Putting A Spin On Electronics

Imagine a personal computer that powers up instantly without the lengthy and bothersome boot-up delay. Think of the convenience offered by a computer that can be switched into a standby mode that requires no electrical power, yet keeps files and data in active memory.

Computers sporting those features aren't available just yet. But someday manufacturers may be able to build electronic devices with those operating characteristics and other sophisticated functions, partly as a result of advances in the emerging field of spintronics (spin-based electronics). A key goal of research in spintronics is, as the name implies, merging the properties of electron spin with the attributes of conventional electronics. According to the field's proponents, achieving such a union holds the promise for developing a novel class of advanced electronic devices with unprecedented capabilities. Meeting that goal, however, requires overcoming major technical challenges, one of which is the design and synthesis of new materials, such as magnetic semiconductors, with ideal properties for spintronics applications.

Fifty-plus years of research and development in microelectronics has led to today's extremely fast, tiny, and densely packed computing circuits. Individual semiconductor chips crammed with many millions of transistors and other components are standard commercial products nowadays. Yet for all the progress in technology, the newest circuit elements operate on the same basic principles as decades-old transistors: by applying voltages to control the flow of electrical current.

As current flows through integrated circuits, transistors sense and respond to a key property of electrons: their charge. In general, though, circuits are indifferent to another fundamental property of electrons: their spin. Closely associated with magnetism, an electron's intrinsic angular momentum (the quantum mechanical property labeled "spin") is typically described by analogy with a rotating charged sphere and is classified as spin-up or spin-down depending on the direction of rotation.

Under most circumstances, electron spins are oriented randomly and function as unread labels that just tag along for the ride through electrical circuits. But in the presence of an external magnetic field, the up and down spin orientations are distinguishable and readily detected as measurable differences in electron energy. That's the property that spintronics researchers seek to exploit. The aim is to develop technologies that can sense and control (read and write) electron spins so that the information contained in the spin labels can be used to complement the charge information that has served as the workhorse of microelectronics for decades.

Ideas for putting electron spin information to use are plentiful. A small number of applications have already been developed and commercialized, while others are seen as technologies for the near future or, in some cases, for the more distant future.

Peek inside a modern computer, and you'll see the first (and to date, the only) large-scale commercially successful example of a spintronics device. Data stored on today's high-density hard discs are read by sophisticated magnetic sensors—"read—heads"-that capitalize on a spintronics effect known as giant magnetoresistance (GMR).

Constructed as a sandwich of thin layers of magnetic and nonmagnetic metallic alloys, GMR read-heads (also known as spin valves) read data by sensing the alignment of magnetization (and hence electron spins) in magnetic bits on a rotating hard disc. One of the spin valve layers consists of a "hard" magnetic material, often cobalt-based, in which the magnetization direction remains fixed (or "pinned"). The other layer is made of a "soft" magnetic material, for example a nickel-iron alloy, in which the polarization of the magnetic field is free to flip its orientation in response to an applied field, such as the fields associated with the bits on a hard disc.

As the micrometer-sized magnetic bits (or domains) race by the read-head, they alter the magnetic orientation of the soft layer. When the magnetizations of the hard and soft layers are aligned in parallel, resistance across the read-head is at a minimum. When the orientations are antiparallel, the resistance increases. Detecting the difference in magnetoresistance is the basis for reading data.

GMR read-heads, which came on the scene in the late 1990s, "brought about a big change in hard-drive storage capacity," says David D. Awschalom, a spintronics pioneer and professor of physics at the University of California, Santa Barbara. Because of the device's high sensitivity to spin orientation, manufacturers have been able to boost data-storage density by shrinking the size of magnetic bits on hard discs without the data signal becoming too small to be detected reliably.

Not only has the technology led to today's 300-gigabyte hard drives, but in Awschalom's view, "it sent a message to the world that there really is a place for spintronics." That message was reinforced earlier this month when the first magnetic random access memory (MRAM) chips, also based on spin technology, were commercialized by Freescale Semiconductor, Austin, Texas.

At the other end of the time scale, a futuristic quantum-computing technology is the focus of some spintronics researchers. Those scientists are exploring ways in which the up and down spins of individual electrons can be coupled with the zeros and ones that represent the charge state of a circuit component such that each data bit can assume four states, not just two as in binary systems.

Somewhere between present-day applications and far-off technologies lies magnetic semiconductor-based spintronics. Building spintronics on a platform of semiconductor technology is seen by many scientists as a viable route to new types of advanced electronics that can be achieved in the near future. But there's a hitch: Unlike the collection of magnetic metallic alloys used, for example, in disc drives and GMR read-heads, magnetic semiconductors are uncommon.

The shortage of materials with potential for a big impact on technology has motivated physicists, chemists, materials scientists, and others to search for ways to prepare novel materials with just the right properties for spintronics applications. One strategy being explored is substituting some of the positive ions in nonmagnetic semiconductors, such as indium arsenide and gallium arsenide, with cobalt, manganese, and other magnetic transition-metal ions. That approach has also been studied with other materials such as gallium nitride and various oxides.

Talk to spintronics enthusiasts, and they'll fire off a list of benefits that may be brought about by magnetic semiconductors. Putting an end to annoying computer boot-ups is one such potential benefit. The idea is that just as magnetic discs hold onto data after power is switched off, computer circuits made from magnetic semiconductors would retain their logic states without power. That feature, which is known as nonvolatile memory, could lead to instant-on computers that use electricity sparingly and are able to save a computer session in active memory even after the user has pulled the plug or removed the battery.

Another key reason to develop magnetic semiconductors is that they may enable the principal functions of a computer—namely, logic operations, communication between circuits, and data storage—to be integrated into a single material. That kind of innovation may lead to smaller and faster computers with data-storage densities that outperform today's machines.

"If you didn't know the history of computer technology and you looked at a modern computer, you might think it's pretty odd that we use magnetic films to store information, semiconductors to do the computation, and optical fibers or metal wires for communication," Awschalom remarks. Why not use just one material?

The UC Santa Barbara researcher explains that the reason for demarcation is that a single material capable of all of those tasks wasn't available when the technologies were developed. But with the advent of magnetically doped semiconductors, in principle, the situation may change, he says.

Given the right set of properties, magnetic semiconductors would be well-suited to hosting logic circuits and storing data. But as Awschalom points out, the potential benefits don't end there. It turns out that the orientations of electron spin and photon polarization are intimately connected through the quantum mechanical selection rules that describe the effects of radiation on semiconductors. For example, left circularly polarized light impinging on a semiconductor liberates spin-up electrons, he says. That correlation can be used to transmit data-including spin-state information-between remote circuits at breakneck speeds. It's easy to do, Awschalom stresses, "because in semiconductor spintronics, integration of logic and communication comes for free."

Cashing in on those promises, however, requires coming up with materials that are endowed with hard-to-find combinations of characteristics. One of the most important requirements is that the semiconductor retain its ferromagnetism, meaning its spins remain aligned, above room temperature. To date, finding or preparing a material with a high Curie temperature (T_C), above which alignment is lost, has dominated much of the research in this area.

Among the most thoroughly studied classes of candidate materials are doped III-V semiconductors (named by group in the periodic table), such as indium arsenide and gallium arsenide. In the late 1980s, researchers at IBM's T. J. Watson Research Center developed molecular-beam-epitaxy (MBE) methods for imparting ferromagnetic character to the otherwise nonmagnetic semiconductors by incorporating manganese into the materials during the growth process. The research group, which included Hideo Ohno, now a professor at Tohuku University, in Sendai, Japan, found that manganese-doped InAs remained ferromagnetic up to roughly 60 K.

Other studies followed, and within a short time researchers were reporting that manganese-doped GaAs (Ga_1-xMn_xAs) retained its ferromagnetism up to approximately 150 K. Those studies and others helped establish the utility of the MBE approach for preparing such materials, which came to be referred to as dilute magnetic semiconductors. But with Curie temperatures down around 100 °C below zero, spintronics applications destined for use somewhere other than, say, the Martian north pole would have to wait.

But not for long. Encouraging news came along a few years later in the form of a theoretical paper by Tomasz Dietl, a physicist and research group leader at the Polish Academy of Sciences, in Warsaw. In that paper, Dietl, Ohno and their coworkers predicted that high Curie temperatures could be reached if magnetic ions were substituted into wide-band-gap semiconductors such as gallium nitride and zinc oxide (Science 2000, 287, 1019).

Ram Seshadri, associate professor of materials at UC Santa Barbara, says Dietl's predictions touched off a wave of research activity as scientists sought to make materials similar to the ones described in the now widely cited paper. Seshadri points out that the study "got a lot of scientists excited about the big materials synthesis challenges in spintronics."

Seshadri explains that the study calls for substituting a few percent of magnetic ions into GaN or ZnO that has been heavily hole-doped, meaning prepared with a high concentration of positive charge carriers (some 10²⁰ per cubic centimeter). Making materials that closely match that description has remained difficult. Nonetheless, researchers in many labs have claimed success in the past few years in preparing high-Curie-temperature ferromagnetic semiconductors.

For example, after the seminal paper by Dietl was published in Science, researchers in Japan reported that manganese-doped GaN films exhibit ferromagnetism at room temperature (J. Cryst. Growth 2002, 237-239, 1358). Likewise, at about the same time, a team of scientists in Sweden, Japan, the U.S., and the U.K. reported above-room-temperature ferromagnetism in manganese-doped ZnO thin films and bulk samples (Nat. Mater. 2003, 2, 673).

It wasn't long before other materials also jumped on the ferromagnetism bandwagon. Titanium dioxide, for example, came aboard when researchers in Japan found that doping thin films of TiO₂ with cobalt produced the effect at room temperature (Science 2001, 291, 854). And even undoped hafnium dioxide, an insulating material without any magnetic ions, was reported to be ferromagnetic above 500 K, an observation that the authors noted was "challenging for the theory of magnetism" (Nature 2004, 430, 630).

The list of high-T_C materials continued to grow quickly-a seemingly positive development. Yet to several scientists, some of the results seemed too good to be true.

"It just didn't add up," Seshadri says. He explains that in many cases, researchers were measuring magnetic signals that were orders of magnitude weaker than what would have been expected given the reported compositions of the test samples. Some of the studies left Seshadri wondering whether the high-T_C claims were in fact due to small magnetic impurities.

Other experts raised the same question. For example, Koji Ando at the Nanoelectronics Research Institute, in Tsukuba, Japan, pointed out in a recent commentary that high-sensitivity magnetic measurements made with a superconducting quantum interference device (SQUID) can readily be dominated by tiny quantities of iron. He notes that iron at a concentration of just 1 part in 2,000 in the sample probed in SQUID measurements can generate signals comparable with the ones reported in the puzzling papers (Science 2006, 312, 1883). Ando's commentary also cites a recent study showing that all it takes "to generate clear ferromagnetic signals" in nonmagnetic HfO₂ is contact between the sample and stainless steel tweezers.

The role of impurities in ferromagnetism was also a hot topic at last year's American Chemical Society meeting in Washington, D.C. At a solid-state inorganic chemistry symposium, Stephanie L. Brock, a chemistry professor at Wayne State University, Detroit, reported that manganese doping in some phosphide semiconductors such as CdSnP₂ leads to formation of MnP "inclusions"—phase-separated crystallites within the bulk material. On the basis of experiments conducted with postdoc Jennifer A. Aitken and coworkers, Brock stressed, these undesirable phases or impurities can give rise to signals that may be incorrectly interpreted as originating from intrinsic ferromagnetism in the host semiconductor.

Brock's cautionary presentation followed reports of above-room-temperature ferromagnetism in a class of manganese-doped compounds known as chalcopyrite semiconductors, of which CdSnP₂ and other compounds with the II-IV-V₂ stoichiometry are members. That class of compounds was the focus of a computational study conducted by Steven C. Erwin and Igor Z utićof the Center for Computational Materials Science at the Naval Research Laboratory (NRL), in Washington, D.C. Using quantum mechanical methods, the NRL team investigated property relationships among 64 members of that class of materials and identified a small subset of promising-looking members that exhibit thermodynamic stability, favorable structural and manganese-doping properties, and other potentially beneficial characteristics (Nat. Mater. 2004, 3, 410).

Theoretical studies can guide experimentalists in selecting worthwhile synthesis targets. But the products need to be characterized thoroughly to interpret magnetic phenomena reliably. That's the take-home message from Seshadri, Aitken, and other solid-state chemists.

Now an assistant chemistry professor at Duquesne University, Pittsburgh, Aitken notes that low concentrations of unwanted phases in amorphous or nanocrystalline form can easily go undetected if samples are examined solely with powder X-ray diffraction methods, as has sometimes been the case. She advocates probing samples with a variety of analytical techniques to determine a dopant's oxidation state, its location in a host lattice, and other basic parameters. Synthesis, too, should follow a multiprong approach, Aitken suggests. Presently, she's exploring traditional high-temperature solid-state methods, low-temperature solvothermal techniques, and other procedures for preparing manganese-doped samples of CuInS₂, AgInSe₂, and other II-IV-V₂ compounds.

Another proponent of using traditional chemical methods to prepare magnetic semiconductors is Daniel R. Gamelin, an associate chemistry professor at the University of Washington, Seattle. By spiking cationic zinc solutions with small quantities of cobalt and then treating the solutions with tetramethylammonium hydroxide, Gamelin and coworkers prepare cobalt-doped ZnO in nanocrystalline (colloidal) form quickly and in large quantities. The group studies the nanocrystals as freestanding quantum dots or collects them as bulk samples on solid supports, modifies them with subsequent reactions, and examines their properties.

Last year, Gamelin and graduate student Kevin R. Kittilstved found that capping manganese-doped crystals with nitrogen from amine ligands, and then briefly heating the samples, leads to strong room-temperature ferromagnetism. In contrast, capping with oxygen does not produce the effect. According to the group, the explanation for the observations is that heating the crystals dopes them with nitrogen atoms, which introduces positive charge carriers, or holes. The holes, in turn, mediate interactions among manganese atoms, which align the magnetic dopants' electron spins, thereby endowing the crystals with ferromagnetic character (J. Am. Chem. Soc. 2005, 127, 5292).

Gamelin's group also applies conventional chemistry methods to the branch of spintronics research focused on exploiting individual electron spins for quantum computation. Just a few months ago, for example, Gamelin, graduate student William K. Liu, and others demonstrated that through the use of ultraviolet laser light, magnetically doped ZnO nanocrystals can be prepared as surprisingly stable charged species. The species, which contain trapped excess electrons, provide an opportunity to probe interactions between the nanocrystals' electrons and magnetic dopants-interactions that lie at the heart of semiconductor spintronics (J. Am. Chem. Soc. 2006, 128, 3910).

One strategy for probing interactions between magnetic dopants is controlling their placement in the host material on an atom-by-atom basis. That's exactly the approach taken by Princeton University physicists Ali Yazdani and Dale Kitchen and their coworkers. Using a scanning tunneling microscope tip, the team nudged one manganese atom at a time into select positions in a frigid GaAs crystal. Then, by measuring scanning tunneling spectra, the group searched for quantum mechanical splitting in the energies of the atoms' electronic states. This splitting is a telltale sign of strong interactions between the atoms.

Indeed, the energy-splitting effect was observed, the team reports, and was found to depend on geometrical placement of the atoms in the crystal. For example, a pair of manganese atoms separated just 5.6 Å along one lattice direction interacted only weakly, whereas a pair separated along a different direction by a wider spacing, 8 Å, coupled strongly (Nature 2006, 442, 436).

In just a few years, spintronics has grown from a relatively obscure topic to a multidisciplinary field that makes science news regularly. The potential for major technological advances has focused researchers from various backgrounds on the challenges of making new types of magnetic semiconductors. The ideal material with just the right combination of properties hasn't yet been found. But many scientists, including UC Santa Barbara's Awschalom, are highly optimistic about spintronics' future. "It's not easy to predict what amazing things will come of this field," he says. "The only thing I know for sure is that something will."