Transistor with a single atom of phosphorous dopant

Capt'n Midnight

http://www.nature.com/nnano/journal/vaop/ncurrent/full/nnano.2012.21.html

Ok they haven't miniaturised it all yet and you need to run it at 20mK above absolute zero, but expect it to lead to better (faster, smaller , lower power) chips

One of the limits at present is that as transistors get smaller you can't rely on having enough special atoms in the active regions as they are randomly dispersed. This means that it may be possible in future to implant atoms where you need them.

krd

The thing is. If you have a transistor that small - will you get weird quantum effects?

Quantum randomness. If you're using the transistor for binary logic, wouldn't it be a case that you would not always get the state you're expecting.

With larger transistors - As I understand it - since there are more molecules, and certain states have a higher probability than others, just by the law of large numbers, the transistors state is highly predictable.

Morbert

krd wrote: »

The thing is. If you have a transistor that small - will you get weird quantum effects?

Quantum randomness. If you're using the transistor for binary logic, wouldn't it be a case that you would not always get the state you're expecting.

With larger transistors - As I understand it - since there are more molecules, and certain states have a higher probability than others, just by the law of large numbers, the transistors state is highly predictable.

Yes, you do get lots of quantum effects. In the mesoscopic scale of current transistors, semiclassical models are used, which incorporate some quantum mechanics. As they get smaller and smaller, quantum effects become more prominent. The latest Intel chips are advertised as having "3D architecture". What this means is the transistors, instead of lying flat on the chip, are stacked like a row of books. These transistors (called finFETs), are on the scale of 22 nanometres, so well into the quantum regime. The next generation will effectively be arrays of quasi-1D wires, which involve pushing electrons through a 10nm wire. So even before we begin to contemplate these new single-atom transistors, quantum mechanics is already important.

A lot of the research goes into characterising how the voltage through a transistor (at the quantum scale, this is often called electron transport) interacts with its surrounding, and how it can be controlled (switched on and off robustly). The research revolves around concepts like Self-Energy and electron correlation.

rgunning

Well, more accurately, you expect to see a lot of quantum effects. And you would expect to see it screw things up transport-wise. However, if you look at a previous paper released by more or less the same group in Australia this year (in Science, around January, Ohm's law survives in the atomic scale - or something like that), you'll see that nanowires down to single atom heights can behave quite reasonably and classical devices can be made out of them.

Also, before you worry about any of this stuff, you worry a lot more about leakage current, losses and all that kind of stuff, I would say.

Morbert

rgunning wrote: »

Well, more accurately, you expect to see a lot of quantum effects. And you would expect to see it screw things up transport-wise. However, if you look at a previous paper released by more or less the same group in Australia this year (in Science, around January, Ohm's law survives in the atomic scale - or something like that), you'll see that nanowires down to single atom heights can behave quite reasonably and classical devices can be made out of them.

Also, before you worry about any of this stuff, you worry a lot more about leakage current, losses and all that kind of stuff, I would say.

You do see a lot of quantum effects, but they don't screw things up because they are already incorporated into the calculations, either directly (DFT or GW approximations) or indirectly (semi-classical models).

krd

Morbert wrote: »

You do see a lot of quantum effects, but they don't screw things up because they are already incorporated into the calculations, either directly (DFT or GW approximations) or indirectly (semi-classical models).

What do you mean. If you're to use the transistor for logic, it's not always going to give the same logical result.

If you used and array of them, you'd get a closer result.

I wonder how durable these transistors could be - and what conditions they need to operate under.

Capt'n Midnight

krd wrote: »

What do you mean. If you're to use the transistor for logic, it's not always going to give the same logical result.

If you used and array of them, you'd get a closer result.

I wonder how durable these transistors could be - and what conditions they need to operate under.

It's not durable at all , it's just a proof of concept , but the discrete quantum levels may mean it could give the same results each time (I haven't read the full paper yet)

Here, we use a combination of scanning tunnelling microscopy and hydrogen-resist lithography to demonstrate a single-atom transistor in which an individual phosphorus dopant atom has been deterministically placed within an epitaxial silicon device architecture with a spatial accuracy of one lattice site. The transistor operates at liquid helium temperatures, and millikelvin electron transport measurements confirm the presence of discrete quantum levels in the energy spectrum of the phosphorus atom

krd

Capt'n Midnight wrote: »

It's not durable at all , it's just a proof of concept , but the discrete quantum levels may mean it could give the same results each time (I haven't read the full paper yet)

I've read the article. I think the whole thing is really interesting.

The other day I did a much longer post to this thread - but it got garbled before it could be posted and I couldn't be arsed to do it again.

I believe even in large scale transistors, the quantum effects do sometimes cause logic errors. Simple by the law of large numbers - for every zillion switches, the logic is wrong. It's also the same law of large numbers ensures most of the time the logic is correct. Individual molecules may be experiencing different things, overall the transistor is behaving predictably.


I'm pretty sure, for an integrated circuit to work, using these transistors, the temperature would have to be in and around liquid helium - permanently, to stop the thing flying apart. At that scale too, the thermionic effect would cause the logic to go haywire.

The use of the scanning tunnelling microscope is interesting. I'd love to see an animation of what they think the transistor will look like when in operation.

Capt'n Midnight

krd wrote: »

I believe even in large scale transistors, the quantum effects do sometimes cause logic errors.
...
I'm pretty sure, for an integrated circuit to work, using these transistors, the temperature would have to be in and around liquid helium - permanently, to stop the thing flying apart. At that scale too, the thermionic effect would cause the logic to go haywire.

there''s always cosmic rays

imagine a transistor with 10 or 20 dopant atoms evenly spread out

Morbert

krd wrote: »

What do you mean. If you're to use the transistor for logic, it's not always going to give the same logical result.

If you used and array of them, you'd get a closer result.

I wonder how durable these transistors could be - and what conditions they need to operate under.

Here is a typical example of the type of calculations I am talking about

http://prb.aps.org/abstract/PRB/v79/i11/e115303

They are very similar to the type of calculations I carry out.

krd

Capt'n Midnight wrote: »

there''s always cosmic rays

Yep. Thought of those too. And there's radioactive decay to consider.

Basically, since day one, with IC logic circuits, they've known about the spontaneous logic errors - that's why it's a little irritating, when people say things like "A computer can't make and error, only a human can". Even testing was carried out behind very heavy shielding - with idea of blocking cosmic rays - and the errors still occurred. Then the idea was it might be radioactive decay - in the IC. Although, I don't have a great understanding of quantum effects - I understand, that they allow for the highly improbable to happen - given long enough, it might be enough for a logic error.

imagine a transistor with 10 or 20 dopant atoms evenly spread out

Yep, kind of mad. ............I think they might need a bigger molecule. I think I've seen suggestions of using a doped buckyball as a transistor - I don't know if anyone has tried it. Buckyballs are very resilient molecules - They're supposed to be very common in chimney soot. If they can survive a chimney - they might be able to operate at a much more practical temp than liquid helium.

I have no doubt someone is going to get this to work. Whatever it's going to be, it's going to look diabolically clever.