MIPT researchers teamed up with their colleagues from the Kotelnikov Institute of Radio Engineering and Electronics (IRE) of the Russian Academy of Sciences (RAS) and the International Associated Laboratory of the Critical and Supercritical Phenomena in Functional Electronics, Acoustics, and Fluidics for a successful demonstration of a new kind of computer memory. Their paper was published in Applied Physics Letters. A transition to the newly demonstrated type of memory could enable a substantial energy saving, as well as the instantaneous startup of devices based on this technology. Random access memory, or RAM, is one of the principal components of any computer or smartphone. The most common type of RAM is known as dynamic random access memory, or DRAM for short. It is a semiconductor memory based on a rather simple principle. In DRAM, each memory cell consists of one capacitor and one transistor. The transistor is used to admit current into the condenser, allowing it to be charged and discharged. The electrical charge of the capacitor stores binary information, which is conventionally represented as zeros (not charged) and ones (charged).
"So far, the RAM technology has been rapidly advancing, with memory modules becoming ever faster. However, this type of memory has one major limitation that cannot be overcome, namely its low energy efficiency," says principal investigator Sergei Nikitov, who is deputy head of MIPT's Section of Solid State Physics, Radiophysics and Applied Information Technologies, corresponding member of RAS, and the director of IRE RAS. "In this paper, we present the magnetoelectric memory cell. It will reduce bit-reading and -writing energy consumption by a factor of 10,000 or more."
A cell in the magnetoelectric memory, also known as MELRAM, consists of two components with remarkable properties. The first of the two is a piezoelectric material. Piezoelectricity is the property of certain materials that are deformed in response to applied voltage and, conversely, generate voltage under mechanical stress. The other MELRAM component is a layered structure characterized by a high magnetoelasticity -- the dependence of magnetization on the elastic strain. Because the structure is anisotropic -- that is, it is organized differently along different axes, -- it can be magnetized along two directions, which correspond to the logical zero and one in the binary code. In contrast to dynamic RAM, magnetoelectric memory cells are capable of maintaining their state: They need not be continually rewritten and do not lose information when power is cut off.
"We built a test piece about 1 millimeter across and showed that it works," says Anton Churbanov, a Ph.D. student at the Department of Physical and Quantum Electronics, MIPT. "It is worth noting that the structures we used could serve as the basis of nano-sized memory cells, whose dimensions are similar to those of regular RAM cells."
At the heart of the study is a novel data reading mechanism, providing an alternative to the sophisticated magnetic field sensors used in earlier MELRAM cells, which do not allow for easy downscaling. As it turned out, there is a simpler way to read information, which does not require such complicated arrangements. When a voltage is applied to the memory cell, the piezoelectric layer of the structure is deformed. Depending on the nature of the strain, magnetization assumes a particular orientation, storing information. The changing orientation of the magnetic field gives rise to increased voltage in the sample. By detecting this voltage, the state of the memory cell can be determined. But the reading operation might affect magnetization; therefore, it is necessary to recommit the value that has been read to the memory cell.
The authors of the paper say their solution can be scaled down without any adverse effect on its efficiency. This makes MELRAM promising for computing hardware applications mandating low energy consumption.