kekulé-type molecules are overwhelmingly singlet species in their ground state. Here, we report the design and synthesis of o-BenD, a 16π-electron Kekulé-type diradical bridged by an ortho-phenylene unit. Despite lacking conventional structural prerequisites, o-BenD exhibits a robust triplet ground state with strong ferromagnetic coupling (J/kB = +320 K). This magnetism originates from pseudodegeneracy of the frontier orbitals, controlled by a simple "frontier-orbital engineering" approach guided by topological charge stabilization. This mechanism bypasses topology-based spin-state prediction, providing a new conceptual framework for stabilizing high-spin states. Furthermore, we demonstrated that o-BenD exhibits ground-state Baird aromaticity, the aromaticity of [4n]π-systems in the lowest triplet state, due to the unique combination of the ground triplet nature and 16π-Kekulé-type conjugation system. While Baird aromaticity is typically restricted to short-lived photoexcited states, this work provides a molecular design to favor the Baird aromatic state over the Hückel antiaromatic singlet state. The realization of such a previously unanticipated electronic state expands the accessible chemical space for the development of organic spintronics and quantum information technologies.

Lithium-ion batteries (LIBs) have become indispensable in present-day energy storage applications, containing portable electronics, electric vehicles, and renewable energy systems. However, rapid growth in LIBs usage has caused a parallel surge in end-of-life batteries, presenting environmental and resource recovery challenges. Among the various components, such as cathode, anode, electrolyte, separators, of LIBs, electrolyte has received minimal attention in recycling efforts. Electrolytes, characterized by their flammable, toxic, and volatile nature, pose significant environmental hazards, including the release of harmful gases and pollutants during disposal. This review focuses on the critical need for efficient recovery and reutilization of electrolytes from spent LIBs. Various recovery methods, including solvent extraction, supercritical fluid extraction, pyrolysis, and freezing, are studied for their effectiveness, efficiency, and environmental impact. Additionally, methods for recycling and regenerating recovered electrolytes into high-purity components for direct reuse are explored, addressing economic and sustainability considerations. Finally, major challenges and research gaps have been discussed. Key research gaps include the degradation of electrolytes during battery operation, complex composition of spent electrolytes, and economic feasibility of large-scale recovery technologies.

Self-assembled monolayers suffer from the insufficient electrical conductivity, stemming from their ultrathin nature and disordered molecular orientation. Here we report a π-skeleton unit of 3,6-dibenzothiophen-9H-carbazol to building a self-assembled multilayer (SAMUL) that exhibits superior carrier transport and outstanding resistance to external stimuli. The π-expanded skeleton effectively enhanced the molecular crystallinity and face-on orientation, which successfully activated a large π-electron conjugated network within SAMULs. This conjugated network structure greatly broadens the delocalization region of free radicals, which not only significantly enhances the electrical conductance and hole-transporting capability, but also reinforces the photochemical stability. Consequently, a record-high efficiency of 21.13% (certified as 20.77%) with a notable fill factor of 83.48% was achieved for binary organic solar cells. This work provides a new inspiration for the molecular skeleton design in organic electronics.

The development of catalytic reactions based on earth-abundant first-row transition metals that use chemicals from renewable feedstocks aligns with current principles of sustainable chemistry. Here, we employ oxovanadium-(IV) salen-type complexes ([VO-(Ln)], n = 1-5: 1-5) as catalysts for the selective epoxidation of biodiesel-derived methyl oleate, where Ln are tetradentate salen-type ligands with different diamine linkers, namely ethylenediamine (1), 1,3-diaminopropane (2 and 5), diaminomaleonitrile (3), and 1,2-diaminocyclohexane (4). The 1,3-diaminopropane system was examined with both unsubstituted (p-H) (2) and substituted (p-OMe) (5) salicylaldehyde-aromatic rings. DFT analysis revealed the influence of the ligand on the electronics of the VO moiety, with [VO-(L3)] exhibiting the most electron-deficient vanadium center. Notably, this complex also proved to be the most efficient epoxidation catalyst under optimized conditions (1 mol % catalyst, 3.5 equiv oxidant-TBHP, no added solvent). UV-Vis spectroscopy monitoring of the reaction between the complexes 1-5 with excess oxidant (pseudo-first order conditions) highlighted pronounced ligand-dependent differences in reactivity. Linear kinetics were observed only for [VO-(L2)] and [VO-(L5)], both containing a 1,3-diaminopropane bridge. In contrast, compounds with saturated two-carbon bridges [VO-(L1)] and [VO-(L4)] reacted slowly with the oxidant, displaying an induction period. Finally, [VO-(L3)] does not react with the oxidant under the same conditions, suggesting an alternative epoxidation mechanism via a V-(IV) center in the initial stage of catalysis. These results demonstrate that vanadium salen-type complexes, although structurally similar, enable epoxidation through either the commonly proposed V-(V)-peroxide pathway or a V-(IV) Lewis-acidic center, depending on the nature of the moiety bridging the two imine nitrogens.

Accurate, non-invasive grading of fish freshness remains challenging in dense and occluded environments, where the fish eye images serve as a reliable and consistently visible indicator of freshness. Loss of structural cues in the fish eye image often limits the existing deep learning models (DLN) from having consistent learning. To address these limitations, this work aims to develop a fusion-based complementary learning network (FCLN) that integrates expert-suggested photometric features with deep visual representations of DLN for the classification of fish freshness. Fish expert knowledge on freshness evaluation is processed using a large language model (LLM) to identify relevant visual cues and derive interpretable numeric features, eliminating the need for manual feature engineering. FCLN is constructed using ResNet50 as a backbone with a numeric feature layer and a fusion layer responsible for integrating both modalities for complementary learning. Complementary learning analysis (CLA) is also formulated to validate the complementary nature of the numeric features and image embeddings of DLNs. The proposed method is evaluated using the Fish Freshness Eye (FFE) dataset, which contains multiple fish species and freshness categories. Experimental results demonstrate a reliable classification accuracy of 81.03% during testing. CLA provides the least average correlation of 0.3755, indicating the complementary nature of the numeric features. Results of ablation studies also highlight the importance of using expert knowledge with deep visual features for fish freshness classification.

The interactions between ionic charge carriers and host framework critically govern electrochemical reactions and ion-storing performance, serving as a pivotal design consideration for energy storage devices. However, the fundamental understanding of the covalent-ionic interactions between anion and oxidized π+-framework remains limited thus far. Here we reveal the covalent-ionic nature of anion-π+ interactions between poly(arylamine)s (PAAs) and Cl- anions. Cl--π+ complexes bearing rigid polymeric aryl-substituted dihydrophenazine (PDPZ) exhibit both electrostatic interaction and distinct Cl-→π+ charge-transfer orbital contribution, confirming the underlying hybrid covalent-ionic nature of Cl--π+ interaction. The synergistic effect of Cl--π+ interaction and high electron delocalization capability of PDPZx+ framework enables reversible Cl- intercalation/deintercalation during multi-electron redox, achieving a high anion storage capacity of 236 mAh g-1 and remarkable energy densities of 175 Wh kg-1 (Zn||PDPZ cell, in 30 m ZnCl2), alongside long calendar life as cathodes for Cl--based dual-ion batteries (Cl-DIBs). Spectroscopic evidence reveals dynamic evolution of vibrational modes and electronic structures from PDPZ to PDPZx+·xCl- complexes, demonstrating the entire π+-framework participation and anion-to-π+ charge transfer during chloride storage. Our mechanistic insights into anion-π+ interactions in Cl-DIBs provide theoretical guidance for designing advanced anion-storage organic cathodes and advance anion coordination chemistry.

Highly sensitive, paper-based tactile sensors utilizing conductive nanomaterials have attracted significant attention due to their porosity, foldability, and mechanical flexibility. However, achieving sensitivity above 10 kPa-1 across a wide pressure range remains a key challenge. In this study, we present a flexible tactile sensor based on stacked mulberry paper coated with Ti3C2Tx MXene. Owing to the hydrophilic nature of mulberry paper, the MXene layers conformally coat the fibrous network via dip coating. The rough, porous surface and multi-layered architecture enhance contact resistance modulation, enabling high sensitivity (>15 kPa-1) over a broad pressure range (1-1000 kPa). We systematically investigate the effects of paper type, stacking configuration, and MXene loading to optimize sensor performance. The resulting device exhibits rapid response, durability over 1000 loading cycles, and consistent reproducibility. Its high flexibility and paper-fabric-based structure allow seamless integration into wearable platforms such as gloves and wristbands, enabling real-time monitoring of finger motion, arterial pulse, and touch intensity. Additionally, the wide detection range supports applications in Morse code signaling and CPR training feedback systems through wireless communication. These findings highlight MXene-coated mulberry paper as a scalable, durable, and cost-effective platform for wearable electronics requiring a balanced combination of high sensitivity and broad-range pressure detection.

Nature has long inspired the design of reversible smart adhesives; however, achieving both high adhesion strength and switchability remains a significant challenge, particularly for emerging applications (e.g., soft robotics and wearable electronics). Herein, we introduce switchable phase-locking-mediated adhesives (SPAs) that leverage high-density hydrogen bonds to deliver ultrahigh adhesion (up to 15 MPa). The underlying mechanism involves dynamic phase-locking mediation that harmonizes interfacial adhesion with bulk cohesion through interfacial mechanical locking and strain-induced phase separation. This strategy optimizes performance across the complete adhesion lifecycle, including spreading, adhering, and debonding via temperature/force-mediated phase-locking. Through detailed molecular analysis of the SPA system, we uncover the mechanistic basis of ultrahigh adhesion and establish design guidelines applicable to future smart adhesive development.

Latent fingerprint development on rough non-porous substrates using fingerprint powders remains challenging because surface microstructures reduce particle-adhesion selectivity and weaken the contrast between ridges and the background. In this study, Cs2NaBi0.6Er0.4Cl6 double-perovskite nanoparticles were prepared by a solvothermal method and investigated as fingerprint-development particles for latent fingerprints on frosted plastic substrates. Structural characterization by X-ray diffraction (XRD), scanning electron microscopy (SEM), Raman spectroscopy, and X-ray photoelectron spectroscopy (XPS) indicated that Er3+ was incorporated into the host matrix and that the product consisted of spherical nanoparticles with smooth surfaces, relatively uniform particle-size distribution, and good dispersibility. Comparative experiments involving 40 categories of latent fingerprint samples showed that the Cs2NaBi0.6Er0.4Cl6 nanoparticles outperformed conventional powders in developing fingerprints on frosted plastic substrates. Quantitative grayscale analysis using Image J 1.53K and Origin 2024 further showed that the development contrast, expressed as the D value, reached 51.21 for sebum-rich fingerprints and 35.87 for oil-contaminated model fingerprints, both of which were higher than those obtained with the other three powders. Because the fluorescence of Cs2NaBi0.6Er0.4Cl6 under UV excitation was weaker than that of the commercial red fluorescent powder, we attribute the improved development performance mainly to selective adhesion of the particles to fingerprint residues rather than to fluorescence intensity alone. In addition, the material maintained good performance for aged fingerprints within 10 days and for developed fingerprints stored for up to 8 days. These results suggest that selective residue-affinitive adhesion, possibly assisted by the hydrophilic or moisture-affinitive nature of the ionic double-perovskite particles, plays an important role in improving fingerprint development on rough non-porous substrates. This study provides a physical perspective for latent fingerprint development on rough non-porous substrates and broadens the forensic-science application of lead-free double-perovskite nanomaterials.

Introduction: Intrauterine growth restriction (IUGR) is a major cause of perinatal morbidity and mortality, often associated with placental insufficiency and progressive alterations in fetal autonomic regulation. Cardiotocography (CTG) represents one of the most widely used tools for fetal monitoring, yet its interpretation remains challenging due to high inter-observer variability and the subtle nature of early pathological patterns. Artificial intelligence approaches have recently shown promising potential for automated CTG analysis, but their development is often limited by the scarcity of large, annotated datasets. Methods: In this study we propose a multidimensional ensemble pipeline for the detection of IUGR from antepartum CTG recordings. The framework integrates two complementary predictive branches: a residual deep learning model (ResNet) operating directly on multivariate temporal sequences, and a hybrid CNN-MLP architecture combining image-based encodings of fetal heart rate signals with physiologically interpretable quantitative descriptors. The outputs of the two models are fused through a logistic regression meta-classifier using a stacking strategy. The pipeline was trained and evaluated using the NAPAMI database, a large clinically curated dataset comprising more than 70,000 CTG recordings collected over a period of 17 years. Results: Both base models (ResNet and CNN + MLP) achieved comparable performance levels. The proposed ensemble approach significantly improved the overall performance, reaching a balanced accuracy of 0.799 and an AUC of 0.868 (95% CI: 0.849-0.885). Statistical comparison using McNemar's test confirmed that the ensemble classifier significantly outperformed the individual models (p < 10-11). Discussion: The results demonstrate that combining complementary representations of fetal heart rate dynamics through an ensemble framework can improve the detection of IUGR from antepartum CTG recordings. The use of a large-scale clinical dataset together with physiologically informed and deep learning-based representations provides a promising direction for the development of AI-assisted decision support tools in prenatal medicine.