arXiv subjects

Nicolas Locatelli

Publications and source records attributed to Nicolas Locatelli.

17 recordsLinked to original sources

arXiv2019

Designing large arrays of interacting spin-torque nano-oscillators for microwave information processing

Arrays of spin-torque nano-oscillators are promising for broadband microwave signal detection and processing, as well as for neuromorphic computing. In many of these applications, the oscillators should be engineered to have equally-spaced frequencies and equal sensitivity to microwave inputs. Here we design spin-torque nano-oscillator arrays with these rules and estimate their optimum size for a given sensitivity, as well as the frequency range that they cover. For this purpose, we explore analytically and numerically conditions to obtain vortex spin-torque nano-oscillators with equally-spaced gyrotropic oscillation frequencies and having all similar synchronization bandwidths to input microwave signals. We show that arrays of hundreds of oscillators covering ranges of several hundred MHz can be built taking into account nanofabrication constraints.

physics.app-ph↗

arXiv2019

Implementing Binarized Neural Networks with Magnetoresistive RAM without Error Correction

One of the most exciting applications of Spin Torque Magnetoresistive Random Access Memory (ST-MRAM) is the in-memory implementation of deep neural networks, which could allow improving the energy efficiency of Artificial Intelligence by orders of magnitude with regards to its implementation on computers and graphics cards. In particular, ST-MRAM could be ideal for implementing Binarized Neural Networks (BNNs), a type of deep neural networks discovered in 2016, which can achieve state-of-the-art performance with a highly reduced memory footprint with regards to conventional artificial intelligence approaches. The challenge of ST-MRAM, however, is that it is prone to write errors and usually requires the use of error correction. In this work, we show that these bit errors can be tolerated by BNNs to an outstanding level, based on examples of image recognition tasks (MNIST, CIFAR-10 and ImageNet): bit error rates of ST-MRAM up to 0.1% have little impact on recognition accuracy. The requirements for ST-MRAM are therefore considerably relaxed for BNNs with regards to traditional applications. By consequence, we show that for BNNs, ST-MRAMs can be programmed with weak (low-energy) programming conditions, without error correcting codes. We show that this result can allow the use of low energy and low area ST-MRAM cells, and show that the energy savings at the system level can reach a factor two.

cs.ET↗

arXiv2018

Use of Magnetoresistive Random-Access Memory as Approximate Memory for Training Neural Networks

Hardware neural networks that implement synaptic weights with embedded non-volatile memory, such as spin torque memory (ST-MRAM), are a major lead for low energy artificial intelligence. In this work, we propose an approximate storage approach for their memory. We show that this strategy grants effective control of the bit error rate by modulating the programming pulse amplitude or duration. Accounting for the devices variability issue, we evaluate energy savings, and show how they translate when training a hardware neural network. On an image recognition example, 74% of programming energy can be saved by losing only 1% on the recognition performance.

cs.ET↗

arXiv2018

Vowel recognition with four coupled spin-torque nano-oscillators

Substantial evidence indicates that the brain uses principles of non-linear dynamics in neural processes, providing inspiration for computing with nanoelectronic devices. However, training neural networks composed of dynamical nanodevices requires finely controlling and tuning their coupled oscillations. In this work, we show that the outstanding tunability of spintronic nano-oscillators can solve this challenge. We successfully train a hardware network of four spin-torque nano-oscillators to recognize spoken vowels by tuning their frequencies according to an automatic real-time learning rule. We show that the high experimental recognition rates stem from the high frequency tunability of the oscillators and their mutual coupling. Our results demonstrate that non-trivial pattern classification tasks can be achieved with small hardware neural networks by endowing them with non-linear dynamical features: here, oscillations and synchronization. This demonstration is a milestone for spintronics-based neuromorphic computing.

q-bio.NC↗

arXiv2018

Nano-oscillator-based classification with a machine learning-compatible architecture

Pattern classification architectures leveraging the physics of coupled nano-oscillators have been demonstrated as promising alternative computing approaches, but lack effective learning algorithms. In this work, we propose a nano-oscillator based classification architecture where the natural frequencies of the oscillators are learned linear combinations of the inputs, and define an offline learning algorithm based on gradient back-propagation. Our results show significant classification improvements over a related approach with online learning. We also compare our architecture with a standard neural network on a simple machine learning case, which suggests that our approach is economical in terms of numbers of adjustable parameters. The introduced architecture is also compatible with existing nano-technologies: the architecture does not require changes in the coupling between nano-oscillators, and it is tolerant to oscillator phase noise.

physics.app-ph↗

arXiv2018

Circuit-Level Evaluation of the Generation of Truly Random Bits with Superparamagnetic Tunnel Junctions

Many emerging alternative models of computation require massive numbers of random bits, but their generation at low energy is currently a challenge. The superparamagnetic tunnel junction, a spintronic device based on the same technology as spin torque magnetoresistive random access memory has recently been proposed as a solution, as this device naturally switches between two easy to measure resistance states, due only to thermal noise. Reading the state of the junction naturally provides random bits, without the need of write operations. In this work, we evaluate a circuit solution for reading the state of superparamagnetic tunnel junction. We see that the circuit may induce a small read disturb effect for scaled superparamagnetic tunnel junctions, but this effect is naturally corrected in the whitening process needed to ensure the quality of the generated random bits. These results suggest that superparamagnetic tunnel junctions could generate truly random bits at 20 fJ/bit, including overheads, orders of magnitudes below CMOS-based solutions.

physics.app-ph↗

arXiv2017

Low-Energy Truly Random Number Generation with Superparamagnetic Tunnel Junctions for Unconventional Computing

Low-energy random number generation is critical for many emerging computing schemes proposed to complement or replace von Neumann architectures. However, current random number generators are always associated with an energy cost that is prohibitive for these computing schemes. In this paper, we introduce random number bit generation based on specific nanodevices: superparamagnetic tunnel junctions. We experimentally demonstrate high quality random bit generation that represents orders-of-magnitude improvements in energy efficiency compared to current solutions. We show that the random generation speed improves with nanodevice scaling, and investigate the impact of temperature, magnetic field and crosstalk. Finally, we show how alternative computing schemes can be implemented using superparamagentic tunnel junctions as random number generators. These results open the way for fabricating efficient hardware computing devices leveraging stochasticity, and highlight a novel use for emerging nanodevices.

physics.app-ph↗

arXiv2017

A Neural Network Based on Synchronized Pairs of Nano-Oscillators

Artificial neural networks are intensively used to perform cognitive tasks such as image classification on traditional computers. With the end of CMOS scaling and increasing demand for efficient neural networks, alternative architectures implementing neural functions efficiently are being studied. This study leverages the demonstrated frequency tuning capabilities of compact nano-oscillators and their synchronization dynamics to implement a neuron using a pair of synchronized oscillators, and which features an unconventional response curve. We show that this compact neuron can naturally implement generic logic gates, including XOR. A simulated oscillator-based neural network is then shown to achieve results equivalent to standard approaches on two reference classification tasks. Finally, the performance of the system is evaluated in the presence of oscillator phase noise, an important issue of oscillating nanodevices. These results open the way for the design of alternative architectures adapted to efficient neural network execution.

cs.ET↗

arXiv2016

Role of spin-transfer torques on synchronization and resonance phenomena in stochastic magnetic oscillators

A theoretical study on how synchronization and resonance-like phenomena in superparamagnetic tunnel junctions can be driven by spin-transfer torques is presented. We examine the magnetization of a superparamagnetic free layer that reverses randomly between two well-defined orientations due to thermal fluctuations, acting as a stochastic oscillator. When subject to an external ac forcing this system can present stochastic resonance and noise-enhanced synchronization. We focus on the roles of the mutually perpendicular damping-like and field-like torques, showing that the response of the system is very different at low and high-frequencies. We also demonstrate that the field-like torque can increase the efficiency of the current-driven forcing, specially at sub-threshold electric currents. These results can be useful for possible low-power, more energy efficient, applications.

cond-mat.mes-hall↗

arXiv2016

Synchronization Detection in Networks of Coupled Oscillators for Pattern Recognition

Coupled oscillator-based networks are an attractive approach for implementing hardware neural networks based on emerging nanotechnologies. However, the readout of the state of a coupled oscillator network is a difficult challenge in hardware implementations, as it necessitates complex signal processing to evaluate the degree of synchronization between oscillators, possibly more complicated than the coupled oscillator network itself. In this work, we focus on a coupled oscillator network particularly adapted to emerging technologies, and evaluate two schemes for reading synchronization patterns that can be readily implemented with basic CMOS circuits. Through simulation of a simple generic coupled oscillator network, we compare the operation of these readout techniques with a previously proposed full statistics evaluation scheme. Our approaches provide results nearly identical to the mathematical method, but also show better resilience to moderate noise, which is a major concern for hardware implementations. These results open the door to widespread realization of hardware coupled oscillator-based neural systems.

cs.ET↗

arXiv2015

Improved spectral stability in spin transfer nano-oscillators: single vortex versus coupled vortices dynamics

We perform a comparative study of spin transfer induced excitation of the gyrotropic motion of a vortex core with either uniform or vortex spin polarizers. The microwave output voltage associated with the vortex dynamics, detected in both cases, displays a strong reduction of phase fluctuations in the case of the vortex polarizer, with a decrease of the peak linewidth by one order of magnitude down to 200kHz at zero field. A thorough study of rf emission features for the different accessible vortex configurations shows that this improvement is related to the excitation of coupled vortex dynamics by spin transfer torques.

cond-mat.mes-hall↗

arXiv2014

Nonlinear behavior and mode coupling in spin transfer nano-oscillators

By investigating thoroughly the tunable behavior of coupled modes, we highlight how it provides new means to handle the properties of spin transfer nano-oscillators. We first demonstrate that the main features of the microwave signal associated with coupled vortex dynamics i.e. frequency, spectral coherence, critical current, mode localization, depends drastically on the relative vortex core polarities. Secondly we report a large reduction of the nonlinear linewidth broadening obtained by changing the effective damping through the control of the core configuration. Such a level of control on the nonlinear behavior reinforces our choice to exploit the microwave properties of collective modes for applications of spintronic devices in novel generation of integrated telecommunication devices.

cond-mat.mes-hall↗

arXiv2014

Noise-enhanced synchronization of stochastic magnetic oscillators

We present an experimental study of phase-locking in a stochastic magnetic oscillator. The system comprises a magnetic tunnel junction with a superparamagnetic free layer, whose magnetization dynamics is driven with spin torques through an external periodic driving current. We show that synchronization of this stochastic oscillator to the input current is possible for current densities below $3 \times 10^6$ A/cm$^2$, and occurs for input frequencies lower than the natural mean frequency of the stochastic oscillator. We show that such injection-locking is robust and leads to a drastic reduction in the phase diffusion of the stochastic oscillator, despite the presence of a frequency mismatch between the oscillator and the excitation.

cond-mat.mtrl-sci↗

arXiv2014

Spin torque building blocks

The discovery of the spin torque effect has made magnetic nanodevices realistic candidates for active elements of memory devices and applications. Magnetoresistive effects allow the read-out of increasingly small magnetic bits, and the spin torque provides an efficient tool to manipulate - precisely, rapidly and at low energy cost - the magnetic state, which is in turn the central information medium of spintronic devices. By keeping the same magnetic stack, but by tuning a device's shape and bias conditions, the spin torque can be engineered to build a variety of advanced magnetic nanodevices. Here we show that by assembling these nanodevices as building blocks with different functionalities, novel types of computing architectures can be envisisaged. We focus in particular on recent concepts such as magnonics and spintronic neural networks.

cond-mat.mtrl-sci↗

arXiv2013

Perfect and robust phase-locking of a spin transfer vortex nano-oscillator to an external microwave source

We study the synchronization of the auto-oscillation signal generated by the spin transfer driven dynamics of two coupled vortices in a spin-valve nanopillar to an external source. Phase-locking to the microwave field hrf occurs in a range larger than 10% of the oscillator frequency for drive amplitudes of only a few Oersteds. Using synchronization at the double frequency, the generation linewidth is found to decrease by more than five orders of magnitude in the phase-locked regime (down to 1 Hz, limited by the resolution bandwidth of the spectrum analyzer) in comparison to the free running regime (140 kHz). This perfect phase-locking holds for frequency detuning as large as 2 MHz, which proves its robustness. We also analyze how the free running spectral linewidth impacts the main characteristics of the synchronization regime.

cond-mat.mes-hall↗

arXiv2013

Origin of spectral purity and tuning sensitivity in a vortex-based spin transfer nano-oscillator

We investigate the microwave characteristics of a spin transfer nano-oscillator (STNO) based on coupled vortices as a function of the perpendicular magnetic field $H_\perp$. While the generation linewidth displays strong variations on $H_\perp$ (from 40 kHz to 1 MHz), the frequency tunability in current remains almost constant (~7 MHz/mA). We demonstrate that our vortex-based oscillator is quasi-isochronous independently of $H_\perp$, so that the severe nonlinear broadening usually observed in STNOs does not exist. Interestingly, this does not imply a loss of frequency tunability, which is here governed by the current induced Oersted field. Nevertheless this is not sufficient to achieve the highest spectral purity in the full range of $H_\perp$ either: we show that the observed linewidth broadenings are due to the excited mode interacting with a lower energy overdamped mode, which occurs at the successive crossings between harmonics of these two modes. These findings open new possibilities for the design of STNOs and the optimization of their performance.

cond-mat.mtrl-sci↗

arXiv2011

Identification and selection rules of the spin-wave eigen-modes in a normally magnetized nano-pillar

We report on a spectroscopic study of the spin-wave eigen-modes inside an individual normally magnetized two layers circular nano-pillar (Permalloy$|$Copper$|$Permalloy) by means of a Magnetic Resonance Force Microscope (MRFM). We demonstrate that the observed spin-wave spectrum critically depends on the method of excitation. While the spatially uniform radio-frequency (RF) magnetic field excites only the axially symmetric modes having azimuthal index $\ell=0$, the RF current flowing through the nano-pillar, creating a circular RF Oersted field, excites only the modes having azimuthal index $\ell=+1$. Breaking the axial symmetry of the nano-pillar, either by tilting the bias magnetic field or by making the pillar shape elliptical, mixes different $\ell$-index symmetries, which can be excited simultaneously by the RF current. Experimental spectra are compared to theoretical prediction using both analytical and numerical calculations. An analysis of the influence of the static and dynamic dipolar coupling between the nano-pillar magnetic layers on the mode spectrum is performed.

cond-mat.mtrl-sci↗