Slashdot Mirror
Aussie, Finnish Researchers Create a Single-Atom Transistor
ACKyushu writes "Researchers from Helsinki University of Technology (Finland), University of New South Wales (Australia), and University of Melbourne (Australia) have succeeded in building a working transistor whose active region comprises only a single phosphorus atom in silicon. The results have just been published in Nano Letters. The working principles of the device are based on sequential tunneling of single electrons between the phosphorus atom and the source and drain leads of the transistor. The tunneling can be suppressed or allowed by controlling the voltage on a nearby metal electrode with a width of a few tens of nanometers."
Well, at least this seems to set an ultimate limit to Moore's law, since it's not very easy to go to less than one atom per transistor.
Intel already has Atom processors.
Perhaps I can reach over 9000 Ghz ...
The devil is in the details. The "Active" region is only 1 atom wide, but the gate is still "10s of nanometers" Last I checked, the gate was still part of a transistor. We're currently mass producing with critical dimensions at 34 nanometers where I work. Granted, this is sweet, knowing that a transistor's active region can be that small. Still, the limit will really be placed on reproducibility. I mean, placing a single phos atom in the middle of a silicon chip at just the right location? That kind of technology being moved into the semiconductor industry for mass reproduction and economies of scale is still a long ways out and I personally think Moore's law will lose steam before then.
We've had single Atom CPUs for some years now... :)
If the state of a gate depends on one electron, it will be highly sensitive to radiation. So what do we do ? Embed these in large blocks of lead ?
Well, at least this seems to set an ultimate limit to Moore's law, since it's not very easy to go to less than one atom per transistor.
Yes, a single atom should be small enough for everyone.
...Wouldn't that be kinda hard to read?
Notice that this discovery was NOT published in Nature. Wonder why? Here's why: http://en.wikipedia.org/wiki/Jan_Hendrik_Sch%C3%B6n . Stay skeptical, wait for replication.
- "Hear that?! The percolations are imminent! Cease your ingress!"
Uncontrolled replication? No, we mustn't wait. Prepare the nanodefenses!
Do people really need to be told where Helsinki, New South Wales, and Melbourne are?
This thingy is just a research device, just good for research. It's not a precursor of anything practical.
It's been known for many, many years that there are serious tradeoffs to be pondered when you shrink transistors (and FETs).
Your basic linear dimension versus surface area versus volume scaling laws are in full play here.
You win at first, as smaller base or gate lengths lead to more speed, and less surface area means less capacitance to charge up.
But below a certain size the rapidly shrinking cross-sectional area reaches its current-carrying capability, while noise and leakage loom large.
Right now the low-level chip designers, with their 10^12 atom transistors are already spending a large part of their time with these issues. The challenges are not going to go away, they just get larger as one attempts to shrink things even more. It's unlikely that these hard challenges can be overcome to span the million-million times distance to a true one-atom transistor.
So don't put any big money on ever having one-atom transistors in any practical device.
Well, at least this seems to set an ultimate limit to Moore's law, since it's not very easy to go to less than one atom per transistor.
Reckon it is time to start working on a Matrioshka brain to ensure our future computing needs are covered.
The Long Now Foundation
Then how would you call a Field Effect Transistor?
No. A valve uses charge carriers (electrons) floating in a vacuum. A transistor uses either electrons or holes in a semiconducting solid as charge carriers. A semi-conducting solid is not as close as possible to a vacuum.
Would be designed around a Higgs Boson which would know when to come back from the future and switch with no gate delay...
Who says? I imagine with certain atoms you could include the functionality of multiple transistors in one atom. Or use smaller atoms. We are getting closer though :p
Not very easy to achieve does not equal "ultimate limit". I though we'd have a major paradigm shift by now, but the current one just won't die. I'm not making any bets anymore...
Then how would you call a Field Effect Transistor?
Nice try with the bolding of F,E,T. Perhaps, with your use of "how" instead of "what", you are not native English, and don't understand the use of "sic". This is often used to denote a particular term/phrase/whatever which the writer considers incorrect, but which is being quoted nevertheless as-is. The fact that a FET is called a FET doesn't mean it is a transistor, any more than the Democratic People's Republic of Korea is democratic, for the people, or a republic. You thus still call the DPRK the DPRK, as you call a FET a FET, because those happen to be the best-known names, but you don't call the DPRK a "democracy" and you don't call the FET a "transistor".
Now, the evidence, from the horses's mouth. Read the quote in the left hand column by the guy who named the transistor. He named it so because it - "it" being the point contact transistor and devices descended from it, such as the modern BJT - had transfer resistance, the dual of the vacuum tube (or FET) which is defined in terms of its transfer conductance. Understand?
A semi-conducting solid is not as close as possible to a vacuum.
A substance comprising one atom is as close as possible to a substance comprising no atoms. My comment was somewhat tongue-in-cheek, designed to illustrate the danger of imprecision. In particular, the discovery is /not/ of a one-atom transistor, or even a one-atom FET, it's of a FET with a single atom channel. The other AC seemed to understand this.
Apart from the implications this might have for classical electronics, the long-term goal here is to build solid-state quantum computing devices. The phosphorus donor has one lonely electron, and that electron's spin is a good candidate for a qubit. One of the good things about P in Si is the long decoherence times -- T2 times of almost one second have been demonstrated. The phosphorus' nuclear spin of 1/2 stays coherent for hours, if we can find a way to get at it.
Of course, the NIST guys with their ion traps have demonstrated several interacting qubits, but perhaps P in Si chips might be useful as a more stable, more scalable, cheaper or smaller alternative.
That kind of technology being moved into the semiconductor industry for mass reproduction and economies of scale is still a long ways out and I personally think Moore's law will lose steam before then.
I would say it's pretty much lost steam already. If you take a function that can't exploit multiple cores, then the single core performance has not improved much in a while. More cores is a "cheat" that extends it somewhat but I doubt 10+ cores makes any sense for end users so it won't scale much further than it already has.
The other bummer is power, even though there's a massive focus on power savings now running a CPU/GPU at 100% draws more and more power. The latest AMD offering, the HD 5950 is bumping the head into the 300W ATX limit, and most agree it's designed to overclock for more. Or it is perhaps the same bummer, since it's the main reason 10GHz+ cores aren't practical.
Fortunately, I think there'll be a lot of innovation in other areas, particularly in networking (fiber, 4G mobile broadband), storage (SSD) and form factor (think iPhone, Wii controller, OLED displays and whatnot).
Live today, because you never know what tomorrow brings
The article states that the goal is to produce a quantum gate out of the single atom. If such a gate could be produced reliably and operate on controllable quantum states, then particular classes of problems could be solved dramatically faster. The promise doesn't lie in pushing existing technologies to be smaller, it's to supplement them with quantum computing capability.
So if this is the future...where's my jet pack?
Mod parent up. mangu has completely missed the point, and so, apparently, have the moderators. A fet is more similar in principle/modelling/use within a circuit to a valve than the transistor, i.e. the BJT (before commercially viable FETs even existed). The only way a FET is more similar to a transistor is the nature of the charge carriers.
Specifically:
gate = gate
cathode = source
anode = drain
A depletion mode FET allows current to flow except when the reverse bias voltage between gate and source is such as to increase the depletion enough to pinch off flow. A triode valve typically allows current to flow except when the bias voltage between gate and cathode is such as to repel all electrons back to the cathode. Either way, input resistance is high. Voltages in between vary current flow.
The basic (BJT) transistor collector-emitter current, meanwhile, is zero unless forward bias is applied on base-emitter. Once your 0.6V is applied, it is mainly base-emitter current which determines collector-emitter current. Input resistance is relatively lower.
Valves and FETs admit voltage controlled currents. They are primarily defined in terms of trans-conductance.
Transistors admit current controlled currents. They are primarily defined in terms of trans-resistance.
I am shocked that /. needs this explained.
Not for my imaginary girl friends, you insensitive clod!
I would say it's pretty much lost steam already. If you take a function that can't exploit multiple cores, then the single core performance has not improved much in a while.
Moore's law does not describe performance.
Right, there's a couple of ways that it could go, one would potentially be to subatomic particles, another would be to using 1 atom as multiple transistors. Or possibly one could start making the chips themselves larger again. Moore's law itself just speaks to the number of transistors on the chip that can be placed inexpensively in an integrated circuit, not to the size itself. The reason for the shrinking has been that for efficiency reasons you want the path of the electrons to be as short as possible and the larger the chip itself typically the more difficult it is to keep the yield up of highly delicate chip designs.
That being said, since Moore's law isn't an actual law, it will eventually come to an end. On the plus side this has gone so far that when it does come to an end it's likely going to be largely a moot point. As we'll probably end up needing a new law to cover the case when we're dealing with quantum computing. But at anyrate, even the cheap computer in front of me can do a hell of a lot.
Just because one researcher commits the "biggest fraud in physics in the last 50 years" that happens to involve transistors doesn't mean all such research is fradulent and there is no reason at all from your link to be more skeptical than normal about this research.
The active part is a single atom of Phosphorus ? While it might sound like you could get high density circuits with that, I hope they plan on using water cooling, given that phosphorus tends so spontaneously combust.
Australia and Finland are planning to take over the world, don't you know. This is just the tip of the ice-burg.
They wanted to attempt a project that none could do alone so they emailed each other and collaborated perhaps? It is very common for Australian researchers to collaborate with scientists from other countries, I don't know why exactly, probably because it makes sense? How did this discussion get off the ground, for all we know the other is in Finland, yet we have a common ground in wanting to explain this?
Really, I'm just astounded by your question, can you give me a good reason why they wouldn't collaborate?
"Just because one researcher commits the "biggest fraud in physics in the last 50 years" that happens to involve transistors doesn't mean all such research is fradulent and there is no reason at all from your link to be more skeptical than normal about this research."
Ahhh, a swing and miss, I felt the wind from that one though! I'm afraid that you (and the schizophrenic raving lunatic AC responding just before you) are making rather unwarranted assumptions about my comment. I never suggested "all such research is fradulent [sic]". I am suggesting that in a field where the science is currently progressing at breakneck speed and is consequently ferociously competitive, at the absolute cutting edge of the cutting edge of materials science and single atom manipulation, a breakthrough discovery as important and consequential as this needs to demonstrate replicability before the provisional acceptance warranted by the evidence in this paper alone can be elevated to the level of confident belief in the phenomenon being demonstrated. So yes, there is rational reason to be a little "more skeptical than normal about this research" because extraordinary claims require demonstration of a commensurate (ie. more than normal) level of extraordinary evidence in their favor before being accepted as solid fact. This is how science and rational skepticism works. Trust, but verify.
- "Hear that?! The percolations are imminent! Cease your ingress!"
10GHz cores are not something you want to sit next to without proper shielding it's possible you could be burned. It was a concern back when we were approaching single core 5GHz before the dual core processors came out and everyone was speculating we would be at 10GHz in 3-5 years. That was about 6 years ago I believe.
Perhaps you could also make a transistor out of nitrogen, which would be smaller than the phosphorus atom. Or maybe they'll discover a way to make a transistor out of sub-atomic particles and even have multiple transistors per atom.
Give me Classic Slashdot or give me death!
Scientists are just as online and wired as the rest of the world... they email each other, read each others papers and attend conferences regardless of nationality. Just imagine one researcher reading the usenet post or paper of someone else doing the same type of work as her... of course they will contact that person.
:) )... A few days after showing him sci.biology(or whatever) and how to post he was so excited that he had already made contact with 4 or 5 people around the globe engaging in similar areas of research as him... It was truly an eye opener for me as to how powerful the Internet was going to be and how it would impact the world some day... alas I did not anticipate rickrolling or LOLcats...
A real story from me was when I was at university and introduced a friend of mine(doign a PhD in microbiology) to USENET... this guy had never used or really heard of the Internet as it was quite new(and I am old
The devil is in the details. The "Active" region is only 1 atom wide, but the gate is still "10s of nanometers" Last I checked, the gate was still part of a transistor. We're currently mass producing with critical dimensions at 34 nanometers where I work.
Yeah, that's the first thing of thought of when I read the title. If I'm understanding correctly the gate is of dimension [1 atom x 10s of nm]. That's still damn impressive. I guess now the work has to come down to dropping that 10s of nm into the handful of atoms range.
"Educate the mind but never at the expense of the soul."~Blessed Basil Moreau
I am an electronic engineer and am perfectly aware of the origin of the word "transistor".
I also happen to know that resistance and conductance are equivalent terms, resistance being the inverse of conductance. You can model vacuum tubes or transistors, both unipolar and bipolar, by either conductance or resistance.
However, if you want to be pedantic about the word "resistance", you should study the behavior of both bipolar transistors and FETs under saturation. Under non-saturated conditions the current in the output side of the circuit depends on the excitation in the input, this excitation being a current in the BJT and a voltage in a FET.
Under saturation the device conducts as much current as the circuit allows, normally this current is limited by an external load resistance. In these conditions a BJT behaves as a resistor whose value does not depend on the excitation current. A FET, OTOH, has a channel resistance that varies with the voltage being applied on the gate. For depletion junction FETs this resistance is almost inversely proportional to the gate voltage and this feature has been used in analogic circuits.
Conclusion: for true gramer nazis, only FETs should be called "transfer resistors".
Just one atom in millions. Magic!
"The results have just been published in Nano Letters."
I'm guessing that'd be a Times New Roman 0.0000001 pt font then? Damn, I left my scanning electron reading glasses at home today.
It gripped her hand gently. 'Regret is for humans,' it said.
They built a few a bombs and reactors.
Secunda the 160,000 barrels of oil a day after a coal into a liquid process.
African science is doing fine.
Domestic spying is now "Benign Information Gathering"
Um, you could still play with the wave function of a electron surrounding the atom. ^^
Or just go to single photons.
Any sufficiently advanced intelligence is indistinguishable from stupidity.
Egocentricity FAIL!
“sic’ is Latin, and known in pretty much every language with an European origin.
Next you tell me, that Edison invented the light bulb.
Any sufficiently advanced intelligence is indistinguishable from stupidity.
I was under the impression that sic was short for "said in context" meaning a mistake that was in the original, quoted element of the article.
:)
Guess I was wrong
"Freedom in the USA is not the ability to do what you want. It is the ability to stop others from doing what THEY want"
The fact that a FET is called a FET doesn't mean it is a transistor, any more than the Democratic People's Republic of Korea is democratic, for the people, or a republic.
Ah, you come from the school of redefinition, so you're trying to change the meaning of the word "transistor" to mean what will support your foolish argument. Yes, there are two types of transistor, each with different characteristics that orient towards particular applications (e.g., high-power for BJT, digital logic for FET) but to define only one of them as a "true" transistor just speaks of your ignorance. You must be a sound engineer?
Next up given how weak your argument is, you'll be raising your voice or writing in capitals.
"Little does he know, but there is no 'I' in 'Idiot'!"
Published in nano letters? I don't have an electron microscope, you insensitive clod!
Wow, you really must be new here.
To have a right to do a thing is not at all the same as to be right in doing it
I believe scientists are at this very moment working on something called the "internet" which allows the transmission of messages and data across vast distances at the speed of light.
It could be huge.
To have a right to do a thing is not at all the same as to be right in doing it