Search

One of the major obstacles to realizing spintronic devices such as MESO logic devices is the small signal magnitude used for magnetization readout, making it important to find materials with high spin-to-charge conversion efficiency. Although intermixing at the junction of two materials is a widely occurring phenomenon, its influence on material characterization and the estimation of spin-to-charge conversion efficiencies is easily neglected or underestimated. Here, we demonstrate all electrical spin-to-charge conversion in Bi$_x$Se$_{1-x}$ nanodevices and show how the conversion efficiency can be overestimated by tens of times depending on the adjacent metal used as a contact. We attribute this to the intermixing-induced compositional change and the properties of a polycrystal that lead to drastic changes in resistivity and spin Hall angle. Strategies to improve the spin-to-charge conversion signal in similar structures for functional devices are discussed., Main text (18 pages, 3 figures, 2 tables) and supporting information (18 pages)

Neuromorphic architectures with low energy barrier nanomagnetic devices have been attracting increasing interest over the past few years. More recently, a low barrier nanomagnet (LBNM)-based probabilistic device (p-bit) has been shown to be the basis of neuronal nodes in deep belief networks (DBNs). The LBNMs with perpendicular magnetic anisotropy (PMA) are analyzed and optimized in the interest of achieving stochasticity present in the learning system. In p-bit-based DBNs, several defects, such as variation of the nanomagnet diameter ( $\sigma d$ ), thickness ( $\sigma t_{f}$ ), and anisotropy field ( $\sigma H_{K}$ ), which results in alteration of the fluctuation speed of the p-bit’s nanomagnet can impair functionality. In this article, the accuracy of p-bit-based DBNs is examined under variation of nanomagnet diameter, thickness, and anisotropy field for various tilt angles and temperatures. As evaluated for the MNIST data set for temperature and tilt angle of 300 K and 25°, accordingly, it is shown that the process variation (PV) of $\sigma d$ , $\sigma t_{f}$ , and $\sigma H_{K}$ can be tolerated up to 8%, 23%, and 25%, respectively. A new method is developed to control the fluctuation frequency of the output of a p-bit device by employing a feedback mechanism. The feedback can alleviate this PV sensitivity of p-bit-based DBNs. The compact and low complexity method, which is presented by introducing the self-compensating circuit, can alleviate the influences of PV in fabrication and practical implementation.

One of the major obstacles to realizing spintronic devices such as MESO logic devices is the small signal magnitude used for magnetization readout, making it important to find materials with high spin-to-charge conversion efficiency. Although intermixing at the junction of two materials is a widely occurring phenomenon, its influence on material characterization and the estimation of spin-to-charge conversion efficiencies are easily neglected or underestimated. Here, we demonstrate all-electrical spin-to-charge conversion in Bi

Bayesian networks are powerful statistical models to understand causal relationships in real-world probabilistic problems such as diagnosis, forecasting, computer vision, etc. For systems that involve complex causal dependencies among many variables, the complexity of the associated Bayesian networks become computationally intractable. As a result, direct hardware implementation of these networks is one promising approach to reducing power consumption and execution time. However, the few hardware implementations of Bayesian networks presented in literature rely on deterministic CMOS devices that are not efficient in representing the inherently stochastic variables in a Bayesian network. This work presents an experimental demonstration of a Bayesian network building block implemented with naturally stochastic spintronic devices. These devices are based on nanomagnets with perpendicular magnetic anisotropy, initialized to their hard axes by the spin orbit torque from a heavy metal under-layer utilizing the giant spin Hall effect, enabling stochastic behavior. We construct an electrically interconnected network of two stochastic devices and manipulate the correlations between their states by changing connection weights and biases. By mapping given conditional probability tables to the circuit hardware, we demonstrate that any two node Bayesian networks can be implemented by our stochastic network. We then present the stochastic simulation of an example case of a four node Bayesian network using our proposed device, with parameters taken from the experiment. We view this work as a first step towards the large scale hardware implementation of Bayesian networks., 9 pages, 4 figures

Employing the probabilistic nature of unstable nano-magnet switching has recently emerged as a path towards unconventional computational systems such as neuromorphic or Bayesian networks. In this letter, we demonstrate proof-of-concept stochastic binary operation using hard axis initialization of nano-magnets and control of their output state probability (activation function) by means of input currents. Our method provides a natural path towards addition of weighted inputs from various sources, mimicking the integration function of neurons. In our experiment, spin orbit torque (SOT) is employed to “drive” nano-magnets with perpendicular magnetic anisotropy (PMA) -to their metastable state, i.e. in-plane hard axis. Next, the probability of relaxing into one magnetization state (+mi) or the other (−mi) is controlled using an Oersted field generated by an electrically isolated current loop, which acts as a “charge” input to the device. The final state of the magnet is read out by the anomalous Hall effect (AHE), demonstrating that the magnetization can be probabilistically manipulated and output through charge currents, closing the loop from charge-to-spin and spin-to-charge conversion. Based on these building blocks, a two-node directed network is successfully demonstrated where the status of the second node is determined by the probabilistic output of the previous node and a weighted connection between them. We have also studied the effects of various magnetic properties, such as magnet size and anisotropic field on the stochastic operation of individual devices through Monte Carlo simulations of Landau Lifshitz Gilbert (LLG) equation. The three-terminal stochastic devices demonstrated here are a critical step towards building energy efficient spin based neural networks and show the potential for a new application space.

Stochastic oscillators based on emerging nanodevices are attractive because of their ultra-low power requirement and the ability to exhibit stochastic resonance, a phenomenon where synchronization to weak input signals is enabled due to ambient noise. In this work, a low barrier nanomagnet-based stochastic oscillator is demonstrated, whose output jumps spontaneously between two states by harnessing the ambient thermal noise, requiring no additional power. By utilizing spin–orbit torque in a three-terminal device configuration, phase synchronization of these oscillators to weak periodic drives of particular frequencies is demonstrated. Experiments are performed to show the tunability of this synchronization frequency by controlling an electrical feedback parameter. The current required for synchronization is more than eight times smaller than that required for the deterministic switching of similar nanomagnetic devices. A model based on Kramers’ transition rate in a symmetric double well potential is adopted and dynamical simulations are performed to explain the experimental results.

Quantum-classical spin hybrids composed of physical system with complimentary characteristics have enabled novel capabilities and functionalities within the realm of existing technology. One half of such hybrid systems is the quantum impurity spin with small spin quantum number such that its description is governed by the counter-intuitive laws of quantum mechanics. The other is a classical magnet with large spin quantum number such that its dynamics can be captured within the framework of classical physics. Such hybrids give rise to possibilities where controlling the degrees of freedom in one system can be leveraged to control dynamics in the other. Leveraging the demonstrated spintronic tools of classical magnet dynamics, we demonstrate two significant steps towards realizing a quantum network for information processing applications. One, a theoretically designed regime where electrical control of non-linear magnetization dynamics of a nanomagnet provides a local, coherent, and low-power drive to manipulate a coupled quantum impurity spin without introducing additional decoherence. Another, where we demonstrate via a joint theoretical and experimental effort, the electrical tuning of interaction between electrically-controlled propagating magnons in an extended magnet and a quantum impurity spin. The merits of such a hybrid system provide pathways to overcome the bottlenecks associated with local controllability of individual quantum spins in a quantum network and modulate the interaction mediating the two subsystems.

Energy-efficient methods are addressed for lever-aging low energy barrier nanomagnetic devices within neuro-morphic architectures. Using a Magnetoresistive Random Access Memory (MRAM) probabilistic device (p-bit) as the basis of neuronal structures in Deep Belief Networks (DBNs), the impact of increasing the Magnetic Tunnel Junction's (MTJ's) energy barrier is assessed and optimized for the resulting stochasticity present in the learning system. A self-compensating circuit is developed herein providing a compact, and low complexity approach to mitigating process variation impacts towards practical implementation and fabrication. As evaluated for the MNIST dataset for energy barriers (E B ) at near-zero kT to 2.0 kT, it is shown that the proposed variation-tolerant circuit can effectively increase the reduced probabilistic fluctuation speed of the nanomagnet in p-bits with E B > OkT achieving approximately 5x improvement in the total energy consumption of a 784 × 200 × 10 DBN.

Spin Orbit Torque Magnetic RAM (SOT-MRAM) is emerging as a promising memory technology owing to its high endurance, reliability and speed. A critical factor for its success is the development of materials that exhibit efficient conversion of charge current to spin current, characterized by their spin Hall efficiency. In this work, it is experimentally demonstrated that the spin Hall efficiency of the industrially relevant ultra-thin Ta can be enhanced by more than 25× when a monolayer (ML) WSe2 is inserted as an underlayer. The enhancement is attributed to spin absorption at the Ta/WSe2 interface, suggested by harmonic Hall measurements. The presented hybrid spin Hall stack with a 2D WSe2 underlayer has a total body thickness of less than 2 nm and exhibits greatly enhanced spin Hall efficiency, which makes this hybrid a promising candidate for energy efficient SOT-MRAM.

Several beyond-CMOS computing technologies have emerged in the recent years to tackle the modern computing tasks that become intractable for Boolean logic based computation, performed on a von Neumann computer. The underlying philosophy in developing such technologies is to harness the natural physics of the computing elements to perform certain specialized computing tasks. One such beyond-CMOS computing paradigm- probabilistic computing is based on a "p-bit" that randomly fluctuates between 0 and 1, a behavior that is naturally mimicked by thermally unstable nanomagnets. A coupled network of such nanomagnets traverses through its collective states and is naturally guided towards the pre-designed low energy states. This property has been shown to be useful in providing hardware acceleration to a wide variety of problems in optimization, invertible logic, inference and machine learning. In order to develop practical circuits with p-bits, an efficient way to implement them in hardware by leveraging spintronics technology is required and forms the subject of this thesis. First, the experiments demonstrating the convergence of a weakly coupled nanomagnet network’s configuration towards the ground state of the associated Hamiltonian is shown. Next, it is demonstrated that by varying the interconnection strength and bias parameters in a two p-bit electrical circuit, Bayesian network building blocks can be implemented in hardware. Following this, a unique p-bit design based on the interaction of spin orbit torque on weak perpendicular anisotropy nanomagnets is presented and its interesting properties such as stochastic resonance, electrically tunable fluctuation rate and correlated fluctuations of two such devices are discussed. As related work, a prototype spin logic device is demonstrated using a composite stack of stable nanomagnets having perpendicular and in plane anisotropies. Finally, the development of a hybrid material stack with greatly improved giant spin Hall efficiency by incorporating WSe2 for energy efficient spin orbit torque switching of nanomagnets is presented.

Low barrier nanomagnets have attracted a lot of research interest for their use as sources of high quality true random number generation. More recently, low barrier nanomagnets with tunable output have been shown to be a natural hardware platform for unconventional computing paradigms such as probabilistic spin logic. Efficient generation and tunability of high quality random bits is critical for these novel applications. However, current spintronic random number generators are based on superparamagnetic tunnel junctions (SMTJs) with tunability obtained through spin transfer torque (STT), which unavoidably leads to challenges in designing concatenated networks using these two terminal devices. The more recent development of utilizing spin orbit torque (SOT) allows for a three terminal device design, but can only tune in-plane magnetization freely, which is not very energy efficient due to the needs of overcoming a large demagnetization field. In this work, we experimentally demonstrate for the first time, a stochastic device with perpendicular magnetic anisotropy (PMA) that is completely tunable by SOT without the aid of any external magnetic field. Our measurements lead us to hypothesize that a tilted anisotropy might be responsible for the observed tunability. We carry out stochastic Landau-Lifshitz-Gilbert (sLLG) simulations to confirm our experimental observation. Finally, we build an electrically coupled network of two such stochastic nanomagnet based devices and demonstrate that finite correlation or anti-correlation can be established between their output fluctuations by a weak interconnection, despite having a large difference in their natural fluctuation time scale. Simulations based on a newly developed dynamical model for autonomous circuits composed of low barrier nanomagnets show close agreement with the experimental results.

Probabilistic spin logic (PSL) is a new paradigm of computing that relies on probabilistic bits (p-bits) that fluctuate randomly between metastable states. PSL may be more efficient than conventional CMOS-based logic in terms of intrinsic optimization, Bayesian inference, invertible Boolean logic, and hardware machine learning. Effectively tunable random number generators, p-bits, can be realized as stochastic nanomagnets that can be made to prefer one state over others by an external input, such as voltage or current. This letter looks at the design of stochastic nanomagnets that is most suitable as p-bits for PSL. Experimental evidence, supported by theory and numerical simulation, shows that the scaling of magnetic anisotropy is more effective than the scaling of the net magnetic moment for voltage-driven PSL applications. A novel system that can be used as a tunable random number generator is demonstrated experimentally and analyzed theoretically: a magnet with perpendicular magnetic anisotropy that is initialized to its hard axis by giant spin Hall effect torque. With zero external input, this system provides a potentially better alternative to other nanomagnet-based random number generators. By tuning the randomness through an external input, this system is suitable for probabilistic networks for Bayesian inference.

In this work, we report the successful demonstration of electrical annealing and stochastic resonance in low barrier nanomagnets with weak perpendicular anisotropy. These novel phenomena are essential to realize oscillatory neural networks with dynamic connectivity, which provide a viable alternative to conventional neural network hardware with full connectivity by reducing the interconnection of synapse hardware from N2 weighted connections to N homogenous connections.

Spin based logic devices have attracted a lot of research interest due to their potential low-power operation, non-volatility and possibility to enable new computing applications. Here we present an experimental demonstration of a novel spin logic device working at room temperature without the requirement of an external magnetic field. Our device is based on a pair of coupled in-plane magnetic anisotropy (IMA) magnet and a perpendicular magnetic anisotropy (PMA) magnet. The information written in the state of the IMA magnet is transferred to the state of the PMA magnet by means of a symmetry breaking dipolar field, while the two layers are electrically isolated. In addition to having the basic tenets of a logic device, our device has inbuilt memory, taking advantage of the non-volatility of nanomagnets. In another mode of operation, the same device is shown to have the functionality of a true random number generator (TRNG). The combination of logic functionality, nonvolatility and capability to generate true random numbers all in the same spin logic device, makes it uniquely suitable as a hardware for many new computing ideas.

Hardware random number generators (RNGs) are specialized circuits that often come with large silicon footprints1. In this work, we experimentally demonstrate a RNG using a single low barrier magnet with perpendicular magnetic anisotropy (PMA) by utilizing its inherent stochasticity. More importantly, we show the tunability of its randomness by applying magnetic field or current, which makes this device an ideal candidate for probabilistic cormutinu-.

In this paper, we have proposed a computationally efficient method to evaluate threshold voltage (V T ) variability due to Line Edge Roughness (LER) in sub-20nm node FinFETs. For channel lengths less than 15 nm, the variability in threshold voltage may be estimated to a great accuracy (error T from LER specifications of a fin patterning technology.

Recently, dopant deactivation (DD) based resistance increase in Si nanowires surrounded by low-k insulator has been experimentally demonstrated. Source/drain extension confined by low-k spacer in scaled fins for sub-20nm FinFET are susceptible to such series resistance increase due to DD. In this paper, we implement DD into experimentally calibrated TCAD simulations of FinFETs and analyze the impact of DD on device performance with scaling for the first time. We show that the performance advantage with scaling is highly degraded in sub-20nm gate length node (e.g. at 8nm node the on-current degrades by more than 50% by adding DD correction). We further propose an optimal high-k spacer due to a DD mitigation vs. gate capacitance reduction trade-off (contrary to the present trend of spacer k-minimization based on only capacitance reduction considerations). We demonstrate that optimal higher-k spacer dielectric improves performance for technology nodes from 22nm to 8nm, e.g. optimized spacer can lead to 116% increase in ON current and 30% reduction in intrinsic delay for 8nm FinFET compared to low-k spacers. In fact, optimized high-k spacer can restore FinFET performance to DD-uncorrected prediction levels at 8nm node.