Making the new silicon | Hacker News Reader

Making the new silicon

newsoffice.mit.edu

51 pointsuser_23571118 comments

"CEI’s GaN transistors have at least one-tenth the resistance of such silicon-based transistors"

To be pedantic, silicon-based transistors also have at least one-tenth the resistance of silicon-based transistors.

What happened to Gallium-Arsenide? Wasn't that going to replace silicon? Did it not pan out? Is it used in some cases?

I really know nothing about the state of semiconductor materials, obviously!

Well Gallium-Arsenide, Indium-Phosphide and all kinds of other semi-conductors. But thing is, Silicon is well entrenched and there are work-arounds for all the different shortcomings of silicone and really a lot of experience working with Si. So the theoretical advantage of different substances where never able to overcome the momentum Si has.

They will eventually, working from the extreme cases on in as price drops, but for the foreseeable future, Si is king.

Most of the microwave amplifiers modules (MMICs), mixers, etc. the company I work for are built on GaAs, and we're starting to use GaN (Gallium Nitride) on one of our Ka band units (28-31GHz). The GaN parts are quite a bit more power efficient.

Not exactly sure why the tech is not more widely used though. Funnily enough, most of the GaN and GaAs devices in general use are LEDs!

In late 1980s, it was expected that by 2000, you'd have a GaAs computer on every desktop (powered by Occam or Prolog, of course ;-)). In the end, it turned out that silicon had more life in it than was expected, while being much cheaper and easier to manufacture.

There was also the fact that silicon was already heavily entrenched in the industry by the time people started to tinker with the possibilities of GaAs seriously. At least that's what I understood to be the main factor after listening to Seymour Cray's talk which was posted here some time ago: https://www.youtube.com/watch?v=xW7j2ipE2Ck He talks about how difficult it was to convince large suppliers to work with GaAs at about 30 minutes into the talk.

Gallium-Arsenisde has a high switching speed because of the higher carrier mobility, but the higher frequency applications it is used in dissipate high power.

This article's transistors will not likely have a high switching speed, just high enough for switching power converters (100kHZ-few MHZ). The important feature of GaN according to the article is that you generate 1/10 the waste heat from current just flowing through the switch.

Power transistors don't perfectly 'turn on' like an ideal switch. They have a resistance to the switched path on the order of milliohms called Rds (resistance source to drain). And, heat dissipated from that is ~ I^2 x Rds.

Right but you can make your on resistance arbitrarily small by speccing a bigger transistor. It might cost more, but there's plenty of ability to reduce the on resistance.

My guess is that this is a breakthrough because it has lower resistance per unit of gate capacitance. It takes a certain amount of energy to switch the transistor on, and when you want to turn it off you generally connect it to ground so that the switch turns off very quickly. That energy is lost.

I've emailed a very experienced electronics engineer friend of mine who has built more switching power supplies than anyone I know. He hasn't gotten back to me yet.

It really has to be something more than the "on" resistance though. It's possible to get arbitrarily large MOSFETs and get micro-ohm resistance, where as the GaN transistor has milli-ohm resistance.

http://www.digikey.com/product-detail/en/TPHR9003NL,L1Q/TPHR...

http://www.gantechnology.com/products.html

What is the difference between silicon and gallium nitride that makes silicon default to off when the circuit is broken and the standard GaN transistors default to on?

Not exactly sure what the article was going for. Maybe the importance of N-type transistors (they are useful just for being more efficient). Silicon transistors are widely known for failing in a shorted state.

I think they actually mean "depletion mode" vs "enhancement mode". It's not so much when the circuit is broken as when the gate is allowed to float.

(The article doesn't even say whether they're talking about FETs or BJTs, but I presume FETs with a super-low RdsOn)

It's mostly a function of how the transistors are made. If a transistor is on by default, they're called "depletion-mode" transistors--this means that the channel between the two sides of the transistors normally conducts, and via the control voltage, you deplete that channel of carriers. This is set up by the "doping" of the channel, which is one of the first steps in the fab process.

Transistors that are off by default are called "enhancement-mode". GaN transistors also come in enhancement-mode flavor, so I'm a little surprised by the article. You can buy them, even as a hobbyist, from EPC. [0]

[0] http://epc-co.com/epc/Products/eGaNFETs/EPC2023.aspx

[1] http://www.digikey.com/product-search/en?WT.z_cid=sp_917_011...

Once it becomes economical, with GaN processes pick up where silicon nodes are right now, or will the whole drive have to start at a less dense node?

Great news, but they didn't say anything about selling transistors to the public. When do you think we can expect to buy those like we do with mosfets?

You can already buy GaN FETs (although not the exact type being researched by MIT, as mentioned in the article):

- http://epc-co.com/epc/Products/eGaNFETs.aspx

- http://www.ti.com/lsds/ti/power-management/gan-overview.page

You could have a pile of GaN FETs on your desk tomorrow morning, if you really wanted: http://www.digikey.com/product-search/en?pv606=18&FV=fff4001...

Would this also significantly reduce the heat produced by data centers as less of the current going through the transistor is converted to heat?

No, it wouldn't (most likely): https://en.wikipedia.org/wiki/Jevons_paradox

Reply on news.ycombinator.com