arXiv subjects

Supriyo Bandyopadhyay

Publications and source records attributed to Supriyo Bandyopadhyay.

At least 37 records · Page 2Linked to original sources

arXiv2023

Increasing Flips per Second and Speed of p-Computers by Using Dilute Magnetic Semiconductors to Implement Binary Stochastic Neurons

Probabilistic computing with binary stochastic neurons (BSN) implemented with low- or zero-energy barrier nanoscale ferromagnets (LBMs) possessing in-plane magnetic anisotropy has emerged as an efficient paradigm for solving computationally hard problems. The fluctuating magnetization of an LBM at room temperature encodes a p-bit which is the building block of a BSN. Its only drawback is that the dynamics of common (transition metal) ferromagnets are relatively slow and hence the number of uncorrelated p-bits that can be generated per second - the so-called "flips per second" (fps) - is insufficient, leading to slow computational speed in autonomous co-processing with p-computers. Here, we show that a simple way to increase fps is to replace commonly used ferromagnets (e.g. Co, Fe, Ni), which have large saturation magnetization Ms, with a dilute magnetic semiconductor like GaMnAs with much smaller saturation magnetization. The smaller Ms reduces the energy barrier within the LBM and increases the fps significantly. It also offers other benefits such as increased packing density for increased parallelization and reduced device to device variation. This provides a way to realize the hardware acceleration and energy efficiency promise of p-computers.

cond-mat.mes-hall↗

arXiv2023

Strained topological insulator spin field effect transistor

The notion of a spin field effect transistor, where transistor action is realized by manipulating the spin degree of freedom of charge carriers instead of the charge degree of freedom, has captivated researchers for at least three decades. These transistors are usually implemented by modulating the spin orbit interaction in a two- or one-dimensional semiconductor structure with an electrostatic potential, which then causes controlled spin precession in the transistor's channel that modulates the current flowing between two ferromagnetic (spin-polarized) source and drain contacts. Here, we introduce a new concept for a spin field effect transistor whose channel is made of a strained topological insulator (strained-topological-insulator-field-effect-transistor or STI-SPINFET), which does not exploit spin-orbit interaction. Instead, the transistor function is elicited by straining the topological insulator (TI) with a gate voltage which modifies the energy dispersion relation, or the Dirac velocity, to vary the interference between the two spin eigenstates on the surface of the TI. This modulates the current flowing between two ferromagnetic source and drain contacts. The conductance on/off ratio of this transistor is too poor to be useful as a switch, but it may have other uses, such as an extremely energy-efficient stand-alone frequency multiplier.

cond-mat.mes-hall↗

arXiv2023

The Strong Sensitivity of the Characteristics of Binary Stochastic Neurons Employing Low Barrier Nanomagnets to Small Geometrical Variations

Binary stochastic neurons (BSNs) are excellent activators for machine learning. An ideal platform for implementing them are low- or zero-energy-barrier nanomagnets (LBMs) possessing in-plane anisotropy (e.g. circular or slightly elliptical disks) whose fluctuating magnetization encodes a probabilistic (p-) bit. Here, we show that such a BSN's activation function, the pinning current (which pins the output to a particular binary state), and the response time - all exhibit strong sensitivity to very slight geometric variations in the LBM's cross section. A mere 1% change in the diameter of a circular nanomagnet in any arbitrary direction can alter the response time by a factor of ~4 at room temperature and a 10% change can alter the pinning current by a factor of ~2. All this causes large device-to-device variation which is detrimental to integration. We also show that the energy dissipation is lowered but the response time is increased by replacing a circular cross-section with a slightly elliptical one and then encoding the p-bit in the magnetization component along the major axis. Encoding the p-bit in the magnetization component along the minor axis has the opposite effect. The energy-delay-product, however, is relatively independent of whether the cross-section is a circle or an ellipse and which magnetization component encodes the p-bit in the case of the ellipse.

cond-mat.mes-hall↗

arXiv2022

A Non-Volatile All-Spin Non-Binary Matrix Multiplier: An Efficient Hardware Accelerator for Machine Learning

We propose and analyze a compact and non-volatile nanomagnetic (all-spin) non-binary matrix multiplier performing the multiply-and-accumulate (MAC) operation using two magnetic tunnel junctions - one activated by strain to act as the multiplier, and the other activated by spin-orbit torque pulses to act as a domain wall synapse that performs the operation of the accumulator. It has two advantages over the usual crossbar-based electronic non-binary matrix multiplier. First, while the crossbar architecture requires N3 devices to multiply two matrices, we require only 2N2 devices. Second, our matrix multiplier is non-volatile and retains the information about the product matrix after being powered off. Here, we present an example where each MAC operation can be performed in ~5 ns and the maximum energy dissipated per operation is ~60Nmax aJ, where Nmax is the largest matrix size. This provides a very useful hardware accelerator for machine learning and artificial intelligence tasks which involve the multiplication of large matrices. The non-volatility allows the matrix multiplier to be embedded in powerful non-von-Neumann architectures, including processor-in-memory. It also allows much of the computing to be done at the edge (of internet-of-things) while reducing the need to access the cloud, thereby making artificial intelligence more resilient against cyberattacks.

cs.ET↗

arXiv2022

Robustness of binary stochastic neurons implemented with low barrier nanomagnets made of dilute magnetic semiconductors

Binary stochastic neurons (BSNs) are excellent hardware accelerators for machine learning. A popular platform for implementing them are low- or zero-energy barrier nanomagnets possessing in-plane magnetic anisotropy (e.g. circular disks or quasi-elliptical disks with very small eccentricity). Unfortunately, small geometric variations in the lateral shapes of such nanomagnets can produce large changes in the BSN response times if the nanomagnets are made of common metallic ferromagnets (Co, Ni, Fe) with large saturation magnetization. Additionally, the response times are also very sensitive to initial conditions. Here, we show that if the nanomagnets are made of dilute magnetic semiconductors with much smaller saturation magnetization, then the variability in their response times (due to shape variations and variation in the initial condition) is drastically suppressed. This significantly reduces the device-to-device variation, which is a serious challenge for large scale neuromorphic systems.

cond-mat.mes-hall↗

arXiv2022

A spintronic analog of the Landauer residual resistivity dipole on the surface of a disordered topological insulator

The Landauer "residual resistivity dipole" is a well-known concept in electron transport through a disordered medium. It is formed when a defect/scatterer reflects an impinging electron causing negative charges to build up on one side of the scatterer and positive charges on the other. This results in the formation of a microscopic electric dipole that affects the resistivity of the medium. Here, we show that an equivalent entity forms in spin polarized electron transport through the surface of a disordered topological insulator (TI). When electrons reflect from a scatterer on the TI surface, a spin imbalance forms around the scatterer, resulting in a spin current that flows either in the same or the opposite direction as the injected spin current and hence either increases or decreases the spin resistivity. It also destroys spin-momentum locking and produces a magnetic field around the scatterer. The latter will cause transiting spins to precess as they pass the scatterer, thereby radiating electromagnetic waves and implementing an oscillator. If an alternating current is passed through the TI instead of a static current, the magnetic field will oscillate with the frequency of the current and radiate electromagnetic waves of the same frequency, thus making the scatterer act as a miniature antenna.

cond-mat.mes-hall↗

arXiv2022

Comment on "Using Dipole Interaction to Achieve Nonvolatile Voltage Control of Magnetism in Multiferroic Heterostructures" [Adv. Mater., 2021, 2105902]

In a recent article, A. Chen et al. claimed that they have switched (rotated by ~90 degrees) the magnetization of the soft layer of a magnetic tunnel junction (MTJ) in a non-volatile way with volatile strain by exploiting dipole interaction with the hard layer. I show that dipole interaction cannot cause this effect and the authors' explanation cannot be correct. An alternate explanation is offered.

cond-mat.mes-hall↗

arXiv2022

Spin Wave Electromagnetic Nano-Antenna Enabled by Tripartite Phonon-Magnon-Photon Coupling

We investigate tripartite coupling between phonons, magnons and photons in a periodic array of elliptical magnetostrictive nanomagnets delineated on a piezoelectric substrate to form a two-dimensional two-phase multiferroic crystal. A surface acoustic wave (phonons) of 5 - 35 GHz frequency launched into the substrate causes the magnetizations of the nanomagnets to precess at the frequency of the wave, giving rise to spin waves (magnons). The spin waves, in turn, radiate electromagnetic waves (photons) into the surrounding space at the surface acoustic wave frequency. Here, the phonons couple into magnons, which then couple into photons. This tripartite phonon-magnon-photon coupling is exploited to implement an extreme sub-wavelength electromagnetic antenna whose measured radiation efficiency and antenna gain exceed the theoretical limits for traditional antennas by more than two orders of magnitude at some frequencies. Micro-magnetic simulations are in excellent agreement with experimental observations and provide insight into the spin wave modes that couple into radiating electromagnetic modes to implement the antenna.

cond-mat.mes-hall↗

arXiv2021

Straintronics: Manipulating the Magnetization of Magnetostrictive Nanomagnets with Strain for Energy-Efficient Applications

The desire to perform information processing, computation, communication, signal generation and related tasks, while dissipating as little energy as possible, has inspired many ideas and paradigms. One of the most powerful among them is the notion of using magnetostrictive nanomagnets as the primitive units of the hardware platforms and manipulating their magnetizations with electrically generated static or time varying mechanical strain to elicit myriad functionalities. This approach has two advantages. First, information can be retained in the devices after powering off since the nanomagnets are non-volatile unlike charge-based devices such as transistors. Second, the energy expended to perform a given task is exceptionally low since it takes very little energy to alter magnetization states with strain. This field is now known as "straintronics", in analogy with electronics, spintronics, valleytronics, etc. We review the recent advances and trends in straintronics, including digital information processing (logic), information storage (memory), domain wall devices operated with strain, control of skyrmions with strain, non-Boolean computing and machine learning with straintronics, signal generation (microwave sources) and communication (ultra-miniaturized acoustic and electromagnetic antennas) implemented with strained nanomagnets, hybrid straintronics-magnonics, and interaction between phonons and magnons in straintronic systems. We identify key challenges and opportunities, and lay out pathways to advance this field to the point where it might become a mainstream technology for energy-efficient systems.

cond-mat.mes-hall↗

arXiv2021

An observable effect of spin inertia in slow magneto-dynamics: Increase of the switching error rates in nanoscale ferromagnets

The Landau-Lifshitz-Gilbert (LLG) equation, used to model magneto-dynamics in ferromagnets, tacitly assumes that the angular momentum associated with spin precession can relax instantaneously when the real or effective magnetic field causing the precession is turned off. This neglect of "spin inertia" is unphysical and would violate energy conservation. Recently, the LLG equation was modified to account for inertia effects. The consensus, however, seems to be that such effects would be unimportant in slow magneto-dynamics that take place over time scales much longer that the relaxation time of the angular momentum, which is typically few fs to perhaps ~100 ps in ferromagnets. Here, we show that there is at least one very serious and observable effect of spin inertia even in slow magneto-dynamics. It involves the switching error probability associated with flipping the magnetization of a nanoscale ferromagnet with an external agent, such as a magnetic field. The switching may take ~ns to complete when the field strength is close to the threshold value for switching, which is much longer than the angular momentum relaxation time, and yet the effect of spin inertia is felt in the switching error probability. This is because the ultimate fate of a switching trajectory, i.e. whether it results in success or failure, is influenced by what happens in the first few ps of the switching action when nutational dynamics due to spin inertia holds sway. Spin inertia increases the error probability, which makes the switching more error-prone. This has vital technological significance because it relates to the reliability of magnetic logic and memory.

cond-mat.mes-hall↗

arXiv2021

Reflection and refraction of an electron spin at the junction between two two-dimensional regions with and without spin-orbit interaction

We derive the reflection and refraction laws for an electron spin incident from a quasi-two-dimensional medium with no spin-orbit interaction on another with both Rashba and Dresselhaus spin-orbit interaction using only energy conservation. We obtain the well-known result that for an incident angle, there can be generally two different refraction angles for refraction into the two spin eigenstates in the refraction medium, resulting in two different 'spin refractive indices' and two critical angles for total internal reflection. We derive expressions for the spin refractive indices, which are not constant for a given medium but depend on the incident electron's energy. If the effective mass of an electron in the refraction medium is larger than that in the incidence medium, then we show that for some incident electron energies and potential barrier at the interface, the spin refractive index of the incidence medium can lie between the two spin refractive indices of the refraction medium, resulting in only one critical angle. In that case, if the incident angle exceeds that critical angle, then refraction can occur into only one spin eigenstate in the refraction medium. If the system is engineered to make this happen, then it will be possible to obtain a very high degree of spin-polarized injection into the refraction medium. The amplitudes of reflection of the incident spin into its own spin eigenstate and the orthogonal spin eigenstate (due to spin flip at the interface), as well as the refraction amplitudes into the two spin eigenstates in the refraction medium are derived for an incident electron (with arbitrary spin polarization and incident energy) as a function of the angle of incidence.

cond-mat.mes-hall↗

arXiv2021

Surface Acoustic Wave induced modulation of tunneling magnetoresistance in magnetic tunnel junctions

We show that a surface acoustic wave (SAW) applied across the terminals of a magnetic tunnel junction (MTJ) decreases both the (time-averaged) parallel and antiparallel resistances of the MTJ, with the latter decreasing much more than the former. This results in a decrease of the tunneling magnetoresistance (TMR) ratio. The coercivities of the free and fixed layer of the MTJ, however, are not affected significantly, suggesting that the SAW does not cause large-angle magnetization rotation in the magnetic layers through the inverse magnetostriction (Villari) effect at the power levels used. This study sheds light on the dynamical behavior of an MTJ under periodic compressive and tensile strain.

cond-mat.mes-hall↗

arXiv2021

Applications of Nanomagnets as Dynamical Systems

When magnets are fashioned into nanoscale elements, they exhibit a wide variety of phenomena replete with rich physics and the lure of tantalizing applications. In this topical review, we discuss some of these phenomena, especially those that have come to light recently, and highlight their potential applications. We emphasize what drives a phenomenon, what undergirds the dynamics of the system that exhibits the phenomenon, how the dynamics can be manipulated, and what specific features can be harnessed for technological advances. For the sake of balance, we point out both advantages and shortcomings of nanomagnet based devices and systems predicated on the phenomena we discuss. Where possible, we chart out paths for future investigations that can shed new light on an intriguing phenomenon and/or facilitate both traditional and non-traditional applications.

cond-mat.mes-hall↗

arXiv2021

Nanomagnetic Boolean Logic -- The Tempered (and Realistic) Vision

The idea of nanomagnetic Boolean logic was advanced more than two decades ago. It envisaged the use of nanomagnets with two stable magnetization orientations as the primitive binary switch for implementing logic gates and ultimately combinational/sequential circuits. Enthusiastic proclamations of how nanomagnetic logic will eclipse traditional (transistor-based) logic circuits proliferated the applied physics literature. Two decades later there is not a single viable nanomagnetic logic chip in sight, let alone one that is a commercial success. In this perspective article, I offer my reasons on why this has come to pass. I present a realistic and tempered vision of nanomagnetic logic, pointing out many misconceptions about this paradigm, flaws in some proposals that appeared in the literature, shortcomings, and likely pitfalls that might stymie progress in this field.

cond-mat.mes-hall↗

arXiv2020

Experimental Demonstration of an Extreme Sub-Wavelength Nanomagnetic Acoustic Antenna Actuated by Spin-Orbit Torque from a Heavy Metal Nanostrip

A novel on-chip extreme sub-wavelength "acoustic antenna" whose radiation efficiency is ~50 times larger than the theoretical limit for a resonantly driven antenna is demonstrated. The antenna is composed of magnetostrictive nanomagnets deposited on a piezoelectric substrate. The nanomagnets are partially in contact with a heavy metal (Pt) nanostrip. Passage of alternating current through the nanostrip exerts alternating spin-orbit torque on the nanomagnets and periodically rotates their magnetizations. During the rotation, the magnetostrictive nanomagnets expand and contract, thereby setting up alternating tensile and compressive strain in the piezoelectric substrate underneath. This leads to the generation of a surface acoustic wave in the substrate and makes the nanomagnet assembly act as an acoustic antenna. The measured radiation efficiency of this acoustic antenna at the detected frequency is ~1%, while the wavelength to antenna dimension ratio is ~ 67:1. For a standard antenna driven at acoustic resonance, the efficiency would have been limited to ~ (1/67)^2 = 0.02%. It was possible to beat that limit (by ~50 times) via actuating the antenna not by acoustic resonance, but by using a completely different mechanism involving spin-orbit torque originating from the giant spin Hall effect in Pt.

physics.app-ph↗

arXiv2020

The effect of material defects on resonant spin wave modes in a nanomagnet

We have theoretically studied how resonant spin wave modes in an elliptical nanomagnet are affected by fabrication defects, such as small local thickness variations. Our results indicate that defects of this nature, which can easily result from the fabrication process, or are sometimes deliberately introduced during the fabrication process, will significantly alter the frequencies, magnetic field dependence of the frequencies, and the power and phase profiles of the resonant spin wave modes. They can also spawn new resonant modes and quench existing ones. All this has important ramifications for multi-device circuits based on spin waves, such as phase locked oscillators for neuromorphic computing, where the device-to-device variability caused by defects can be inhibitory.

cond-mat.mes-hall↗

arXiv2019

Robustness and scalability of p-bits implemented with low energy barrier nanomagnets

Probabilistic (p-) bits implemented with low energy barrier nanomagnets (LBMs) have recently gained attention because they can be leveraged to perform some computational tasks very efficiently. Although more error-resilient than Boolean computing, p-bit based computing employing LBMs is, however, not completely immune to defects and device-to-device variations. In some tasks (e.g. binary stochastic neurons for machine learning and p-bits for population coding), extended defects, such as variation of the LBM thickness over a significant fraction of the surface, can impair functionality. In this paper, we have examined if unavoidable geometric device-to-device variations can have a significant effect on one of the most critical requirements for probabilistic computing, namely the ability to "program" probability with an external agent, such as a spin-polarized current injected into the LBM. We found that the programming ability is fortunately not lost due to reasonable device-to-device variations. The little variation in the probability versus current characteristic that reasonable device variability causes can be suppressed further by increasing the spin polarization of the current. This shows that probabilistic computing with LBMs is robust against small geometric variations, and hence will be "scalable" to a large number of p-bits.

cond-mat.mes-hall↗

arXiv2019

Error Probability in Magneto-elastic Switching of Non-ideal Nanomagnets with Defects: A Case Study for the Viability of Straintronic Logic and Memory

Magneto-elastic (straintronic) switching of bistable magnetostrictive nanomagnets is an extremely energy-efficient switching methodology for (magnetic) binary switches that has recently attracted widespread attention because of its potential application in ultra-low-power digital computing hardware. Unfortunately, this modality of switching is also error very prone at room temperature. Theoretical studies of switching error probability of magneto-elastic switches have predicted probabilities ranging from 10E-8-10E-3 at room temperature for ideal, defect-free nanomagnets, but experiments with real nanomagnets show a much higher probability that exceeds 0.1 in some cases. The obvious spoilers that can cause this large difference are defects and non-idealities. Here, we have theoretically studied the effect of common defects (that occur during fabrication) on magneto-elastic switching probability in the presence of room-temperature thermal noise. Surprisingly, we found that even small defects increase the switching error probabilities by orders of magnitude. There is usually a critical stress that leads to the lowest error probability and its value increases enormously in the presence of defects. All this could limit or preclude the application of magneto-elastic (straintronic) binary switches in either Boolean logic or memory, despite their excellent energy-efficiency, and restrict them to non-Boolean (e.g. neuromorphic, stochastic) computing applications. We also studied the difference between magneto-elastic switching with a stress pulse of constant amplitude and sinusoidal time-varying amplitude (e.g. due to a surface acoustic wave) and found that the latter method is more reliable and generates lower switching error probabilities in most cases, provided the time variation is reasonably slow.

cond-mat.mes-hall↗