CHAPTER 1
Introduction to Spintronic and Nanomagnetic Computing Devices

Jayasimha Atulasimha1 and Supriyo Bandyopadhyay2

1Department of Mechanical and Nuclear Engineering, Virginia Commonwealth University, Richmond, VA, US

2Department of Electrical and Computer Engineering, Virginia Commonwealth University, Richmond, VA, USA

This book focuses on recent developments in two important and interrelated information processing device concepts and related phenomena: "spintronic devices" and "nanomagnetic devices." In the former, individual electron spins are coherently manipulated as they flow through the active region of a device to elicit device functionality. In the latter, an ensemble of spins in a nanostructure acts collectively as a giant classical spin (a single domain nanomagnet) owing to mutual exchange coupling, and the giant spin polarization (or the magnetization of the nanomagnet) is switched between stable orientations to store and/or process binary data. These information processing paradigms have attracted attention because of their low energy dissipation, nonvolatility and relatively fast speed of operation.

1.1 Spintronic Devices

An iconic device in the field of spintronics is the Datta-Das [1] Spin Field Effect Transistor (SPINFET) in which the current flowing between two of the terminals (source and drain) is modulated with a gate potential that does not change the carrier concentration in the channel of the transistor, but instead changes the spin polarization of the carriers. The source and drain contacts are ferromagnets that act as spin polarizers and analyzers. The source injects spin polarized electrons, the gate voltage precesses the spins in the channel owing to Rashba spin-orbit interaction [2] and the drain selectively transmits electrons depending on the degree of precession they have undergone in the channel. Thus, by varying the gate voltage, one can vary the source-to-drain current and realize transistor action. The operation of the transistor is briefly explained in Figure 1.1.

Figure 1.1 Operation of a Datta-Das SPINFET. The source injects spin polarized electrons, polarized in the direction of source-to-drain current (x-direction). When the gate voltage is zero, the spins do not precess and are fully transmitted by the drain resulting in maximum (on) current. When the gate voltage is turned on, it produces an electric field Ey in the y-direction due to Rashba spin-orbit interaction that results in an effective magnetic field of flux density Bz in the z-direction. This field causes the electrons to precess about itself. The left panel shows a one-dimensional SPINFET and the right panel a two-dimensional SPINFET.

There are several impediments to practical room temperature implementation of the Datta-Das SPINFET. Foremost among them is the inefficiency of the spin polarizer and analyzer. The inability of ferromagnet/semiconductor interfaces to inject and detect spins with high efficiency results in low on-off ratios of the drain current [3]. The on-off ratio is also reduced significantly if the channel of the SPINFET is not strictly one-dimensional [4], that is, if it is not a quantum wire with only the lowest carrier subband occupied. Finally, coherent transportation and manipulation of spins over the length of the channel at room temperature is challenging. The channel has to be sufficiently long to allow at least one-half period of spin precession and retaining spin coherence over that length is difficult at room temperature. Recently, coherent spin transport was demonstrated in a strictly one-dimensional InSb nanowire at room temperature [5], raising hopes for the Datta-Das transistor. That, together with the vast improvement in spin injection and detection efficiencies made possible by the use of quantum point contacts [6] as source and drain, has made a significant advance toward the demonstration of the Datta-Das device. Very significant steps in that direction have been reported recently involving spin injection, detection and manipulation with quantum point contacts as well as spin manipulation using spin orbit coupling to realize all-electric and all-semiconductor spin field effect transistors [7].

Chapter 2 discusses the use of quantum point contacts (QPC) with lateral spin-orbit coupling (LSOC) to create a strongly spin-polarized current by tuning the asymmetric bias voltages on the side gates in the absence of any applied magnetic field. By injecting this strongly spin-polarized current into the channel of a SPINFET, high injection efficiency can be obtained. This chapter also explores the different regimes of operation of all-electric spin valves made of quantum point contact and quantum dots, with spin-orbit coupling, and the ramification of an all-electric spin valve for future spin-based devices, circuits, and architectures.

Chapter 3 explores and surveys interesting variations of the Datta-Das spin transistor by proposing devices that do not rely on the gate voltage controlled precession of spins in the channel. Instead it surveys two other devices:

1. "Spin MOSFET," comprising a regular MOSFET with ferromagnetic contacts whose magnetizations can be switched from parallel or antiparallel configuration, thereby turning the transistor on and off.
2. "Pseudo-spin-MOSFET," which is essentially a MOSFET with a magneto-tunneling junction, or MTJ connected to either the source or the drain.

This chapter further discusses the use of these two devices for energy-efficient (low power) nonvolatile logic circuits. Since the gating action and the parallel/antiparallel orientation of the magnetizations of two ferromagnetic contacts (in case of Spin-MOSFET) or the MTJ's magnetic layers (in case of Pseudo spin-MOSFET) can be independently controlled, these devices are well suited for nonvolatile bistable circuits. Finally, implementation of nonvolatile memory elements based on these devices is also discussed. In some sense, these devices are closer to "nanomagnetic devices" as the magnetic states of the MTJ/ferromagnetic contacts (nanomagnets) encode information.

1.2 Nanomagnetic Devices

Inherent advantages: Nanomagnets have two inherent advantages over transistors as binary switches: nonvolatility (or the ability to store information without any standby power dissipation) and the potential to switch from one stable state to another with extremely small energy dissipation. These are explained below:

1. Consider an elliptical Terfenol-D nanomagnet as shown in Figure 1.2 (rightmost figures) with major axis, minor axis and thickness respectively 110 nm, 90 nm and 6 nm. These dimensions ensure that the nanomagnet has a single domain [11] and that the shape anisotropy energy barrier (Eb), which separates the two degenerate minima in the potential energy profile of the nanomagnet (these minima correspond to the two stable magnetization orientations that are mutually antiparallel and aligned along the long axis), is 2.2 eV (85.12 kT at room temperature). That makes the probability of spontaneous magnetization flipping between the two stable orientations due to thermal agitations equal to e-Eb/kT = e-85 per attempt [8]. Therefore, if binary bit information has been written into the magnetization orientation of the nanomagnet, then that information is retained for a time of (1/f0) e85 = 2.6 × 1017 years, if we assume the attempt frequency f0 to be 1 THz [9]. In other words, the nanomagnet is nonvolatile. If we "write" binary information in the nanomagnet by orienting the magnetization along one of the two stable states, that information stays uncorrupted almost in perpetuity, even when no energy is supplied to the nanomagnet to retain the information.

   Figure 1.2 Transistor, single-spin and single-domain nanomagnet encoding logical "0" and "1" states.

2. The nanomagnet can not only retain information but also process it in a very energy-efficient way. The minimum energy dissipated in switching a charge-based device like a transistor at a temperature T is NkTln(1/p) independent of the switching speed [10] where N is the number of information carriers (electrons) in the transistor, k is the Boltzmann constant, and p is the bit error probability. This happens because the charges act independently of each other and there is no collective dynamics when switching takes place, resulting in N degrees of freedom for the charge ensemble. In c6ontrast, the minimum energy dissipated to switch a single-domain nanomagnet's magnetization is only ~ kTln(1/p), since the exchange interaction between the many spins comprising a single domain nanomagnet makes all of them behave collectively like a giant single spin and rotate in unison [10, 11], resulting in a single degree of freedom. The collective dynamics of spins - absent among charges - make the nanomagnet a far more energy-efficient switch than a transistor. If we assume the same number of information carriers in a transistor and in a single domain nanomagnet in Figure 1.2, then for the same bit error probability, the ratio...