Technical / Research - Page 4

Researchers from Northwestern University developed a new method of building artificial neural networks using MRAM-based stochastic computing units. The researchers say that this design could enable AI devices that are highly energy efficient.

Embedded MRAM technologies are being adopted at major foundries, which enable the use of these technologies for unconventional computing architectures that use the stochasticity of MRAM cells (rather than their nonvolatility), to perform energy-efficient computing operations. MRAM cells exhibit stochastic switching characteristics, which is a challenge for reliable memory devices. But for neural networks, this can be taken advantage of if the MTJs are appropriately designed.

Researchers from Northwestern University and the University of Messina in Italy developed a new MRAM memory device composed of antiferromagnetic materials, which could be beneficial for use in AI systems and cryptocurrency mining.

Antiferromagnetic materials (AFM), offer inherently faster dynamics than ferromagnetic materials (FM), have no macroscopic magnetic poles and can be scaled much better. AFM-based memory cannot be erased with external magnetic fields which could prove to be a major security advantage.

IBM announced that during the 2020 IEEE International Electron Devices Meeting (IEDM 2020), that is now being held virtually, its researchers will reveal the first 14 nm node STT-MRAM. IBM says that efficient and high-performance STT-MRAM systems will help to address memory-compute bottlenecks in hybrid cloud systems.

IBM says that the 14 nm node embedded MRAM which will be revealed is the most advanced MRAM demonstrated to date. It features circuit design and process technology that could soon enable system designers to replace SRAM with twice the amount of MRAM in last-level CPU cache.

Researchers from Seoul's National University and Pohang's University of Science and Technology (POSTECH) report that a 2D iron germanium telluride (Fe₃GeTe₂, or FGT) layer is an excellent candidate to be used as the basis SOT-MRAM material.

An SOT-MRAM based on FGT is highly energy-efficient, in fact the researchers say that the measured magnitude of SOT per applied current density is two orders of magnitude larger than the values reported previously for other candidate materials.

Researchers from Tohoku University say they have developed the world's smallest (2.3 nm) high-performance magnetic tunnel junctions (MTJs).

The design is based on the Shape-anisotropy MTJ (developed by the same researchers in 2018) in which thermal stability is enhanced by making the ferromagnetic layer thick. In this new research, the scientists used a new structure that uses magnetostatically coupled multilayered ferromagnets - which enabled the scaling down to 2.3 nm diameters.

Researchers from the National Taiwan University developed an ultra-high performance MTJ, using a superlattice barrier and half-metallic magnets. The so-called superlattice-MTJ can be the basis of a new class of STT-MRAM (which the researcher call SS-MRAM) that achieves ultra-low power RA and write operations, high writing speed and unlimited endurance.

SS-MRAM adopts a superlattice barrier that replaces the MgO layer used in common STT-MRAM. The MgO layer is unstable and also suffers from a very large RA which results in high power consumption for writing operations. The superlattice has higher spin polarization than MgO and so the SS-MARM can provides not only ultra-high MR ratio but also ultra-low RA for high-speed and low power writing.

A team of researchers from the National Tsing Hua University (NTHU) in Taiwan have discovered that by adding a layer of platinum only a few nanometers thick, one can switch the pinned magnetic moments at MRAM cells at will. This was never achieved before, and can lead to faster and more efficient MRAM devices.

The platinum layer is placed between the two layers of the MRAM device - the upper layer, a freely flipping magnet used for data computation and the bottom layer that consists of a fixed magnet, responsible for data storage. Due to spin-orbit interactions, the electric current drives the collective motion of electron spins first. The spin current then switches the pinned magnetic moment effectively and precisely.

An international team of researchers led by Israel's Bar Ilan University have developed a new type of MTJ that has four resistance states. The researchers also demonstrated how it is possible to switch between these four different states using spin currents.

The four-state design uses a structure that is in the form of two crossing ellipses instead of one of the standard magnetic layers of the MTJ. Such a design could be used to create multi-level MRAM which stores data more densely compared to current MRAM memories which only have two states in each MTJ.

Researchers from the Korea Institute of Science and Technology (KIST) have managed to drastically incraese the speed of MRAM devices while reducing the power consumption by adding a layer of yttria-stabilized zirconia (YSZ) to MRAM devices.

The YSZ layer, which has high ion conductivity helps inject hydrogen ions into the MRAM cell. This resulted in an increase in the speed of the spin alignment direction changes 100-fold.

This is a sponsored post by Integral Solutions Int'l

MRAM is likely to be the most promising next-generation non-volatile memory technology today. Toggle MRAM and STT-MRAM are already entering the market, gaining market share in many applications. Next-generation MRAM technologies, such as SOT-MRAM could enable the replacement of even the fastest SRAM applications, with higher densities.

Source: Coughlin Associates, 2019

The MRAM production process has many stages, as device architecture is relatively complex, with a magnetic cell (frontplane) fabricated on top of a CMOS backplane. (use Figure 2 or Figure 3 from Coughlin). Measurement and characterization of devices are highly important, and the production of MRAM memories depend on measurement tools are are specialized for MRAM and STT-MRAM measurements.