Slashdot Mirror
Single-Atom Layer of Tin May Be a New Wonder Conductor
At Kurzweil AI, an article proclaims that the next wonder material for computer chips may be an unexpectedly common one:
"Move over, graphene. 'Stanene' — a single layer of tin atoms — could be the world’s first material to conduct electricity with 100 percent efficiency at the temperatures that computer chips operate, according to a team of theoretical physicists led by researchers from the U.S. Department of Energy’s (DOE) SLAC National Accelerator Laboratory and Stanford University." (Original paper is available here, but paywalled.)
Well if we are talking about power transmissions then superconductors are 100% efficient. Nil resistive losses. You still have capacitive and inductive losses you cant get rid of when dealing with AC or DC ramp up, ramp down. You also have external costs like keeping the superconductor cooled, but that is system efficiency, not semiconductor efficiency, that is cooling cost is not dependent on power transmitted. So if you are looking at time invariant current and exclude cooling costs then superconductors are 100% efficient in current transmission.
Arxiv to the rescue: http://arxiv.org/abs/1306.3008 (This may lack editorial changes etc. made by the journal, but should be factually complete.)
Stefan Axelsson
For those of you not in the semiconductor business, the fact that these conductive strips is pretty important too. Most of the capacitance (that has to be charged and discharged whenever a node switches, causing losses in the transistors driving the node) is sidewall capacitance: capacitance between adjacent lines on the same level. Single-layer conductors won't completely do away with lateral capacitance (fringing, for instance) and the vertical capacitance will still be there -- but there's going to be a big reduction in power if they can get this to work. My guess is that by the time it reaches production it won't exactly be one layer, either -- it'll be a laminate with multiple redundant layers.
Always assuming the predictions play out.
Lacking <sarcasm> tags,
What is the maximum current that can be transported through strips of various widths?
Mostly moot. The really nasty consequence of conductive losses in semiconductors is that it degrades signals traveling across the chip. We insert buffers along the route to restore signal amplitude and reduce delays (those RC delays are ugly). This would zero the resistance and reduce the capacitance, which is a big deal. Also, for reliability reasons, we'd probably build laminates with multiple layers separated by dielectrics.
How sensitive to defects is the process?
Depends on the width of the path. The usual solution is to add redundancy, multiple single-atom layers separated by dielectric. Vertical space on chips is relatively cheap, as long as you don't need to use extra mask layers or move the material from one process stage to another.
Tin is going to be a major problem for much semiconductor processing - as it means you basically now can't solder the chip, or do any even 'low' temperature processing after it's deposited - it has to be the last layer.
We don't solder the devices directly anyway -- the organic dielectrics used in advanced processes like the old metal-melting temperatures even less than tin does.
Lacking <sarcasm> tags,
the losses are capacitive, you can reduce the losses by making smaller transistors but you really cant affect it by material selection
Yeah, the reason why material selection doesn't matter in capacitors is precisely why many of them are being manufactured using the fairly rare element named tantalum. It's just for the fun of it. ;-) Perhaps you're right about the interconnect material selection but there's a lot of material selection going on in modern ICs beyond that.
Ezekiel 23:20
These guys are talking about a 2D topological insulator. This is the current hot area of research in condensed matter physics, and is absolutely not a superconductor.
A topological insulator is best described as an insulator, which for very particular types of conduction (direction, location and energy limited) acts like a very good metal. It's really interesting, and scientists are trying to show it will have practical use, and these materials might end up in a computer chip in a few years, but...
There is a big difference between a lab effect and the real world. Carbon nanotubes have most of the same "non scattering" effects you'd hope to find in a topological insulator. Yet, in most actual devices, they do not conduct in bulk the way theory would suggest. For nanoscale systems (these are nanoscale systems) the environment around the material is nearly as important as the material itself, and scattering from the environment (oxides, metals, air) drastically reduces the performance of the material. There are ways around that, but the additional costs and engineering difficulty are generally enough to prevent any practical commercialization.