New magnetoresistance effect leads to four-state memory device

multistate memory
(Left) With a single ferromagnetic layer, the system has two resistance levels. (Right) Adding another ferromagnet to the system creates four levels of resistance, corresponding to the four different magnetic states indicated by the arrows. Credit: Avci et al. ©2017 American Institute of Physics

(Phys.org)—In 2015, scientists discovered a new magnetoresistance effect—that is, a new way in which magnetization affects a material's electric resistance—but hadn't yet found a promising application for the discovery, beyond the existing technologies. Now in a new paper, the same researchers have demonstrated that the effect can be used to design memories with four distinct stable magnetic states, allowing the memories to store four bits of information in a single magnetic structure.

The researchers, Can Onur Avci et al., at MIT and ETH Zürich, have published a paper on the new concept in a recent issue of Applied Physics Letters.

"With some and structural optimization, the bit density of existing random access memory devices may be increased by several factors, with the possibility of all-electrical operation," Avci told Phys.org.

Magnetoresistance effects date back to around 1850, when Lord Kelvin demonstrated that applying a magnetic field to a metal object increases the object's electric resistance in one direction and decreases it in the perpendicular direction. Since then, several other types of magnetoresistance have been discovered. Most notably, Albert Fert and Peter Grünberg won the 2007 Nobel Prize in Physics for their discovery of giant magnetoresistance, which is used to make magnetic field sensors that are found in many of the hard disk drives in today's computers.

In 2015, scientists discovered the newest magnetoresistance , called unidirectional spin Hall magnetoresistance. This effect differs from other kinds of magnetoresistance in that the change in resistance depends on the direction of either the magnetization or the electric current. As the scientists explain, this direction-dependent effect occurs because the spin-polarized electrons created by the spin Hall effect in a nonmagnetic layer are deflected in opposite directions by the magnetization of the adjacent .

Previously, this new effect was demonstrated in two-layer structures consisting of a nonmagnetic and a magnetic layer. But by adding another magnetic layer, the researchers achieved a great potential advantage for memories: the ability to distinguish between not just two, but four magnetic states. Other types of magnetoresistance effects are only sensitive to the relative orientation of the magnetizations (parallel or antiparallel), although it's possible to have four distinct magnetic states. Because the new effect is sensitive to the magnetization direction of individual layers, it can distinguish between all four states.

The researchers then demonstrated four distinct resistance levels corresponding to the four different in their three- device. They showed that the four resistance levels can be read out by a simple electric measurement, paving the way for the development of an all-electrical multi-bit-per-cell memory device.

The researchers expect that it will be possible to scale up this memory device to higher bit densities by adding more layers, which could realistically enable eight different magnetization states, each with its own unique resistance level. In the future, the researchers also plan to look for materials that exhibit a larger unidirectional spin Hall effect, which would further enhance the performance of these memory devices.


Explore further

Smart multi-layered magnetic material acts as an electric switch

More information: Can Onur Avci et al. "A multi-state memory device based on the unidirectional spin Hall magnetoresistance." Applied Physics Letters. DOI: 10.1063/1.4983784

ABSTRACT
We report on a memory device concept based on the recently discovered unidirectional spin Hall magnetoresistance (USMR), which can store multiple bits of information in a single ferromagnetic heterostructure. We show that the USMR with possible contribution of Joule heating-driven magnetothermal effects in ferromagnet/normal metal/ferromagnet (FM/NM/FM) trilayers gives rise to four different 2nd harmonic resistance levels corresponding to four magnetization states (⇉⇉, ⇄⇄, ⇆⇆, ⇇⇇) in which the system can be found. Combined with the possibility of controlling the individual FMs by spin-orbit torques, we propose that it is possible to build an all-electrical lateral two-terminal multi-bit-per-cell memory device.

Journal information: Applied Physics Letters

© 2017 Phys.org

Citation: New magnetoresistance effect leads to four-state memory device (2017, June 5) retrieved 18 July 2019 from https://phys.org/news/2017-06-magnetoresistance-effect-four-state-memory-device.html
This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no part may be reproduced without the written permission. The content is provided for information purposes only.
767 shares

Feedback to editors

User comments

Jun 05, 2017
four distinct stable magnetic states, allowing the memories to store four bits of information in a single magnetic structure.


Nope. Four states is two bits. 00, 01, 10, 11.

Doubling the number of distinct states or levels increases the number of bits by one.


Jun 05, 2017
They have four states, so they can store two bits of information, not four, as stated in the article.

One bit = two states.
Two bits = four states
Three bits = eight states
Four bits = Sixteen states, etc.

Jun 05, 2017
Yes, that's correct, four states is two bits. Every CS on the site, and lots of the EEs as well, are going to point this out immediately. It leapt off the page at me.

Do better, physorg.

Jun 05, 2017
Ok, we've always had this. How about non-linear material, i.e. switchable light, frequency, modulation ... every "object" can know be defined within the HW. Get it? Optically. This, meh

Jun 05, 2017
Four states, yes, no, I don't know, and hey baby?

Jun 05, 2017
This comment has been removed by a moderator.

Jun 05, 2017
No voltage equals a zero does it not ?
Four resistances = four voltages so therefore 4 "on" bits ?.
or 0000 through to 1111 on one cell

Jun 05, 2017
Back in the day (early sixties) I remember russians were running terniary? tri-bit computers? I would assume,as opposed to binary digits (bits) they may have called them trits. So in our binary computers on/off represented bit 0 or bit 1 (bulb on or off) each bit could represent two states. Russian computers probably needed say off/red/green bulb to represent 'tri' state for each 'bit' or 'trit'. - three 'states' 0,1,2. Now for 4 states or 'qits' we may have off/red/green/blue - so each 'bit' now represents 4 states, as in agreement with the article. On a per bit readout we may see say 302103 instead of 011011 - the first number spans 4 power six, the second 2 power six. I don't want to think of building a '4 bit-state system' - probably just convert 'qits' to bits and leave it at that

Jun 05, 2017
eight states if you include 0 or no voltage, and it's still binary as well

Jun 05, 2017
Actually, while four states can be represented by two bits it's natively a quaternary number system, not binary. Two bits is two digits, which requires more complex circuitry. Real improvement will harness the native quaternary states as a quaternary symbol. Lower power, less complexity, smaller circuits, higher density.

And since memory cells are the basis for logic, it will be interesting to see what kind of quaternary logic circuits can be made. CPUs are even more in need of lower power, less complexity, smaller circuits and higher density per performance than memory is.

Jun 06, 2017
@Emcee, so you're thinking of making registers with these? Not a bad idea. The challenge might be the microcode.

Keep in mind as well that these aren't replacements for transistors. And transistors are pretty much two-state devices.

Jun 06, 2017
Something quite similar to another implementation with MO materials:
http://www.mdpi.c...8/4/1976

Jun 06, 2017
eight states if you include 0 or no voltage, and it's still binary as well

magnetic field or no magnetic field, I think you might mean. And it's more if you can combine each of the different states with any other in a binaural fashion. Not even to mention the results if you combined THREE or FOUR states...

Jun 06, 2017
Back in the day (early sixties) I remember russians were running terniary? tri-bit computers? I would assume,as opposed to binary digits (bits) they may have called them trits. So in our binary computers on/off represented bit 0 or bit 1 (bulb on or off) each bit could represent two states. Russian computers probably needed say off/red/green bulb to represent 'tri' state for each 'bit' or 'trit'. - three 'states' 0,1,2. Now for 4 states or 'qits' we may have off/red/green/blue - so each 'bit' now represents 4 states, as in agreement with the article. On a per bit readout we may see say 302103 instead of 011011 - the first number spans 4 power six, the second 2 power six. I don't want to think of building a '4 bit-state system' - probably just convert 'qits' to bits and leave it at that

I'm of German/Russian/Scottish heritage - it was "tits" n "quits". Layered. Very quantum in nature...

Jun 06, 2017
The linked article specifically states individual switching of bits.
A binary cell representing 0 to 15.

Jun 06, 2017
Da Schneib:
@Emcee, so you're


Well, registers should probably be fast, and no performance data from this basic materials research. Except that they tried but couldn't rule out magnetothermal effects, which can set upper bounds on access frequencies.

The basic devices studied are not direct replacements for transistors, but engineers have created magnetoresistive FETs. Perhaps these new devices could produce quaternary transistors.

Integrating quaternary storage and logic with the conventional binary circuits that make up the rest of any useful device is a real challenge. Software that exploits quaternary states too. If quaternary MR components aren't cheap and simple to exploit compared to their performance benefits, they'll never be used. Except perhaps academically, which could eventually deliver useful devices and uses of them.

Jun 06, 2017
OK, 2, 4, or more bits per signal is a function of the signal and the processing. With a nonlinear medium, there may be multiple bits and multiple transactions that take place at the same time.

The materials are rare, so what is required are new optical materials. Note: with optics, this will be ... dunno

Jun 06, 2017
Da Schneib:
@Emcee, so you're
Well, registers should probably be fast, and no performance data from this basic materials research.
In general reading and writing magnetoresistive memory should be on very sub-nanosecond timescales; one of the big advantages is you don't need to move electrons; you only need to change their spins. This also helps with:
Except that they tried but couldn't rule out magnetothermal effects, which can set upper bounds on access frequencies.


The basic devices studied are not direct replacements for transistors, but engineers have created magnetoresistive FETs. Perhaps these new devices could produce quaternary transistors.
As far as I know, this will still make two-state transistors. We can discuss the active region of transistors as opposed to the rails if you like. I'd want to see more detail on these magnetoresistive FETs before commenting further. I don't deny it but it seems unlikely.
[contd]

Jun 06, 2017
[contd]
Integrating quaternary storage and logic with the conventional binary circuits that make up the rest of any useful device is a real challenge.
It's still easily convertible to binary; I don't see a problem making registers in a microprocessor be quaternary and having them interact with binary circuits. And there are optimizations in transistor circuits that will make them much faster; but there's spintronics for that too. I haven't started to probe the state of the art there yet.

Software that exploits quaternary states too. If quaternary MR components aren't cheap and simple to exploit compared to their performance benefits, they'll never be used. Except perhaps academically, which could eventually deliver useful devices and uses of them.
Meh, once you got the hardware you write the compiler for it. We been doin' this for decades.

Please sign in to add a comment. Registration is free, and takes less than a minute. Read more