NYU physicists recognized for discovering novel spin-based memoryOctober 12th, 2012 in Physics / General Physics
A discovery by NYU physicists that has potential to significantly enhance computer memory has been cited by Applied Physics Letters. Their work creates a new type of “spin-based memory” that has the potential to replace all conventional memory, such as the semiconductor memory in computers and portable devices. The above image shows the magnetic layer that stores information; it can be magnetized to the left or the right to represent one bit of information. Layer P provides spins while layer SAF is used to read out the information. The forces on the magnetization, shown below, drive it to reverse on sub-nanosecond time scales. These are presented by arrows on a sphere.
A discovery by New York University physicists that has potential to significantly enhance computer memory has been cited by Applied Physics Letters as "one of the most notable" articles the journal has published over the past four years. The work appears in the journal's "50th Anniversary Collection," which includes the most noteworthy articles it has published over the last 50 years.
The article, authored by NYU's Huanlong Liu, a doctoral student, post-doctoral researchers Daniel Bedau and Dirk Backes, and Physics Professor Andrew Kent, along with researchers at HGST and Singulus Technologies in Kahl am Main, Germany, may be downloaded here.
"Spin-based memory" seeks to manipulate magnetism of different materials in response to electric currents and fields. The magnetic coupling occurs through the exchange or flow of electron "spin angular momentum"—the fundamental property of electrons that gives rise to magnetism in materials. When a current flows in a magnetic material, the spins of the electrons move and can transport spin angular momentum from one region to another. This transport of spins can cause the magnetization to rotate. As an analogy, linear momentum, like that in a breeze on a windy day, can cause a wind turbine to rotate. Linear momentum is transferred into angular momentum—the rotation of the turbine blades.
In the Applied Physics Letters article, the research team, led by NYU physicist Kent, describes how it sought to store information using nanomagnets—a billionth of a meter in size—in order to write information with spin-current pulses.
This approach improves upon what is typically used in computers and portable devices, semiconductor random access memory (RAM), which involves storing information by charging a capacitor. However, this charge leaks away and thus the device needs to be read and refreshed periodically. For example, Dynamic RAM (DRAM), currently the fastest type of computer memory, is refreshed 1,000 times per second, consuming a great deal of energy. By contrast, nanomagnets retain their direction of magnetization without the need for a source of energy. In fact, energy is needed only to write or read the information, not to retain it.
In the Applied Physics Letters article, the researchers describe the key to their discovery: in order to switch a nanomagnet's magnetization quickly, a memory device should use electron spins oriented orthogonally—at a 90-degree angle—to the nanomagnet's magnetization direction. The magnetization then rotates rapidly about a direction set by the injected spin direction—and, in doing so, takes the fastest possible path from its initial to final orientation.
The NYU device is called an orthogonal spin-transfer magnetic RAM—OST-MRAM. The group recently built and tested OST-MRAM devices and demonstrated that their performance is far superior to conventional magnetic memory devices, both in terms of write speed and energy consumption.
"The memory device is both significantly faster and requires much less energy than a conventional memory, offering the potential for more energy-efficient computing devices," explained Kent.
Provided by New York University
"NYU physicists recognized for discovering novel spin-based memory." October 12th, 2012. http://phys.org/news/2012-10-nyu-physicists-spin-based-memory.html