Boffins build the smallest transistor, controlled by an atom

The world’s smallest transistor can be controlled by a single atom, according to a scientists from the Karlsruhe Institute of Technology in Germany. Unlike more traditional transistors it isn’t made from semiconductors. Instead, it’s crafted from metal and the gate - the part that acts as a switch turning a transistor on or …

[Anonymous Coward]

Scaling up the manufacturing process.

You can make a heat engine like this: make a disc on an axle and around the periphery place little pieces of cerium, then bring up a magnet (which attracts the nearest piece of cerium) and place a heat source under the disc on the same side as the magnet.

The cerium heats up above the Curie point, ceases to be magnetic, and the next piece is attracted causing a slow rotation of the disc.

You could build a cerium magnet heat engine of several thousand horsepower if you could only scale up the manufacturing process enough. It is however unlikely to replace existing heat engines. Ever.

[Yet Another Anonymous coward]

Re: Scaling up the manufacturing process.

Unfortunately it only rotates once - unless you put energy into re-magnetise the magnets

[Anonymous Coward]

Re: Scaling up the manufacturing process.

YAAC - no, I have built one and it works. It's quite simple; once one of the cerium pieces (lighter flints work) leaves the flame, it starts to cool and is soon below the Curie point, ready for the next cycle. You need to be a little careful with the design to get the dimensions and the flame size right, but once going it continues to rotate. The thermodynamics are obvious; some heat is transported from the flame to the air via the pieces of cerium, giving a heat source and sink.

[John Smith 19]

1 atom switched junction <> 1 atom wide junction.

So still a ways to go before the end of Moore's law.

But I do like the thinking behind it.

And metal is a better conductor than Silicon will ever be (the clue is in the name. semi-conductor).

Now, can they do a normally on transistor that switches off when you add an atom?

[bombastic bob]

Re: 1 atom switched junction <> 1 atom wide junction.

I'm actually curious about switching speeds here. If we can't go faster, what's the point?

Current tech limits us to under 4Ghz for the most part (yeah overclockers, but still). I always figured it would be 3Ghz, but it's gone all the way to 4Ghz. 4Ghz has a wavelength of less than 8cm and so in theory for a single clock cycle read transaction, you need a distance of 2cm or less between whatever on either end [to allow for switching and propogation etc.]. You mght get away with 4cm, but that's the speed of light we're talking about, and you can't exceed it. So yeah clocking along at 4Ghz generally means no more than 2cm between any 2 points that have to talk to each other at that speed.

Atom-sized transistors, therefore, could speed this up by compressing the size significantly. You'll still need a way to get the RAM physically closer to the L2 cache, though, so you can at least read it at the full clock rate. 2cm isn't very big, and that's only 4Ghz.

Now, HOW fast can this gel-substance switch at?

[Wellyboot]

Re: 1 atom switched junction <> 1 atom wide junction.

@Bob, You beat me to it :)

If (big if) the gel can reliably operate at these speeds, I was thinking discrete I/Cs operating perhaps 10x faster and a few magnitudes smaller with the lower energy requirement.

How far the atom actually needs to move in a transition will be the limiting factor, lets reconvene in 10 years and see how things are progressing.

[Yet Another Anonymous coward]

Re: 1 atom switched junction <> 1 atom wide junction.

>means no more than 2cm between any 2 points that have to talk to each other at that speed.

Only in DC, busses are transmission lines.

We 'talk' across oceans at GHz with light that takes 30ms to get to the other side.

[isogen74]

Re: 1 atom switched junction <> 1 atom wide junction.

> We 'talk' across oceans at GHz with light that takes 30ms to get to the other side.

... but 30ms is an eternity; if you want to clock at 4GHz you only get 250 *picoseconds* to get from A to B ...

[John Brown (no body)]

Re: 1 atom switched junction <> 1 atom wide junction.

"Now, HOW fast can this gel-substance switch at?"

And if the gel starts to slosh around and cause data errors, are you holding it wrong?

[Mayday]

I love it when...

...the people smarter than me discover things like this.

Keep up the great work!

[DJV]

Excellent news

But, they'd better watch out they don't attract the bullshitters and new-age nutters by mentioning "quantum" twice in the abstract...

[Tom 7]

Re: Excellent news

Its already cured my yoga injuries!

[Pascal Monett]

Re: Excellent news

Yeah, it's a transistor, or it's a quantum thingy ?

And that picture is quite an achievement : getting that good a shot of a single-atom object is impressive. I can't believe there's not more than one atom in there.

[Crisp]

Quantum Gel

Both apply and do not apply to affected areas.

[Kanhef]

Of course it has potential

they applied voltage to it.

[TonyJ]

Re: Of course it has potential

"...they applied voltage to it..."

<cough> across it...

I'll get my coat now.

[pɹɐʍoɔ snoɯʎuouɐ]

Re: Of course it has potential

pedant alert...

a potential difference which is measured in volts across it !!!

[ibmalone]

Re: Of course it has potential

Glad someone spotted the potential pun.

[Korev]

Re: Of course it has potential

>they applied voltage to it.

Did they resist it?

[ravenviz]

Re: Of course it has potential

Currently not.

[5p0ng3b0b]

Re: Of course it has potential

Faraday or two

[Schultz]

They show quantized conduction as expected for a single/few atom junction...

but the gate is 100 nm wide. Compare this to the scale of modern silicon gates before getting too excited about atomic-scale transistors!

[werdsmith]

Re: They show quantized conduction as expected for a single/few atom junction...

It's useful as a switch where we might want to measure the state of the junction and use it to hold information I suppose.

The article mentions on and off but nothing about the gradual increase in flow with forward bias that you would get around the knee of a doped layer silicon transistor.

[Wellyboot]

Re: They show quantized conduction as expected for a single/few atom junction...

@Schultz >>gate is 100 nm wide<<

Thats not shabby for a prototype proof of concept unless it turns out to be difficult to reduce in this design. The gel used here will very likely be improved over time.

[Mike 137]

The REAL real challenge

"The real challenge to making them applicable in the real world lies in the scaling up the manufacturing process."

The real challenge will actually be maintaining predictable behaviour in the face of the incredible amount of interference now present pretty much everywhere on Earth. Any control system that operates at the quantum scale intrinsically has very little if any noise immunity, so its reliable operation is problematic in real-world environments.

Some years back researchers (at Cambridge if I remember correctly) demonstrated a transistor that could be switched by a single electron. The essential question of course would be "which electron - the one we provided, or a randomly passing one?"

[John Brown (no body)]

Re: The REAL real challenge

"The real challenge will actually be maintaining predictable behaviour in the face of the incredible amount of interference now present pretty much everywhere on Earth. Any control system that operates at the quantum scale intrinsically has very little if any noise immunity, so its reliable operation is problematic in real-world environments."

Shirley the answer is to coat the device in wonder material graphene to shield it.

[sitta_europea]

Slight numerical error.

"It requires millions and billions of these transistors to build anything useful..."

Don't you just love it when journalists get hold of technical stuff and then fall flat on their faces?

I still have what we used to call a "transistor radio". This was, for those of you who think a smartphone is a pretty neat idea, the most popular communication device in history.

It has seven transistors in it. The makers were so proud of that, they put it on the box.

https://www.radiomuseum.org/r/orion_seven_transistor.html

https://en.wikipedia.org/wiki/Transistor_radio

[Grooke]

Re: Slight numerical error.

In attempted defence of the author: can you make a radio with 7 microAmp transistors?

I'd imagine the transistors used if frequency filtering and modulation would have to be a lot bigger than what we use today in digital circuits.

[Mage]

Re: Slight numerical error.

Yes, you can make a radio with about four 7 µA transistors. You'd need a bit more for the audio drive, though you can use a ceramic disk as used in a beeper as the earphone. A single 1.5V button cell to power it.

Or you can make an entire radio in an IC using DSP with only a low pass filter on the aerial (up to 30MHz or 40MHz), or a bandpass filter (probably any 40MHz band from 40MHz to 2GHz) using the ADC as an aliasing mixer.

[Anonymous Coward]

Re: Slight numerical error.

"I'd imagine the transistors used if frequency filtering and modulation would have to be a lot bigger than what we use today in digital circuits."

You probably have a few little boxes around the place that do frequency filtering and modulation of various complex kinds in the UHF and even SHF bands. I have several, and they consume an average of a few tens of milliamps including running the control computers and occasionally lighting up a screen.

So I wouldn't bank on it.

[Simon Harris]

Re: Slight numerical error.

I remember those days when radio manufacturers would advertise the transistor count like mechanical watch manufacturers would advertise how many 'jewel' bearings the watch had.

[Mage]

Re: Slight numerical error.

Valve (tube count) earlier.

Why when transistor usable by 1949 that there were almost no transistor sets till 1959?

There were some prototypes in 1948 or 1949. First commercial set was the Regency TR1 in 1954. It used a 22V battery (developed for valve hearing aids).

The first transistors about $18 each compared with 50c for some valves (tubes). The battery valves by 1953 had got to 1.4V and 25mA filaments. Some 0.7V for hearing aids (two in series for a 1.5V battery). Eventually the Russians had tubes with 1.2V 11mA filaments only about x4 length and similar diameter to transistors that could work to over 100MHz.

Though Sony changed their name to Sony to sell transistor radios made in Japan in USA in latter half of the 1950s, some US and many Japanese makers made pocket valve (tube) sets using 22V or 45V and 1.5V cell, using military & hearing aid tubes, with one regular battery tube for speaker or a pair of transistors. These even looked like 1960s plastic transistor sets.

[werdsmith]

Re: Slight numerical error.

If this is just an on off switch as described in the article, then it won’t even be capable of even a simple class A amplifier circuit. It will only be useful for logic and so millions will be required.

[Toastan Buttar]

More akin to a relay than a transistor

When the gate atom is in place and closes the circuit, is it persistent until moved back into the 'off' position? This could be a new type of non-volatile memory.

[annodomini2]

Re: More akin to a relay than a transistor

Was about to say the same thing, but for the majority of uses of transistors today (by count) most are used as switches anyway.

The major issue I see with it is the removal of the solid state functionality, just like relays, making the thing reliable over billions if not trillions of actions, in wide environmental conditions is the challenge.

I could only see this being developed if we had room temperature superconductors.

Any data on the switching speed (I respect it's a prototype)?

[Steve Knox]

"It has potential..."

Wouldn't that be a battery?

[Mage]

within the aqueous electrolyte

Curiously the first attempts to make a transistor, before the name was even invented, had the crystal immersed in an aqueous electrolyte,

[Arthur the cat]

"Since it’s made from metal"

I'm not sure a single atom of silver is a metal in the electronic sense of the word because metals are pretty much defined to have their conduction electrons smeared out across the bulk material, and one atom doesn't have any bulk for smearing across.

[Camilla Smythe]

Re: "Since it’s made from metal"

The article mentions that the metallic strips which the single atom contacts are silver so there is your 'bulk' metal. As Toastan Buttar suggests it sounds more like a relay than a transistor.

[steward]

The tech to save blockchain?

" This quantum electronics element enables switching energies smaller than those of conventional silicon technologies by a factor of 10,000."

Which, given the anticipated power consumption of blockchain, would help greatly...

[adam 40]

Potential breakdown

With the switch gap only an atom across or thereabouts I am wondering what the source-drain breakdown voltage will be in the 'off' state. I am guessing microvolts? So we might be looking at a whole new generation of core supply dc converters....

[Anonymous Coward]

Terry Pratchett said its

Turtles all the way down. (RIP)

But at a small enough scale, the latest thinking is that beneath quarks there are vertices.

Similar idea to superstrings but there are differences in the math.

I am intrigued, but would need to study this further to be sure it is correct and not merely another

mathematical anomaly which has little or no bearing on reality(s)