Probabilistic machine learning enabled by the Bayesian formulation has recently gained significant attention in the domain of automated reasoning and decision-making. While impressive strides have been made recently to scale up the performance of deep Bayesian neural networks, they have been primarily standalone software efforts without any regard to the underlying hardware implementation. In this paper, we propose an "All-Spin" Bayesian Neural Network where the underlying spintronic hardware provides a better match to the Bayesian computing models. To the best of our knowledge, this is the first exploration of a Bayesian neural hardware accelerator enabled by emerging post-CMOS technologies. We develop an experimentally calibrated device-circuit-algorithm co-simulation framework and demonstrate 24× reduction in energy consumption against an iso-network CMOS baseline implementation.
We use spin torque ferromagnetic resonance and ferromagnetic-resonance-driven spin pumping to detect spin-charge interconversion at room temperature in heterostructure devices that interface an archetypal Dirac semimetal, Cd 3 As 2 , with a metallic ferromagnet, Ni 0.80 Fe 0.20 (permalloy).Angle-resolved photoemission directly reveals the Dirac semimetal nature of the samples prior to device fabrication and high-resolution transmission electron microscopy is used to characterize the crystalline structure and the relevant heterointerfaces. We find that the spin-charge interconversion efficiency in Cd 3 As 2 /permalloy heterostructures is comparable to that in heavy metals and that it is enhanced by the presence of an interfacial oxide. Spin torque ferromagnetic resonance measurements reveal an in-plane spin polarization regardless of an oxidized or pristine interface.We discuss the underlying mechanisms for spin-charge interconversion by comparing our results with first principles calculations and conclude that extrinsic mechanisms dominate the observed phenomena. Our results indicate a need for caution in interpretations of spin transport and spincharge conversion experiments in Cd 3 As 2 devices that seek to invoke the role of topological Dirac and Fermi arc states.
Devices based on the unique phase transitions of phase change materials (PCMs) like GeTe and Ge2Sb2Te5 (GST) require low-resistance and thermally stable Ohmic contacts. This work reviews the literature on electrical contacts to GeTe, GST, GeCu2Te3 (GCuT), and Ge2Cr2Te6 (GCrT), especially GeTe due to the greater number of studies. We briefly review how the method used to measure the contact resistance (Rc) and specific contact resistance (ρc) can influence the values extracted, since measurements of low contact resistances are susceptible to artifacts, and we include a direct comparison of Au-, Pt-, Ni-, Mo-, Cr-, Sn-, and Ti-based contacts using a systematic approach. Premetallization surface treatment of GeTe, using ex situ or in situ approaches, is critical for minimizing contact resistance (Rc). Transmission electron microscopy reveals that interfacial reactions often occur and also clearly influence Rc. The lowest Rc values (∼0.004 ± 0.001 Ω mm) from the direct comparison were achieved with as-deposited Mo/Ti/Pt/Au (Ar+ plasma treatment) contacts and annealed Sn/Fe/Au (de-ionized H2O premetallization treatment). In the case of Sn-based contacts, low Rc was attributed, in part, to the formation of SnTe at the contact interface; however, for Mo-based contacts, no such interfacial reaction was observed. Comparing all contact metals tested beneath a cap of at least 100 nm of Au, Mo/Ti/Pt/Au offered the lowest contact resistance as-deposited, even though the work function of Mo is only 4.6 eV, and the low contact resistance remained stable even after annealing at 200 °C for 30 min. This trend is surprising, as high work function metals, like Ni and Pt, would be expected to provide lower Rc values when they are in contact with a p-type semiconductor like GeTe. Through materials’ characterization, an inverse relationship between the metal work function and Rc for higher work function metals can be attributed to the reactivity of many of the metals with GeTe. Studies of contacts to GST in the literature involve only a small number of contact materials (Ti, TiN, TiW, W, Pt, and graphene) and employ varied geometries for extracting contact resistance. For hexagonal GST, TiW is reported to provide the lowest ρc of ∼2 × 10−7 Ω cm2, while TiN provided the lowest reported ρc of ∼3 × 10−7 Ω cm2 to cubic GST. For the ternary PCMs GCuT and GCrT, contact resistance studies in the literature are also limited, with W being the only metal studied. While more extensive work is necessary to draw wider conclusions about trends in current transport at metal/GST, metal/GCuT, and metal/GCrT interfaces, reduction of Rc and high thermal stability are critical to engineering more efficient and reliable devices based on these materials.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.