This is pretty huge. The last major break thru was getting superconductivity in Y1B2C3O at liquid nitrogen temperatures (-200 C) back in the 1980s. I worked in a lab with the discoverer of this phenomena, Dr. Paul Chu.
While -70 is still damn cold, it should be achievable a lot more cheaply than having to cool things with LN2, so hopefully-- cross our fingers-- this will lead to the breakout of superconductivity into broad industrial use (and maybe the mainstream consumer market) that people have sought for decades.
BTW- while in that lab I used a variation of the meissner effect to design a memory circuit that was theoretically buildable at the time- static RAM that was superconducting. (and basically, the essentials of a transistor were there so logic gates could be built from super conductors, that was the thought experiment I was doing.)
Since heat is a major concern in CPUs, the ability to be superconducting (and thus producing no heat) would be a huge boon for computation. Of course initially this wold be at scales much larger than current lithography and thus only for specialized applications. But who knows.
With effort the cost of keeping a CPU at -70 should come down dramatically the way battery capacity per dollar has... or dare we hope the way flash density has.
Your math is incorrect. This is a common mistake. Tonnes is not a unit of force although many laymen believe it is.
90 GPa is 90x10^9 Pa which is 90x10^9 N/m^2
90x10^9 N/m^2 x 1m^2/1^6mm^2 = 90x10^3 N/mm^2
90x10^3 N/mm^2 is the equivalent to weight of 9,174 kg at sea level (9x10^10 / 9.81 = 9.17x10^9 kg/m^2) which is equivalent to 9.174 tonnes per square millimeter.
As an aside, to make this more relatable: 90GPa is equivalent to ~3x10^7 feet worth of water pressure or roughly 5700 miles.
> Since heat is a major concern in CPUs, the ability to be superconducting (and thus producing no heat) would be a huge boon for computation.
But (if the superconducting transistors worked like current CMOS, which is a big if): While there would be no power draw or heat dissipation at steady-state, there would still be some when gates switched, correct? And if so, you could still wind up with pretty significant heat dissipation...
SQUID devices can be made switchable by the presence of an applied magnetic field, if I recall correctly. So you could build your superconducting computer entirely out of superconducting components.
For low values of "earth temperature". Apparently, when hydrogen sulfide gas is compressed to very high pressures and cooled to -70C, it becomes a superconductor. That's an interesting physical phenomenon, but lots of materials have unusual physical properties under extreme pressures. Chemicals can be forced into a crystal lattice even when they don't stably bond that way.
Still, it's a new data point. Maybe someone will find something that's both stable under standard conditions and superconductiong.
> That's an interesting physical phenomenon, but lots of materials have unusual physical properties under extreme pressures.
This is not interesting for its practical applications.
Rather, this is interesting for the new science and new ideas it illuminates. By understanding how this works, the ideas can maybe be replicated by other materials that don't require extreme pressure.
Thermoelectric cooling ([1]) can reach such temperatures, which will not require any gas consumables. However, it could be more energy-efficient to cool with CO2. I have not done math for that.
Most biology labs have a -70C or -80C freezer (or several) in their lab. As far as I know, they work the same way as the freezer in your kitchen. Since they're so widespread, I'm assuming this is probably the most economical way to achieve temperatures of -70C. But that might change depending on what volume needs to be cooled to that temperature.
That's a great choice and I applaud going in that direction. It makes HN headlines more trustworthy and, in a way, more interesting than vague ones are (myself I was all like "holy hell!" when I saw the -70 figure).
Note that hydrogen sulfide is in the realm of hydrogen cyanide in terms of toxicity. (Unfortunately, Wikipedia doesn't have comparable LC50 times for humans, but for H2S we're talking 600 ppm for 30 min and HCN is 357 ppm for 60 min.)
We're unlikely to see high-pressure H2S lines running through residential neighborhoods any time soon.
Probably much less than 1% - you can smell H2S at incredibly low concentrations. Single digits parts per million AFAIR.
Edit: at 0.00047 ppm you know something is there. 2 ppm is dangerous. At 10–20 ppm you would experience eye irritation. At 100–150 ppm your nose goes numb. All far below 1%.
Edit 2: the concentration of farts would obviously be higher than, uh, "outside," but as much as 1% would probably cause harm to yourself and anyone around you.
Funny, H2S is a natural product of sewage. And on vessels sewage lines are frequently pressurized. Maybe not to your house, but maybe there are some limited uses.
Being natural doesn't make H2S safe, especially at 1,000,000 ppm under high pressure.
There's nothing funny about it. In many areas sewer workers are required to check H2S levels above manhole covers before opening, because it sometimes kills sewage workers.
If the sewer smells, the H2S level is probably below 150 ppm. Starting at 100 ppm, H2S begins to paralyze the olfactory nerve and the sewer stops stinking.
Maybe I should have said "interesting" instead of "funny". Of course, the reason I know about H2S in sewage is that I was in charge of all the water and sewage piping systems on my first ship. Including safe re-entry and emergency response to leaks. So, yes, I'm quite aware of the hazard presented by H2S. I wouldn't advocate running it residentially, but perhaps in high power low voltage situations like inside a power plant, or major substations, it might make sense.
> We're unlikely to see high-pressure H2S lines running through residential neighborhoods any time soon.
That wouldn't be necessary to make the technology commercially viable, though. Forty years ago Cray sold over eighty Cray-1s, initially priced at the equivalent of about USD 35 million today, and there were probably much larger numbers of mainframes, scientific computers like the PDP-10 and so on that could go for about $1m in today's money. If anyone today could offer much faster single-core execution speed at comparable costs I'm sure they could sell similar numbers of systems, especially since the infrastructure for remote timesharing and job execution is much better and more widespread nowadays. It's the titanic pressures needed to keep the material solid at all that make the idea infeasible at present I assume (how would you even whittle a circuit out of it?)
To my untrained eye, the Meissner effect appears to be the most obvious connection between the previous high-temperature superconductors and this finding.
If intense pressure causes superconduction, then perhaps the Meissner effect is a manifestation of the same phenomenon. If the Meissner effect expels magnetic fields from the material, then perhaps this effect also causes the material itself to be compressed physically in response, or something like that.
Disclaimer: I know nothing about superconducting physics, but I did stay at a Holiday Inn Express last night.
Disclaimer: I'm a physicist but have no expertise in solid state physics. I remember when the whole high-Tc thing started. My recollection is that the discovery broke what was at the time considered to be the mainstream understanding of superconductivity. And, a more comprehensive and satisfactory theory has not yet emerged. As a result, we're in a sort of tortoise-and-hare race between theory and experiment, where each one advances a bit when the other one catches up and makes a new discovery.
Disclaimer: I'm (technically) an engineer, not a physicist, but I took a course on this back in uni, and as far as I can remember, you're correct. The gist of it is that BCS theory, which satisfactorily explains conventional superconductivity, fails to predict the behaviour of high-temperature superconductors, and there simply isn't a better model available at the moment.
For any sane individuals reading this, the answer is "We're trying really hard but so far we haven't been able to deduce a formula for the highest temperature."
I'm not very optimistic with regards to this ever happening. I remember reading a paper regarding one of the many high-temperature superconductors, BSCCO; it has a crazy crystaline structure, it's quite unlikely that we'll ever come up with an analytical model describing its superconductive behaviour.
>we're in a sort of tortoise-and-hare race between theory and experiment, where each one advances a bit when the other one catches up and makes a new discovery.
Nitpick: not much heat generated, but not no heat. There are fundamental lower bounds on the entropy increase caused by doing irreversible computations (Landauer's principle), superconductors or no superconductors. TAANSTAAFL: the universe won't let you compute for free.
As for what limits the clock speed: for one thing the speed of light - information still has to get from one part of a CPU to another.
There is no dry-ice necessary. You can cool something down to -70C and lower with a freezer as used in biology or medical labs. Such a freezer isn't even that fancy, they look and afaik work pretty much like the household appliance variant.
While -70 is still damn cold, it should be achievable a lot more cheaply than having to cool things with LN2, so hopefully-- cross our fingers-- this will lead to the breakout of superconductivity into broad industrial use (and maybe the mainstream consumer market) that people have sought for decades.
BTW- while in that lab I used a variation of the meissner effect to design a memory circuit that was theoretically buildable at the time- static RAM that was superconducting. (and basically, the essentials of a transistor were there so logic gates could be built from super conductors, that was the thought experiment I was doing.)
Since heat is a major concern in CPUs, the ability to be superconducting (and thus producing no heat) would be a huge boon for computation. Of course initially this wold be at scales much larger than current lithography and thus only for specialized applications. But who knows.
With effort the cost of keeping a CPU at -70 should come down dramatically the way battery capacity per dollar has... or dare we hope the way flash density has.