Hacker News Reader

Comment by phegler | Hacker News Reader

Okay this is cool research

just one tiny tiny microscopic question...

why would someone use diamond wafer costing 1000x more than a silicon wafer??!

If I'm reading this correctly it's two or threefold. Faster switching times, which both increases max speed and reduces power consumption. Avoids flaws due to silicon doping, presently we add ions to silicon to get P or N type junctions and they've found a way to get P type junctions using "2D Hole Gas" which is basically a thin layer of voids inside the bulk diamond crystal after treatment and avoids a category of flaws that affect silicon yields. Third, and this is speculation, but diamond has a way higher thermal conductivity and lower thermal expansion coefficient, so it can likely withstand higher temps and conduct heat away faster.

I think the better question is: what real world applications need these qualities enough to justify the cost? They mention cooling application, I don't buy that anyone is that interested in reducing hotspots in their chips even if it makes them last longer.

I had a family friend working on diamond semiconductors in the 90s. The application at that time was for sensors located in the hot section of a jet engine.

For high-voltage applications (like power supplies) or high-power high-frequency applications (like radio transmitters) or high temperature applications (e.g. inside ovens or engines) silicon transistors are already in the process of being completely replaced by silicon carbide or gallium nitride transistors.

Despite the fact that those are much more expensive than silicon transistors, they are so much better that their use saves money in other parts of the system in which they are used.

Diamond transistors could be even better than SiC and GaN transistors, but there are many difficult technological problems that must be solved until being able to make commercial products.

The linked page reports some significant advances, which avoid a source of problems in the past, which were caused by the poor behavior of dopant impurities inside diamond crystals.

> the cost?

What cost? The lab cost today, or the cost after we figure out how to make thousands a day in a diamond wafer fab?

If this were to become practical, operating high speed circuits in space might become a great deal easier due to higher temperature differential. High frequency RF in space is valuable enough that this application is certainly a candidate for such devices, even at costs far greater than Si/SiC/GaN.

Space. They are more resilient to radiation, faster, don't generate as much heat and are more tolerant of heat. That pretty much tells you where they're going to be used.

I would like to clarify your second point. The "2D Hole Gas" is not a thin layer of voids in the diamond. It is rather an absence of electrons in that layer. Holes are a pseudo particle with positive charge and a different mass from electrons (effective mass). All semiconductors have them when you excite electrons, but they're generally not used as the main charge carriers, as their mass tends to be larger than electrons, making them less mobile

gotchya

Thank you for helping me understand

> why?

>> 50x faster switching

It's hard to say which applications will benefit the most, but orders of magnitude improvement tend to enable qualitative shifts downstream.

Not to mention, 1000x is just the cost right now while the tech is new.

yeah, I get it now. I was a bit too narrow minded

The moment it becomes useful its cost will fall down to silicone prices really fast.

Right now you can buy 100g of diamonds for low teens in $ from China. Diamonds are produced in enormous quantities as abrasives. And if you do something on large scale you become better at it.

So far we haven't invested in big monocrystals because there was not much use for them aside from jewelry and probably other small niches. The moment some R&D makes something useful out of them - they will be Jevoned into oblivion.

The higher the voltages it can handle, the faster it can switch efficiently, and the better the thermal performance, the wider its potential applications in power electronics.

Hmm, power electronics is often so big, that might need a lot of diamond and could get pricy, who knows.

Very low power chips might be the main application, as there is already a huge market (laptops and phones).

Or maybe there will be a whole set of new rf chips, profiting from the clock speed.

Power electronics isn't just grid scale equipment.

Indeed, variable frequency drives for three-phase AC induction motors are available for single horsepower motors. A VFD rectifies the incoming AC to DC which then flows through a capacitor and into an inverter which converts the DC to quasi-sine (stepped) three-phase AC for use by the induction motor.

Possibly on a par with the other exotic materials currently in use.

Reply on news.ycombinator.com