Xinman Chen

Spin–orbit torque (SOT) technology stands at the forefront of next‐generation non‐volatile memory and logic computing, promising superior endurance and ultrafast dynamics compared to conventional spin‐transfer torque (STT) architectures. However, the persistent requirement of an external magnetic field for deterministic switching of perpendicular magnetization severely impedes high‐density integration and commercial viability. This review systematically evaluates the evolutionary trajectory of symmetry‐breaking strategies developed to overcome this bottleneck. The analysis progresses from early macroscopic structural engineering, such as wedged geometries and lateral asymmetries, to mesoscopic built‐in effective fields driven by interfacial exchange coupling and chiral spin textures. Crucially, we identify a defining paradigm shift in the field: the transition from complex geometric modifications toward intrinsic material‐level innovations, including low‐symmetry crystals, vertical composition gradients (VCG), and synergistic dual‐torque (STT‐SOT) architectures. By resolving the trade‐offs among switching efficiency, thermal stability, and structural complexity, these intrinsic and hybrid approaches offer a practical roadmap for integrating scalable, all‐electrical SOT‐MRAM with standard CMOS back‐end‐of‐line (BEOL) processes. Finally, we establish a dual‐track strategic roadmap: while long‐term breakthroughs require parallel exploration of new exotic physics, we underscore the importance of focusing on mass production and foundry scalability at the current stage.

Energy-level-modulated organic memristive synapses integrate electrical and optical modulation, exhibiting low power consumption, multimodal operation, and strong synaptic plasticity for neuromorphic learning.

The rapid growth of edge artificial intelligence (AI) and the escalating computational demands of multimodal perception systems accentuate the bottlenecks of von Neumann architecture and the high costs of training. Therefore, to streamline system construction, reduce energy consumption, and improve the processing efficiency of artificial perception systems, the advancement of emerging device technologies and next-generation computing architecture is urgently required. Here, we present a brain-inspired reservoir computing (RC) framework employing dynamic memristors with the Ag/WOx/ITO structure, enabling efficient audio-visual emotion recognition. The developed WOx-based memristor demonstrates key biological synaptic functionalities, including short-term plasticity, paired-pulse facilitation/depression, and nonlinear temporal dynamics, enabling its use as an effective physical reservoir. By exploiting the device's transient response to electrical pulses and its nonlinear mapping capability, complex spatiotemporal signals from auditory and visual modalities are transformed into separable high-dimensional reservoir states through simple 4-bit pulse sequences. The implemented dual-modal RC system achieves remarkable classification accuracies of 95.46% for speech-based emotion recognition and 91.08% for facial expression recognition, with an energy cost of only ∼3.9 nJ per pulse operation. This work provides a compelling hardware paradigm for low-power, real-time audio-visual perception on edge devices, paving the way for advanced neuromorphic computing.

In an underwater wireless sensor network (UWSN), autonomous underwater vehicle (AUV) path planning faces challenges of high energy consumption, path instability, and limited adaptability in dynamic 3-D underwater environments. To overcome these limitations, a hybrid deep Q-network and ant colony optimization (D-ACO) algorithm is proposed, integrating the adaptive learning of deep Q-network (DQN) with the global search ability of ant colony optimization (ACO) to enhance stability and multiobjective optimization. In particular, a multiobjective reward mechanism is designed to jointly optimize energy consumption, task timeliness, and data acquisition efficiency, enabling the AUV to achieve effective obstacle avoidance while minimizing redundant movements. To validate its effectiveness, the D-ACO, ACO, DQN, PSO, GWO, and artificial potential field (APF) algorithms are applied in a 3-D grid environment for comparative analysis, focusing on coverage, task execution time, energy consumption, and path stability. Simulation results demonstrate that the D-ACO algorithm exhibits superior performance in balancing multiple objectives, achieving better convergence. This study delivers an efficient solution for underwater data acquisition and provides new insights into applying deep reinforcement learning (DRL) to autonomous system path planning.

Altermagnets unify the merits of ferromagnets (strong magneto responses) and antiferromagnets (zero stray fields), demonstrating time-reversal symmetry breaking, intrinsic demagnetization effects, and terahertz-range spin dynamics. These unique properties have attracted significant attention for spintronic applications requiring both robust spin control and minimal magnetic interference. Recently, KV2Se2O was identified as a metallic room-temperature altermagnetic material, but the origin of its high magnetic transition temperature remains elusive. In this work, it was found that both antiferromagnetic and ferromagnetic exchange coupling jointly play a crucial role in achieving the high magnetic transition temperature. Furthermore, K atoms enhance the magnetic moments of V atoms through charge transfer, thereby enhancing magnetic exchange energy and elevating the magnetic transition temperature. Additionally, strain effects can modulate the magnetic properties and the magnetic transition temperature, highlighting their potential for tailoring altermagnetism. These results not only motivate further exploration of room-temperature altermagnets but also pave the way for optimizing KV2Se2O in future magnetic and spintronic nanodevices.

Perovskite memristors have attracted considerable attention for their potential in emulating artificial synapses. However, the widespread use of the toxic lead-based perovskites poses significant challenges to futural application. In this work, we developed a lead-free memristor with an Ag/PMMA/Cs3Cu2I5/ITO architecture, in which the Cs3Cu2I5 halide perovskites functional layer was fabricated by physical vapor deposition. The memristor demonstrated a data retention time of 104 s and stable resistive switching behavior over 100 cycles under electrical pulses stimulation. Furthermore, it emulated a range of biologically relevant synaptic functions, including paired-pulse facilitation, short-term plasticity, long-term plasticity, spike-amplitude-dependent plasticity, spike-number-dependent plasticity, and spike-duration-dependent plasticity. Capitalizing on its nonlinear dynamics and short-term memory characteristics, the device was further integrated into a reservoir computing (RC) system. The RC system demonstrated strong recognition performance and robustness when tested with distorted digital image datasets. These results suggest that the Cs3Cu2I5-based memristor provides a promising, environmentally friendly platform for next-generation artificial intelligence hardware.

Fe3GaTe2 has been recently identified as a potential van der Waals (vdW) ferromagnetic material for spintronic devices, owing to long-range ferromagnetic order, strong perpendicular magnetic anisotropy, and high Curie temperature (TC) above room temperature. The band structure and electronic transitions are crucial for understanding magnetic properties of Fe3GaTe2, requiring a comprehensive investigation of the electronic behavior of Fe3GaTe2 under external magnetic fields. In this study, magnetic circular dichroism (MCD) spectroscopy was employed to examine the electronic transitions in Fe3GaTe2 at room temperature. Three distinct MCD peaks are clearly observed under the applied magnetic fields, which could correspond to three electronic transitions determined by first-principles density functional theory calculations of the band structure of Fe3GaTe2. Furthermore, the three transition bands would be correlated with the Fe d orbitals, as supported by the calculated orbital-resolved band structure of Fe3GaTe2. These findings offer insights into the electronic transitions and the underlying electronic structure in Fe3GaTe2, providing a basis for further fundamental research and potential applications in spintronic devices.

In this work, we report an optoelectronic synapse based on the Bi2O2Se/Cs3Cu2I5 heterojunction formed by consequently depositing Bi2O2Se and Cs3Cu2I5 thin films by chemical vapor deposition methods. The fabricated Au/Bi2O2Se/Cs3Cu2I5/Au devices exhibit the light-tunable photoresponse under visible light illumination. Synaptic functions including paired-pulse facilitation, the transition from short-term plasticity to long-term plasticity, associative learning and optical encoding, are conveniently reproduced on one single synaptic device by utilizing the light-tunable evolution of photocurrent. Furthermore, the front-end image preprocessing to enhance the accuracy and efficiency of image recognition was demonstrated based on the constructed device array (5 × 7). This indicates the potential application of Bi2O2Se/Cs3Cu2I5 optoelectronic synapses for the forthcoming neuromorphic computing.

In this work, we report the bidirectional photoresponse of the carbon-based graphdiyne/graphene (GDY/Gr) heterojunction photodetector toward the optical logic gates. The uniform GDY film was facilely grown on the surface of Gr serving as the template. Under mV-level bias, the resultant Au/GDY/Gr/Au devices manifest the bidirectional photoresponse to a broad wavelength range with rapid response time, namely, positive and negative photoresponse under negative and positive bias, respectively. It can be understood from the viewpoint of the divergence and coincidence between in-plane and external potential. Furthermore, the optically operated logical gates including logic AND, OR, and Exclusive-OR gates have been realized by integrating the fabricated Au/GDY/Gr/Au devices, converting optical signals to electrical signals in real-time. It suggests the potential application of GDY/Gr heterojunction toward visible-NIR detector as well as the stable optical logic units.