MIT model explains how the brain can learn novel tasks while still remembering what it has already learned.
A new computer memory device developed by researchers at the Francis Bitter Magnet Laboratory has the potential to vastly increase the data storage capabilities of computers.
Dr. Jagadeesh S. Moodera, a research scientist at the Magnet Lab, along with a team of undergraduates and senior scientist Dr. Robert H. Meservey, have fabricated a "spin-dependent tunnel junction device" whose feasibility was first predicted by researchers about 25 years ago. Its primary uses are in magnetic read-heads and as magnetic random-access memory in computers, but it also has applications as sensors in devices ranging from automobiles to medical diagnostic tools.
Computers record data in tracks on the circular surface of hard disk drives that spin like phonograph records, only much faster (about 7,000 rpm). A write-head records new information, while a read-head searches for and reads those bits of data. Both heads work by detecting and interacting with the small magnetic fields on the disk drive. The smaller and more sensitive the read-head sensors are, the more data can be stored on the disk.
The state of the art for read-heads now found in computers uses a property of magnetic materials called anisotropic magnetoresistance or AMR. More recently, devices using giant magnetoresistive (GMR) materials are being developed in which the read-head sensor is composed of about 60 layers of alternating ferromagnetic and nonmagnetic metal thin films. During manufacturing, the thicknesses of these layers must be controlled very precisely; erring by the thickness of even one layer of atoms will degrade performance.
In contrast, the tunnel junction magnetoresistance (JMR) device developed by the MIT researchers consists of just three layers: two ferromagnetic films on either side of an ultrathin insulating material, Dr. Moodera explained. This inner layer or tunnel barrier must be extremely thin (1-2 nanometers), though the thickness of the outer layers needn't be as carefully controlled as with GMR devices. Because the three-layer assembly is thinner (at less than 50 nanometers) than a 60-layer GMR read-head, it is more sensitive, just as a phonograph can "read" more information from a record if the needle is the width of a pin rather than a ballpoint pen tip. Thus, the amount of information a disk can hold is, in effect, significantly increased.
Although the intervening layer is made of insulating material, it is so thin that when voltage is applied between the outer layers, some current can still pass through because of the wave nature of electrons. They are able to do this because of a purely quantum phenomenon called tunneling, Dr. Moodera said. When an electron's spin (its magnetic orientation) is parallel to the direction of the magnetic orientation in the adjacent film, it can slip across the barrier "like a member entering a members-only club," he said.
However, because of the inner layer's insulating properties, the resistance is high, which means voltage signals are high and thus easier to detect, requiring much less current to do so. Consequently, "the power required is almost negligible," he said. This could lead to use in magnetic RAM for computers; a JMR device could allow a computer to write and retain information with a fraction of the battery power now required, thus increasing computer portability.
The JMR device has many advantages over current technology: it is smaller, economical to produce, relatively temperature-insensitive and highly resistant to radiation, Dr. Moodera said. Most importantly, it has the potential to raise the magnetic storage density of hard drives to tens of gigabits per square inch while using less power than AMR or GMR devices, which can store one to 10 gigabits per square inch.
Dr. Meservey and Dr. Paul Tedrow of the Magnet Lab showed many years ago in other experiments that such a device was theoretically possible, but it wasn't created until this year. "I didn't realize myself at first that this was such a big issue," Dr. Moodera said. "It just opened up the whole field. this could change the digital storage industry in a dramatic way."
He and his colleagues have applied for patents on their invention, which was initially described in Physical Review Letters in April 1995 (further results are being published this month). Companies including IBM, Hewlett-Packard and Motorola are working on commercial development of JMR devices, said Dr. Moodera, who received the 1995 IBM Research Partnership Award for his work. Among those assisting in the research are postdoctoral fellow Janusz Nowak; Clifford Tanaka, a graduate student in materials science and engineering; Terrilyn Wong, a junior in materials science and engineering; Patrick LeClair, a sophomore in physics, and freshman Lisa Kinder. The research was sponsored by the Office of Naval Research, the National Science Foundation and UROP.
A version of this article appeared in MIT Tech Talk on May 1, 1996.