Spin-torque nano-oscillators (STNOs) inherently exhibit thermally driven phase fluctuations that render their dynamics truly stochastic. Here, we demonstrate that, despite this intrinsic randomness, the probability of occupying each phase state can be deterministically and continuously programmed. We experimentally investigate a vortex-based STNO operating under second-harmonic injection-locking, where the oscillator phase settles into two degenerate attractors separated by π and undergoes thermally activated phase jumps. By applying a weak radio-frequency perturbation at the free-running frequency, we tune the phase-jump rates between the two attractors without suppressing the fluctuations, achieving continuous probability control from the unbiased limit to values approaching 0 or 1. The bias phase selects which attractor is favored while the bias amplitude sets the strength of the imbalance, providing two complementary control knobs within a single nanoscale device. A phase-reduced description based on an effective quasipotential quantitatively accounts for the observations. These results establish injection-locked STNOs as programmable stochastic elements and provide a hardware primitive for probabilistic computing, Ising machines, and brain-inspired computing architectures.

Magnetic tunnel junction (MTJ)-based magnetic random-access memory (MRAM) is a promising platform for neuromorphic and in-memory computing owing to its non-volatility, high endurance, fast switching dynamics and CMOS compatibility. However, conventional spin-transfer torque and spin-orbit torque MRAM implementations for neural networks often suffer from high critical switching currents, large latency, thermal instability and significant read-write overheads. Here, we demonstrate a unified multistate MRAM-spin-torque nano-oscillator (STNO) architecture that integrates synapses and neurons on a single chip for convolutional neural network (CNN) applications. The system employs 1x8 multistate MRAM arrays as programmable synapses coupled with a vortex-based STNO neuron, enabling both individual and collective programming through fieldline-driven write channels. Multiple configurable resistance states are achieved by tuning internal and external magnetic fields together with bias currents, allowing quantized positive and negative synaptic weights for configurable kernel and pooling operations. The proposed architecture is evaluated through simulation on MNIST, SVHN, CIFAR-10, Google Speech Commands (GSC) and RadioML datasets, achieving accuracy of 99.76

Implantable and wearable devices require antennas that are both miniaturized and efficient, yet conventional designs are constrained by narrow bandwidth and orientation sensitivity. We report overtone ultrawideband magnetoelectric (OUWB-ME) antennas that exploit higher-order acoustic modes in polished silicon substrates to achieve a 22.6-gigahertz -10-decibel bandwidth and overtone capability in the 3- to 4-gigahertz range. Packaged into "μBots," these magnetoelectric heterostructures bonded with silver nanoparticle inks maintain stable operation under biological loading. In vitro assays confirm the biocompatibility of aluminum nitride and the protective role of parylene encapsulation for iron-gallium. Ex vivo rat and human tissues reshape transmission and reflection spectra, with reproducible frequency windows near 3.3 and 3.9 gigahertz. μBots enable real-time audiovisual telemetry using software-defined radios and exhibit compatibility with 7-tesla magnetic resonance imaging. By combining wideband response, robustness to misalignment, and biocompatible packaging, OUWB-ME μBots provide a scalable platform for wireless bio-integrated communication and telemetry.

This paper presents novel GPU implementation strategies that effectively exploit the available parallelism, based on the "more work per thread" approach, for three machine learning design exploration tasks: Multiple K-means evaluation, dimensionality reduction through parallel K-means encoding for XGBoost trees, and XGBoost tree pruning optimization. Our experimental results demonstrate significant performance improvements across all implementations. For multiple K-means evaluations, we achieve substantial speedups of 20 to 70 compared to Nvidia's cuML library. The dimensionality reduction approach, which uses parallel K-means encoding, achieves encoding reductions of up to two orders of magnitude while preserving classification accuracy within 1%-2% of the original performance. Additionally, we propose a Gini coefficient-based design exploration optimization that greatly reduces the number of models to be trained during the design space optimal search, achieving a 400 speedup. Furthermore, a parallel post-pruning evaluation framework for XGBoost demonstrates the ability to remove up to 50% of tree nodes without significant loss of accuracy.

This work presents the design, fabrication, and experimental characterization of a hybrid spintronic MEMS device for large-bandwidth nanoscale displacement sensing. The developed system integrates a micromagnet on a movable mass with Magnetic Tunnel Junctions (MTJs) in the fixed frame that detect variations in the magnetic field induced by motions on the mass. The system achieves a sensitivity of $0.606 \% / \mu \mathrm{m}$ over a $5 \mu \mathrm{m}$ linear range, and dynamic measurements show that sub-micron displacements can be detected up to hundreds of kilohertz, confirming compatibility with high-frequency MEMS/NEMS operation. These results show that spintronic transduction can be effectively integrated into microscale and nanoscale mechanical devices, offering a compact alternative to conventional readout approaches.

The human brain achieves exceptional energy efficiency by co-locating memory and processing, yet reproducing this principle in hardware remains challenging because many neuromorphic devices require standby power, offer limited programmability, or separate state storage from nonlinear computation. Here we demonstrate a multifunctional spintronic platform based on storage-layer-enabled vortex magnetic tunnel junctions (MTJs) that unifies non-volatile weight storage, optoelectrically driven nonlinear computation, and multilevel readout within a single nanopillar. A thermally programmable FM/AFM storage layer retains analog synaptic weights with zero standby power and enables non-volatile tuning of the vortex gyrotropic resonance over ∼15 MHz. Under optoelectrical operation, combined laser heating and dc bias drive the junction into the bias-enhanced tunnel magneto-Seebeck (bTMS) regime, where the thermoelectric response exhibits a pronounced cubic nonlinearity providing a compact, hardware-native transfer function for weighted analog computation. The electrical and thermoelectric channels switch at matched coercive fields but with distinct amplitudes, yielding an effective four-level readout space. Crossbar-array simulations parameterized by measured device response maps evaluate two neuromorphic modes – a bTMS mode (optical input, dc-bias weights) and a spin-diode mode (RF-frequency input, RF-power weights) – achieving image-classification accuracies of 95.4% and 94.9%, comparable to a digital single-layer network with sigmoid activations. Smaller 600 nm devices consistently outperform larger ones, identifying nonlinear-response engineering as a key device-level lever. Because bTMS and spin-diode rectification coexist in the same junction, a combined regime could enable nonlinear multi-input interactions, including quadratic cross-terms, within a single nanoscale element.

Spintronic nano-neurons offer a promising route towards energy-efficient, high performance hardware neural networks thanks to their inherent low-input nonlinear dynamics. However, training such networks remains a major bottleneck as it depends on oversimplified models of device behaviour and is highly sensitive to device variability. Here, we introduce a hardware architecture that overcomes these limitations by enabling on-device generation of gradients. First, we introduce theoretically and demonstrate experimentally that magnetic tunnel junctions can generate tunable and complex nonlinear responses. Building on this, we implement an analogue finite-difference approach to enable on-chip training in spintronic neural networks with one and two hidden layers. We experimentally implemented device-in-the-loop backpropagation in a magnetic tunnel junction-based neural network, achieving a classification accuracy of 93.3% despite pronounced device variability. During training, the gradients generated by the proposed analog neurons closely match the values derived numerically, without incurring in computational overhead. Via physical simulations, we also demonstrate that this approach can be scaled up to support training in deep architectures. Our results pave the way for reliable, trainable and fully analogue spintronic neural networks, opening up new possibilities for next-generation, energy-efficient artificial intelligence hardware.

In this study, the synchronization ability of vortex-based spin-torque nano-oscillators is investigated for three different dynamical regimes: the fundamental gyrotropic mode, the dynamic C-state, and the transition regime characterized by stochastic switching between the gyrotropic mode and the dynamic C-state. By combining injection locking at 2f and mutual synchronization experiments between two oscillators, it is shown that the ability to synchronize is larger in the transition regime than in the gyrotropic mode. By slightly tuning the injected dc current, this transition regime, which is highly efficient at synchronization, evolves into a dynamic state with no ability to synchronize. Thus, the synchronization range can be tuned, and the synchronized state can be easily switched on and off by selecting the dynamic regime. These results are promising for applications requiring large-scale networks of synchronized oscillators, where tuning the synchronization range and controlling the synchronized state are important features, such as neuromorphic computing and broadband microwave communication

This paper presents the design and implementation of a miniaturized, low-noise Magnetic Tunnel Junction (MTJ)-based proximity sensor with a high-performance readout channel. The MTJ-based proximity sensor consists of 1102 circular pillars of 100um diameter arranged in series, providing accurate detection of subtle interactions, such as a finger approaching the sensor. The system exhibits a 54 dB gain and a bandwidth of 1 kHz, with a noise power density of less than 30 nV/root Hz at 100 Hz, ensuring high precision. The proximity sensor demonstrated linear behavior for distances from 18 mm to 45 mm, with a sensitivity sufficient to detect low magnetic field variations. Experimental validation of the sensor shows a high degree of accuracy (R-2 = 0.9715), confirming its potential for use in touchless control, mobile technology, and industrial applications.

This paper presents a novel approach to reservoir computing (RC) using Granular Vortex-Based Magnetic Tunnel Junctions (GV-MTJs) for temporal applications. GV-MTJs, with their unique magnetic domain configurations and granular structures, provide the necessary fading memory and non-linear dynamics essential for RC. The vortex core’s oscillatory motion within the device allows for temporal correlation of inputs, giving fading memory, while grain-induced non-linear resistance and frequency variations enhance data dimensionality. Our findings indicate that varying device parameters can affect the relaxation time and gyrotropic frequency in both simulation and experiments. Relaxation times range from 100-140 ns and frequencies from 250-100 MHz. Through experiments, the classification error was reduced by 27% for the best sample, others showed limited potential. Due to signal application speed constraints, the fading memory is not fully utilized. However, the inherent RC capabilities of GV-MTJs are validated. This paper highlights the promise of GV-MTJs in neuromorphic computing and suggests avenues for future research to optimise their use in practical applications.

Neuromorphic computing, inspired by the brain's parallel and energy-efficient processing, offers a transformative approach to artificial intelligence. In this study, we fabricated optimized spin-transfer torque nano-oscillators (STNOs) and investigated their dynamic behaviors using a hybrid excitation scheme combining AC laser illumination and DC bias currents. Laser-induced thermal gradients generate pulsed thermoelectric voltages (V_AC) via the Tunnel Magneto-Seebeck (TMS) effect, while the addition of bias currents enhances this response, producing both V_AC and a DC component (V_DC). Magnetic field sweeps reveal distinct switching between parallel (P) and antiparallel (AP) magnetization states in both voltage components, supporting multistate memory applications. Millivolt-range thermovoltage signals in open-circuit conditions demonstrate CMOS compatibility, enabling simplified, scalable neuromorphic systems. Under biased conditions, enhanced thermovoltage outputs exhibit intriguing phenomena, including spikes correlated with Barkhausen jumps and double-switching behavior, offering insights into magnetization dynamics and vortex transitions. These features resemble neural spiking behavior, suggesting applications in spiking neural networks, reservoir computing, multistate logic, analog computing, and high-resolution sensing. By bridging spintronic phenomena with practical applications, this work provides a versatile platform for next-generation AI technologies and adaptive computing architectures.

This work proposes a Spintronics-based Hopfield oscillatory neural network (HONN) that leverages dynamic frequency-encoded electrical synchronization between two spin-torque vortex nano-oscillators (SVNOs) as oscillatory neurons, with a non-volatile memristor as a coupling element (synaptic connection). The frequency synchronization mechanism, inspired by the brain's oscillatory dynamics, enables the synchronization of SVNOs, facilitating efficient information processing of the dynamic oscillatory signals within the network. This coupling mechanism has been investigated to design SVNOs-based neural circuit design topology for enhanced frequency-encoded computing using SVNOs neurons and memristive coupling synapses. The proposed transmission gate-based SVNO oscillatory neural circuit has been implemented, offering efficient frequency synchronization, non-linearity, and a less complex neural circuit design. Further, a hybrid Spintronic/complementary metal oxide semiconductor 16-SVNOs HONN is designed, and circuit-based simulations are performed, which offer a promising solution for building robust and scalable HONNs. We achieve fast computation (similar to 4 ns) and offer significantly lower energy consumption (similar to 24 fJ/neuron) as compared to VO2-based ONN architectures (8x faster and 4x reduced power/neuron). Finally, we demonstrate an image denoising application on the proposed SVNO-based HONN hardware-compatible accelerator using an image-splitting approach with parallel processing. The 32 x 32 street view house number image dataset is efficiently split into blocks and processed through the 16-SVNOs HONN design, dividing the image into 4 x 4 blocks. Lastly, we examined the peak signal-to-noise ratio and structural similarity index measure for denoising the images with an efficient splitting approach for scalability. The network effectively denoises images while maintaining image quality, demonstrating the potential of the HONN hardware-compatible architecture for large-scale and real-time applications.

This work investigates the temperature effects on spin-torque vortex nano-oscillators (SVNOs) and demonstrates temperature robustness in SVNO coupling-based Hopfield oscillatory neural networks (HONNs) for image denoising tasks. We analyze how temperature variations influence the oscillation frequency and output power of SVNOs, highlighting their potential in spintronics-based HONNs. Temperature variations ranging from 300 K to 390 K is analyzed for the SVNO, revealing that fluctuations alter the magnetization dynamics, leading to shifts in power spectral density (PSD) and oscillation frequency. The frequency and output power exhibit a −9.19 dBm peak at 240 MHz under 300 K, which shifts to −10.77 dBm at 225 MHz at 350 K, indicating a reduction in frequency accompanied by an increase in output power. Moreover, the energy landscape is studied to investigate the variation in energy component interplay, which plays a major role in the formation of the magnetic vortex core, which has been analytically studied. It illustrates the increase in exchange and thermal energy due to temperature effects, while demagnetization and Zeeman energy remain nearly unchanged. As a result, the total minimized ground energy of vortex core increases from $3.90 \times 10^{-18} ~\mathrm{J}$ without temperature to $232 \times 10^{-18} ~\mathrm{J}$. Furthermore, we utilize the in-house physics-based Verilog-A model of SVNO, incorporating temperature effects to implement SVNO coupling-based HONN for image denoising applications. By integrating device-level insights, we develop efficient SVNO-based HONNs for image denoising, leveraging temperature-robust frequency-encoded coupling. Despite frequency shifts under varying temperatures (from 300K to 390K), coupling remains stable, enabling robust denoising through four-SVNO coupling. This makes the system temperature-resilient HONNs well-suited for edge applications like drones and autonomous systems operating in ambient temperature variations.

This paper presents a miniaturized, low-noise Magnetic Tunnel Junction (MTJ)-based gyroscope, designed for high-precision rotation sensing. The sensor leverages MTJ technology, utilizing a Wheatstone bridge configuration to measure angles in the XY plane. A highly linear readout channel with an adjustable high-side biasing current circuit was designed to ensure accurate signal detection. The readout channel includes a 41.5 dB gain and a 10 kHz bandwidth, achieving a remarkably low input-referred noise power density of 96nV/vHz at 10 Hz. The offset cancelation loop effectively manages an input offset of +/- 250mV. Measurement results demonstrate the system's capability to capture angular rotations from 0 to 360 degrees with high fidelity, with the absolute maximum error being less than 0.5 degrees. The MTJ-based gyroscope's specifications meet tactical-grade requirements, highlighting its potential for integration into applications where size, precision, and stability are critica

Spin-torque nano-oscillators (STNOs) are promising nanoscale microwave sources for spintronic applications, serving as signal generators or elements in neuromorphic computing systems. In this paper, we investigate the experimental realization of an oscillator based on a magnetic tunnel junction (MTJ) comprising two magnetic layers: a reference layer (RL) and a free layer (FL). We demonstrate that when magnetic vortices with opposite chirality and polarity are formed in the layers, the application of a current induces auto-oscillations even in the absence of external magnetic fields. This effect is observed in devices with diameters ranging from 800 to 1000 nm, exhibiting oscillation frequencies between 110 and 60 MHz. The underlying mechanism is attributed to the action of a spin current with vortex-like polarization injected from the RL, interacting with the magnetic vortex in the FL. This interaction generates a local out-of-plane effective field due to spin-transfer torque, which acts on the vortex core and initiates its motion. The observed mechanism differs qualitatively from the case of uniformly polarized spin currents perpendicular to the plane, where the resulting in-plane field acts on the planar components of the vortex magnetization.

This paper introduces Go-Fast, a novel approach that accelerates the simulation and optimization of LUT-based FPGA circuits using GPUs. Unlike previous GPU-based simulators that target general digital circuits with event-driven approaches, Go-Fast employs batch simulation techniques specifically optimized for approximate computing scenarios. The system utilizes both data parallelism to execute testbenches across numerous threads and structural parallelism for simultaneous simulation and logic pruning. Go-Fast fully exploits GPU architectural features by maximizing register usage, minimizing memory access, and allocating more work per thread. This domain-specific approach generates efficient source-to-source CUDA code that achieves significant performance improvements: five orders of magnitude over the Verilator simulator and two to three orders over optimized multi-core implementations.

This work presents the development of reconfigurable processing units capable of encapsulating various operations to design new domain-specific reconfigurable accelerators. These processing units are known as coarse-grained operators because they can execute multiple operations. We validated the new operators within the HPCGRA framework, a design environment for creating custom coarse-grained reconfigurable arrays that run as a virtual layer on commercial field-programmable gate arrays. The environment is parameterized and employs an intermediate portable format that abstracts away low-level specific bitstream details, enabling hardware-agnostic reconfiguration. In this paper, we present three case studies. The first is a domain-specific accelerator for the K-means algorithm, which can be entirely reconfigured in less than 2.34 ms to explore various clustering values and attributes, achieving up to 159 Gop/s performance for K=16$$ K=16 $$ considering 8 features. The reconfigurability of our accelerator has no impact on performance compared to static versions implemented directly in HLS and RTL. The second case study extends K-means with a Gini calculation operator for dimensionality reduction, quantization, and quality classification, achieving up to 278 Gops/s. The third case study presents a systolic matrix multiplier, demonstrating the versatility of the environment for designing different architectural domains.

Magnetic tunnel junctions (MTJs) offer a promising pathway toward energy-efficient neuromorphic computing due to their nanoscale footprint, nonvolatile switching, and intrinsic nonlinear dynamics that emulate synaptic behavior. However, generating large thermoelectric voltages with bias-tunable nonlinearities for neuromorphic use remains largely unexplored. Here, we introduce a hybrid opto-electrical excitation scheme-combining pulsed laser heating with DC bias-to drive MTJs into the nonlinear bias-enhanced tunnel magneto-Seebeck regime. This regime yields thermoelectric voltages in the tens of millivolts with a strong contrast between magnetic states, while also revealing spiking and double-switching behavior linked to vortex dynamics and fixed-layer depinning. The thermovoltage exhibits cubic dependence on bias current, enabling tunable synaptic weights. We simulate a single-layer neuromorphic network using optically encoded inputs and achieve 93.7% classification accuracy on handwritten digits. These results establish hybrid-driven MTJs as a compact, CMOS-compatible platform for neuromorphic computing, integrating optical input with spintronic functionality.

The human brain is competent in information processing and learning, largely due to the coordinated activity of neuronal populations. These populations exhibit rhythmic fluctuations in activity known as oscillatory dynamics, observed across different brain regions. These oscillations are crucial in various cognitive functions, such as memory consolidation, sensory processing, and motor control. This work proposes a novel neuromorphic computational (NC) model comprising analytical derivation for a spintronics-based Hopfield oscillatory neural network (HONN) employing frequency synchronization inspired by the brain's oscillatory mechanism of pattern recognition and associative memory. Kuramoto model, which represents the phase-based coupling of two oscillators, this work presents an extended mathematical analysis (modeling) of frequency and phase-based dual coupling mechanism. Further, this model is translated to the Theile equation which governs the dynamics of magnetic vortex core (such as core position and oscillation frequency) applicable to spintronic-based HONN. This approach uses spin-torque vortex nanooscillators (SVNOs) as neurons connected via resistive synapses representing synaptic coupling strength. The model describes the analytical relation between the SVNOs and the synaptic element, and how it influences the synchronized frequency ( f(sync)). Stronger coupling aligns f(sync )closer to the higher gyrotropic frequency of the magnetic vortex core within the network, while weaker coupling promotes f(sync) closer to the lower one. The synaptic connections transition between low resistance state (LRS) and high resistance state (HRS) mimicking biological brain plasticity. This NC model is then utilized to design a spintronic-based HONN circuit to illustrate oscillatory properties, frequency-based synchronization, and resistive coupling to create an energy-efficient and scalable architecture. The proposed NC model represents a promising approach to advancing large-scale HONN hardware architecture, particularly by enabling the integration of SVNOs as neurons, resistive memories as synapses, and conventional electronics as peripherals. This model serves as a foundational framework for exploring the feasibility, functionality, and reliability of advanced neural network architectures, crucial for evaluating the potential of these hybrid systems in practical, large-scale applications.