arXiv subjects

Neil Smith

Publications and source records attributed to Neil Smith.

17 recordsLinked to original sources

arXiv2020

Ultrathin perpendicular free layers for lowering the switching current in STT-MRAM

The critical current density $J_{c0}$ required for switching the magnetization of the free layer (FL) in a spin-transfer torque magnetic random access memory (STT-MRAM) cell is proportional to the product of the damping parameter, saturation magnetization and thickness of the free layer, $\alpha M_S t_F$. Conventional FLs have the structure CoFeB/nonmagnetic spacer/CoFeB. By reducing the spacer thickness, W in our case, and also splitting the single W layer into two layers of sub-monolayer thickness, we have reduced $t_F$ while minimizing $\alpha$ and maximizing $M_S$, ultimately leading to lower $J_{c0}$ while maintaining high thermal stability. Bottom-pinned MRAM cells with device diameter in the range of 55-130 nm were fabricated, and $J_{c0}$ is lowest for the thinnest (1.2 nm) FLs, down to 4 MA/cm$^2$ for 65 nm devices, $\sim$30% lower than 1.7 nm FLs. The thermal stability factor $\Delta_{\mathrm{dw}}$, as high as 150 for the smallest device size, was determined using a domain wall reversal model from field switching probability measurements. With high $\Delta_{\mathrm{dw}}$ and lowest $J_{c0}$, the thinnest FLs have the highest spin-transfer torque efficiency.

cond-mat.mtrl-sci↗

arXiv2019

Origin of the resistance-area product dependence of spin transfer torque switching in perpendicular magnetic random access memory cells

We report on an experimental study of current induced switching in perpendicular magnetic random access memory (MRAM) cells with variable resistance-area products (RAs). Our results show that in addition to spin transfer torque (STT), current induced self-heating and voltage controlled magnetic anisotropy also contribute to switching and can explain the RA dependencies of switching current density and STT efficiency. Our findings suggest that thermal optimization of perpendicular MRAM cells can result in significant reduction of switching currents.

cond-mat.mes-hall↗

arXiv2019

Learning a Controller Fusion Network by Online Trajectory Filtering for Vision-based UAV Racing

Autonomous UAV racing has recently emerged as an interesting research problem. The dream is to beat humans in this new fast-paced sport. A common approach is to learn an end-to-end policy that directly predicts controls from raw images by imitating an expert. However, such a policy is limited by the expert it imitates and scaling to other environments and vehicle dynamics is difficult. One approach to overcome the drawbacks of an end-to-end policy is to train a network only on the perception task and handle control with a PID or MPC controller. However, a single controller must be extensively tuned and cannot usually cover the whole state space. In this paper, we propose learning an optimized controller using a DNN that fuses multiple controllers. The network learns a robust controller with online trajectory filtering, which suppresses noisy trajectories and imperfections of individual controllers. The result is a network that is able to learn a good fusion of filtered trajectories from different controllers leading to significant improvements in overall performance. We compare our trained network to controllers it has learned from, end-to-end baselines and human pilots in a realistic simulation; our network beats all baselines in extensive experiments and approaches the performance of a professional human pilot. A video summarizing this work is available at https://youtu.be/hGKlE5X9Z5U

cs.RO↗

arXiv2018

Latent Filter Scaling for Multimodal Unsupervised Image-to-Image Translation

In multimodal unsupervised image-to-image translation tasks, the goal is to translate an image from the source domain to many images in the target domain. We present a simple method that produces higher quality images than current state-of-the-art while maintaining the same amount of multimodal diversity. Previous methods follow the unconditional approach of trying to map the latent code directly to a full-size image. This leads to complicated network architectures with several introduced hyperparameters to tune. By treating the latent code as a modifier of the convolutional filters, we produce multimodal output while maintaining the traditional Generative Adversarial Network (GAN) loss and without additional hyperparameters. The only tuning required by our method controls the tradeoff between variability and quality of generated images. Furthermore, we achieve disentanglement between source domain content and target domain style for free as a by-product of our formulation. We perform qualitative and quantitative experiments showing the advantages of our method compared with the state-of-the art on multiple benchmark image-to-image translation datasets.

cs.CV↗

arXiv2018

OIL: Observational Imitation Learning

Recent work has explored the problem of autonomous navigation by imitating a teacher and learning an end-to-end policy, which directly predicts controls from raw images. However, these approaches tend to be sensitive to mistakes by the teacher and do not scale well to other environments or vehicles. To this end, we propose Observational Imitation Learning (OIL), a novel imitation learning variant that supports online training and automatic selection of optimal behavior by observing multiple imperfect teachers. We apply our proposed methodology to the challenging problems of autonomous driving and UAV racing. For both tasks, we utilize the Sim4CV simulator that enables the generation of large amounts of synthetic training data and also allows for online learning and evaluation. We train a perception network to predict waypoints from raw image data and use OIL to train another network to predict controls from these waypoints. Extensive experiments demonstrate that our trained network outperforms its teachers, conventional imitation learning (IL) and reinforcement learning (RL) baselines and even humans in simulation. The project website is available at https://sites.google.com/kaust.edu.sa/oil/ and a video at https://youtu.be/_rhq8a0qgeg

cs.CV↗

arXiv2017

Teaching UAVs to Race: End-to-End Regression of Agile Controls in Simulation

Automating the navigation of unmanned aerial vehicles (UAVs) in diverse scenarios has gained much attention in recent years. However, teaching UAVs to fly in challenging environments remains an unsolved problem, mainly due to the lack of training data. In this paper, we train a deep neural network to predict UAV controls from raw image data for the task of autonomous UAV racing in a photo-realistic simulation. Training is done through imitation learning with data augmentation to allow for the correction of navigation mistakes. Extensive experiments demonstrate that our trained network (when sufficient data augmentation is used) outperforms state-of-the-art methods and flies more consistently than many human pilots. Additionally, we show that our optimized network architecture can run in real-time on embedded hardware, allowing for efficient on-board processing critical for real-world deployment. From a broader perspective, our results underline the importance of extensive data augmentation techniques to improve robustness in end-to-end learning setups.

cs.CV↗

arXiv2017

Sim4CV: A Photo-Realistic Simulator for Computer Vision Applications

We present a photo-realistic training and evaluation simulator (Sim4CV) with extensive applications across various fields of computer vision. Built on top of the Unreal Engine, the simulator integrates full featured physics based cars, unmanned aerial vehicles (UAVs), and animated human actors in diverse urban and suburban 3D environments. We demonstrate the versatility of the simulator with two case studies: autonomous UAV-based tracking of moving objects and autonomous driving using supervised learning. The simulator fully integrates both several state-of-the-art tracking algorithms with a benchmark evaluation tool and a deep neural network (DNN) architecture for training vehicles to drive autonomously. It generates synthetic photo-realistic datasets with automatic ground truth annotations to easily extend existing real-world datasets and provides extensive synthetic data variety through its ability to reconfigure synthetic worlds on the fly using an automatic world generation tool. The supplementary video can be viewed a https://youtu.be/SqAxzsQ7qUU

cs.CV↗

arXiv2010

Measurement of Gilbert damping parameters in nanoscale CPP-GMR spin-valves

In-situ, device level measurement of thermal mag-noise spectral linewidths in 60nm diameter CPP-GMR spin-valve stacks of IrMn/ref/Cu/free, with reference and free layer of similar CoFe/CoFeGe alloy, are used to simultaneously determine the intrinsic Gilbert damping for both magnetic layers. It is shown that careful alignment at a "magic-angle" between free and reference layer static equilibrium magnetization can allow direct measurement of the broadband intrinsic thermal spectra in the virtual absence of spin-torque effects which otherwise grossly distort the spectral line shapes and require linewidth extrapolations to zero current (which are nonetheless also shown to agree well with the direct method). The experimental magic-angle spectra are shown to be in good qualitative and quantitative agreement with both macrospin calculations and micromagnetic eigenmode analysis. Despite similar composition and thickness, it is repeatedly found that the IrMn exchange pinned reference layer has ten times larger intrinsic Gilbert damping (alpha ~ 0.1) than that of the free-layer (alpha ~ 0.01). It is argued that the large reference layer damping results from strong, off -resonant coupling to to lossy modes of an IrMn/ref couple, rather than commonly invoked two-magnon processes.

cond-mat.mtrl-sci↗

arXiv2009

Tensor damping in metallic magnetic multilayers

The mechanism of spin-pumping, described by Tserkovnyak et al., is formally analyzed in the general case of a magnetic multilayer consisting of two or more metallic ferromagnetic (FM) films separated by normal metal (NM) layers. It is shown that the spin-pumping-induced dynamic coupling between FM layers modifies the linearized Gilbert equations in a way that replaces the scalar Gilbert damping constant with a nonlocal matrix of Cartesian damping tensors. The latter are shown to be methodically calculable from a matrix algebra solution of the Valet-Fert transport equations. As an example, explicit analytical results are obtained for a 5-layer (spin-valve) of form NM/FM/NM'/FM/NM. Comparisons with earlier well known results of Tserkovnyak et al. for the related 3-layer FM/NM/FM indicate that the latter inadvertently hid the tensor character of the damping, and instead singled out the diagonal element of the local damping tensor along the axis normal to the plane of the two magnetization vectors. For spin-valve devices of technological interest, the influence of the tensor components of the damping on thermal noise or spin-torque critical currents are strongly weighted by the relative magnitude of the elements of the nonlocal, anisotropic stiffness-field tensor-matrix, and for in-plane magnetized spin-valves are generally more sensitive to the in-plane element of the damping tensor.

cond-mat.mtrl-sci↗

arXiv2008

Co-resonant enhancement of spin-torque critical currents in spin-valves with synthetic-ferrimagnet free-layer

It is experimentally shown that the critical current for onset of spin-torque instability in current-perpendicular-to-plane spin-valves can be strongly enhanced using "synthetic ferrimagnet" free-layers of form FM1/Ru/FM2 (FM=ferrromagnet). However, this enhancement occurs for only one polarity of bias current. A two-macrospin model is shown to reproduce the observations. The model suggests that this phenomenon is related to a polarity-dependent, spin-torque induced co-resonance between the two natural dynamic modes of the FM1/FM2 couple. The resonance condition facilitates energy transfer out of the spin-torque destabilized mode into the other stable mode whose effective damping is actually enhanced by spin-torques, thereby delaying the onset of instability of this coupled system to larger critical currents.

cond-mat.other↗

arXiv2008

High-output CPP-GMR sensor with synthetic-ferrimagnet free layer and enhanced spin-torque critical currents

It is shown that the maximum stable output of a CPP-GMR sensor is increased significantly by using a synthetic ferrimagnet free layer, provided the electron current flows from free layer to reference layer. This free layer allows a larger magnetoresistance ratio for a given free layer magnetic moment, and in addition results in a greater than three-fold increase in the critical current above which spin-torque instability of the free layer occurs. In read heads with net free layer moments equivalent to only 4.5nm of Ni80Fe20, this effect is shown to result in sustainable sense current densities above 2e8 A/cm2.

cond-mat.mtrl-sci↗

arXiv2008

Audit and Change Analysis of Spreadsheets

Because spreadsheets have a large and growing importance in real-world work, their contents need to be controlled and validated. Generally spreadsheets have been difficult to verify, since data and executable information are stored together. Spreadsheet applications with multiple authors are especially difficult to verify, since controls over access are difficult to enforce. Facing similar problems, traditional software engineering has developed numerous tools and methodologies to control, verify and audit large applications with multiple developers. We present some tools we have developed to enable 1) the audit of selected, filtered, or all changes in a spreadsheet, that is, when a cell was changed, its original and new contents and who made the change, and 2) control of access to the spreadsheet file(s) so that auditing is trustworthy. Our tools apply to OpenOffice.org calc spreadsheets, which can generally be exchanged with Microsoft Excel.

cs.HC↗

arXiv2008

TellTable Spreadsheet Audit: from Technical Possibility to Operating Prototype

At the 2003 EuSpRIG meeting, we presented a framework and software infrastructure to generate and analyse an audit trail for a spreadsheet file. This report describes the results of a pilot implementation of this software (now called TellTable; see www.telltable.com), along with developments in the server infrastructure and availability, extensions to other "Office Suite" files, integration of the audit tool into the server interface, and related developments, licensing and reports. We continue to seek collaborators and partners in what is primarily an open-source project with some shared-source components.

cs.SE↗

arXiv2007

Gilbert and Landau-Lifshitz damping in the presense of spin-torque

A recent article by Stiles et al. (cond-mat/0702020) argued in favor of the Landau-Lifshitz damping term in the micromagnetic equations of motion over that of the more commonly accepted Gilbert damping form. Much of their argument revolved around spin-torque driven domain wall motion in narrow magnetic wires, since the presence of spin-torques can more acutely draw a distinction between the two forms of damping. In this article, the author uses simple arguments and examples to offer an alternative point of view favoring Gilbert.

cond-mat.other↗

arXiv2006

Comment on "Fluctuation-dissipation considerations and damping models for ferromagnetic thin-films"

In a recent article (Phys. Rev. B 71, 224402 (2005)), Safonov and Bertram claim "inconsistent" results can occur when applying the "Callen-Welton fluctuation dissipation theorem" to magnetic systems with dissipation. This author strongly disputes these claims, and instead will show that the inconsistencies claimed by Safonov and Bertram stem solely from a failure to correctly complete a linear transformation of variables, and a concomitant invalid application of well known fluctuation dissipation relations. When used correctly, such fluctuation-dissipation relations most certainly will apply to magnetic (or other physical) systems with dissipation.

cond-mat.mtrl-sci↗

arXiv2005

Angular Dependence of Spin-Torque Critical Currents in CPP-GMR Read Heads

This paper derives expressions for the critical current at the onset of spin-transfer-torque (STT) instability in CPP-GMR read heads, as a function of the relative angle (theta) between the free and reference layer magnetizations. Including a general angular dependent STT coefficient B(q=cos(theta)) exclusive of the angular momentum conserving sin(theta) factor, the critical current is found to depend on both B(q) and dB/dq in the non-collinear case |q| < 1. The paper also details the experimental measurement of the angular dependent critical currents on 50-nm sized CPP-GMR devices with synthetic antiferromagnet pinned layers, and fabricated using e-beam lithography. The measurements are consistent with prior theoretical models of the form B(q) ~ 1/[1+cnst*q], and indicate perhaps unanticipated implications for read head operation due to the critical current dependence on dB/dq.

cond-mat.mtrl-sci↗

arXiv2004

Basic considerations for magnetization dynamics in the combined presence of spin-transfer torques and thermal fluctuations

This article reviews basic theoretical features of Gilbert magnetization dynamics of a single domain magnetic film in the presence of Slonczewski spin-transfer torques, with and without thermal fluctuations taken into account. Rather than showing results of detailed numerical calculations, the discussion here is restricted to basic analytical results and conclusions which can mostly be derived from simply the form of the equations of motion, as well as elementary considerations based on classical stability analysis and the fluctuation-dissipation theorem. The presents work describes how interesting features of spin-transfer may be viewed as arising from non-equilibrium thermodynamics that are a direct consequence of the nonreciprocal nature of spin-transfer torques. The present article discusses fairly general results for spin-torque induced instability without thermal fluctuations, as well as the case of thermally activated magnetization reversal in uniaxial devices in the combined presence of external fields, thermal fluctuations, and spin-transfer torques. The results will be discussed and briefly compared and contrasted with that of prior work.

cond-mat.mtrl-sci↗