Multi-level, forming and filament free, bulk switching trilayer RRAM for neuromorphic computing at the edge.

CMOS-RRAM integration holds great promise for low energy and high throughput neuromorphic computing. However, most RRAM technologies relying on filamentary switching suffer from variations and noise, leading to computational accuracy loss, increased energy consumption, and overhead by expensive program and verify schemes. We developed a filament-free, bulk switching RRAM technology to address these challenges. We systematically engineered a trilayer metal-oxide stack and investigated the switching characteristics of RRAM with varying thicknesses and oxygen vacancy distributions to achieve reliable bulk switching without any filament formation. We demonstrated bulk switching at megaohm regime with high current nonlinearity, up to 100 levels without compliance current. We developed a neuromorphic compute-in-memory platform and showcased edge computing by implementing a spiking neural network for an autonomous navigation/racing task. Our work addresses challenges posed by existing RRAM technologies and paves the way for neuromorphic computing at the edge under strict size, weight, and power constraints.

Reconfigurable Cascaded Thermal Neuristors for Neuromorphic Computing.

While the complementary metal-oxide semiconductor (CMOS) technology is the mainstream for the hardware implementation of neural networks, an alternative route is explored based on a new class of spiking oscillators called "thermal neuristors", which operate and interact solely via thermal processes. Utilizing the insulator-to-metal transition (IMT) in vanadium dioxide, a wide variety of reconfigurable electrical dynamics mirroring biological neurons is demonstrated. Notably, inhibitory functionality is achieved just in a single oxide device, and cascaded information flow is realized exclusively through thermal interactions. To elucidate the underlying mechanisms of the neuristors, a detailed theoretical model is developed, which accurately reflects the experimental results. This study establishes the foundation for scalable and energy-efficient thermal neural networks, fostering progress in brain-inspired computing.

Detection of electromagnetic phase transitions using a helical cavity susceptometer.

Fast and sensitive phase transition detection is one of the most important requirements for new material synthesis and characterization. For solid-state samples, microwave absorption techniques can be employed for detecting phase transitions because it simultaneously monitors changes in electronic and magnetic properties. However, microwave absorption techniques require expensive high-frequency microwave equipment and bulky hollow cavities. Due to size limitations in conventional instruments, it is challenging to implement these cavities inside a laboratory cryostat. In this work, we designed and built a susceptometer that consists of a small helical cavity embedded into a custom insert of a commercial cryostat. This cavity resonator operated at sub-GHz frequencies is extremely sensitive to changes in material parameters, such as electrical conductivity, magnetization, and electric and magnetic susceptibilities. To demonstrate its operation, we detected superconducting phase transition in Nb and YBa2Cu3O7-δ, metal-insulator transitions in V2O3, ferromagnetic transition in Gd, and magnetic field induced transformation in meta magnetic NiCoMnIn single crystals. This high sensitivity apparatus allows the detection of trace amounts of materials (10-9-cc) undergoing an electromagnetic transition in a very broad temperature (2-400 K) and magnetic field (up to 90 kOe) ranges.

Stochastic transition in synchronized spiking nanooscillators.

This work reports that synchronization of Mott material-based nanoscale coupled spiking oscillators can be drastically different from that in conventional harmonic oscillators. We investigated the synchronization of spiking nanooscillators mediated by thermal interactions due to the close physical proximity of the devices. Controlling the driving voltage enables in-phase 1:1 and 2:1 integer synchronization modes between neighboring oscillators. Transition between these two integer modes occurs through an unusual stochastic synchronization regime instead of the loss of spiking coherence. In the stochastic synchronization regime, random length spiking sequences belonging to the 1:1 and 2:1 integer modes are intermixed. The occurrence of this stochasticity is an important factor that must be taken into account in the design of large-scale spiking networks for hardware-level implementation of novel computational paradigms such as neuromorphic and stochastic computing.

Probing FeSi, a d-electron topological Kondo insulator candidate, with magnetic field, pressure, and microwaves.

Recently, evidence for a conducting surface state (CSS) below 19 K was reported for the correlated d-electron small gap semiconductor FeSi. In the work reported herein, the CSS and the bulk phase of FeSi were probed via electrical resistivity ρ measurements as a function of temperature T, magnetic field B to 60 T, and pressure P to 7.6 GPa, and by means of a magnetic field-modulated microwave spectroscopy (MFMMS) technique. The properties of FeSi were also compared with those of the Kondo insulator SmB6 to address the question of whether FeSi is a d-electron analogue of an f-electron Kondo insulator and, in addition, a "topological Kondo insulator" (TKI). The overall behavior of the magnetoresistance of FeSi at temperatures above and below the onset temperature TS = 19 K of the CSS is similar to that of SmB6. The two energy gaps, inferred from the ρ(T) data in the semiconducting regime, increase with pressure up to about 7 GPa, followed by a drop which coincides with a sharp suppression of TS. Several studies of ρ(T) under pressure on SmB6 reveal behavior similar to that of FeSi in which the two energy gaps vanish at a critical pressure near the pressure at which TS vanishes, although the energy gaps in SmB6 initially decrease with pressure, whereas in FeSi they increase with pressure. The MFMMS measurements showed a sharp feature at TS ≈ 19 K for FeSi, which could be due to ferromagnetic ordering of the CSS. However, no such feature was observed at TS ≈ 4.5 K for SmB6.

Thermal Management in Neuromorphic Materials, Devices, and Networks.

Machine learning has experienced unprecedented growth in recent years, often referred to as an "artificial intelligence revolution." Biological systems inspire the fundamental approach for this new computing paradigm: using neural networks to classify large amounts of data into sorting categories. Current machine-learning schemes implement simulated neurons and synapses on standard computers based on a von Neumann architecture. This approach is inefficient in energy consumption, and thermal management, motivating the search for hardware-based systems that imitate the brain. Here, the present state of thermal management of neuromorphic computing technology and the challenges and opportunities of the energy-efficient implementation of neuromorphic devices are considered. The main features of brain-inspired computing and quantum materials for implementing neuromorphic devices are briefly described, the brain criticality and resistive switching-based neuromorphic devices are discussed, the energy and electrical considerations for spiking-based computation are presented, the fundamental features of the brain's thermal regulation are addressed, the physical mechanisms for thermal management and thermoelectric control of materials and neuromorphic devices are analyzed, and challenges and new avenues for implementing energy-efficient computing are described.

Unusual Magnetic Hysteresis and Transition between Vortex and Double Pole States Arising from Interlayer Coupling in Diamond-Shaped Nanostructures.

Controlling the magnetic ground states at the nanoscale is a long-standing basic research problem and an important issue in magnetic storage technologies. Here, we designed a nanostructured material that exhibits very unusual hysteresis loops due to a transition between vortex and double pole states. Arrays of 700 nm diamond-shaped nanodots consisting of Py(30 nm)/Ru(tRu)/Py(30 nm) (Py, permalloy (Ni80Fe20)) trilayers were fabricated by interference lithography and e-beam evaporation. We show that varying the Ru interlayer spacer thickness (tRu) governs the interaction between the Py layers. We found this interaction mainly mediated by two mechanisms: magnetostatic interaction that favors antiparallel (antiferromagnetic, AFM) alignment of the Py layers and exchange interaction that oscillates between ferromagnetic (FM) and AFM couplings. For a certain range of Ru thicknesses, FM coupling dominates and forms magnetic vortices in the upper and lower Py layers. For Ru thicknesses at which AFM coupling dominates, the magnetic state in remanence is a double pole structure. Our results showed that the interlayer exchange coupling interaction remains finite even at 4 nm Ru thickness. The magnetic states in remanence, observed by magnetic force microscopy (MFM), are in good agreement with corresponding hysteresis loops obtained by the magneto-optic Kerr effect (MOKE) and micromagnetic simulations.

The impact of the suppression of highly connected protein interactions on the corona virus infection.

Several highly effective Covid-19 vaccines are in emergency use, although more-infectious coronavirus strains, could delay the end of the pandemic even further. Because of this, it is highly desirable to develop fast antiviral drug treatments to accelerate the lasting immunity against the virus. From a theoretical perspective, computational approaches are useful tools for antiviral drug development based on the data analysis of gene expression, chemical structure, molecular pathway, and protein interaction mapping. This work studies the structural stability of virus-host interactome networks based on the graphical representation of virus-host protein interactions as vertices or nodes connected by commonly shared proteins. These graphical network visualization methods are analogous to those use in the design of artificial neural networks in neuromorphic computing. In standard protein-node-based network representation, virus-host interaction merges with virus-protein and host-protein networks, introducing redundant links associated with the internal virus and host networks. On the contrary, our approach provides a direct geometrical representation of viral infection structure and allows the effective and fast detection of the structural robustness of the virus-host network through proteins removal. This method was validated by applying it to H1N1 and HIV viruses, in which we were able to pinpoint the changes in the Interactome Network produced by known vaccines. The application of this method to the SARS-CoV-2 virus-host protein interactome implies that nonstructural proteins nsp4, nsp12, nsp16, the nuclear pore membrane glycoprotein NUP210, and ubiquitin specific peptidase USP54 play a crucial role in the viral infection, and their removal may provide an efficient therapy. This method may be extended to any new mutations or other viruses for which the Interactome Network is experimentally determined. Since time is of the essence, because of the impact of more-infectious strains on controlling the spread of the virus, this method may be a useful tool for novel antiviral therapies.

Generation of Tunable Stochastic Sequences Using the Insulator-Metal Transition.

Probabilistic computing is a paradigm in which data are not represented by stable bits, but rather by the probability of a metastable bit to be in a particular state. The development of this technology has been hindered by the availability of hardware capable of generating stochastic and tunable sequences of "1s" and "0s". The options are currently limited to complex CMOS circuitry and, recently, magnetic tunnel junctions. Here, we demonstrate that metal-insulator transitions can also be used for this purpose. We use an electrical pump/probe protocol and take advantage of the stochastic relaxation dynamics in VO2 to induce random metallization events. A simple latch circuit converts the metallization sequence into a random stream of 1s and 0s. The resetting pulse in between probes decorrelates successive events, providing a true stochastic digital sequence.

Wireless Force-Inducing Neuronal Stimulation Mediated by High Magnetic Moment Microdiscs.

Noninvasive manipulation of cell signaling is critical in basic neuroscience research and in developing therapies for neurological disorders and psychiatric conditions. Here, the wireless force-induced stimulation of primary neuronal circuits through mechanotransduction mediated by magnetic microdiscs (MMDs) under applied low-intensity and low-frequency alternating magnetic fields (AMFs), is described. MMDs are fabricated by top-down lithography techniques that allow for cost-effective mass production of biocompatible MMDs with high saturation and zero magnetic magnetic moment at remanence. MMDs are utilized as transducers of AMFs into mechanical forces. When MMDs are exposed to primary rat neuronal circuits, their magneto-mechanical actuation triggers the response of specific mechanosensitive ion channels expressed on the cell membranes activating ≈50% of hippocampal and ≈90% of cortical neurons subjected to the treatment. Mechanotransduction is confirmed by the inhibition of mechanosensitive transmembrane channels with Gd3+ . Mechanotransduction mediated by MMDs cause no cytotoxic effect to neuronal cultures. This technology fulfills the requirements of cell-type specificity and weak magnetic fields, two limiting factors in the development of noninvasive neuromodulation therapies and clinical equipment design. Moreover, high efficiency and long-lasting stimulations are successfully achieved. This research represents a fundamental step forward for magneto-mechanical control of neural activity using disc-shaped micromaterials with tailored magnetic properties.

Imaging the itinerant-to-localized transmutation of electrons across the metal-to-insulator transition in V2O3.

In solids, strong repulsion between electrons can inhibit their movement and result in a "Mott" metal-to-insulator transition (MIT), a fundamental phenomenon whose understanding has remained a challenge for over 50 years. A key issue is how the wave-like itinerant electrons change into a localized-like state due to increased interactions. However, observing the MIT in terms of the energy- and momentum-resolved electronic structure of the system, the only direct way to probe both itinerant and localized states, has been elusive. Here we show, using angle-resolved photoemission spectroscopy (ARPES), that in V2O3, the temperature-induced MIT is characterized by the progressive disappearance of its itinerant conduction band, without any change in its energy-momentum dispersion, and the simultaneous shift to larger binding energies of a quasi-localized state initially located near the Fermi level.

Transverse barrier formation by electrical triggering of a metal-to-insulator transition.

Application of an electric stimulus to a material with a metal-insulator transition can trigger a large resistance change. Resistive switching from an insulating into a metallic phase, which typically occurs by the formation of a conducting filament parallel to the current flow, is a highly active research topic. Using the magneto-optical Kerr imaging, we found that the opposite type of resistive switching, from a metal into an insulator, occurs in a reciprocal characteristic spatial pattern: the formation of an insulating barrier perpendicular to the driving current. This barrier formation leads to an unusual N-type negative differential resistance in the current-voltage characteristics. We further demonstrate that electrically inducing a transverse barrier enables a unique approach to voltage-controlled magnetism. By triggering the metal-to-insulator resistive switching in a magnetic material, local on/off control of ferromagnetism is achieved using a global voltage bias applied to the whole device.

Inherent stochasticity during insulator-metal transition in VO2.

Vanadium dioxide (VO2), which exhibits a near-room-temperature insulator-metal transition, has great potential in applications of neuromorphic computing devices. Although its volatile switching property, which could emulate neuron spiking, has been studied widely, nanoscale studies of the structural stochasticity across the phase transition are still lacking. In this study, using in situ transmission electron microscopy and ex situ resistive switching measurement, we successfully characterized the structural phase transition between monoclinic and rutile VO2 at local areas in planar VO2/TiO2 device configuration under external biasing. After each resistive switching, different VO2 monoclinic crystal orientations are observed, forming different equilibrium states. We have evaluated a statistical cycle-to-cycle variation, demonstrated a stochastic nature of the volatile resistive switching, and presented an approach to study in-plane structural anisotropy. Our microscopic studies move a big step forward toward understanding the volatile switching mechanisms and the related applications of VO2 as the key material of neuromorphic computing.

Spatiotemporal characterization of the field-induced insulator-to-metal transition.

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.

A quantum material spintronic resonator.

In a spintronic resonator a radio-frequency signal excites spin dynamics that can be detected by the spin-diode effect. Such resonators are generally based on ferromagnetic metals and their responses to spin torques. New and richer functionalities can potentially be achieved with quantum materials, specifically with transition metal oxides that have phase transitions that can endow a spintronic resonator with hysteresis and memory. Here we present the spin torque ferromagnetic resonance characteristics of a hybrid metal-insulator-transition oxide/ ferromagnetic metal nanoconstriction. Our samples incorporate [Formula: see text], with Ni, Permalloy ([Formula: see text]) and Pt layers patterned into a nanoconstriction geometry. The first order phase transition in [Formula: see text] is shown to lead to systematic changes in the resonance response and hysteretic current control of the ferromagnetic resonance frequency. Further, the output signal can be systematically varied by locally changing the state of the [Formula: see text] with a dc current. These results demonstrate new spintronic resonator functionalities of interest for neuromorphic computing.

Energy-efficient Mott activation neuron for full-hardware implementation of neural networks.

To circumvent the von Neumann bottleneck, substantial progress has been made towards in-memory computing with synaptic devices. However, compact nanodevices implementing non-linear activation functions are required for efficient full-hardware implementation of deep neural networks. Here, we present an energy-efficient and compact Mott activation neuron based on vanadium dioxide and its successful integration with a conductive bridge random access memory (CBRAM) crossbar array in hardware. The Mott activation neuron implements the rectified linear unit function in the analogue domain. The neuron devices consume substantially less energy and occupy two orders of magnitude smaller area than those of analogue complementary metal-oxide semiconductor implementations. The LeNet-5 network with Mott activation neurons achieves 98.38% accuracy on the MNIST dataset, close to the ideal software accuracy. We perform large-scale image edge detection using the Mott activation neurons integrated with a CBRAM crossbar array. Our findings provide a solution towards large-scale, highly parallel and energy-efficient in-memory computing systems for neural networks.

Operando characterization of conductive filaments during resistive switching in Mott VO2.

Vanadium dioxide (VO2) has attracted much attention owing to its metal-insulator transition near room temperature and the ability to induce volatile resistive switching, a key feature for developing novel hardware for neuromorphic computing. Despite this interest, the mechanisms for nonvolatile switching functioning as synapse in this oxide remain not understood. In this work, we use in situ transmission electron microscopy, electrical transport measurements, and numerical simulations on Au/VO2/Ge vertical devices to study the electroforming process. We have observed the formation of V5O9 conductive filaments with a pronounced metal-insulator transition and that vacancy diffusion can erase the filament, allowing for the system to "forget." Thus, both volatile and nonvolatile switching can be achieved in VO2, useful to emulate neuronal and synaptic behaviors, respectively. Our systematic operando study of the filament provides a more comprehensive understanding of resistive switching, key in the development of resistive switching-based neuromorphic computing.

Driving magnetic domains at the nanoscale by interfacial strain-induced proximity.

We investigate the local nanoscale changes of the magnetic anisotropy of a Ni film subject to an inverse magnetostrictive effect by proximity to a V2O3 layer. Using temperature-dependent photoemission electron microscopy (PEEM) combined with X-ray magnetic circular dichroism (XMCD), direct images of the Ni spin alignment across the first-order structural phase transition (SPT) of V2O3 were obtained. We find an abrupt temperature-driven reorientation of the Ni magnetic domains across the SPT, which is associated with a large increase of the coercive field. Moreover, angular dependent ferromagnetic resonance (FMR) shows a remarkable change in the magnetic anisotropy of the Ni film across the SPT of V2O3. Micromagnetic simulations based on these results are in quantitative agreement with the PEEM data. Direct measurements of the lateral correlation length of the Ni domains from XMCD images show an increase of almost one order of magnitude at the SPT compared to room temperature, as well as a broad spatial distribution of the local transition temperatures, thus corroborating the phase coexistence of Ni anisotropies caused by the V2O3 SPT. We show that the rearrangement of the Ni domains is due to strain induced by the oxide layers' structural domains across the SPT. Our results illustrate the use of alternative hybrid systems to manipulate magnetic domains at the nanoscale, which allows for engineering of coercive fields for novel data storage architectures.

Controlling Metal-Insulator Transitions in Vanadium Oxide Thin Films by Modifying Oxygen Stoichiometry.

Vanadium oxides are strongly correlated materials which display metal-insulator transitions (MITs) as well as various structural and magnetic properties that depend heavily on oxygen stoichiometry. Therefore, it is crucial to precisely control oxygen stoichiometry in these materials, especially in thin films. This work demonstrates a high-vacuum gas evolution technique which allows for the modification of oxygen concentrations in VOX thin films by carefully tuning the thermodynamic conditions. We were able to control the evolution between VO2, V3O5, and V2O3 phases on sapphire substrates, overcoming the narrow phase stability of adjacent Magnéli phases. A variety of annealing routes were found to achieve the desired phases and eventually control the MIT. The pronounced MIT of the transformed films along with the detailed structural investigations based on X-ray diffraction measurements and X-ray photoelectron spectroscopy show that optimal stoichiometry is obtained and stabilized. Using this technique, we find that the thin-film V-O phase diagram differs from that of the bulk material because of strain and finite size effects. Our study demonstrates new pathways to strategically tune the oxygen stoichiometry in complex oxides and provides a road map for understanding the phase stability of VOX thin films.

Nanoscale Imaging and Control of Volatile and Non-Volatile Resistive Switching in VO2.

Control of the metal-insulator phase transition is vital for emerging neuromorphic and memristive technologies. The ability to alter the electrically driven transition between volatile and non-volatile states is particularly important for quantum-materials-based emulation of neurons and synapses. The major challenge of this implementation is to understand and control the nanoscale mechanisms behind these two fundamental switching modalities. Here, in situ X-ray nanoimaging is used to follow the evolution of the nanostructure and disorder in the archetypal Mott insulator VO2 during an electrically driven transition. Our findings demonstrate selective and reversible stabilization of either the insulating or metallic phases achieved by manipulating the defect concentration. This mechanism enables us to alter the local switching response between volatile and persistent regimes and demonstrates a new possibility for nanoscale control of the resistive switching in Mott materials.