Neuromorphic devices, inspired by the human brain's efficiency and adaptability, hold great potential for artificial intelligence (AI) hardware to overcome the limitations of traditional von Neumann architecture. As a subclass, multimodal and multifunctional neuromorphic devices have recently gained a lot of attention due to their advantages in in‐sensor computing and sophisticated behaviors. In this review, recent advances in materials, device structures, and applications in this field are systematically presented. It includes optical, electrical, mechanical, and chemical sensing in multimodal neuromorphic device, which enable in‐sensor computing to minimize energy consumption and enhance real‐time decision‐making. The materials applied in this field such as phase‐change, 2D materials, and ferroelectrics are summarized for their roles in achieving synaptic plasticity, nonvolatile memory for multifunctional neuromorphic devices. Structural innovations, including reconfigurable, multi‐terminal, and 3D‐integrated designs, further optimize parallel processing and multifunctional integration. Besides, application scenarios of multimodal and multifunctional neuromorphic devices and their advantages for improving the efficiency of AI are reviewed. Finally, challenges in material stability and commercialization are discussed, it emphasizes the need for interdisciplinary efforts to bridge the gap. This review provides critical insights and future directions for developing brain‐inspired, energy‐efficient AI hardware.

Rapid and dense DNA functionalization of upconversion nanoparticles (UCNPs) remains a critical bottleneck, as conventional covalent and electrostatic methods suffer from low labeling efficiency and time-consuming processing. We report a microwave-assisted rapid dehydration method for efficient DNA immobilization on UCNPs within 3 min, drastically outperforming conventional strategies. This rapid dehydration method demonstrates broad applicability across diverse UCNP compositions, morphologies, and different DNA chain lengths. Molecular dynamics simulations confirm the underlying mechanism: strong noncovalent interactions between DNA phosphate groups/bases and lanthanide ions. Utilizing this approach, we constructed a FRET biosensor by integrating the DNA-functionalized UCNPs with Nb2CTx MXene nanosheets. This platform achieved an ultralow limit of detection of 8.5 pM for SARS-CoV-2 oligonucleotides, representing a 107.5-fold improvement. Importantly, the biosensor was successfully validated using real COVID-19 clinical samples, effectively distinguishing between negative and positive results. The presented DNA functionalization strategy holds immense potential for advancing the design and synthesis of diverse nucleic acid-functionalized nanomaterials for advanced therapeutic and diagnostic applications.

Artificial intelligence (AI)-assisted workflows have transformed materials discovery, enabling rapid exploration of chemical spaces of functional materials. Endowed with extraordinary optoelectronic properties, two-dimensional (2D) hybrid perovskites represent an exciting frontier, but current efforts to design 2D perovskites rely heavily on trial-and-error and expert intuition approaches, leaving most of the chemical space unexplored and compromising the design of hybrid materials with desired properties. Here, we introduce an inverse design workflow for Dion-Jacobson perovskites that is built on an invertible fingerprint representation for millions of conjugated diammonium organic spacers. By incorporating high-throughput density functional theory (DFT) calculations, interpretable machine learning, and synthesis feasibility screening, we identified promising organic spacer candidates with deterministic energy level alignment between the organic and the inorganic motifs in the 2D hybrid perovskites. These results highlight the power of integrating invertible, physically meaningful molecular representations into AI-assisted design, streamlining the property-targeted design of hybrid materials.

Multicomponent nanoparticles with tunable phase and composition are of great interest due to their unique electronic structures and size-dependent properties. Yet, a rapid, universal, and air-compatible route to phase-pure metal compounds remains challenging. Here, we develop a reaction microenvironment engineering-assisted pulsed laser shock (RME-PLS) strategy that expands laser-based synthesis from metals and alloys to a broad family of transition metal compound (TMC) nanoparticles, including oxides, phosphides, sulfides, and selenides. By regulating the local chemical microenvironment, this approach enables phase-selective synthesis in ambient air with high purity and short reaction time. Ligand-controlled PLS induces metal-to-oxide phase evolution, while precursor introduction of P/S/Se sources further extends the strategy to a universal RME platform. As a demonstration, carbon-coated TMC nanoparticles are used as electrocatalysts for water splitting. Notably, the laser-derived Fe2P@C reconstructs into α-FeOOH as the true active phase, which delivers exceptional catalytic performance and maintains stable operation at industrial current densities for more than 1000 h. In situ characterization reveals the structural evolution of the catalyst during water oxidation. In addition, the RME-PLS strategy enables tunable particle size control, highlighting its broad adaptability. This work establishes a general, scalable, and air-compatible platform for the rapid and phase-selective synthesis of transition metal compound nanoparticles.

Controlling crystallographic orientation in quasi-van der Waals (vdW) epitaxy remains a fundamental challenge, especially for material systems located near the boundary between weakly and strongly coupled growth regimes. In such marginal systems, epitaxial selection is governed by a delicate thermodynamic competition between surface-energy penalties and interfacial interaction gains, giving rise to two archetypal limits: vdW-dominated free-epitaxy and strong interfacial coupling dominated locked-epitaxy. However, dynamically driving transitions between these regimes has remained elusive. Here, we demonstrate that external light irradiation can deterministically induce such a transition. Using the thermodynamically frustrated Fe4N/mica interface as a model system, we show that photo-excited carriers act as a chemical potentiator, significantly enhancing the interfacial chemical affinity. Within a quantitative thermodynamic description, this optical modulation increases the locking criterion (I_lock)-defined as the ratio of interfacial energy gain to surface-energy cost-beyond its critical threshold. As a result, the system switches from vdW-dominated free-epitaxy with (001) orientation to chemically locked-epitaxy with (111) orientation. Our findings establish light as a non-invasive and switchable control knob to dynamically reconfigure the interfacial energy landscape in quasi-vdW epitaxy, enabling programmable access to distinct epitaxial states beyond intrinsic material limitations.

Controlling the epitaxial state in quasi-van der Waals heterostructures remains challenging because weak interfacial guidance often favors free epitaxy. Here, we show that light irradiation during sputtering drives Fe4N/mica from a free-epitaxy state to a locked-epitaxy state. Under dark growth, the film adopts the Fe4N(001) with a bulk-like lattice constant and rotationally degenerate in-plane alignment, characteristic of free-epitaxy. Under illumination, the system follows a locked-epitaxy pathway and forms Fe4N(111)/mica with a specific in-plane registry and finite in-plane strain. Timing-, wavelength-, and intensity-dependent experiments show that this transition is established during the earliest stage of growth. Density functional theory calculations further reveal that the Fe4N(001)/mica interface is mainly vdW-dominated, whereas the Fe4N(111)/mica interface is stabilized by a much stronger non-vdW contribution, dominated by chemically specific interaction. These results establish optical modulation of interfacial coupling as a practical route toward free-to-locked epitaxy in quasi-vdW heterostructures.

Spatial-light computing requires rewritable, multilevel and persistent optical weights, but many implementations rely on volatile modulators or static power. Here we report a CMOS-imaged luminescent memristor array in which ferroelectric domain switching in Er/Yb-doped PMN-PT single crystals programs non-volatile photoluminescence (PL) states. Domain reconfiguration tunes the local crystal-field symmetry of lanthanide emitters, enabling 16 analogue levels, retention over 27 h, endurance beyond 105 cycles and microsecond-scale state programming with zero electrical standby power for state retention. An 8 × 8 array, addressed by a diffractive optical element and read by a proximal CMOS sensor, converts stored emissive states into a single-frame intensity map and achieves 94.02% pixel-wise state recognition using calibration-aware decoding. The array implements single-step optical linear weighting, while the same decoded weights support hybrid optical-electrical handwritten-digit inference approaching a 32-bit floating-point software baseline. These results establish a non-volatile, image-addressable emissive weight element for photonic computing. Spatial-light computing is limited by the lack of compact, non-volatile, rewritable analog devices. Wen et al. report a ferroelectric crystal that converts lanthanide emission into a rewritable, multilevel memory, allowing stored optical states to be read and used as persistent weights for neuromorphic computing.

Parkinson's disease (PD) is a progressive neurodegenerative disorder with incompletely understood pathophysiology, necessitating advanced tools for long-term dynamic biomarker monitoring. Current techniques lack noninvasive platforms capable of real-time multiparametric monitoring with high stability, sensitivity, and responsiveness. Here, we engineered core-shell-shell lanthanide-doped upconversion nanoparticles (cssUCNPs) through controlled Tm3+/Er3+ gradient doping and crystal field symmetry modulation. This "sandwich" heterostructure design suppressed cross-relaxation, while Ca2+ doping enhanced radiative transition efficiency. These modifications yielded a quantum yield improvement from 0.35 +/- 0.014% to 3.27 +/- 0.067%, representing a 9.3-fold enhancement, alongside a fluorescence intensity increase of 425.5-fold. The cssUCNPs exhibited exceptional photostability (>10,000 excitation cycles) and biocompatibility (>95% cell viability). Upon integration of Fluo4 with UCNPs, the system harnessed the luminescence resonance energy transfer effect to enable dual-responsive sensing of Ca2+ (detection limit: 1 mu M) and temperature (sensitivity: 0.5 K-1), thereby facilitating long-term tracing of Ca2+ and temperature dynamics across PD pathological stages. Our results revealed that early-phase PD is characterized by rapid Ca2+ and temperature surges, followed by oscillatory fluctuations in the middle phase, and eventual stabilization at elevated levels in the terminal phase. These dynamic profiles establish a direct link between mitochondrial dysfunction and Ca2+ dysregulation, highlighting Ca2+ flow-sensitive and thermally sensitive therapeutic targets for PD intervention.

ABSTRACT MXenes, a family of two‐dimensional transition metal carbides and nitrides, have garnered significant attention in multifunctional sensing electronics due to their unique properties and versatile functionalization strategies. Despite rapid advancements in MXene‐based sensing materials, a comprehensive understanding of the role of multi‐scale interface engineering in designing MXene‐based multifunctional sensory systems with diverse sensing functionalities remains lacking. In this review, the recent advances of multi‐scale interface engineering technologies applied to MXene‐based multifunctional sensors and the advanced signal processing technologies for building intelligent sensory systems are reviewed. Specifically, the multi‐scale interface engineering strategies that endow MXene‐based multifunctional sensors with multiple sensing capabilities, superior sensing performance, and system compatibility are systematically analyzed. We further highlight the latest progress of data processing technologies for achieving seamless integration of perception, storage, and processing in a single sensory system. Finally, the opportunities and challenges faced by MXene‐based multifunctional sensory systems are discussed. This review will provide an in‐depth perspective on interface‐engineered MXenes for advanced multifunctional sensors and outlines prospects for next‐generation MXene‐based multifunctional sensory systems.

Hydrogels, with their hydrophilicity, flexibility, and environmental friendliness, are highly desirable for moisture-electric generators (MEGs) that harness ubiquitous moisture to generate electrical energy. As the active material layer in MEGs, hydrogels play a crucial role in absorbing atmospheric moisture and converting chemical potential energy into electricity. However, the relatively low output current of the device and the instability of hydrogels pose challenges to the development of high-performance hydrogel-based MEGs. Herein, we introduce a straightforward, feasible, cost-effective, and versatile two-step solvent displacement strategy to overcome the barrier associated with the development of MEGs. Through tunable solvent interactions of glycerol and water, the moisture absorption capability and stability of the hydrogel can be improved, while promoting favorable ion migration. Such an effective processing route not only significantly boosts the output performances but also greatly improves the long-term durability of hydrogel-based MEGs. Notably, the current output and power density of the treated MEGs can increase by up to two orders of magnitude. The mechanisms behind the intriguing observation are investigated by various characterizations and theoretical calculations. This universal strategy holds promise to be extended to various hydrogel-based MEGs. Moreover, the MEGs can be used for energy harvesting, self-powered respiratory monitoring, and non-contact humidity detection. This work offers new opportunities for advancing green energy and self-powered technologies.

Traditional deep-brain stimulation via implanted electrodes can effectively treat neurological disorders, but surgical injury limits its clinical application. Here, we developed ultrasound-responsive piezoelectric nanoparticles for minimal-invasive and wireless neuromodulation. In the 6-OHDA-induced Parkinson's disease (PD) mouse model, these nanoparticles are injected into the subthalamic nucleus (STN) of the mouse brain. After ultrasound stimulation for several days, the motor behavior, particularly gait abnormalities and nonmotor symptoms such as pain and anxiety, in PD mice is alleviated without detectable toxicity. The piezoelectric nanoparticles can activate the STN area of the mouse after ultrasound stimulation. Our results demonstrate that piezoelectric-mediated neuromodulation of the STN reverses motor deficits in PD by modulating neural signals, thereby protecting dopaminergic neurons and enhancing levels of the neurotransmitter dopamine. This process can rescue and mitigate mitochondrial dysfunction and neuroinflammation in the nigrostriatal pathway. Our approach enables STN neuronal activation with minimal invasiveness, offering a promising strategy for treating neurodegenerative diseases.

Abstract Biological visual systems with polarization sensitivity enable perception in complex environments beyond the capability of human vision. The realization of polarization‐sensitive visual devices with such integrated spatial and temporal perception and nonvolatile modulation remains challenging. Here, drawing inspiration from ocular function of mantis shrimp, we combine intrinsic anisotropy of palladium diselenide (PdSe 2 ) and interfacial ferroelectric field from Poly(vinylidene fluoride‐trifluoroethylene) (P(VDF‐TrFE)) to drive polarized and nonvolatile synaptic weight modulation. The proposed architecture enables high‐resolution imaging with a wide grayscale range based on underwater environment by spatially distributed polarization illumination. Polarization‐resolved imaging yields graded recognition accuracies ranging from 66.2% to 93.9% without multiframe collection or off‐chip processing, while polarization‐dependent kernels facilitate fuzzy image sharpening and feature extraction. We further demonstrate vehicular temporal‐evolved direction identification within reduced visibility and disruptive interference conditions. Owing to the specific polarized illumination and distinguishable temporal encoding, the visual system realizes high‐accuracy direction recognition of 96.3% based on in‐sensor reservoir computing (RC) and mitigates contrast loss from the foggy weather. This work advances polarization‐enhanced machine vision and provides a viable pathway toward multidimensional perceptual processing through optoelectronic systems. image

The International Roadmap for Devices and Systems (IRDS) has identified the tunnel field-effect transistor (TFET) as the most promising next-generation logic device that enables sustainable downscaling in driving voltage and power consumption. Demonstrating an acceptable sub-Boltzmann-limit ON current (namely I60, the current level when a TFET switches to a subthreshold swing level of 60 millivolts per decade) and current-switching ratio has presented a formidable challenge. We report a TFET based on a bismuth/indium selenide (Bi/InSe) heterostructure that exhibits an I60 of up to ~10 microamperes per micrometer and a current-switching ratio of >107. We attribute such promising TFETs to precise material design, clean interfaces fabricated under vacuum, and band engineering based on subthreshold swing physics. Our results demonstrate a high-performance basic building block that meets the IRDS requirements for next-generation integrated circuits.

For decades, the integration of power handling and nonvolatile memory has been fundamentally impeded by the incompatibility between wide-bandgap semiconductors and ferroelectric materials. We resolve this challenge by demonstrating robust room-temperature ferroelectricity in epitaxial metastable κ-Ga2O3, grown via industry-compatible metal-organic chemical vapor deposition, creating an intrinsically ferroelectric wide-bandgap semiconductor. Through systematic characterization including piezoresponse force microscopy, polarization hysteresis measurements, and positive up-negative down tests, we provide conclusive evidence of stable ferroelectric switching down to 5-nanometer thickness-exceeding conventional ferroelectric limits-via a unique octahedral-tetrahedral transformation. Ferroelectric tunnel junctions achieve giant tunneling electroresistance exceeding 105. This fundamental discovery in a mainstream semiconductor challenges conventional materials paradigms and enables monolithic integration of power and memory functionalities on a unified platform.

Deep brain stimulation (DBS) is an established therapeutic approach for treating various neurological disorders, including Parkinson's disease, epilepsy, etc. Traditional DBS systems rely on implanted batteries, which pose challenges such as limited lifespan and the need for replacement surgeries. Transducer materials have provided new opportunities for developing DBS technology in recent years. These materials can convert remotely delivered energy forms, such as light, ultrasound, or magnetic fields, into electrical, thermal, light, or mechanical energy that can interface with neural signals. By injecting these materials into effective DBS targets of neurological disease and applying remote stimulation, they can generate signals such as electric, heat, or light that can interface with neurons, thus effectively regulating neural signal disturbances in the disease and treating disorders related to motor or emotional. This review offers insights into developing a class of materials to advance DBS technology for related neurological disorders. It provides a promising approach to replacing conventional electrodes and inducing neural stimulation in a noninvasive way. Future research should focus on optimizing material performance, ensuring biocompatibility, accurately modulating neural signals, and conducting clinical trials to advance this innovative field.

The conventional Lewis acid molten salt etching approach for synthesizing MXenes typically necessitates elevated temperatures exceeding 550 °C, in addition to the use of inert gas protection to prevent oxidation. Additionally, delamination of molten salt‐etched MXenes typically requires hazardous intercalating agents. Herein, a scalable and non‐toxic low‐temperature shielded salt (LSS) approach for synthesizing MXene in air is reported, with the use of only a small portion of salts and a low reaction temperature down to 150 °C. Especially, the synergistic effect of the increased diffusion rate by Li + ions and the phase change of magnesium chloride hexahydrate (MgCl 2 ·6H 2 O) enables a redox‐controlled A‐site etching of the parent MAX phase, which even facilitates the synthesis of hard‐to‐etch MXenes. As a proof‐of‐concept demonstration, several theoretically hard‐to‐synthesize MXenes including Cr 2 CT x and Nb 2 CT x are successfully prepared through the LSS technique, where Cr 2 CT x is not achieved by Lewis acidic molten salt yet. Compared to conventional techniques, this low‐temperature shielded salt etching method exhibits unconventional molten behavior while offering several advantages, including a non‐oxidizing environment, shorter processing time, and elimination of highly corrosive washing agents and organic intercalants. These advances render the LSS approach a promising route for synthesizing MXenes with applications toward diverse practical applications.

In-sensor computing paradigm holds the promise of realizing rapid and low-power signal processing. Constructing crossmodal in-sensor computing systems to emulate human sensory and recognition capabilities has been a persistent pursuit for developing humanoid robotics. Here, an artificial mechano-optical synapse is reported to implement in-sensor dynamic computing with visual-tactile perception. By employing mechanoluminescence (ML) material, direct conversion of the mechanical signals into light emission is achieved and the light is transported to an adjacent photostimulated luminescence (PSL) layer without pre- and post-irradiation. The PSL layer acts as a photon reservoir as well as a processing unit for achieving in-memory computing. The approach based on ML coupled with PSL material is different from traditional circuit-constrained methods, enabling remote operation and easy accessibility. Individual and synergistic plasticity are elaborately investigated under force and light pulses, including paired-pulse facilitation, learning behavior, and short-term and long-term memory. A multisensory neural network is built for processing the obtained handwritten patterns with a tablet consisting of the device, achieving a recognition accuracy of up to 92.5%. Moreover, material identification has been explored based on visual-tactile sensing, with an accuracy rate of 98.6%. This work provides a promising strategy to construct in-sensor computing systems with crossmodal integration and recognition.

Two-dimensional (2D) nanomaterials hold immense application potentials such as in high-performance nano-electronics, and asymmetric 2D structures with inherent electric dipoles will extend the application promises. Yet synthesizing asymmetric 2D structures remains challenging. Herein, we report the first synthesis of single-layer (SL) hexagonal (H-) phase polar Janus MoSeN via nitrogen-plasma-assisted molecular beam epitaxy. This is a significant achievement given the incommensurate valence between Mo, Se, and N, and the inherent strain from the Janus architecture. Using an array of compositional and structural characterization methods, we establish the atomic configurations of the synthesized MoSeN SL, confirming that they are 2D Janus transition-metal chalcogen-nitrides rather than alloys. By employing density functional theory calculations and transport measurements, we explore the structural feasibility and offer insights into its electronic properties, demonstrating its metallic behavior with ohmic contact characteristics. Piezoresponse force microscopy measurements reveal vertical piezoelectricity and ferroelectric potentials from the Janus MoSeN SL. Therefore, it exhibits great potential for applications in, e.g. piezoelectric and ferroelectric devices, sensing technologies, and optoelectronic devices. This work not only addresses existing challenges in 2D nanomaterial research but also opens new avenues for the development of advanced functional materials.

The application of zero-emission passive radiative coolers is a crucial step toward global carbon neutrality. However, a single radiative cooling function cannot meet the thermal requirements under various weather conditions. We present a dual-mode thermal management film that integrates passive radiative cooling and heating functions through its porous polymer surface for cooling and a light-to-heat conversion surface enabled by graphene and carbon nanotubes for heating. The surfaces of the dual-mode film were physically flipped, positioning the corresponding surface toward solar radiation to obtain the desired functionality. In the cooling surface, the film achieves sub-ambient cooling of approximate to 13.3 degrees C under 853.88 W m-2 of sunlight, thanks to its high solar reflectance (0.92) and mid-infrared emissivity (0.95). In the heating surface, it uses high solar absorption (0.90) to increase the temperature by 11.4 degrees C and generates Joule heating at various voltage levels. According to EnergyPlus software estimates, buildings with roofs covered in the film could reduce CO2 emissions by 1.109 billion metric tons, equivalent to 3% of current global CO2 emissions. This study offers a promising solution to climate challenges and holds great potential for energy savings and carbon reduction.