Search for: All records

In-memory computing with emerging non-volatile memory devices (eNVMs) has shown promising results in accelerating matrix-vector multiplications. However, activation function calculations are still being implemented with general processors or large and complex neuron peripheral circuits. Here, we present the integration of Ag-based conductive bridge random access memory (Ag-CBRAM) crossbar arrays with Mott rectified linear unit (ReLU) activation neurons for scalable, energy and area-efficient hardware (HW) implementation of deep neural networks. We develop Ag-CBRAM devices that can achieve a high ON/OFF ratio and multi-level programmability. Compact and energy-efficient Mott ReLU neuron devices implementing ReLU activation function are directly connected to the columns of Ag-CBRAM crossbars to compute the output from the weighted sum current. We implement convolution filters and activations for VGG-16 using our integrated HW and demonstrate the successful generation of feature maps for CIFAR-10 images in HW. Our approach paves a new way toward building a highly compact and energy-efficient eNVMs-based in-memory computing system.

Synchronization of electrical oscillators is a crucial step toward practical implementation of oscillator-based and bio-inspired computing. Here, we report the emergence of an unusual stochastic pattern in coupled spiking Mott nanodevices. Although a moderate capacitive coupling results in a deterministic alternating spiking, increasing the coupling strength leads counterintuitively to stochastic disruptions of the alternating spiking sequence. The disruptions of the deterministic spiking sequence are a direct consequence of the small intrinsic stochasticity in electrical triggering of the insulator–metal transition. Although the stochasticity is subtle in individual nanodevices, it becomes dramatically enhanced just in a single pair of coupled oscillators and, thus, dominates the synchronization. This is different from the stochasticity and multimodal coupling, appearing due to collective effects in large oscillator networks. The stochastic spiking pattern in Mott nanodevices results in a discrete inter-spike interval distribution resembling those in biological neurons. Our results advance the understanding of the emergent synchronization properties in spiking oscillators and provide a platform for hardware-level implementation of probabilistic computing and biologically plausible electronic devices.

Resistive switching can be achieved in a Mott insulator by applying current/voltage, which triggers an insulator-metal transition (IMT). This phenomenon is key for understanding IMT physics and developing novel memory elements and brain-inspired technology. Despite this, the roles of electric field and Joule heating in the switching process remain controversial. Using nanowires of two archetypal Mott insulators—VO₂and V₂O₃we unequivocally show that a purely non-thermal electrical IMT can occur in both materials. The mechanism behind this effect is identified as field-assisted carrier generation leading to a doping driven IMT. This effect can be controlled by similar means in both VO₂and V₂O₃, suggesting that the proposed mechanism is generally applicable to Mott insulators. The energy consumption associated with the non-thermal IMT is extremely low, rivaling that of state-of-the-art electronics and biological neurons. These findings pave the way towards highly energy-efficient applications of Mott insulators.

Machine learning imitates the basic features of biological neural networks at a software level. A strong effort is currently being made to mimic neurons and synapses with hardware components, an approach known as neuromorphic computing. While recent advances in resistive switching have provided a path to emulate synapses at the 10 nm scale, a scalable neuron analogue is yet to be found. Here, we show how heat transfer can be utilized to mimic neuron functionalities in Mott nanodevices. We use the Joule heating created by current spikes to trigger the insulator-to-metal transition in a biased VO₂nanogap. We show that thermal dynamics allow the implementation of the basic neuron functionalities: activity, leaky integrate-and-fire, volatility and rate coding. This approach could enable neuromorphic hardware to take full advantage of the rapid advances in memristive synapses, allowing for much denser and complex neural networks.

Ramp‐reversal memory has recently been discovered in several insulator‐to‐metal transition materials where a non‐volatile resistance change can be set by repeatedly driving the material partway through the transition. This study uses optical microscopy to track the location and internal structure of accumulated memory as a thin film of VO₂is temperature cycled through multiple training subloops. These measurements reveal that the gain of insulator phase fraction between consecutive subloops occurs primarily through front propagation at the insulator‐metal boundaries. By analyzing transition temperature maps, it is found, surprisingly, that the memory is also stored deep inside both insulating and metallic clusters throughout the entire sample, making the metal‐insulator coexistence landscape more rugged. This non‐volatile memory is reset after heating the sample to higher temperatures, as expected. Diffusion of point defects is proposed to account for the observed memory writing and subsequent erasing over the entire sample surface. By spatially mapping the location and character of non‐volatile memory encoding in VO₂, this study results enable the targeting of specific local regions in the film where the full insulator‐to‐metal resistivity change can be harnessed in order to maximize the working range of memory elements for conventional and neuromorphic computing applications.

The emergence of spin‐orbit torques as a promising approach to energy‐efficient magnetic switching has generated large interest in material systems with easily and fully tunable spin‐orbit torques. Here, current‐induced spin‐orbit torques in VO₂/NiFe heterostructures are investigated using spin‐torque ferromagnetic resonance, where the VO₂layer undergoes a prominent insulator‐metal transition. A roughly twofold increase in the Gilbert damping parameter, α, with temperature is attributed to the change in the VO₂/NiFe interface spin absorption across the VO₂phase transition. More remarkably, a large modulation (±100%) and a sign change of the current‐induced spin‐orbit torque across the VO₂phase transition suggest two competing spin‐orbit torque generating mechanisms. The bulk spin Hall effect in metallic VO₂, corroborated by the first‐principles calculation of the spin Hall conductivity , is verified as the main source of the spin‐orbit torque in the metallic phase. The self‐induced/anomalous torque in NiFe, with opposite sign and a similar magnitude to the bulk spin Hall effect in metallic VO₂, can be the other competing mechanism that dominates as temperature decreases. For applications, the strong tunability of the torque strength and direction opens a new route to tailor spin‐orbit torques of materials that undergo phase transitions for new device functionalities.

Many correlated systems feature an insulator-to-metal transition that can be triggered by an electric field. Although it is known that metallization takes place through filament formation, the details of how this process initiates and evolves remain elusive. We use in-operando optical reflectivity to capture the growth dynamics of the metallic phase with space and time resolution. We demonstrate that filament formation is triggered by nucleation at hotspots, with a subsequent expansion over several decades in time. By comparing three case studies (VO₂, V₃O₅, and V₂O₃), we identify the resistivity change across the transition as the crucial parameter governing this process. Our results provide a spatiotemporal characterization of volatile resistive switching in Mott insulators, which is important for emerging technologies, such as optoelectronics and neuromorphic computing.