Introduction

List of topics investigated in this chapter

The topics covered in this chapter and summarized in Table 7 are similar to those covered in Chapter 6. Here, however, we deal with a double- rather than a singleinterface structure which consists of identical lossless dielectric (DPS-type) cover and substrate bounding a lossy metallic (ENG-type) guide, as shown in Fig. 7.1. Such a symmetric structure produces two solutions to the mode equation that are distinct from each other for a finite guide thickness. As the thickness of the guide increases, the two solutions converge to that of a single-interface DPS-ENG-type structure. The propagation constants of these modes are calculated for the freely propagating case and for the case where the modes are excited and loaded by a prism coupler. For each mode the electric and magnetic fields are evaluated, together with the local power flow, wave impedance and surface charge density at each guide interface. In Chapter 6 we dealt with a single-interface structure and used the Otto (O) or Kretschmann (K) configurations to excite a mode. Here, because we have a double-interface structure, we use the general prism coupling configuration (G). The reflectivity of an incident electromagnetic wave off the base of the prism, ℛ, is calculated for this configuration. The theory of double-interface surface plasmons was adapted from Refs. [1] to [4] and the concept of short-range (SR) and long-range (LR) SPs and recent reviews are from Refs. [5] to [11].

In 1952 Pines and Bohm discussed a quantized bulk plasma oscillation of electrons in a metallic solid to explain the energy losses of fast electrons passing through metal foils [1]. They called this excitation a “plasmon.” Today these excitations are often called “bulk plasmons” or “volume plasmons” to distinguish them from the topic of this book, namely surface plasmons. Although surface electromagnetic waves were first discussed by Zenneck and Sommerfeld [2, 3], Ritchie was the first person to use the term “surface plasmon” (SP) when in 1957 he extended the work of Pines and Bohm to include the interaction of the plasma oscillations with the surfaces of metal foils [4].

SPs are elementary excitations of solids that go by a variety of names in the technical literature. For simplicity in this book we shall always refer to them as SPs. However, the reader should be aware that the terms “surface plasmon polariton” (SPP) or alternately “plasmon surface polariton” (PSP) are used nearly as frequently as “surface plasmon” and have the advantage of emphasizing the connection of the electronic excitation in the solid to its associated electromagnetic field. SPs are also called “surface plasma waves” (SPWs), “surface plasma oscillations” (SPOs) and “surface electromagnetic waves” (SEWs) in the literature, and as in most other technical fields, the acronyms are used ubiquitously. Other terms related to SPs which we will discuss in the course of this book include “surface plasmon resonance” (SPR), “localized surface plasmons” (LSPs), “longrange surface plasmons” (LRSPs) and of course “short-range surface plasmons” (SRSPs).

Introduction

The first part of this chapter describes the edge effects of patterned ferromagnetic films. The edge pole of the ferromagnetic film plays a significant role in the film properties, in particular, the magnetic energy state. Since the magnetic memory cells are made of tiny pieces of patterned film or film stack, the effects due to the end poles become a dominant factor governing the stability and switching behavior of the memory cell. The second part of this chapter deals with the switching properties of a small patterned film under an external magnetic field. A coherent switching model is introduced to describe the switching properties of the film. This is the basis of the write operation of the field-MRAM cells.

Edge poles and demagnetizing field

When a ferromagnetic thin film is patterned and etched into shapes, the magnetic poles at the edge of the film are exposed. Like the end of a bar magnet, magnetic flux emits from the poles at the edge of the thin film. Inside the film, the flux points in a direction opposite to the magnetization. The magnetic field associated with the end poles is called the demagnetizing field, HD. The magnitude of HD is position-dependent.

Consider a semi-infinite film of thickness t and saturation magnetization MS. The film extends from y = 0 in the +y-direction toward infinity and extends in both the +x- and –x-direction toward infinity (see Fig. 3.1).

Introduction

A short time after the discovery of magnetic tunneling devices, tunneling magnetoresistance (TMR) replaced giant magnetoresistance (GMR) read sensor in the hard disk drive. This marked the first successful commercialization of magnetic tunnel junction technology. The first mass production of the magnetic recording head based on a MgO tunnel barrier took place in 2006. In the same year, 4 Mb MRAM chips were commercialized, and it was the first field-MRAM product working in the toggle-write mode. The viability of MTJ technology at the product level is proven. Subsequently, electronic system designers started to consider seriously how to take advantage of this technology. Many new applications of MTJ technology begin to emerge. One of the new circuit elements is the non-volatile magnetic flip-flop device, which is used for the reduction of VLSI chip power as well as for run-time system re-configuration. Such new applications can only be realized with the unique properties of magnetic tunnel junction devices.

Other new applications are being explored in the field of healthcare. GMR and TMR chips are used for detecting biological molecules labeled with magnetic particles, and this could be a powerful platform for next-generation diagnostics. The sensitivity achievable with simple portable instrumentation can be orders of magnitude better than the current methods. Since this application is still in its infancy, it will not be discussed further here. Interested readers are referred to the references above.

SI magnetic units are easily related to the current, voltage and energy in MKS units, since the SI system was originally developed under the assumption that magnetism is originated from electric current. The dimensions of magnetic units are shown below; A = amperes, s = seconds, kg = kilograms and m = meters.

1. (1) newton (N) = kg m/s2;

2. (2) joule (J) = kg m2/s2;

3. (3) magnetic field (H) = A/m;

4. (4) henry (h) = kg m2/s2A2;

5. (5) tesla (T) = kg/s2A;

6. (6) weber (Wb) = kg m2/s2A.

Magnetism in cgs units is less transparent. The unit of magnetic moment m is the emu. The density of the magnetic moment MS is emu/cm3. The magnetic induction is given by B = H + 4πMS, where the magnetic field H is given in units of oersted (Oe) and B is given in gauss (G). (1 Oe = 1 G in air, since M = 0 in air.) Thus, the “4π” in 4πMS is not dimensionless. Its dimension is [G/(emu/cm3)], and it is equivalent to the inverse of susceptibility, so that the unit of 4πMS is [G/(emu/cm3)] · [emu/cm3] = G.

The dimension of “emu” can be understood from the dimension of energy. In Chapter 2, we discussed the magnetostatic energy per unit volume of magnetic material under a magnetic field H to be ∼MS · H. The dimension of MS · H is [emu · cm−3] · [Oe], which should be equivalent to [erg · cm−3]. Therefore, the dimension of emu is [erg/Oe] or [emu/G] in air.

Introduction

Magnetic ferrite core memory was invented and produced in the 1960s, prior to semiconductor memory. Ferrite cores are made from a paste of ferrite powers, which are sintered at high temperature. The process of forming a discrete core is not as scalable as the integrated circuit process on a silicon wafer. The product life of a magnetic core was short, and in the 1970s this technique was replaced by semiconductor memory. A similar fate happened to magnetic bubble memory, another type of magnetic memory, which was built on a magnetic garnet material substrate (gadolinium gallium garnet, Gd3Ga2(GaO4)3). The bit density of bubble memory technology is scalable since it is made with a planar process, similar to the silicon integration circuits. However, because it is on garnet, it is passive and cannot perform logic functions (such as address decoding), and it requires a companion silicon chip to provide the logic function to complete the memory access function. Even with better memory performance, magnetic bubble memory could not compete against magnetic hard disk and semiconductor memory, which continue to show a clear path of scaling for a lower cost. By the mid 1980s, commercial magnetic bubble production had ended.

Subsequent efforts in the development of magnetic memory have been focused on the integration of magnetic thin-film memory devices into silicon wafer processes. Magnetic memory devices exist in the form of thin-film stacks, which can easily be integrated into the back-end metal wiring metallurgy process.

The advent of semiconductor technology has impacted the lives of many of us since the 1970s. Silicon CMOS (complementary metal-oxide-semiconductor) devices are practically ubiquitous, and by the year 2000, the value of the semiconductor industry exceeded that of the automobile industry. The magnetic industry, on the other hand, is much smaller than the semiconductor industry. Engineering schools of universities rarely cover any courses in this discipline. Nonetheless, a tiny magnetic recording device is in the hard disk of every computer. Like CMOS devices, magnetic recording technology is being scaled down from generation to generation. At the time of writing, the physical size of the magnetic bit remains smaller than a DRAM bit on silicon chips.

Researchers working in these two communities had little in common until the development of the modern magnetic random access memory, or MRAM. A MRAM chip is built by integrating magnetic tunneling junction (MTJ) devices onto the silicon CMOS circuits. The research activity of MTJs in academia and industry, both hard disk and semiconductor, has been very active since it first showed signs of technology implication in the mid 1990s. That effort led to the mass production of the MTJ recording head in hard disk in 2006. In the same year, the semiconductor industry announced the first successful introduction of an MTJ memory product. The viability of MTJ technology is proven. It is expected that research activities will develop further, which will increase cooperation between these two research communities.