MRAM (magnetoresistive random access memory) is a method of storing data bits using magnetic states instead of the electrical charges used by dynamic random access memory (DRAM).
Scientists define a metal as magnetoresistive if it shows a slight change in electrical resistance when placed in a magnetic field. By combining the high speed of static random access memory (SRAM) and the high density of DRAM, proponents say MRAM could be used to significantly improve electronic products by storing greater amounts of data, enabling it to be accessed faster while consuming less battery power than existing electronic memory.
The U.S. Defense Advanced Research Projects Agency (DARPA) provided funding to help private industries conduct research into the potential of MRAM. Beginning in 1995, DARPA began funding three private consortia researching the viability of making MRAM a general-purpose memory with high density, high speed and low power usage. Leading the three consortia were IBM, Motorola and Honeywell. Hewlett-Packard Enterprise, Matsushita, NEC, Fujitsu, Toshiba, Hitachi and Siemens also have invested in MRAM research.
How MRAM works
Unlike DRAM, which uses electrical charge to determine if a bit is a binary 1 or 0, magnetoresistive memory uses a pair of ferromagnetic metal plates separated by a thin insulating material layer. One plate is a permanent magnet, always magnetized, and the other can be magnetized. The orientation of the two magnetic fields defines the 1 or 0 in a binary bit. This basic structure is called a magnetic tunnel junction (MTJ). Arrays of such magnetic tunnel junctions make up the memory device, like arrays of transistors in an integrated circuit make up random access memory (RAM).
The MTJ works because of a quantum phenomenon called electron tunneling. The insulating layer is only a few nanometers thick and that allows electrons to tunnel through it from one plate to the other. Depending on if the magnetic fields of each plate are oriented in parallel or not will determine how much tunneling will occur, and that changes the electrical resistance of the magnetic tunnel junction, which determines the 1 or 0 in a binary bit.
Early MRAM systems used electrical currents to induce the magnetic field in the MTJ to read or write the cell, but that requires more power than is ideal for modern computer systems. Today, most effort is being placed in using something called spin transfer torque (STT) to read or write an MRAM cell.
Since magnetoresistive random access memory can perform read and write operations faster than DRAM using less power, while being a nonvolatile memory, it is considered a "universal memory" that can be applied to any use, from system computing to storage. Unlike the current most popular nonvolatile memory type, flash memory, MRAM doesn't wear out and can be read from and written to theoretically until the physical material degrades.
Spin transfer torque doesn't look at the charge induced by the flow of electrons from one layer of an MTJ to another, but at the torque applied to the changeable layer by the angular momentum -- or spin -- of the electrons in the flow. Enough spin in the electron flow from one plate in the MTJ to the other can change the magnetic orientation on the second plate.
A magnetic tunnel junction that uses STT draws much less power than an electrically induced MTJ. In addition, STT-MRAM can be made much smaller, allowing for greater memory density in the memory devices that use it. However, to perform at high speeds, even STT-MRAM requires more power than is commercially feasible to make it a replacement for most uses of DRAM.
MRAM vs. DRAM
Conventional RAM computer chips store information as long as electricity flows through them. Once power is turned off, the information is lost unless it has been copied to a hard drive or USB flash drive.
In contrast, MRAM retains data after the power supply is cut off. Replacing DRAM with MRAM could prevent data loss and enable computers to start instantly without waiting for software to boot up.
Current state of MRAM
There are a few commercially available MRAM products on the market. In 2012, Everspin Technologies Inc. introduced a 64 Mb STT-RAM module -- Everspin uses the acronym ST-MRAM in its marketing materials -- that was fully compatible with the then-standard DDR3 RAM interface and packaged in a typical DDR3 form factor.
In 2013, Buffalo Technology became the first company to announce a commercial application of Everspin's ST-MRAM module, using it in a Serial Advance Technology Attachment (SATA) III solid-state drive as cache memory.
IBM and Samsung made a joint announcement in July 2016 that they could make MRAM at an 11-nanometer scale. In April of that same year, Samsung announced it would have MRAM products very soon.