The discovery in 1988 of GMR in structures of alternating magnetic and nonmagnetic thin layers (Baibich et al. 1988) was the accumulation of several decades of intensive research in thin film magnetism (Shinjo and Takada 1987) and improvements in epitaxial growth techniques developed mainly in semiconductor materials. Not surprisingly, the first GMR structure was fabricated using molecular beam epitaxy (Baibich et al. 1988). The high quality magnetic and nonmagnetic metallic films provide electrons with a mean free path exceeding 100 Å; on the other hand, the epitaxial growth allows for each constituent layer of the structure to be as thin as a few atomic layers. The greatly enhanced spin-dependent scattering in these multilayered structures provides magnetoresistance changes as high as 50%. Table 5.3 shows some of the institutions involved in GMR research and development, based on various publications, patents, and WTEC visits.
Two subsequent major developments from IBM enabled the application of GMR materials to hard disk heads, RAM, and sensors. The first development was the demonstration of equally good or better GMR materials using high throughput and production-worthy magnetron sputtering systems (Parkin et al. 1990). The other development was the invention of magnetically soft spin-valve structures, which allow low field and low power operation (Dieny et al. 1991a; Dieny et al. 1991b).
Giant Magnetoresistance Activities
Industrial R&D efforts on GMR materials initially focused on high density read heads. The major U.S. players are IBM, Seagate, Quantum, ReadRite, and Applied Magnetics. In Japan, all the semiconductor companies are involved, in addition to strong magnetic media powerhouses such as TDK and Yamaha. Korea's Samsung is also actively involved in the GMR race. In Europe, Thomson CSF, Philips, and Siemens seem to have fallen behind. All in all, IBM is in a commanding position to reap the benefits of the GMR phenomenon. In November 1997, it announced the volume production of the first generation of GMR read heads.
In 1995, a different class of high magnetoresistive materials was discovered in which the nonmagnetic layer separating the two ferromagnetic layers is made with an ultrathin insulating material, such as an aluminum oxide layer < 20 Å thick (Miyazaki and Tezuka 1995; Moodera et al. 1995). With the switching of magnetization of the two magnetic layers between parallel and antiparallel states, the differences in the tunneling coefficient of the junction and thus the magnetoresistance ratio have been demonstrated to be more than 25%. A distinctive feature of this MTJ class of materials is its high impedance (> 100 kW -Ám2), which allows for large signal outputs.
The gradual improvement of GMR and MTJ materials have made them attractive for nonvolatile magnetic random access memory (MRAM) applications. The potential to make MRAM a high density, high speed, and low power, general purpose memory prompted the Defense Advanced Reseat Projects Agency to fund three MRAM consortia beginning in 1995, led by IBM, Motorola, and Honeywell, respectively. Other companies engaged in MRAM research are Hewlett-Packard, Matsushita, NEC, Fujitsu, Toshiba, Hitachi, and Siemens.
The key for a competitive MRAM technology is the fabrication of deep submicron-dimension magnetic cells. The further development of lithography tools utilizing e-beam and deep ultraviolet radiation is essential. Magnetic storage elements as small as 0.25 Ám have been demonstrated by both Motorola and IBM (Tehrani et al. 1996; Chen et al. 1997; Gallagher et al. 1997). Among the steps of MRAM fabrication that are not yet compatible with semiconductor processing is the ion milling of the magnetic cells. The possibility of dry etching the magnetic materials has, nevertheless, been demonstrated (Jung et al. 1997). Figures 5.12 through 5.15 summarize the major directions in memory R&D. Arrows denote ferromagnetic layers.
The ability to fabricate submicron magnetic elements has opened a very rich and fascinating area of micromagnetics research. Characterization techniques having nanoscale resolution have been utilized and improved to measure and image the complex magnetization patterns in order to understand the magnetization switching characteristics. Examples of such techniques include the following:
Another essential tool is micromagnetics modeling, which is used to predict complex magnetic domain configurations in patterned magnetic elements and to generate transient pictures that demonstrate the process of forming complex domain configurations (Zheng and Zhu 1997).
By combining MFM, SQUID magnetometry, SEMPA, and micromagnetics modeling, researchers at Motorola have conducted a systematic study of the switching characteristics of single-layer and multilayer submicron magnetic structures. Three different phases of the magnetization phase diagram have been identified with regards to material composition, dimension, shape, thickness, and other structural parameters: (1) the quasi-single domain phase can be well described by the coherent rotation model (Sakaki et al. 1995); (2) the end-domain phase is dominated by the nonuniform regions of magnetization at the two ends of the element-the magnetization switching process can either be rotational or through domain wall nucleation and propagation (Shi et al. 1998); and (3) the trapped magnetization vortice (TMV) phase, which is characterized by the presence of magnetization vortices. Nucleation from TMV sites requires lower reversal fields than coherent rotation, but a high field is needed to drive out TMVs in the element. When the driving field is not high enough, TMVs can cause unusually large fluctuations in the switching field (Shi et al. n.d.).
As the size of magnetic elements scales below 20 nm, a superparamagnetic phase emerges in which the room temperature thermal energy overcomes the magnetostatic energy well of the element, resulting in zero hysteresis (Hylton 1993). In other words, although the element itself is a single-domain ferromagnet, the ability of an individual magnetic "dot" to store magnetization orientation information is lost when its dimension is below a threshold. On the other hand, suitably prepared alloys of immiscible ferromagnetic and nonmagnetic metals that contain single-domain ferromagnetic grains in a nonmagnetic matrix have been shown to exhibit GMR characteristics. The moments of the magnetic grains are aligned at high fields and random at the coercive field, leading to GMR characteristics. In such "granular" metals, GMR has been reported for sputtered alloy films of CoCu (Berkowitz et al. 1992; Xiao et al. 1992), FeCu (Xiao et al. 1992), NiFe/Ag (Jiang et al. 1992), and CoAg (Carey et al. 1992; Tsoukatos et al. 1992). GMR values as high as 55% at 4.2 K and 20% at room temperature have been observed. The granular films require magnetic fields of the order of 10 kOe to achieve such a change in electrical resistance.
A very exciting consequence of ultrascaled magnetic particles is quantum tunneling of the magnetization direction of a collection of spins. There is no simple Schrödinger equation that describes this process, since it is not an elementary particle that is tunneling but a collective coordinate. Below its "blocking temperature," at which thermally assisted hopping between magnetic orientations becomes frozen out, magnetic particles of TbCeFe at sizes around 15 nm have been observed to behave independent of temperature and with no freeze-out magnetic relaxation (Barbara et al. 1993). Because of the coherent tunneling of the magnetization orientation between the symmetric double-well potential, a resonance line in the magnetic susceptibility and noise spectra has been observed at temperatures below 200 mK in zero applied magnetic field (Awschalom et al. 1992). This work has stimulated a number of theoretical investigations into the effects of dissipation and the feasibility of producing quantum effects in larger magnetic structures (Prokofev and Stamp 1993; Gaarg 1994; Braun and Loss 1994).
Another interesting type of nanomagnetic structure is nanometer ferromagnetic wires fabricated using conventional nanolithography (Adeyeye et al. 1997; Chou 1997), nanoimprint lithography (Chou et al. 1995), AFM/MFM direct writing (Kong et al. 1997), groove deposition (Hong and Giordano 1995), and electrodeposition into pores of template polymer membranes (Piraux et al. 1994; Blondel et al. 1994). Such nanowires of either single layer or multilayers may provide new approaches to very small magnetoresistive sensors, ultrahigh-density hard disks (Chou et al. 1994), and other extensions of conventional applications. Another intriguing possibility is the suggestion to use heterostructure nanowires to investigate single electron tunneling (Cavicchi and Silsbee 1984; Kumzerov and Poborchii 1994).
Recently, molecular magnetism has received much attention with the development of a variety of synthesis techniques largely adapted from biology and chemistry (Kahn 1993). Natural and artificial ferritin proteins are examples of systems obtained using these methods (Gatteschi et al. 1994). The ability to add one magnetic ion at a time has resulted in nanoscale magnets precisely defined by atomic weight. Ferritin consists of a segmented protein shell in the shape of a hollow sphere, with an outer diameter of 12.5 nm and an inner diameter of 7.5 nm. In vivo, the inner space is normally filled with a crystal of an iron oxide that is antiferromagnetic below 240 K. The empty protein shells can also serve as vessels for the synthesis of ferrimagnetic magnetite and maghemite. Thus, there exists a system in which its size as well as the nature of its magnetic interactions can be varied.
Another example of molecular magnetism is a cobalt-iron-cyanide-based Prussian blue analog (Sato et al. 1996a). In the ground state the Fe+2 and Co+3 ions are low-spin and diamagnetic, and there is no interaction between them. Red light excitation transfers one electron from an iron site to a cobalt site, resulting in high-spin Fe+3 and Co+2 ions and magnetic interactions between them. The application of a blue light causes a transition back to the initial state and switches off the Fe+3-Co+2 interactions. In addition to this kind of photochemically controllable magnets, electrochemically controllable magnets have also been reported (Sato et al. 1996b).