Slashdot Mirror
Law-Defying Transistor Smashes Industry 'Limit', Measures Just 1nm (thestack.com)
An anonymous reader quotes a report from The Stack: U.S. researchers have unveiled the world's smallest transistor reported to date, combining a new mix of materials, which makes even the tiniest silicon-based transistor appear big in comparison. The team, led by the U.S. Department of Energy's Lawrence Berkeley National Laboratory, designed the minuscule transistor with a working one-nanometer gate -- far surpassing any industry expectation for reducing transistor sizes. In the scientific study, MoS2 transistors with 1-nanometer gate lengths, published today in the journal Science, the researchers describe a prototype device which uses a novel semiconductor material known as transition metal dichalcogenides (TMDs). The transistor structure uses a single-walled carbon nanotube as the gate electrode and molybdenum disulfide (MoS2) for the channel material, rather than silicon. "The semiconductor industry has long assumed that any gate below 5 nanometers wouldn't work, so anything below that was not even considered. This research shows that sub-5-nanometer gates should not be discounted. Industry has been squeezing every last bit of capability out of silicon. By changing the material from silicon to MoS2, we can make a transistor with a gate that is just 1 nanometer in length, and operate it like a switch," explained study lead Sujay Desai.
At what cost?
Call Chuck Norris - he'll kick that transistor's ASS!
On a more serious note, wouldn't devices with such small geometries need some really heavy shielding to prevent destruction by cosmic particles? Heck, I have to wonder if at that size even background radiation would be a risk factor.
'The Economy' is a giant Ponzi scheme whose most pitiable suckers are the youngest among us and the yet-unborn.
... when they actually have a working product. These lab projects don't quality as realizable, I remember the same promises were made about CPU's getting to 10+GHZ that never happened and CPU speeds hit a brick wall around 2006 because heat and leakage became too much which meant going much beyond 5Ghz became a pipe dream.
Good thing it deals with very low energy levels. Because progress in computing does not translate to progress in other fields. In ten years, jet airplanes will still be the same and take 6 hours to cross the Atlantic.
This..
>The semiconductor industry has long assumed that any gate below 5 nanometers wouldn't work, so anything below that was not even considered.
This is simply false.
I should use this sig to advertise my book ISBN-13 : 978-1501515132.
Best news in a long time.
This is great news! The information density with these 1 nanometer transistors should be such that I can simulate simulate the universe with enough accuracy for sentient beings to eventually come into existence. Eventually they will advance sufficiently enough to question whether or not they are in a simulation, and they will begin efforts to test how accurate my simulation is in order to determine its existence. Then, just when they discover that the simulation is flawed in some way, and thus detectable, I'll pull the plug and start a fresh simulation.
Better known as 318230.
Let me use it to make and play the world's smallest electric violin.
Those who do not learn from commit history are doomed to regress it.
... Apple should be able to knock at least another 2-3mm of thickness off the iPhone with these things.
Log in or piss off.
Muon-catalyzed transistors!
Will this result in higher possible clock speeds?
For what at least looks like a possible solution to the ever-present electronic challenges of shrinking geometry, yielding the usual benefits of speed, size and density, I'm seeing some really ridiculous comments. Recall that the first transistor "...was about the size of the palm of a hand, with a depth of two matchbooks stacked on top of each other." Basic technological breakthroughs do not automagically turn into "thinner iPhones" or other crapola for a while, but they do break new ground.
1nm is the gate length, not the size of the entire transistor. Typically-quoted transistor sizes are actually the process nodes, which are half of the distance between the same feature in neighboring transistors, so they're not comparable to a measurement of an individual transistor. That said, I seem to recall a story from over 10 years ago, about someone creating a single 1nm transistor. The trick, now as then, is to use lithography to create billions of them connected to one another to form integrated circuits, and the main limitation in size reductions has been lithography tech rather than transistor tech.
Corruption is convincing someone that the selfless ideal is the same as their selfish ideal.
from the paper's abstract: "Simulations show an effective channel length of ~3.9 nm in the Off state". what does this mean? that the gate, in it's off state, needs 4nm or it will start interfering with nearby gates?
The word Gate is not referring to a logic gate (which is what it sounds like you're inferring), but to the Gate terminal of the transistor. When the correct polarity of voltage is applied to the Gate, the field effect causes a channel of charge carriers to form between 2 other terminals, the Source and Drain, allowing current to flow between them. The channel length refers to the distance between the Source and Drain terminals.
The channel length (as well as other parameters like the width, charge carrier mobility, etc.) determines how much current can flow between the Drain and Source when a given voltage is applied (i.e. resistance). By applying higher voltage to the Gate, you are narrowing the "effective" channel length (lowering the resistance).
When you switch transistors on and off, you are basically charging and discharging capacitors, which takes time. How much time is determined by the time constant, RC (resistance x capacitance). So, shorter channel length = lower resistance = smaller time constant = faster charge/discharge = higher speeds. That's why we make transistors smaller to make computers faster.
Only crack the nuts that crack. You don't put the ones that don't crack in the sack.
With normal transisitors, they degrade over time as electrons/atoms get knocked off the C/B/E. For big transistors, this degradation or wear is no big deal but how long can miniturization go on for before this does become a problem?
I was noticing that also, struck me as interesting, but I'm to u-educated about the exact nature of molybdenum disulfide to know why.
I've decided to Diversify my Holdings. I've divided my cash between my left and right pockets, instead of all in one.
Is this IBM vs US Lab?
Electron driving down a single lane that's very narrow might hop over to the next lane & go the other way.
Transistors depend on type of materials.
*Some transistors are slow to wake up (type of materials)
*Some transistors are fast to wake up (type of materials)
*Fat old transistors can switch on fast due to High wattage & materials warn.
*Feed the old fat transistor slowly with wattage and it will wake up slowww.
Channel length is a physical measurement and it does not change with applied bias. The "effective" is in contrast with "drawn". Ldrawn is the dimension as measured on the masks; Leff is the dimensions between the actual edges of the diffusion.
The industry is secretly hoping for a must-use technology that will break in 5 years. Does this accomplish that business goal?
That's an ancient dream. According to an old-time engineer who was an early transistor user in the early 60s, that was the industry's goal back then: purposefully only slightly longer life than vacuum tubes.
He told me a story of meeting strong criticism from the semiconductor vendor when they found out his company was dipping transistors in paint to color-code parts that they (the customer) tested as better or worse. Device quality/performance was very uneven back then. The vendor's opposition to dipping didn't quite hold water. Turned out that the hollow metal packages weren't airtight as to allow in oxygen that would eventually degrade them. Dipping in paint sealed them. Foiled again!!
Channel length is a physical measurement and it does not change with applied bias.
This is only true with "long" channel MOSFETS. Once you get down below a certain size, you get "short channel effects" including the narrowing of the inverted channel region with increased bias, called channel length modulation.
Otherwise this sentence from the abstract, which is what the OP was referring to, makes no sense:
Simulations show an effective channel length of ~3.9 nm in the Off state and ~1 nm in the On state
Only crack the nuts that crack. You don't put the ones that don't crack in the sack.