Researchers take a step toward valleytronics

Apr 27, 2011
The band structure of graphene with its two valleys is shown in blue and red. Credit: Naval Research Laboratory

Valley-based electronics, also known as valleytronics, is one step closer to reality. Two researchers at the Naval Research Laboratory (NRL) have shown that the valley degree of freedom in graphene can be polarized through scattering off a line defect. Unlike previously proposed valley filters in graphene, which rely on confined structures that have proven hard to achieve experimentally, the present work is based on a naturally occurring line defect that has already been observed.

The discovery was published in on March 28, 2011 and was also the subject of a separate Viewpoint article in Physics.

Information in , either classical or quantum, is generally carried by electrons and holes. The information can be encoded in various degrees of freedom such as charge or spin. Charge representations, for example the absence or presence of an electron in a quantum dot, are attractive as they are easily manipulated and interrogated through electric fields. The advantage of spin representations, used in the field of , is their superior shielding from undesired electric in the environment, making the information in these latter representations more robust. In the future, there might be a third middle-ground alternative in the valley degree of freedom that exists in certain , including graphene.

This image shows an extended line defect in graphene. Credit: Naval Research Laboratory

The valley degree of freedom in graphene gained attention in 2007 when it was proposed that electrons and holes could be filtered according to which valley they occupy. Unfortunately, the structures required for this and subsequent valley filters are difficult to fabricate, and as a result a valley filter has yet to be demonstrated experimentally. The present study from NRL shows that an extended line defect in graphene acts as a natural valley filter. "As the structure is already available, we are hopeful that valley-polarized currents could be generated in the near future" said Dr. Daniel Gunlycke who made the discovery together with Dr. Carter White. Both work in NRL's Chemistry Division.

Valley refers to energy depressions in the band structure, which describes the energies of electron waves allowed by the symmetry of the crystal. For graphene, these regions form two pairs of cones that determine its low-bias response. As a large crystal momentum separates the two valleys, the valley degree of freedom is robust against slowly varying potentials, including scattering caused by low-energy acoustic phonons that often require low-bias electronic devices to operate at low temperatures typically only accessible in laboratories.

Owing to symmetry, the electrons and holes in the blue (red) valley transmit (reflect). Credit: Naval Research Laboratory

Valley polarization is achieved when and holes in one valley are separated spatially from those in the other valley, but this is difficult to do as the two valleys have the same energies. It was found, however, that this spatial separation can be obtained in connected graphene structures that possess reflection symmetry along a particular crystallographic direction with no bonds crossing the reflection plane. This property turns out to be present in a recently observed line defect in graphene. The reflection symmetry only permits electron waves that are symmetric to pass through the line defect. Anti-symmetric waves are reflected. By projecting an arbitrary low-energy wave in graphene onto its symmetric component, one gets the transmission amplitude through this defect, which is strongly dependent on the valley. Electron and hole waves approaching the line defect at a high angle of incidence results in a polarization near 100%.

There is a long way to go before valleytronics can become a viable technology, explains Gunlycke. The recent advance, however, provides a realistic way to reach a crucial milestone in its development. This research was supported by the Office of Naval Research, both directly and through the Naval Research Laboratory.

Explore further: New 'topological insulator' could lead to superfast computers

Related Stories

Graphene and 'spintronics' combo looks promising

Jan 25, 2011

A team of physicists has taken a big step toward the development of useful graphene spintronic devices. The physicists, from the City University of Hong Kong and the University of Science and Technology of China, present ...

Creating a pure spin current in graphene

Feb 07, 2011

(PhysOrg.com) -- Graphene is a material that has the potential for a number of future applications. Scientists are interested in using graphene for quantum computing and also as a replacement for electronics. However, in ...

Two graphene layers may be better than one

Apr 27, 2011

(PhysOrg.com) -- Researchers at the National Institute of Standards and Technology have shown that the electronic properties of two layers of graphene vary on the nanometer scale. The surprising new results ...

Scientists make magnetic new graphene discovery

Apr 14, 2011

(PhysOrg.com) -- University of Maryland researchers have discovered a way to control magnetic properties of graphene that could lead to powerful new applications in magnetic storage and magnetic random access ...

Recommended for you

Uncovering the forbidden side of molecules

Sep 21, 2014

Researchers at the University of Basel in Switzerland have succeeded in observing the "forbidden" infrared spectrum of a charged molecule for the first time. These extremely weak spectra offer perspectives ...

How Paramecium protozoa claw their way to the top

Sep 19, 2014

The ability to swim upwards – towards the sun and food supplies – is vital for many aquatic microorganisms. Exactly how they are able to differentiate between above and below in often murky waters is ...

User comments : 14

Adjust slider to filter visible comments by rank

Display comments: newest first

maxcypher
5 / 5 (1) Apr 27, 2011
Is there a "Valleytronics for Dummies"?
eachus
3.4 / 5 (5) Apr 27, 2011
Is there a "Valleytronics for Dummies"?


Or for that matter, a "Spintronics for Dummies"? Not really, but I can explain why there is so much interest in both.

Decades ago, a some computers used asynchronous logic. What this means is that instead of looking at one signal to see whether it is high or low, each bit requires two wires, one representing "1" and one "0". Why go to all this "extra" effort? In synchronous computers--the type you are using now--there is a clock signal which is distributed everywhere in the chip. I won't go into the difficulties with clock skew, etc., but there are two problems with synchronous circuits that cannot be worked around.

The first is that you have to wait for the worst case timing on an adder or a multiplier. Since bits are represented by the presence or absence of a signal at the clock "edge," you have to allow for the worst case cycle times.
eachus
3.4 / 5 (5) Apr 27, 2011
The other problem is power, not only do the clock signals use a significant portion of the computer's entire power budget, but they use that power whether or not anything is happening. Thus modern CPU chips have special ways to shut off the clock signals to all or part of the chip to save power. But the circuitry shut off that way can't resume instantly. A few nanoseconds is not something that you will notice, but it means that aggressively shutting down clocks slows the overall performance of the chip.

Phew! That's the hard part--understanding why there is a problem. With spintronics and anything that allows for two different signals to travel on the same wire, designers can go back to asynchronous designs, but with one wire per bit, not two. Even better, all of the clock synchronization and distribution issues go away. Since you have one wire per bit, there are no skew issues, and circuits run as fast as they can, but no faster. ;-)
jjoensuu
not rated yet Apr 27, 2011
So echus, why can the designers go to one wire per bit with spintronics? Is it because the single wire can carry both the voltage and the spin?

Also will there be (or is there) some sort of law comparable (in importance) to Ohms law also for spintronics?

And does spintronics also have components? Perhaps "spinresistors" and/or "spincapacitors"?
eachus
3.4 / 5 (5) Apr 28, 2011
So eachus, why can the designers go to one wire per bit with spintronics? Is it because the single wire can carry both the voltage and the spin?


You could say it that way. When electrons arrive at every bit of a register the answer is encoded in the spin of the electrons (spin up or spin down). Reading the spins is, for the most part, not necessary. You add two numbers, multiply by a third, store the result and so on. If you are really tricky, you can keep the indicator bits as normal electronic signals, and the quantities as spins.

Also will there be (or is there) some sort of law comparable (in importance) to Ohms law also for spintronics?


Spintronics is regular electronics except that you align the spins (possibly from two different current sources) and occasionally you use the magnetic field from the spins to determine what number you have. Logic circuits are like any other Boolean logic, except that XOR is easier and NOT is harder. ;-)
eachus
3.7 / 5 (6) Apr 28, 2011
Hmm. For people who haven't studied digital logic or computer design the above is a bit sparse. If you didn't study Boolean algebra in school, find a web page. But computer design usually works at a higher level of abstraction, the register level.

A register is an ordered set of bits (ones and zeros) that will often have a fixed number of data bits, such as 32 or 64, and a name. (Yes, modern OoO CPUs cheat by renaming registers dynamically, but that is a detail.) A register will also have associated flag (or indicator) bits and may have associated parity bits. The important thing is that designers can specify what the various registers look like, plus specific design blocks like adders, multipliers, and so on. "Silicon compilers" take these register level designs convert them to the gate level, then place individual transistors and wires on the chip.

The important point is that spintronics, if the spins can be kept stable enough, can simplify--and speed up--processors.
jjoensuu
3.7 / 5 (3) Apr 28, 2011
Thank you eachus, your explanation is very clear.
AngryMoose
4 / 5 (4) Apr 28, 2011
Thank you eachus one of the best posts I've read on this site, was actually more interesting than the article.
Bonkers
not rated yet Apr 28, 2011
Hmm, spintronics - looks to more than double the complexity in order to halve the gatecount. I can't see it ever working, doubling density only offers 9 months advantage against Moore's Law, yet requires fundamental redesign and re-qualification. At present Silicon is pretty much unaffected by magnetic fields, spintronics will need to be proven against this new susceptibility.
On a similar note, we could move from binary electronics to a 3 or 4-level approach, much like multi-bit flash, OK we'd need to redesign every logic element (like spintronics) and the noise margins would reduce significantly (like..) - yet we haven't bothered.
So, I can't see a compelling reason for this technology.
Quantum computing? - that's a totally different story.
Bog_Mire
3 / 5 (2) Apr 28, 2011
Eachus, if you could prove you where born in the USA, I would vote you for Prez. And as long as you don't have suspicious head scars.
maxcypher
not rated yet Apr 28, 2011
Eachus, could you explain how this "valleytronics" works?
FrankHerbert
1 / 5 (5) Apr 28, 2011
@bog_mire

Just don't post here, okay?
Eikka
not rated yet Apr 28, 2011
The thing is, asynchronous logic requires up to twice the number of transistors to implement, and it's not just a matter of having the 1s and 0s on two wires. Each piece of logic has to internally separate the control signal from the data signal, and then implement a synchronization mechanism so the operations are done in the correct order and correct times relative to each other, so that one bit doesn't fall off the calculation because it wasn't there on time.

So, it will have to save over 50% of power per transistor to be worthwhile, but it still takes twice the area on silicon, which is a bit difficult in the cost and yield perspective.

Spintronics will presumably make the multiplexing and separating of the control and data in one signal easier, but it won't make the logic itself any simpler.
Eikka
not rated yet Apr 28, 2011
And the fact that you need to add catches and arbiters to prevent random glitches and unlucky flips of bits in the circuit from propagating also adds complexity and increases the number of transistors.

Synchronous logic avoids this problem by only doing anything on the clock, so it doesn't matter if the circuits flutter randomly out of order in the mean time as long as they settle down when it's time to pass the data.

Thirdly, there's no way of knowing accurately how long it takes for anything to happen in an asynchronous circuit, so real time control systems require additional circuitry to keep synchronization and time events. There is no fully asynchronous chip that could control e.g. a robot deterministically, when the chip can speed up or slow down depending on things like how hot it is, or if the supply voltage creeps up a little bit.