Graphene May be the New Silicon

esocid writes to share that University of Maryland physicists have demonstrated that the material of the future may be graphene rather than silicon. Electricity conduction through graphene is about 100 times greater than that of silicon and could offer many improvements to things like computer chips and biochemical sensors. "Graphene, a single-atom-thick sheet of graphite, is a new material which combines aspects of semiconductors and metals. [...] A team of researchers led by physics professor Michael S. Fuhrer of the university's Center for Nanophysics and Advanced Materials, and the Maryland NanoCenter said the findings are the first measurement of the effect of thermal vibrations on the conduction of electrons in graphene, and show that thermal vibrations have an extraordinarily small effect on the electrons in graphene."

The "100 times greater"...

...refers to electron mobility [wikipedia.org], a concept I hadn't previously encountered. But it's easy enough to understand: if I apply a unit electric field to a material, how fast does it make the electrons drift? This is the mobility.

Apparently graphene (also new to me ... a single-atom layer of carbon) is exciting because it has much higher electron mobility than silicon. Which leads to faster switching times, although they don't explain that part.

All this seems to be theoretical at the moment, due to insufficiently pure graphene. Still, 100th the switching delay is not a bad target to be aiming at... 100Ghz processing!

Re:The "100 times greater"...

Graphene has been studied for a few years now, even longer if you count it as rolled into a nanotube.

What took awhile (and was solved with a fairly low-tech solution : scotch tape) was how to make a single layer of graphene to measure, whereas graphite usually rolled off into multi-layer pieces.

Graphene is interesting for a number of reasons. Primarily is it's Minkowski lightcone-like density of states. The Fermi level lies right at the cone vertex, which makes this material a "zero-bandgap insulator", which brings about a huge number of interesting properties in itself.

Anyway, graphene has been hugely popular in condensed matter physics for a few years now, and people have studied the phonon spectra, I remember going to a seminar about the modes of graphene in a carbon nanotube a few years ago.

However, don't get your hopes up for mass-produced graphene tech anytime soon. While people will probably demonstrate small-scale single-electron transistors or other interesting graphene devices (if they haven't already), the ability to deposit and pattern graphene is still very crude, and it's hard to do anything other than one-off devices at this point.

Re:

It's also very hard to "solder" interconnects on a single layer sheet. Alnd, due to the 2 dimensional nature of the graphene sheet you can't easily take advantage of modern multilayer silicone processing. Making a true device from this will be challengin

Re:The "100 times greater"...

I remember reading about how physicists are running into the limitations of "C" (speed of light) with regards to signal propagation across the CPU die. Even though something measuring 143 mm^2 is small, at speeds of 100GHz (or was that 1Thz), I doubt your processing can remain symmetrical. If that's true, such fast CPUs will need to be engineered for asymmetrical processing instead.

Re:The "100 times greater"...

On the other side of the coin, the design for an original Pentium had around 5 million transistors. Modern processors have more like 300 million. What's changed? Well, dual-core, and 64-bit, sure. But a lot of those extra transistors are to create extra pipelines or additional specialized instructions or even specialized pipelines that only run specialized instructions to compensate for the fact that the clock speeds just won't ramp up as quickly as designers want. Perhaps if we were able to start cranking up the clock speeds again, it would be possible to start streamlining those pipelines and instruction sets into something more manageable for keeping your signals properly synchronized.

Re:The "100 times greater"...

Cache, bloody huge cache. 6 transistors per bit, 48 per byte, 49152 per KiB, 50,331,648 per MiB. If you have 4 MiB cache, it's 201,326,592 transistors.

Re:

Still, 100th the switching delay is not a bad target to be aiming at... 100Ghz processing!

I will physically reach out and strangle the first person to make a joke relating 100GHz to the system requirements of Windows...

Re:

I'm so tempted by the idea of seeing a pair of disembodied hands lunging through my screen, but I'd better not risk it...

Re:The "100 times greater"...

I think you mean silicone.

The "e" is very important. (As the raver said to the priest).

Re:The "100 times greater"...

I think you mean silicone.

Okay then, Graphenee boobs then.

Re:The "100 times greater"...

Thanks, now I don't have to RTFA. I was wondering why pure conductivity improvements are good for gates. Semiconductors are used for a reason. :-)

The increased mobility has little to do with gates. In fact, you want gates (in MOSFETs [wikpedia.org]) to be as resistive as possible, but still not attenuate the electric field that results from the gate voltage, hence the use of Halfnium dioixde instead of silicon dioxide (you can make it thicker, (and thus more resistive) while still having a strong enough field.)

Mobility results from the equation v=(mu)E, where mu is the mobility and v is the velocity of an charge carrier (electron or hole) The reason we use semiconductors is that we can easily control the number of electrons or holes. But by increasing the speed of electrons, we can allow them to switch faster since they will be able to cross the channel more quickly. That's why smaller transistors can switch more quickly, the channel length is shorter so it takes less time for carriers to traverse them.

I'm not sure why it's considered so amazing to discover that graphene has a good electron mobility. Since, the entire structure consists of delocalized pi orbitals, you would expect electrons to easily travel through graphene. I'm not sure how graphene would be doped either. I suppose you could use boron and phosphorous like in silicon, but it remains to see if they will still bond appropriately. Ah well, there's a reason, they're professors and I'm a student.

Re:The "100 times greater"...

"Something new and nifty and important has been discovered! But it's too complicated to explain it to you, so we'll spare you the boring, complicated details."

Your in-depth analysis intrigues me, and I wish to subscribe to your newsletter.

Unfortunate name

He must get "Herr Fuhrer" jokes all fricking day

Re:Unfortunate name

He must get "Herr Fuhrer" jokes all fricking day

Is it actually possible to Godwin a thread about microprocessor engineering?

Re:

Not in public, at least.

Re:

Germany's current president does not use that title

In part because she's a woman...

I, for one...

I, for one, welcome our new carbon based overlords :-)

Seriously, however, I don't expect to see a CPU based on this anytime soon.

Would oxidation be a problem?

While you could coat it with a hard protective layer like aluminum oxide, I think it would be hard to protect it well enough to prevent oxidation from degrading a layer only one atom thick.

I recall that early compact discs had this problem, in which oxygen trapped in the plastic would oxidize the aluminum and reduce its reflectivity.

Questions to ask proponents of new semiconductors

1) Do you have a decent quality oxide for it?
2) Can you make good low resistance contacts?
3) Can it be doped?

Graphene probably fails 1 and 2 at this point. I'm not sure about 3.

2D (graphene) and 1D (carbon nanotube) semiconductor systems have a lot of trouble with surface effects ruining your ability to make decent devices.

Re:

Yes, using conventional e-beam lithography.

link to the paper

Here's the press release with a link to the paper.
http://www.thinkgene.com/um-physicists-show-electrons-can-travel-over-100-times-faster-in-graphene-than-in-silicon/ [thinkgene.com]

let me clear up some confusion...

Graphene is certainly a lot like carbon nanotubes, but is much easier to work with. Where you have to hope to get a semiconducting crystal structure in a nanotube (or make crappy transistors based on defects), you can pattern graphene to make a transistor. Which directions you cut the 2D sheet determine whether it is metallic or semiconducting. There are some problems with this, and practically speaking any small channel (10 nm, I think) of graphene is semiconducting. Fuhrer has shown (along with other people) that graphene can make pretty good transistors (very fast switching, thermally stable and I'm sure I'm missing some stuff).

It can be doped. This is another thing Fuhrer has done (as well as other people... but this is his article we're talking about). You don't want to insert something into the crystal structure (that ruins it), but you can layer the top of it with potassium ions (about 1 per 1000 carbons), which dopes it just fine. This isn't a bulk semiconductor though, and the addition of charged impurities (dopants) decreases device performance (in bulk, it's a metal). You can very easily electrostatically gate graphene in any direction you want; transistors and PN junctions are easy to make this way.

It is not hard to make graphene. The "scotch tape" method from Manchester is widely used, but there are a number of other ways to do it which may be commercially viable: oxidizing graphite, ultrasounding graphite with special polymers (Dai's method), growing it from SiC wafers. Of course, none of these really work yet, and may never be economical.

Graphene is stable in air (almost all devices are measured in air at some point), and liquids. It's not going to spontaneously dissolve on you just because it's only 1 atomic layer thick. It's actually very robust.

It can be used with silicon processing techniques. People are using SiO2, HfO2 and all the usual silicon processing with it.

Big companies are looking at this material. IBM has already reported results on their work at physics conferences, I'm fairly sure that the more secretive companies (Intel) are also working with graphene... just like they worked with nanotubes.

Re:

So, when do I get my 360 GHz sixteen core processor?

The Sony Playstation 36, Holodeck Edition.

Re:

16-core processor?
You must be extremely conservative...

I'm waiting for my one MegaCore processor with 1,048,576 cores, while mocking the market-war with MegiCore processors who only have 1,000,000 cores, but perform better at rendering realistic 3D mode

Re:

Chemistry 101:

Silica : crystalline silicon dioxide aka sand
Silicon : element # 14, greyish semimetallic crystalline
Silicone : Inorg. polymer typ. -Si(CH3)2-O- Liquid or can be rubber if crosslinked. Using for boob jobs.