Tikalon Header Blog Logo

Spintronics

February 1, 2011

My first job in industrial research was the development of better magnetic bubble materials using the liquid phase epitaxy crystal growth process. My objective was to design materials that would support stable circular magnetic domains of smaller and smaller diameters. In fact, I was doing nanotechnology in the early 1980s when I was working with magnetic bubbles in the 800 nm range. These magnetic domains needed to be stable not just at room temperature; they needed to be stable over a wide range of temperatures. As anyone who's worked with any sort of magnetic material will tell you, magnetic properties change a lot with temperature. When telecommunications satellites still used traveling wave tube amplifiers, the magnets that made this type of vacuum tube work were made from an expensive platinum-cobalt alloy, simply because it was the only magnet that kept the same properties over the expected range of temperature.

The application for the materials that I developed was bubble memory, but it was possible to make logic circuits from these same materials using the fact that magnetic bubbles will repel each other. Semiconductor memory technology advanced, and magnetic bubble work ceased, as I mentioned in a recent article (Culturomics, January 13, 2011). Well, the bubbles may have been retired, but the idea of magnetic logic lives on in spintronics.

Photograph of an Intel Magnetics magnetic bubble memory module (c. 1982)

Photograph of an Intel Magnetics magnetic bubble memory module (c. 1982) with a US quarter dollar coin for size comparison. (Photo by author).


As the spintronic name implies, spintronics attempts to utilize electron spin to do things normally associated with electronics. Magnetism, of course, derives from electron spin, so spintronics is a form of magnetic memory and logic. It could be viewed as the successor to magnetic bubbles, brought down to a size that makes magnetics competitive with semiconductor circuitry. Just as with magnetic bubbles, we need a means of producing a stream of electrons with a certain magnetic polarity. These would be spin-polarized electrons; that is, electrons with spin mostly aligned in one direction. We also need a spin detector, which is just a magnetic field detector. Unlike magnetic bubbles, which are discrete units of a magnetic state, spintronic electrons are an excitation wave in a material. They can be created by application of a magnetic field, and detected by device structures based on the giant magnetoresistance effect.

There are two factors driving spintronic research - Low power and non-volatility. These goals are so desirable that DARPA has budgeted $18 million for a four-year spintronics logic project. UCLA and the University of Notre Dame will share the funding about equally.[1] The project goal, which will be attempted by a different method by each team, is a two bit adder that uses just 10 attojoules of energy (10-18 joules) per computation. A CMOS equivalent circuit consumes a hundred times as much energy. Not only that, but the spintronic adder will maintain its state without power, which is something the CMOS adder can't do (Well, it could, with additional circuitry, but that would mean much more power).

Although the UCLA team is using a traditional spin wave approach, the Notre Dame team is basing its design on nano-magnets. Magnetic state would be transported between 60-90 nm nano-magnets that are separated by only 20 nm, somelike a ball knocking over bowling pins. The magnetic state of a nano-magnet can affect the state of adjacent magnets. In this way, the Notre Dame approach is much like magnetic bubble logic, but instead of the bubbles moving, it's the magnetic field. In my mind, it seems much simpler and less energy-intensive that the UCLA approach, but I'm an old bubble guy. UCLA defends its approach as being potentially denser. DARPA made the right decision by dividing the eggs into two baskets.

The present contender in the spintronic memory area is magnetoresistive random access memory (MRAM), which is already being sold by Everspin Technologies, of Chandler, Arizona.[2] Everspin sells a 35 nsec 16 Mbit MRAM operable from -40 to +125oC. Doubling-up on the spin concept, Everspin is a spin-off from Motorola. MRAM is, of course, non-volatile, and this is its biggest advantage, but it is also radiation-hardened. Presently, MRAM is not nearly as dense as the slower flash, and it's much more expensive, but there are no physical reasons why the bit density can't be very large.

Not surprisingly, IBM has been doing some spintronics research of its own. IBM has invented what they call a "racetrack memory," in which magnetic domains in wires are the memory elements, and electrical currents transport them through the wire to be read.[3-5] The bit density for this technology is potentially a hundred time greater than current memory technologies. The IBM research team reported on their racetrack memory in a recent issue of Science.[5] The IBM team demonstrated magnetic domain motions over several micrometers, domain speeds of hundreds of miles an hour, and precise positioning of domains in a wire by controlled current pulses. The ability to position the domains precisely was aided by the fact that the inertia of setting a domain in motion was balanced by the inertia of motion once the current is removed. Yes, the domain walls were found to have inertia, which means they also have mass, which is another interesting discovery, although it's not the gravitational kind of mass.[4]

If you want to generate a lot of press, combine two interesting technologies. This is what scientists at Seagate Technology did when they made a spintronic memristor.[6] The memristor, or memory resistor, hailed as the fourth fundamental circuit element to join the resistor, capacitor and inductor, was demonstrated by Hewlett Packard in May, 2008. The property of a memristor is that its resistance is determined by the quantity of charge that's flowed through it, and it will remember its last resistance state. Seagate's spin on the memristor (pun intended) was to describe a device for which the resistive element has a domain wall in it. As the domain wall is shifted, the resistance changes, and the resistance state is non-volatile. The Seagate researchers note that their devices are easy to make. Seagate's spintronic memristor was announced just a year after HP's memristor. Both technologies are still in their infancy, and I expect a lot of interesting memristor papers in 2011.

References:

  1. Joseph Calamia, "Magnetic Logic Attracts Money," IEEE Spectrum, January, 2011.
  2. Salah M. Bedair, John M. Zavada and Nadia El-Masry, "Spintronic Memories to Revolutionize Data Storage," IEEE Spectrum, November, 2010.
  3. Super Memory Breakthrough: Store Every Movie Made This Year on Your Phone (With Room to Spare), Extremetech, December 27, 2010.
  4. IBM Almaden Research Center, Spintronics Devices Research
  5. Luc Thomas, Rai Moriya, Charles Rettner and Stuart S.P. Parkin, "Dynamics of Magnetic Domain Walls Under Their Own Inertia," Science, vol. 330, no. 6012 (December 24, 2010), pp. 1810-1813
  6. Neil Savage, "Spintronic Memristors," IEEE Spectrum, March, 2009.

Permanent Link to this article

Linked Keywords: Applied physics; industrial research; magnetic bubble; liquid phase epitaxy; crystal growth; magnetic domains; magnetic; telecommunications satellite; traveling wave tube; vacuum tube; platinum; cobalt; bubble memory; semiconductor memory; spintronics; electron spin; electronics; magnetic polarity; spin-polarized electrons; giant magnetoresistance; non-volatile memory; DARPA; UCLA; University of Notre Dame; adder; atto; joule; CMOS; nanotechnology; nano-magnets; bowling pins; dividing the eggs into two baskets; magnetoresistive random access memory; MRAM; Everspin Technologies; Chandler, Arizona; Motorola; radiation-hardened; IBM; racetrack memory; Science; inertia; effective mass; equivalence principle; Seagate Technology; memristor; resistor; capacitor; inductor; Hewlett Packard; electric charge; domain wall.