ORNL Review Home Page
Featured in This Edition
Previous Article
Next Article
Search the ORNL Review Site
Comment on this article

Modeling Magnetic Materials for Electronic Devices

Digital cameras, computer-game-playing stations, laptop computers, and portable devices that record music and play back television programs use magnetic disk drives as small as a fat credit card—taking up much less space than the shoebox-sized drives in desktop computers. The number of bits of information (1's and 0's) per square inch is doubling every year as magnetic disk drives shrink yet store and access more data in less time. But in its quest to make disk drives that are extremely fast and small, the magnetic recording industry will soon be hitting a wall. Currently used materials will eventually reach fundamental limits in performance as disk drives are further downsized.

ORNL and the Department of Energy's Brookhaven National Laboratory (BNL) are collaborating with industry to address the fundamental scientific problems confronted by the magnetic recording industry. The industrial partners in this Laboratory Technology Research (LTR) project, funded by DOE's Office of Science, are Seagate Recording Heads, Inc., the world's largest manufacturer of computer hard disk drives; the Almaden Research Laboratory of IBM, the inventor of the magnetic hard disk drive (in 1956) as a direct access storage device; and Imago Scientific, Inc., a startup company in Madison, Wisconsin, that can analyze magnetic materials using its patented local electrode atom probe (LEAP) technology. The ORNL researchers involved in this LTR project are Bill Butler, Malcolm Stocks, Mike Miller, and Balazs Ujfalussy, all of the Metals and Ceramics (M&C) Division; Thomas Schulthess of the Computer Science and Mathematics Division; and Xiaoguang Zhang of the Computational Physics and Engineering Division.

"Designers of the next generation of magnetic readers and magnetic media are working with badly blurred vision," Butler says. "They are unable to determine the structure of magnetic material at the necessary atomic scale. We plan to use atomic probe technology and computational simulation to better understand the limits of currently used magnetic disk drive materials and to identify new material structures and deposition processes that may overcome these limits."

To understand the problems, first it is necessary to understand how today's small magnetic disk drives work. Most of these drives read the stored data using giant magnetoresistive (GMR) read heads.

"The read head is like a jet flying at supersonic speeds a meter above a pasture and counting the grass blades pointed in two different directions," Schulthess says. "Think of each grass blade as a magnetic particle, or region of magnetization, pointing in one direction—or the opposite direction—like a compass needle. The directions, or magnetic moments, represent bits of information, either a 1 or 0."

The reader is a spin-valve device consisting partly of layers of nickel-iron and cobalt (separated by copper spacers). These layers are free to rotate, or spin, in response to an applied magnetic field. The total thickness of these free magnetic layers may be only 10 nanometers. Based on its orientation, the small magnetic field of each magnetic particle on the disk affects the electrical resistance of the read head.

In operation, the magnetic moments of one of the "pinned" magnetic layers are held fixed by an adjacent magnet, which has an internal magnetic field that does not destroy data on the magnetic media; it is made of an antiferromagnetic (AFM) material, such as a manganese alloy. A current is passed between the pinned and free magnetic layers. When the free layer senses a magnetic moment signifying "1," it stays or becomes parallel to the pinned layer; when it senses a "0," it adopts a position perpendicular to the pinned layer. In one position, the electrical resistance is high between the magnetic layers (because the magnetic moments of the free and pinned layers are not parallel), so the current passing through them is low; in the other position, the reverse is true. The shifting strengths of an electrical signal as a result of changes in resistivity allow the read head to copy stored data to a computer.

Schematic diagram of a GMR read sensor used in a disk drive. The sensor responds to the changing magnetic fields on the disk as it spins beneath it. The blocks indicate different layers of magnetic and nonmagnetic material, which may be only a few nanometers in thickness. A resistance change occurs when the magnetic fields from the bits on the disk cause a change in the relative orientation of the magnetic moments in the two magnetic layers. (Illustration enhanced by Jane Parrott)

What are the problems that will result from downsizing disk drives made of currently used materials? According to Butler, "As the disk drive density gets higher, the magnetic particles representing each bit get smaller. So the read head must be shrunk to several hundred angstroms. Because Seagate wants the fixed AFM magnet to be much thinner, the material structure may have to be changed to make it work effectively.

"Secondly, the smaller read head must have a higher sensitivity to detect smaller bits. As the density of the regions of magnetization become smaller, the magnetic fields that must be sensed to read the data get weaker. We need to obtain the largest possible change in resistance for relatively small changes in the magnetic field. One solution may be to grow smooth films with the proper interfacial structure to optimize the magnetic performance of the device."

The third problem is that, as bits get smaller, the superparamagnetic limit is reached. "When magnetic particles become small enough, their magnetism may be affected by changes in temperature," Butler says. "These thermal fluctuations can upset a tiny magnetic particle's magnetization, causing loss of valuable data. To solve this problem, a magnetic disk could be made out of a magnetically hard material, such as neodymium iron boride. But that would create a new problem: You could not write new information on such a disk without using higher magnetic fields. To generate these higher magnetic fields, the write head must be made of material that has a higher saturation magnetization."

Another LTR objective is to create a structured material that has a higher number of unpaired electron spins per unit volume, or a higher saturation (density) of magnetization. Iron has a magnetic moment of 2, or 2 unpaired spins per atom; manganese has a moment of 5 and some rare earths, 7, but their spins are often in the opposite rather than the same direction. Using computational simulation at the IBM supercomputer at DOE's Center for Computational Sciences at ORNL, Butler, Stocks, Schulthess, Ujfalussy, and Zhang hope to design artificial, layered structures for smaller read heads. These structures may contain iron, manganese, and rare-earth thin films that offer a higher magnetization saturation, as well as magnetic softness and corrosion resistance. Such structures may also be used for write heads for magnetic recording.

Atom probe image of a magnetic multilayer. The individual atomic planes can be clearly discerned.

The LTR project also has an objective to use computational simulation to help enable atomic probe field ion microscope tools (such as those employed by Mike Miller of the M&C Division, as well as the LEAP technology developed by Imago Scientific) to better image metallic multilayers with atomic-scale resolution. Dave Larson, formerly of the M&C Division and now with Seagate, has already done atomic probe studies of a structure containing layers of nickel, cobalt, and copper. His images show some diffusion of nickel atoms into the cobalt layer and diffusion of cobalt atoms into the nickel layer. Such interdiffusion could make differences in resistivity impossible to detect.

"Atom probe information may allow us to compare deposition processes to determine which ones make thin-film structures that work best as read and write heads in very small disk drives," Butler says.

Plot of scattering intensity caused by cobalt impurities in copper.

Using computer simulation at ORNL, Stocks and Schulthess recently obtained insights about antiferromagnetism in iron manganese (FeMn), an alloy used in the fixed AFM magnet of spin-valve GMR read heads. They studied the alloy's non-collinear magnetic structure, in which the magnetic moments point at angles to each other rather than up or down.

"We did calculations using constrained density functional theory and spin dynamics," Stocks says. "Our goal was to understand the alloy's noncollinear antiferromagnetic 3Q magnetic structure, which has the lowest known energy level for a system of atoms and electrons in a crystal, compared with the 1Q and 2Q structures. We predicted there is a more relaxed, even lower-energy-level magnetic state that we call 3QR in which R stands for relaxed. In this state, the magnetic moment orientations, shown as arrows, are pointed in a slightly different direction, according to our simulation. This insight could lead to a design of improved GMR devices."

Detailed magnetic structure of an iron manganese alloy showing small relaxations of the magnetic moment directions relative to the 3Q structure.

Stocks and his colleagues have also been involved in another problem in which a material currently being used in a basic electronic device will hit a wall as the device is downsized. As transistors are shrunk to increase a chip's computing power, the use of silicon dioxide to block or allow electron flow will limit transistor performance as an on-off switch, or memory unit, for storing a bit (1 or 0). When silicon dioxide becomes too thin, electrons will leak from it by quantum tunneling, making it useless as a gate electrode material.

Rodney McKee of the M&C Division and his colleagues have shown that silicon dioxide can be replaced by certain crystalline oxides whose superior electrical properties will allow reduction in transistor size without loss of performance; more research is being done on this project by ORNL in collaboration with Motorola and DOE's Pacific Northwest Laboratory under a cooperative research and development agreement. Stocks and his colleagues are using the IBM supercomputer to model the interfaces between silicon and silicon dioxide and between layers on silicon. These layers consist of strontium disilicide, strontium oxide, titanium dioxide, and strontium titanate (to make the gate electrode ferroelectric).

"We are simulating the electronic structure of strontium disilicide and trying to determine how many layers of strontium oxide are needed to make it a good insulator," Stocks says. "Sometimes the experimental observations do not agree with our calculations because we find electrons diffusing between layers, while the observations suggest that the interfaces are clean." The simulations could aid the experimenters in improving their deposition processes to make a more effective thin-film structure as an advanced transistor gate material.

When a material's performance in an electronic device is about to hit a wall as the device gets very small, some ORNL materials scientists focus on designing advanced materials to allow the electronic industry to leap to the other side.

Beginning of Article

Related Web sites

Brookhaven National Laboratory
ORNL Metals and Ceramics Division
ORNL Computer Science and Mathematics Division
ORNL Computational Physics and Engineering Division
Center for Computational Sciences