A Semiconductor is for Ever...

[nelg]

An interesting read on the use of synthetic diamonds for economically viable semiconductors (and engagement rings).
http://www.wired.com/wired/archive/11.09/diamond.html?pg=1&topic=&topic_set=

The third big challenge has been the most daunting for materials scientists: To form microchip circuits, positive and negative conductors are needed. Diamond is an inherent insulator - it doesn't conduct electricity. But both Gemesis and Apollo have been able to inject boron into the lattice, which creates a positive charge. Until now, though, no one had been able to manufacture a negatively charged, or n-type, diamond with sufficient conductivity. When I visit Butler in Washington, he can barely contain his glee. "There's been a major breakthrough," he tells me. In June, together with scientists from Israel and France, he announced a novel way of inverting boron's natural conductivity to form a boron-doped n-type diamond. "We now have a p-n junction," Butler says. "Which means that we have a diamond semiconductor that really works. I can now see an Intel diamond Pentium chip on the horizon."

 

[KimB]

A couple of nice things about diamond:

1. Carbon is a smaller atom than Silicon, which, I think, means smaller semiconductor devices are possible (I'd guess about 10%-20% smaller, though I haven't looked at the actual numbers).
2. Diamond has tremendous thermal conductivity (Copper, widely used in heatsinks for its high thermal conductivity properties, has a thermal conductivity of about 4 W/cm°C. Diamond's can be as high as 24 W/cm°C).
3. Diamond has tremendous heat tolerance. A Silicon chip will burn out at around 100 degrees Celsius (if I remember correctly). Diamond should be able to exceed 1000 degrees Celsius.

Anyway, if synthetic diamonds can be made much more economically, it would make an excellent heatsink material, not just as a semiconductor.

And, of course, its use as a semiconductor would be heavily dependent upon its electrical properties. Unfortunately, I'm not so sure those would be as good as already available semiconductors (like Gallium Arsenide).

 

[Dio]

In fact, diamond semiconductors might be operated at substantially higher temperatures than now if at all possible.

Of course, the point where metal connects to the diamond may be a limitation .

Partly this is because it increases charge mobility, and partly this is because a higher temperature would pump heat away more efficiently.

 

[3dilettante]

I wonder to what extent diamond's thermal tolerance can be exploited as a semiconductor. In high heat situations, it might be that various dopants and interconnects could fail or migrate at much lower temperatures.

Should they get manage to make large enough wafers of diamond, how will they cut the individual dies out of it? Current wafers are cut with diamond-tipped saws, but what do you do when what you want to cut is just as hard as your saw?

 

[KimB]

I wonder to what extent diamond's thermal tolerance can be exploited as a semiconductor. In high heat situations, it might be that various dopants and interconnects could fail or migrate at much lower temperatures.

The dopants themselves wouldn't fail, but they may migrate. So yes, the actual temperature tolerance of a diamond chip would depend on what other components are there to make the thing work.

Should they get manage to make large enough wafers of diamond, how will they cut the individual dies out of it? Current wafers are cut with diamond-tipped saws, but what do you do when what you want to cut is just as hard as your saw?

They'd probably need to grow the wafers individually.

 

[g__day]

Adds mental caution - never touch a diamond chip with fingers to see how hot it is!

 

[Tahir2]

Interesting factoid: Ruby is 50x rarer than natural Diamond.

 

[PC-Engine]

You can cut diamond with high pressure water or highpower laser.

 

[Saem]

There is actually some very incredible results that have been achieved with manufactured diamonds, which are incredibly pure.

 

[Simon F]

Interesting factoid: Ruby is 50x rarer than natural Diamond.

Aren't natural diamonds meant to be relatively common? (The price being completely artificially inflated). There are huge deposits in Western Australia for example.

 

[Dio]

Adds mental caution - never touch a diamond chip with fingers to see how hot it is!

It'll be glowing if it's really dangerous.

 

[horvendile]

I'm not so sure those would be as good as already available semiconductors (like Gallium Arsenide).

Is gallium arsenide ever used in consumer products with substantial transistor counts? I've been given the impression that it's a bit on the expensive side.

 

[Althornin]

Interesting factoid: Ruby is 50x rarer than natural Diamond.

Aren't natural diamonds meant to be relatively common? (The price being completely artificially inflated). There are huge deposits in Western Australia for example.

yep.

 

[KimB]

Is gallium arsenide ever used in consumer products with substantial transistor counts? I've been given the impression that it's a bit on the expensive side.

Yes, it is. And so it isn't used much. I meant available as in the technology exists to make use of it.

Anyway, diamond semiconductors may be interesting, but I'm not convinced the electron mobility would be sufficient enough to make them very useful.

Additionally, I still think that GaAs wouldn't be much more expensive than silicon if just as many resources were poured into developing GaAs wafers and chips. Within the next few years, we may well find that going for higher-mobility chips will be the only way to keep reducing the size of microprocessors (as higher mobility means less heat).

 

[arjan de lumens]

AFAIK, the limiting factor of silicon temperature is not so much the silicon itself as it is packaging and interconnect layers. (When forming the transistors, silicon wafers are routinely exposed to temperatures around 1000 Celsius, whereas much lower temperatures are used when adding the metal layers.) So I'm not really convinced that Diamond will be that much more heat resistant in practical use (and at ~800 Celsius it will catch fire unless completely insulated from air).

As for gallium arsenide, the main problems are wafer brttleness (=> low yield) and low hole mobility, forcing you to use logic families with awful leakage currents if you want good performance; also IIRC the element gallium is scarce enough to impact the price of the final chip as well. Diamond IIRC has much better hole mobility than both Si and GaAs, and so lends itself much better to CMOS-like logic families.

 

[rwolf]

How about a diamond heat sink?

 

[Thowllly]

How about a diamond heat sink?

I was thinking the same thing. A heatsink made from pure diamond out to be really efficient. Or diamond waterblock. That ought to be cool too...

 

[KimB]

How about a diamond heat sink?

If synthetic diamond was ever cost-effective, and could be made with a flat surface (for contact with the core) and fins (or something else to increase surface area, either for contact with air or a metal, or water), then yeah, diamond would be a great heatsink material.

But, I think that using diamond dust in heatsink paste might be a significantly more efficient use for the time being (of course, its conductivity in the paste would depend on a large number of factors, but if the diamond dust is in granules large enough to be the main contact between the core and the heatsink, then perhaps the interaction between the diamond and the paste would be immaterial...with the paste just existing to keep the diamonds in place).

 

[Althornin]

How about a diamond heat sink?

If synthetic diamond was ever cost-effective, and could be made with a flat surface (for contact with the core) and fins (or something else to increase surface area, either for contact with air or a metal, or water), then yeah, diamond would be a great heatsink material.

But, I think that using diamond dust in heatsink paste might be a significantly more efficient use for the time being (of course, its conductivity in the paste would depend on a large number of factors, but if the diamond dust is in granules large enough to be the main contact between the core and the heatsink, then perhaps the interaction between the diamond and the paste would be immaterial...with the paste just existing to keep the diamonds in place).

you dont want large granules, you dont want diamond to be the main contact betweent he core and heatsink.
you ideally want the main contact between core and heatsink to be direct, and you want as fine a granularity as possible in your paste so that it only fills those small places where core and heatsink dont directly touch. Large particle size is 100% the wrong direction to go.
http://www.buffalo.edu/reporter/vol34/vol34n32/articles/Chung.html
Read that, then go and download her research here:
http://zzz.com.ru/chung.zip
then read the paper, and understand.
Thermal pastes of today focus too much on thermal conductivity. But thermal pastes need to perform one function - gap filling - and spreadability and small particle size are paramount in that. Thus, with a material (Chung uses carbon black) with a far lower thermal conductivity but a vastly smaller particle size (diamonds effective limit is about 25 microns), more efficient thermal transfer is enabled.

TIM joints and the like is a hobby of mine.

 

[KimB]

I was thinking along the lines of the diamond actually cutting into the metal, creating a larger total contact surface than is possible with a "softer" paste, and would obviously work best if the diamond crystals themselves were in contact with both surfaces.

I suppose you wouldn't want to also sacrifice the thermal conductivity of the paste itself, so you wouldn't want them to be too big, and you definitely wouldn't want the heatsink to move once it is placed (even a fraction of a micron could drastically reduce the conductivity, though it *might* be recovered if the heatsink/paste settles...).

And this would definitely require exacting specifications on actually placing the heatsink. Might be better for factory-applied heatsinks than user-applied ones...