Rashba-type spin-orbit coupling in two-dimensional electron gases (2DEGs) is of great interest for efficient charge-to-spin conversion. Ultrathin LaTiO3 films on SrTiO3 (LTO/STO) give rise to a 2DEG and have previously been shown to exhibit giant Rashba spin-orbit coupling, making this a promising platform for demonstrating efficient charge-to-spin conversion. For this reason, we fabricated CoFeB/LaTiO3/SrTiO3 heterostructures to understand how spin-charge interconversion occurs as a function of temperature in these systems. The spin-to-charge conversion in LTO/STO can be understood in terms of Rashba spin-orbit coupling and thermoelectric contributions. Ferromagnetic resonance shows spin-current absorption by the LTO/STO interface, which is reduced as the LTO layer thickness increases. Through second-harmonic Hall measurements, we observe a damping-like torque efficiency of approximately 0.30 at 50 K, which increases sharply to approximately 8.6 at 10 K. Our results characterize the factors contributing to spin-charge conversion in these heterostructures, which is important for the incorporation of 2DEGs with strong Rashba-type spin-orbit coupling into spintronic systems.

Nanoscale transistors demand aggressive scaling of all channel dimensions—length, width and thickness. Two-dimensional semiconductors (2DS) provide the ultimate thickness limit, yet good device performance has largely remained restricted to micrometre-wide channels. Here we report monolayer 2DS nanoribbon transistors with both n- and p-type operation, fabricated by a top-down multipatterning process that includes ‘anchored’ contacts to limit nanoribbon delamination. This approach achieves channel lengths and widths down to 25–30 nm, with minimal edge degradation confirmed through nanoscale characterization, including tip-enhanced photoluminescence. Integrated with thin high- κ gate dielectrics, the devices deliver on-state currents up to 560, 420 and 130 µA µm −1 at a drain-to-source voltage of 1 V for n-type MoS 2 , n-type WS 2 and p-type WSe 2 , respectively. These results exceed prior single-gated 2DS nanoribbon reports, with WS 2 improving by more than two orders of magnitude, even for normally off (enhancement-mode) operation. Overall, these findings position top-down patterned 2DS nanoribbons as promising building blocks for future nanosheet transistor architectures.

Magnetoresistive random-access memory (MRAM) has become a leading candidate for next-generation non-volatile memory and energy-efficient computing, driven by the growing demand for low-power, scalable, and artificial intelligence (AI) capable hardware. Beyond conventional spin-transfer torque (STT) and three-terminal spin-orbit torque (SOT) MRAM, several emerging technologies have recently demonstrated promising advantages in switching speed, energy efficiency, and functional flexibility. In this review, we present a unified overview of six emerging MRAM technologies: hybrid STT-SOT MRAM, STT-assisted SOT (SAS) MRAM, two-terminal SOT-MRAM, voltage-controlled magnetic anisotropy (VCMA)-MRAM, voltage-gated spin-orbit torque (VG-SOT)-MRAM and voltage-controlled exchange coupling (VCEC)-MRAM. In the first part, we examine the distinct physical mechanisms underpinning these devices, along with recent experimental results. For each technology, we summarize key figures of merit, materials considerations, and scalability challenges. In the second part, we shift toward computing applications, highlighting how these devices expand the design space of non-von-Neumann architectures. SAS-MRAM supports compute-in-memory (CIM) primitives with low energy consumption and compact peripheral circuitry, enabling efficient vector-matrix multiplication and multiply-accumulate operations for online learning in edge-AI systems. In contrast, the voltage-tunable stochasticity and low-energy precessional dynamics of VCMA-MRAM naturally suit stochastic computing, enabling hardware acceleration of diffusion models, probabilistic inference, and reservoir computing. Overall, this review bridges device physics with system-level opportunities, outlining the potential of next-generation MRAM technologies to enable future edge-AI and non-von-Neumann computing platforms.

The development of the sixth generation of wireless communications technology requires terminals that can operate at frequencies above 100 GHz. For human-centric applications, these terminals should also be flexible and have low power. However, current flexible radio-frequency transistors typically have lower maximum frequencies, in part due to the poor thermal conductivity of flexible substrates. Here we report radio-frequency transistors that are based on aligned carbon nanotube arrays on flexible substrates, having current-gain cut-off frequencies (f(T)) and power-gain cut-off frequencies (f(max)) above 100 GHz. This is achieved by using electrothermal co-design to improve the heat dissipation and radio-frequency performance of the devices. The transistors exhibit an on-state current of 0.947 mA & micro;m(-1), a transconductance of 0.728 mS & micro;m(-1), a peak extrinsic f(T) of 152 GHz, a peak extrinsic f(max) of 102 GHz and a power consumption under 200 mW mm(-1). We also show that the devices can be used to create flexible radio-frequency amplifiers with an output power of 64 mW mm(-1) and an 11-dB power gain in the K band.

Kidney stones are highly prevalent and frequently necessitate surgical treatment, with ureteroscopic laser lithotripsy being the most common procedure. The removal of stone fragments after lithotripsy remains inefficient and often leads to incomplete clearance. We demonstrated the magnetization of fragments using a magnetic hydrogel followed by magnet capture. Herein, we translated the concept of magnetic removal into a medical device capable of in situ magnetic hydrogel delivery, stone magnetization, and capture with a specifically configured magnetic wire, all under direct ureteroscopic visualization. In porcine models, we demonstrated the feasibility and safety of the device and the visible complete removal of the magnetic hydrogel, without signs of toxic reactions or unexpected device-related injury. In 1-week survival studies, we found that hydrogel intentionally left in porcine kidneys is cleared through physiologic urination without adverse effects.

We present spin–orbit torque (SOT) field free magnetization switching and the detection of magnetization at room temperature using Pt and Sn alloys. Observations of the planar Hall effect and weak antilocalization provide evidence of topological features present in the PtSn4. The figures of merit of the spin-torque efficiency and spin-to-charge conversion (SCC) were estimated to be as large as 0.31 ± 0.02 and 0.47 ± 0.08, respectively, in Pt1Sn1. High SOT efficiency, large SCC signal, low magnetization switching current density, and industry compatibility for large-scale production have led to the application of PtxSn1−x alloys in magnetic memory and logic devices.

Magnetization switching driven by spin–orbit torque could be used to create an energy-efficient form of magnetic random-access memory. Tungsten is a promising heavy metal for such applications and can generate large spin–orbit torques when stabilized in its β-phase. However, the α-phase, which has a lower spin-Hall angle, is more thermodynamically stable. It is thus challenging to integrate metastable β-tungsten into complementary metal–oxide–semiconductor processes while maintaining phase stability under the back-end-of-line thermal constraints (400 °C for extended durations). Here we show that the insertion of thin layers of cobalt can be used to stabilize β-tungsten under back-end-of-line-compatible thermal conditions. Our composite β-tungsten layers can maintain their phase up to 400 °C for 10 h and can withstand 700 °C for 30 min. The film stacks exhibit a spin-Hall conductivity of around 4,500 Ω−1 cm−1, which we measure by means of spin-torque ferromagnetic resonance and harmonic Hall resistance measurements. Using the tungsten composite film stacks, we fabricate a 64-kb spin–orbit torque magnetic random-access memory that offers a spin–orbit torque switching of 1 ns, data retention of more than 10 years and a tunnelling magnetoresistance of 146

Amorphous oxide semiconductors are gaining interest for logic and memory transistors compatible with low-temperature fabrication. However, their low thermal conductivity and heterogeneous interfaces suggest that their performance may be severely limited by self-heating, especially at higher power and device densities. Here, we investigate the high-field breakdown of ultrathin (∼4 nm) amorphous indium tin oxide (ITO) transistors with scanning thermal microscopy (SThM) and multiphysics simulations. The ITO devices break irreversibly at channel temperatures of ∼180 and ∼340 °C on SiO2 and HfO2 substrates, respectively, with failure primarily caused by thermally-induced compressive strain near the device contacts. Combining SThM measurements with simulations allows us to estimate a thermal boundary conductance of 35 ± 12 MWm-2K-1 for ITO on SiO2 and 51 ± 14 MWm-2K-1 for ITO on HfO2. The latter also enables significantly higher breakdown power due to better heat dissipation and closer thermal expansion matching. These findings provide insights into the thermo-mechanical limitations of indium-based amorphous oxide transistors, which are important for more reliable and high-performance logic and memory applications.

With the rapid advancement of DNNs, numerous Process-in-Memory (PIM) architectures based on various memory technologies (NonVolatile (NVM)/Volatile Memory) have been developed to accelerate AI workloads. Magnetic Random Access Memory (MRAM) is highly promising among NVMs due to its zero standby leakage, fast write/read speeds, CMOS compatibility, and high memory density. However, existing MRAM technologies such as spin-transfer torque MRAM (STT-MRAM) and spin-orbit torque MRAM (SOT-MRAM), have inherent limitations. STT-MRAM faces high write current requirements, while SOT-MRAM introduces significant area overhead due to additional access transistors. The new STT-assistedSOT (SAS) MRAM provides an area-efficient alternative by sharing one write access transistor for multiple magnetic tunnel junctions (MTJs). This work presents the first fully digital processing-in-SASMRAM system to enable 8-bit floating-point (FP8) neural network inference with an application in on-device session-based recommender system. A SAS-MRAM device prototype is fabricated with 4 MTJs sharing the same SOT metal line. The proposed SAS-MRAMbased PIM macro is designed in TSMC 28nm technology. It achieves 15.31 TOPS/W energy efficiency and 269 GOPS performance for FP8 operations at 700 MHz. Compared to state-of-the-art recommender systems for the same popular YooChoose dataset, it demonstrates a 86x, 1.8x, and 1.12x higher energy efficiency than that of GPU, SRAM-PIM, and ReRAM-PIM, respectively.

Rapid, sensitive, multiplexed antibody detection technologies are essential for assessing vaccine efficacy and recipient immunity against viruses. As severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) variants continue to emerge and novel viruses may pose potential threats in the near future, personalized vaccines can be precisely tailored to an individual's measured immunity to help mitigate the symptoms of infection and viral spread. Here, we present giant magnetoresistive (GMR) biosensor microarrays for the quantitative multiplexed detection of antibodies against SARS-CoV-2 variants. The multiplexed GMR biosensor microarrays demonstrated high sensitivity, comparable to that of the single-plex enzyme-linked immunosorbent assay, enabling multiplexed measurements using blood obtained from finger pricks. Additionally, the GMR biosensor microarrays demonstrated compatibility with clinical samples and could detect variant-specific antibodies in serum, allowing for the assessment of current immunity status. Notably, an increased concentration of antibodies against the Omicron variant was clearly observed two weeks after receiving an Omicron-based booster vaccination. These results indicate that GMR biosensor microarrays offer a practical point-of-care tool for monitoring humoral immunity and guiding personalized vaccination schedules, thereby supporting immunity management against SARS-CoV-2 and future viral threats.

We demonstrate the unconventional and conventional spin-to-charge conversion (SCC) in MnPd3 using local SCC device at room temperature. The low crystal symmetry along the (114) growth direction allows unconventional SCC in MnPd3. The figure of merit of in-plane spin polarization in unconventional SCC is determined to be 4.70%. Unconventional and conventional SCCs are promising for magnetization detection in spintronics and magnetoelectric devices.

We carry out the first thermal simulations and comparison of channel-all-around (CAA) and gate-all-around (GAA) indium tin oxide (ITO) vertical channel transistors (VCTs), rooted in the experimental measurements. Despite potential electrical advantages and scalability, such CAA or GAA ITO device may face thermal challenges due to the low thermal conductivity of ITO and their structure, with ITO channel deeply buried in a vertically-etched trench. Our quantitative thermal analysis reveals that CAA architecture can achieve similar to 48% reduction in the maximum device temperature rise (Delta T-max) compared to the GAA counterpart. Moreover, by adopting molybdenum (Mo) as source/drain material, Delta T-max further decreases, resulting in similar to 56% total reduction of both Delta T-max and thermal resistance (R-th). This thermal analysis of advanced device architectures lays the groundwork for future vertically-stacked technologies, where effective thermal management is expected to become a key enabler of system scalability and stability.

We present the first comparative thermal analysis of gate-all-around (GAA) nanosheet (NS) transistors with 2D semiconductors vs. Si channels. Despite potential electrical advantages, replacing Si- with 2D-NS may introduce thermal challenges due to the low thermal boundary conductance (TBC) between the 2D material and its environment. Building on our thermal measurements of monolayer MoS2 and ultrathin HfO2, we use simulations to compare the thermal behavior of 2D- vs. Si-NS transistors at 6-12 nm gate lengths. In this regime, the thermal role of the contacts dominates, and 2D-NS transistors are at a disadvantage due to their small contact area. We find 2D wrap-around contacts (WAC) are thermally superior to edge contacts (EC), and an even better choice are metallized multilayer 2D contacts. These bring the thermal resistance of 2D-NS within 50% that of Si-NS transistors. This study provides key thermal insights into 2D-NS technology, where electro-thermal co-design is expected to be essential.

Transformers face scalability challenges due to the quadratic cost of attention, which involves dense similarity computations between queries and keys. We propose CAMformer, a novel accelerator that reinterprets attention as an associative memory operation and computes attention scores using a voltage-domain Binary Attention Content Addressable Memory (BA-CAM). This enables constant-time similarity search through analog charge sharing, replacing digital arithmetic with physical similarity sensing. CAMformer integrates hierarchical two-stage top-k filtering, pipelined execution, and high-precision contextualization to achieve both algorithmic accuracy and architectural efficiency. Evaluated on BERT and Vision Transformer workloads, CAMformer achieves over 10x energy efficiency, up to 4x higher throughput, and 6-8x lower area compared to state-of-the-art accelerators–while maintaining near-lossless accuracy.

Nanoscale biosensors for sensitive DNA detection require advanced and precise fabrication techniques, which make them highly expensive and result in low yield rates. For such DNA biosensors, sensor regeneration is highly desirable. In this study, we investigated the effectiveness of various denaturants, including ultrapure water, urea solution, tris-ethylenediaminetetraacetic acid buffer, and dimethyl sulfoxide (DMSO), for the denaturation of target DNAs hybridized to probe DNAs on sensors. We used giant magnetoresistive (GMR) biosensors equipped with a temperature control unit in conjunction with magnetic nanoparticles. To examine the effect of DNA sequence on denaturation efficiency, 14 orthogonal DNA pairs were designed and tested. Furthermore, to maintain a consistent sensitivity in subsequent measurements, we evaluated the integrity of the probe DNAs on the sensors after denaturation. Among all the denaturants tested, 40% DMSO demonstrated excellent performance in the denaturation of probe DNAs covalently bonded to the sensors, without any heating process. This optimal denaturant can be applied to other planar DNA biosensor systems; moreover, GMR biosensors can facilitate the evaluation of newly developed denaturants.

Energy-efficient computing is essential for addressing the rising power demands of modern data-intensive applications and ensuring sustainable technology advancement. Magnetoresistive random access memory (MRAM) has emerged as a pivotal technology in this domain, offering nonvolatile memory solutions that combine low power consumption with high performance. Spin-orbit torque (SOT) MRAM (SOT-MRAM) and its variants stand out for its potential to deliver SRAM-like performance at a higher bit-cell density. In this article, we present a novel high-density STT-assisted SOT-MRAM (SAS-MRAM) technology designed for energy-efficient artificial intelligence (AI) applications. SAS-MRAM capitalizes on the advantages of both spin-transfer torque (STT) and SOT mechanisms, utilizing a multi-bit-shared SOT line to achieve high-speed, high-density, and high-endurance memory performance. Our experimental results validate the potential of SAS-MRAM to address the limitations of current memory technologies. An AI application of ResNet-18 deployed in SAS-MRAM shows similar to 32.7x energy-delay-product (EDP) benefits compared to that in SRAM, presenting a promising solution for future AI hardware implementations, especially at edge where low-power training and inference of AI models are necessary.

BACKGROUND:As anticoagulants are widely used to treat patients with atrial fibrillation (AF) and other thrombotic conditions, it is necessary for physicians to determine whether the medication has been taken in emergencies. Among many anticoagulants, rivaroxaban has attracted attention due to its safety and convenience. Since rivaroxaban inhibits activated coagulation factor X (factor Xa), measuring factor Xa activity can determine the presence of rivaroxaban. RESULTS:We report a magnetic biosensing platform capable of measuring the activity of factor Xa using peptide substrates conjugated with magnetic nanoparticles (MNPs). Due to the size of factor Xa, a new method of solution-phase assays was proposed for magnetic biosensing platforms to address issues with immobilized peptides on the sensors. This method was validated with factor Xa and trypsin, both of which are serine proteases. In the solution-phase assays, samples with either the enzymes of interest or no enzyme were simultaneously measured, and the activity of the enzyme was estimated using the difference between the measurements. Unlike conventional optical methods, our platform was capable of detecting the activity of factor Xa at 2 μg mL-1 with a 30 min sample incubation. SIGNIFICANCE:The assay using giant magnetoresistive biosensors outperformed conventional optical techniques. This platform can facilitate the determination of the presence of rivaroxaban and assist physicians in deciding on appropriate treatments for patients.

Magnetic tunnel junctions (MTJs) with ultrathin MgO tunnel barriers are at the heart of magnetic random-access memory (MRAM) and exhibit potential for spin caloritronics applications due to the tunnel magneto-Seebeck effect. However, the high programming current in MRAM can cause substantial heating which degrades the endurance and reliability of MTJs. Here, we report the thermal characterization of ultrathin CoFeB/MgO multilayers with total thicknesses of 4.4, 8.8, 22, and 44 nm, and with varying MgO thicknesses (1.0, 1.3, and 1.6 nm). Through time-domain thermoreflectance (TDTR) measurements and thermal modeling, we extract the intrinsic (∼3.6 W m-1 K-1) and effective (∼0.85 W m-1 K-1) thermal conductivities of annealed 1.0 nm thick MgO at room temperature. Our study reveals the thermal properties of ultrathin MgO tunnel barriers, especially the role of thermal boundary resistance, and contributes to a more precise thermal analysis of MTJs to improve the design and reliability of MRAM technologies.

The magnetization switching driven by spin-orbit torque (SOT) has garnered significant interest due to its potential for realizing Spin-Orbit Torque Magnetic Random-Access Memory (SOT-MRAM). This design features distinctly separated read and write paths, promising enhanced device reliability and a more favorable window for minimizing read/write interference. Among many explored heavy metals which possess strong spin-orbit coupling, tungsten stands out as a particularly intriguing material, exhibiting substantial spin–orbit torques in thin films stabilized in the A15 (β-phase) structure. However, challenges arise from the low spin Hall angles (~ 0.01) observed in the energetically favorable α-phase tungsten. Integration of β-W with modern CMOS processes, particularly under the back-end-of-line (BEOL) thermal budget (400℃ 30 mins), remains problematic. In this study, we report a design strategy for achieving BEOL thermal budget in tungsten layers, focusing on β-tungsten (β-W) as a promising material for efficient spin-orbit torques (SOTs) with a recorded high spin Hall conductivity of approximately 4500 Ω-1cm-1 measured by spin-torque ferromagnetic resonance (ST-FMR) and Harmonic Hall resistance. Finally, we demonstrate 1 ns SOT switching with 146% tunneling magnetoresistance based on the proposed β-W film stack. This comprehensive investigation provides a manufacturable and CMOS comparable path for next-generation low-power MRAM and spintronics.

Due to the separate memory and computation units in traditional von Neumann architecture, massive data transfer dominates the overall computing system's power and latency, known as the "Memory-Wall" issue. Especially with ever-increasing deep learning-based AI model size and computing complexity, it becomes the bottleneck for state-of-the-art AI computing systems. To address this challenge, in-memory computing (IMC)-based Neural Network accelerators have been widely investigated to support AI computing within memory. However, most of those works focus only on inference. The on-device training and continual learning have not been well explored yet. In this work, for the first time, we introduce on-device continual learning with STT-assisted-SOT (SAS) magnetoresistive random-access memory (MRAM)-based IMC system. On the hardware side, we have fabricated a STT-assisted-SOT MRAM (SAS-MRAM) device prototype with 4 magnetic tunnel junctions (MTJs, each at 100 nm x50 nm) sharing a common heavy metal layer, achieving significantly improved memory writing and area efficiency compared to traditional SOT-MRAM. Next, we designed fully digital IMC circuits with our SAS-MRAM to support both neural network inference and on-device learning. To enable efficient on-device continual learning for new task data, we present an 8-bit integer (INT8)-based continual learning algorithm that utilizes our SAS-MRAM IMC-supported bit-serial digital in-memory convolution operations to train a small parallel reprogramming network (Rep-Net) while freezing the major backbone model. Extensive studies have been presented based on our fabricated SAS-MRAM device prototype, cross-layer device-circuit benchmarking and simulation, as well as the on-device continual learning system evaluation.