bit-tech.net

IBM: Graphene as it is won't replace silicon in CPUs

IBM: Graphene as it is won't replace silicon in CPUs

A single graphene sheet measures just one atom-thick, potentially paving the way tiny transistors.

IBM has revealed that graphene can't yet fully replace silicon inside CPUs, as a graphene transistor can't actually be completely switched off.

In an interview for a forthcoming Custom PC feature about chip-building materials, Yu-Ming Lin from IBM Research - Nanometer Scale Science and Technology told us that 'graphene as it is will not replace the role of silicon in the digital computing regime.'

Last year, IBM demonstrated a graphene transistor running at 100GHz, claiming that the technology could be used to manufacture 'zippy computer chips' in the years to come. The story, along with news that researchers at the UCLU had produced a graphene transistor with a cut-off frequency of 300GHz, prompted all sorts of predictions of silicon marching towards its demise, making way for a graphene-based future with 1THz (one terahertz, or 1,000GHz) CPUs.

However, Lin says that 'there is an important distinction between the graphene transistors that we demonstrated, and the transistors used in a CPU. Unlike silicon, 'graphene does not have an energy gap, and therefore, graphene cannot be “switched off," resulting in a small on/off ratio.'

However, he also pointed out that graphene 'may complement silicon in the form of a hybrid circuit to enrich the functionality of computer chips.' He gives the example of RF circuits, which aren't dependent on a large on/off ratio.

Graphene is a two-dimensional atomic-scale material, made from a hexagonal lattice of carbon atoms and their bonds. A single graphene sheet measures just one atom-thick, potentially paving the way for the smallest transistors physically possible.

'In principle, there's no limit to the size of a graphene transistor,' says Lin. 'Compared to silicon, graphene is more robust in terms of device scaling, because it has the thickness of a single layer of atoms, while it's known that the quality of silicon will suffer significantly once it's thinned down.'

Graphene also offers much higher potential clock speeds over comparable silicon transistors. 'Graphene transistors can achieve a higher clock speed than those made of silicon with the same gate length,' explains Lin, 'because the electrons in graphene can move at a higher speed than those in silicon.'

Intel's director of components research, Mike Mayberry, concurred. 'Graphene is still very much in the research phase,' Mayberry told us, but 'researchers are predicting a number of interesting properties for it and now experimentalists are trying to confirm them.'

Mayberry also added that 'silicon’s properties make it a nearly ideal material. The industry has so much experience with it that there are no plans to move away from silicon as the substrate for chips.'

Basically, graphene has lots of potential advantages over silicon, but it's not going to replace it as the main material for building CPUs, at least not in its current research state.

Look out for a full feature on the future of silicon and chip-building materials in Issue 91 of Custom PC magazine, on sale from 17 Feb and online at Zinio.

24 Comments

Discuss in the forums Reply
trig 21st January 2011, 20:00 Quote
can someone explain "small on/off ratio" to me. does "can't be switched off" mean that once electricity is applied to it the charge is never dissipated?
Picky88 21st January 2011, 20:06 Quote
It means a small ratio between the current flowing through the transistor when it is on, and the current flowing when it is off. In a normal transistor that is turned off, only a tiny amount of "leakage" current can flow, meaning that the ratio between on and off is much higher. A chip made from transistors that leak alot of current when they are turned off would waste alot of energy, and therefore produce alot of heat.
Cthippo 21st January 2011, 20:22 Quote
The bigger issue than heat is signalling. If a computer can't tell the difference between a 0 (gate closed) and a 1 (gate open) then the whole thing breaks down. in silicon gates the difference a 1 and a 0 is a lot, with graphene it is not as much and so the bits are harder to read.
KidMod-Southpaw 21st January 2011, 20:27 Quote
Graphene will also run cooler than silicon. A hybrid chip sounds like a very good idea.
Penfolduk01 21st January 2011, 21:10 Quote
Hybrid chips sound a good idea for most applications. Although for the high-end computing IBM makes it's money in, having mainly graphene chips would make sense.

Can't help think what Intel could do with this and its turboboost technology. Maybe for the Core i13 maybe? :-)
schmidtbag 21st January 2011, 23:07 Quote
Quote:
Originally Posted by Cthippo
The bigger issue than heat is signalling. If a computer can't tell the difference between a 0 (gate closed) and a 1 (gate open) then the whole thing breaks down. in silicon gates the difference a 1 and a 0 is a lot, with graphene it is not as much and so the bits are harder to read.

thats exactly right. what picky88 said is true, but what you said is the more significant problem.

however, shouldn't the solution be as simple as just decreasing the wattage of a cpu? i don't see how graphene can even be considered a transistor if it doesn't even turn off. as another question, how would you manage to do a hybrid of graphene and silicon? to me thats like mixing water with vegetable oil - it CAN be done but they just don't mix because of how different they are, even though they are both liquids. silicon and graphene can both be transistors but they operate so differently that i don't understand how a hybrid could be possible.
TWeaK 21st January 2011, 23:11 Quote
@KM-S the trouble is you can't just do a straight hybrid chip - it's not like a hybrid car where both systems do the same thing. As Cthippo said, with a low ratio the computer won't be able to differentiate between the two signals, thus it cannot (at least right now) be used for a lot of the functions silicon is used for, in particular pretty much everything silicon does in computer IC's.

It sounded like a dream when I first read about this. The cynic in me said there must be a catch and I guess this is it. I have a feeling now, while getting graphene transistors usable for logic circuitry might be cracked eventually, it just won't be all that practical and end up being on the expensive shelf along with gallium-arsenide and other exotic material substrates.
benji2412 22nd January 2011, 00:25 Quote
I'm sick of hearing about how 'great' graphene is, when in most applications it's just a fad.
jihadjoe 22nd January 2011, 13:06 Quote
Quote:
Originally Posted by schmidtbag
thats exactly right. what picky88 said is true, but what you said is the more significant problem.

however, shouldn't the solution be as simple as just decreasing the wattage of a cpu? i don't see how graphene can even be considered a transistor if it doesn't even turn off.

A transistor doesn't need to be able to output a zero to be a transistor, it just needs to be able to output a variable amount of current, depending on the input.

Anyways, please see this illustration from wikipedia regarding bandgap:

http://upload.wikimedia.org/wikipedia/commons/thumb/4/46/Bandgap_in_semiconductor.svg/220px-Bandgap_in_semiconductor.svg.png

The bandgap represents input voltage range where the semiconductor is able to output a "0". In silicone, that is pretty large at 1.1eV. Graphene's bandgap, even in the best dual-layer designs is from 0-0.25eV.

Currently, the bandgap for graphene is smaller than a single unbound electron (1eV), meaning it's effectively useless for making computer chips because we can't even make it send a zero.
j_jay4 22nd January 2011, 14:12 Quote
It did win a Nobel prize this year, there must be some substance in it.

New technologies regularly start out with a notion that seems crazy and unworkable to most, but with such huge amounts of time and money going into this research the benefits of graphene will eventually be harnessed and made into a practical application.

I'm really looking forward to this article, Custom PC have done some brilliant features recently, I especially liked all the broadband companies being named and shamed in issue 90.
Repo 22nd January 2011, 18:18 Quote
Quote:
Originally Posted by j_jay4
It did win a Nobel prize this year, there must be some substance in it.
Obama won one for world peace... 'nuff said...
Mraedis 22nd January 2011, 18:35 Quote
Not world peace, just peace.
Fizzban 22nd January 2011, 23:29 Quote
This still has a lot of potential. Hopefully they will find a way to turn them off, but even if they don't it's not the end for Graphene. Many things have been thought up with a specific use in mind, only for it to be later discovered that they can actually be used for something else. So even if this can't help much with chips someone will eventually find a good use for it.

Can't wait to read the article in Custom PC!
Cthippo 23rd January 2011, 03:02 Quote
I'm wondering if instead of being able to make them more "off", if it's possible to make whatever they hook into more sensitive to the differences between "on" and "off".
msh317 23rd January 2011, 14:27 Quote
One of the primary values of Carbon is that it can be programmability changed from a P junction to N junction it is well known that high temperature super conductivity requires materials that provide an electron gap for example Iridium Oxide - I agree that graphene will not directly replace silicon in current chip designs however no one said we must use carbon by itself. For example Iridium Oxide provides the necessary electron gap to completely switch off the electron flow.

I suspect we will begin to see a new basis for chip design; carbon is only one of the key elements necessary to make it happen.
TheStockBroker 23rd January 2011, 16:21 Quote
Quote:
Originally Posted by Cthippo
I'm wondering if instead of being able to make them more "off", if it's possible to make whatever they hook into more sensitive to the differences between "on" and "off".

Of course!

The voltage/current is taken as an analogue > converted to digital.

e.g. Digital is either 1 or 0. With the analogue interpretation you could take anything up to 85% as a 0. anything greater than 85% as a 1.

or does that negate the effect of the CPU - having to have some form of secondary (software) detection?
ssj12 24th January 2011, 06:39 Quote
the could just find a new way to shut a graphene cpu off... i mean it can't be to hard to invent a different process or have a start-up processor that activated a graphene processor?
Doomah 24th January 2011, 13:08 Quote
To everyone saying that 'they should just change the design'; That is not really possible. In my own measurements it shows that graphite (multiple layers of graphene) doesn't have a real non-0 voltage, 0 ampere area.
http://i55.tinypic.com/4vlqir.jpg
(my measurement on a student version STM)
while for silicon it does have that area.
http://turboblogsite.com/wp-content/uploads/2010/05/20.jpg
(not my measurements)

This basically means that with silicon you can have an area that outputs (near) 0 ampere over a large voltage range, while for graphene it is way harder to make a near 0 output.
benji2412 24th January 2011, 17:21 Quote
Quote:
Originally Posted by msh317
One of the primary values of Carbon is that it can be programmability changed from a P junction to N junction it is well known that high temperature super conductivity requires materials that provide an electron gap for example Iridium Oxide - I agree that graphene will not directly replace silicon in current chip designs however no one said we must use carbon by itself. For example Iridium Oxide provides the necessary electron gap to completely switch off the electron flow.

I suspect we will begin to see a new basis for chip design; carbon is only one of the key elements necessary to make it happen.

I assume you mean carbon can be either n- or p-doped? The same way that Si can? Even if you did dope graphene to be a p- or n- material, you'd seriously affect the properties that make it so special. For example the fact that it's one atom thick....

Also you can't change from a p-junction to an n-junction, you have p and n doped semi-conductors that form a p-n junction .

Please do tell me how you'd use IrO2 to provide the band-gap to essentially form the p-n junction and thus the overall transistor? More to the point the cost of IrO2 can't be relied upon and as it stands, Si works just fine and the manufacturing processes are just dandy for what we need.
LightningPete 24th January 2011, 22:42 Quote
Can anyone see Nvidia looking at this for its next GPU, that runs at stupid amounts of Ghz and this time with much less heat. I do!

lol

Obviously im talking in the next 50 years when we have CPU's even turning the window wipers on our cars and the ironing been automatically done when we think about it because of the Intel stamped chip lodged into our heads, and the words GOO and GLE appearing in both our eyes.

Ohh did i forget to mention we wear and eat and live in a house full of Tesco branded stuff?

Yep you guessed it.

TESINTEGOOO-gle .... the new earth
[USRF]Obiwan 25th January 2011, 15:19 Quote
Quote:
Originally Posted by Picky88
It means a small ratio between the current flowing through the transistor when it is on, and the current flowing when it is off. In a normal transistor that is turned off, only a tiny amount of "leakage" current can flow, meaning that the ratio between on and off is much higher. A chip made from transistors that leak alot of current when they are turned off would waste alot of energy, and therefore produce alot of heat.

Basicly you have a Quantum computer like switch it can both be on and off. sort of Qubit on/off/on&off ..
:+)
chwynn 4th February 2011, 15:04 Quote
Whatever happened to optical computing?
ribrin55 14th February 2011, 20:14 Quote
I am late to this but I was wondering if the switching issue could be handled by a form of carbon (pyrolytic graphite) that has severe anisotropy in its directional plane? i.e. one plane has resistance of 500 µÎ©cm and the other 0.5 Ωcm? It also has the same effects for thermal conductivity with 440 W/mk in one plane and 1.4 W/mK in another plane?
cgthomas 1st September 2011, 22:35 Quote
The Earth used to be flat for several centuries. However, later on man learnt this isn't the case.
My point is just because you don't have the knowledge now to turn off the transistor; it doesn't actually mean that it can't be done.

What IBM mean to say is they're not smart enough (yet) to figure out how to achieve the dream chip, made of graphene.

Or it could be a business strategy not to introduce graphene yet, as they want to profit as much as possible from the production of traditional silicon chips.

Also, bear in mind that graphene chip production requires new equipment, new breed of scientists, OS developers need to know more about chips. It's a multi-billion dollar step for an industry that likes to grow in doubles each year so I don't think we'll see anything special in the near future.

Chip producing companies are like pharmaceutical firms, they request a premium for products which don't cost much to produce on a large scale. Making money from dirt is a booming business!
Log in

You are not logged in, please login with your forum account below. If you don't already have an account please register to start contributing.



Discuss in the forums