Это видео недоступно.
Сожалеем об этом.
LK-99 Wouldn't Have Changed Semiconductors Anyway
HTML-код
- Опубликовано: 17 авг 2024
- Links:
- The Asianometry Newsletter: www.asianometr...
- Patreon: / asianometry
- Threads: www.threads.ne...
- Twitter: / asianometry
Well Mr. John Asianometry, I'll have you know I was promised laser weapons, heatless batteries and a woman to finally look in my direction by people who equate anything above a bronze age technological level to magic, so I must say that yes, it would've changed a lot of things.
I mean, technically, laser weapons will probably be a thing soon, but they will be much less cool than something from a movie haha. They will mainly see use as part of anti-missle/anti-drone defense, and are more about disrupting the circuitry than actually melting them.
@@MFMegaZeroX7Laser weapons are already in use, have been for over a decade and do not in any way "disrupt circuitry". They super heat the target, causing it to explode.
@@ARCSYS4049depends there is research on microwave lasers that disrupt circuitry. However the system in play currently try to spoof sensors and misdirect missiles. In the future the superheating lasers may be used to explode incoming missiles but that requires a ton of energy.
@@ARCSYS4049 I thought the most common use of lasers was to confuse the heat sensors on missiles, causing them to miss.
@@totallynotai7131 I think that happens in some ECM systems, though in that case you'd not usually really call it a weapon.
True weapon lasers are a thing lately though, and increasingly are seen in proposed and demonstrated in military prototypes, for instance to deal with drones, missiles, etc. They aren't instant and take some seconds to overheat a target and destroy, but as such we can probably expect such point defense lasers to become more common in future.
There actually is an area where the Josephson junction revolutionized electronics - the standards grade voltage reference. It used to be that voltage references were basically just special batteries. But once the Josephson junction came around, we realized we could use it to build a very, very accurate frequency to voltage converter. And since we have really, really good frequency references, we now have really, really good voltage references. In fact, the change was so dramatic, meteorologists actually changed the definition of the volt because of it. These days, Josephson junctions themselves aren't very common because they are big, expensive, and not very portable (tanks of liquid helium will do that), but virtually all modern electrical test equipment, especially production test equipment, can literally trace its calibration back to one of them. And that's not just a one off thing. Electrical test equipment is required to undergo periodic recalibration to stay in spec. The equipment that does that has a documented, periodically updated calibration train back to a national standards lab, ultimately culminating in a Josephson junction array.
I did two years at a NIST lab, one day we had a retirement party for one of the old people, big boss director comes in, notices we have like 5 plain balloons floating next to the cake. "Oh how nice, wait, you didn't use the lab helium did you?" Silence. "Oh good, cause that would be like $5k worth of balloons right there right, haha." I feel one drop of sweat on my forehead and say "Oh no sir, haha, right, those balloons came with the cake."
meteorologists = metrologists?
@@Graham_Wideman Yes - typo/autocorrect.
@JoaquinElf Well, it's some pretty high level volt nut stuff, so you have to wade in pretty deep to find it and it can get pretty deep into precision analog electrical engineering. Here's a video talking about the Josephson Arrays themselves:
ruclips.net/video/VoRab8U2eS0/видео.html
And here's a video talking about how multimeters get calibrated from primary standards (like a Josephson array): ruclips.net/video/lnh87hXNcfo/видео.html
Once you have a meter calibrated against a primary standard (the term sometimes used is that it becomes a "transfer standard"), then it gets used to calibrate other equipment, like this video shows: ruclips.net/video/JID8bS2-skg/видео.html
If you like some of this stuff, I recommend Marco Reps channel: www.youtube.com/@reps . I also recommend the xDevs site (can't link it. Search for "xdevs"). He talks alot about how the "transfer standards" these high level multimeters work and repairing electrical test equipment.
@JoaquinElf If you're an electrician, no it won't be explained, other than maybe in passing, probably not even that. The closest you'll get is having to have your handheld multimeter sent for calibration periodically. The stuff I linked to is basically the infrastructure behind getting that meter calibrated. This stuff is a very specialized niche within electrical engineering.
"At the risk of being made an idiot later, I am saying that I don’t believe that any superconductor-based computer will become commercially competitive at any time in the future."
We'll see if your quote ages better than Lord Kelvin's quote: "I think it cannot be done. No balloon and no aeroplane will ever be practically successful."
How long did it take though after Kelvin said that :(
@@TheAwkwardGuy Apparently he said it in an interview in 1902.The first powered flight by the Wright brothers was in late 1903.
Damn nice @@denodoko
@@denodokodamn lol!
He lived to the WW1 planes?
Capacitive and Inductive “losses” aren’t losses as such. They’re just energy storage. The magnetic fields created in a superconductor don’t dissipate anything. They are just energy stored in the wire, and trying to stop the current will release it. That being said, high frequency current through a superconducting wire will still emit EM radiation to the environment. Good circuit design should mitigate this, but I don’t believe it can be eliminated altogether. To summarize: there should be losses, but only as radiation - which is a few orders of magnitude less than what the inductance and capacitance values would imply otherwise.
In this sense it's a speed bump.. For example if a signal is going from high to low, it has to fight the inductance trying to disallow the change. The resistance of the material doesn't matter there..
When a gate output goes from low to high, the gate has to charge the output capacitance. Then, when it goes back to low, it has to discharge it. The output capacitance is just energy storage, but the energy is still lost, because the capacitance is charged from the positive rail and discharged into the negative rail.
The difference between the amount of energy stored in the two states is important. The capacitance itself will not dissipate it, but the transistors that charge and discharge the output will, as will any resistances in the current path on the way, etc.
Capacitive/inductive loss is not joule effect loss (heat loss), but it is still an energy loss. The energy that is used on the transient state (aka while these components are charging) are used to load the capacitive/inductive coupling, and not to transfer energy between devices.
Only after the capacitor/inductor is charged (the steady state) we can say we don't lose any energy.
@@possible-realities but the capacity of a wire is determined by its gauge, no? If the wire has no resistance, then it needs very little diameter, which means it would have very little capacitance.
@davidgervais5974 You're right, making the wire thinner should lower the capacitance. But I believe that it would also increase the inductance, so that you would need to put more energy into getting a current going through the wire.
Still waiting for my cold fusion powered graphene magic carpet...
Sorry, but we need room temperature superconductors to make that work.
I'm just waiting for my memristor
@@laurelkaye6313
And, of course, dilithium.
I just wish to live to see the day we finally find a room temperature, ambient pressure, semi conductor.
@@ivoryas1696 Naturally. We can't forget the most vital component of the warp core!
In fact, in high speed circuitry, the shape of the data lines and the material around them is far more important than its own material.
Lol... When dealing with silicon ya
@@VincentAnzalone Capacitance is much more important when it comes to rapidly changing signals
@@personzorz tell that to the future
That's like saying ram is more important than the speed of the processor. It's connected but ...
@@VincentAnzaloneit's not
No surprise there. LK-99 study was released without consent of all of the authors, had no peer review, didn't measure resistance and hasn't been replicated. It was always going to be a long shot.
Microscopic uses will be hard with any superconductor. So think of big applications instead?
I can't speak to feasibility, but I am interested to see if rail guns and mass drivers might find some benefit from high temperature superconductors.
Cheers!
Railguns won't find a use any time soon, because that's not the problem. Coil guns _might_ find a use, but also might not.
The biggest use there is (power grids) are quoting 5% transmission and distribution loss. How much of it can we replace with superconductors?
yea, the story on how that paper got released in the wilds is... wild.
A guy (Kwon) who got fired 4 months ago happened to catch a wind of the team planning to do a peer review for LK-99. So Kwon rushed to make a paper with basically everything he remembered, wrote it extremely sloppily, and just push it out as-is with his name on it. Of course, many replication efforts following that sloppy paper got nowhere, and a lot of people calling it BS right out of the gate while calling the authors idiots.
But yeah, the original authors are now on it, and maybe we'll see something in the next few weeks, maybe a revised paper, a proper sample or something, or maybe we won't, idk.
what about particle accelerators and fusion reactors? those two things would benefit from this i believe
@@xureality at our current level of mastery of superconductors, none of it.
It's a dream.
It may well be a pipe dream. At least, that's what my intuition says. Not that I trust my intuition very far.
I think our chances of advancing solar and battery tech will offer a better payback than sinking R&D money into superconductor long-haul transmission lines.
The problem with solar today is, it's not very good across vast swaths of land with largely unfavorable sunlight potential, and a lack of economical power storage solutions. More efficient and cheaper solar panels and more efficient and cheaper batteries could nibble away at those disadvantages. Or so I hope.
If that works out, more power generation will be local. Less need for lossy long-haul transmission lines.
Batteries are improving at maybe 5% capacity every year, not counting generational jumps which come along maybe once every three decades or so.
Solar is harder. More efficient panels can be demonstrated, but the path to commercializing them and beating current-generation costs is difficult. Very difficult.
Tech is not standing still. I am cautiously optimistic than we will muddle through without betting on superconductor power transmission lines. Or fusion power. Or any other pie-in-the-sky advances. Though of course if any hair-on-fire commercially-viable tech appears, we will gladly take advantage.
For quantum photonics, it would have made an impact. Superconducting nanowire single photon detectors are bulky because of the cooling and everyone would love to integrate them on-chip.
Lk99 claims to work by using tunneling supercurrents between quantum wells not cooper pairs so the working temperature calculations might be different
the working temp calcs are wrong to begin with looking at existing high temp super conductors.
Apparently tunneling super currents are unrelated to temperature
@@vyor8837 : The initially claimed operating temperature was high enough that we should assume the existing formulas are wrong for it anyways (which, yes, there is precedence for: already known "high temperature" superconductors don't follow exactly the same formulas as the originally discovered batch of superconductors). Whether LK-99 is a room-temperature superconductor I don't know, but we should expect that _if_ it is, then it _won't_ be of any known type.
I know a cooper.
Great guy.
Very strong thighs.
People just need to wait until those Korean show up and proof or disproof whole thing. I don't understand all of this replication frenecy. Every scummy nation is going to claim that they have it. We just need to wait the oficial announcement
Thank you for doing all of the math on the temperatures I was lost until you got to Wedgewood😁. As always your videos and research are great stuff.
Wedgewood temperatures are also close to my scale
I had not heard of that scale, but it's actually quite interesting. It was made by *Josiah Wedgwood* who was a prominent abolitionist fighting slavery and also the grandfather of Charles Darwin.
12:59 😂. Yeah, there are those many units of temperatures. For those curious, Rankine is the equivalent of Kelvin in imperial units.
Imperial units need to go extinct.
Even way back with vacuum tube circuits we had to deal with Miller Capacitance that could limit both frequency and gain factor. That's what Pentode tubes were invented to improve upon, the extra elements at different voltage potentials lower the Miller Effect and allow better control of the electrons flowing inside the tube. The new Pentode tubes worked at a lot higher frequencies compared to the original triodes, and they could be built to control larger currents.
Very cool! Especially enjoyed the bit of history on IBMs research on Superconducting computers based on Josephson junctions. Interestingly the same tech is now the basis of qubits in IBMs Quantum computers. Wonder what they are thinking about LK99...
they're probably thinking like the rest of us are thinking, that LK-99 will turn out to be not a superconductor
Surprised that you missed the most important use of the Josephson Junction, and that is as a voltage standard. Gone are day the days of Watson cells. In fact, that could make for an interesting video (if you haven't looked into it already), the history of standards of measurement and the quest to move them from being based on artifacts to mathematics.
12:59 I'm disappointed with the glaring omission of the Réaumur and Rømer scales. Personally, I only use temperature scales containing only the consonants r-m-r in my daily life.
Back when I was working on Thin Film HTSC there was talk of using them for interconnects. Our current densities for TBCCO were in the 10 E6 A/cm^ range if I remember correctly. We were at 3" wafers when I left. We never made one, but we did mount a SUN microsystems processor on a closed cycle Stirling Gooler. This sped up the chip significantly, but the whold thing wasn't really practical. We also made an HTSC SQUID with HTSC Josephsen Junctions.
I genuinely can't tell if anything you just said actually means anything or if you're trolling. I'm getting fairly strong turboencabulator vibes.
@@jlco : There were typos, but it's legit. It all boils down to "none of it ever reached production."
My first thought when reading about LK99 was that if it worked, it could be used to create tiny MRI machines and super powerful and small electromagnets for all kinds of research and industry use. The reason MRIs are huge is because of the cooling requirements to keep the superconducting magnets at operating temperature. The reason most powerful electromagnets are huge is because of heat generation.
Sad it hasn't panned out. But we may still see a room temperature superconductor material as computing power increases and AI can be used to explore options faster.
another is that you need a person to fit inside it 😉
Perhaps but current MRI scanner magnet design uses niobium wire with a dual helium / nitrogen cooling system that (at least in fixed facilities) has a liquefier unit on site. Ceramic wire is apparently too flimsy & prone to failure, which is dangerously catastrophic in magnets that size (that is, anything suddenly stopping the current including drifting above Tc will cause the back-EMF to physically rip the magnet apart & damage anything nearby, much less inside; there are videos of this)
for starters there have been some superconductive effects reported by some labs. For instance some labs got flakes that were able to stay in suspension over a magnet. Needless to say, you miiiiight be able to explain why a lab would fake a breakthrough to get headlines, but there is much less of a reason why other labs would repeat the experiments and get somewhat similar results but not quite exactly the same. So, point is, the jury is still out eitherway, chances are there is *_something_* going on still, we just don't know what.
As for your point about supermagnets, that's also a large issue with fusion power. And, unfortunately, the public is bloody brain dead and needs some miracle breaktrhough like that before they'll actually accept nuclear power as an option. (despite the fact that nuclear fission is already the cheapest source of power when factoring in plant lifetime extensions, kills less people than solar per MW produced, pollutes 4x less than solar per MW produced, doesn't require grid scale lithium ion batteries, etc.) I mean fission as is might as well be a miracle cure, but evidently people won't take anything until they get a miracle vaccine that prevents all illnesses and gives you super powers and a room temp super conductor could do basically that. (speaking metaphorically sitll of course. My point is that if you could get a super magnet then containing plasma for a fusion reaction would be far more doable and if we managed to crack fusion power it would basically be completely free energy that not even the most brain damaged of laymen could spin as dangerous and evil.)
Plus I vaguely recall someone mentioning that a room temp super conductor could be used to make super capacitors that could store enough power to basically act like batteries while charging near instantly, discharging however fast they need to, etc.
@@ms_nopI can't seem to find these videos...
I doubt that a broken wire will make the whole thing self destruct in the described way.
@@robonator2945The gods of the universe gave us 100 trillion tons of uranium, and we spit in their faces.
Informative thank you ! The current density limit of cuprate discuted at 6:00 has to be the reason why new MIT SPARC Tokamak Toroidal magnets have such a huge thickness. Add this to what we learned from your first video (cuprates are ceramics less convenient than metallic Nb-Ti superconductor) and one can realize that current HTS are not the industrial and economical revolution some are pretending.
Awesome video! Honestly surprised to see the subject of superconducting ICs on RUclips, but if it's anywhere it's gotta be here.
As someone who works in the field, I think you did a good job highlighting the issues with superconducting circuits. Fabrication, memory density, and scalability are huge obstacles to the technology. It definitely isn't the most accessible technology or field, but it's really interesting stuff, with huge potential if it can overcome its problems.
"I was heartbroken when I learned that it wasn't so." Feel ya, bro. Seems like all of these issues would've gotten out sooner, so the volume of hype around every superconductor discovery wouldn't have reached such a pitch as it has. Maybe those Nobel Prizes could be better distributed now.
dang I've never found an asianometry video this fast- I'm sure it'll be good :D
I always wondered why we do not use superconducting ICs for certain applications and now I know why. Turns out they would be too big for what we are used to by today's standards.
It would be great to have another Ice Age to cool down our hot transistors
Good video, but as a physicist I need to point out one thing: Josephson junctions ARE NOT the ONLY thing in which you see quantum mechanics in action.
The description of what we call the "electronic band structure" of crystalline solids, like SEMICONDUCTORS, can only be described and understood because of quantum mechanics. It is because of the development of quantum mechanics and its application to solid state physics, that transistors came to be. As a matter of fact, John Bardeen is the only person with two Nobel prizes in physics because of his participation in the invention of the transistor (along with Brattain, and Shockley), and his co-creation of the theory of (ordinary) superconductivity (along with Cooper, and Schriefer). (The high temperature ceramic superconductors haven't been fully understood, as far as I know, that's why I mention "ordinary" superconductors, which are all low temperature.)
Also, the energy consumed through transferring data between the processor and memory can also be largely addressed through other things, like in-memory computing. It feels like if superconductors are ever used in semiconductors at all, the benefit will be very marginal.
I clicked thinking it was going to parrot Thunderf00t but I was pleasantly surprised.👍
Actually, we can't conclude anything about LK-99... the leaked papers appear to have been incomplete, so no one trying to replicate from them could likely succeed because of it, and iron contaminants and other ferromagnetic effects in the replicated (leaked) samples were responsible for levitation effects in tests. The supposed paper "debunking" LK99 was also based on the replicated samples, and while it may have made valid statements about those samples, the only people who can make any statements about the actual LK99 material are the folks at the University of Science and Technology in South Korea. They have the original samples from the engineering team that made them and the presently final version of the paper. The only report they have released is that they have verified the structure of the samples matches that in the (final) paper. That's good, because it means the paper has at least some validity. The simulation performed by Dr. Griffin at LBL also matches with that, I believe. So, the whole world waits....
Thanks for the interesting discussion! Generation of magnetic and electric fields doesn’t mean energy loss. The energy is stored in electromagnetic fields and can be recovered. RF circuit design takes these into account, and I think they can be managed:) I wouldn’t rule out the usefulness just yet.
You still get the propagation delay, reflection loss in the switching device resistance and the hysteresis loss in the dielectric still acts as resistance at AC.
The term 'Loss' means potential energy that is put in is not used for the process or transferred. They are inherent, whether stored, radiated by fields of energy or heat. Recovering or limiting such losses in itself makes for more complex resolutions?
you still get delay, and you don't get 100% recovery, there is still losses there.
As another commentor said, they claim to use quantum wells, not cooper pairs. If it is only a matter of development time, they will come up with working specimen in some time. So the title should be "LK-99 wouldn't have changed traditional superconductors anyway". I'm going the optimistic line here, believing their model works as proposed and they just didn't wait long enough before having sufficient specimen, model and data.
I had a busy week and had seen the headlines but had no time to follow up. The delay was well worth it as it have time for this video to be produced
I don’t think you’re seeing other applications that would happen immediately. Power generation, power storage and power transmission. Power use would be a distant concern since power would basically be limitless and nearly free.
Power generation and transmission?
I dunno about you chief, but I don't think you could easily make wires with it.
And how could you use it for storage?
Free such as beer and women? 😅
LK-99 is not only produced in the form of ceramic chunks. Kim HT has developed a form of doped thin PVD film of LK-99. They mentioned such a film in their Korean paper in April, and included the film technique in their patent application in Korea.
It’s almost like some people don’t want it to be true. I don’t understand why there’s so much negativity.
I’m guessing some people don’t want to be let down again, so they demean it until proven otherwise.
@@jalanmcraeand hell yeah... Weren't they right?
You know, one of my professors, the late Viktor Baranauskas, was trying to make a film of diamond to be used as insulator in IC packages between the die and the metal. Diamond is a gread insulator and also great conductor - of heat. Any thoughts about the field? I mean, not exactly about the diamond film but about the technology to remove heat from the die. Thanks!
Firstly, using superconducting material could allow us to choose different materials in between interconnect, which would decrease leakage and improve chips (there's also the possibility we'd need to choose a worse material).
Secondly RC delay is based on time constant Tau, which is resistance times capacitance, so RC delay would decrease a lot with near-0 resistance aided by superconductance.
Thirdly (and this is the longest one), the josephson junction problem was likely at least in part due to the issue of cooling. The superconductors IBM used needed liquid helium to cool to near-absolute zero (4 Kelvin). IBM cited Josephson junctions consuming 1/1000th the power of a traditional semiconductor, and I'd imagine if the claims of the superconductor were true, it would be significantly easier to cool the area around superconductor which hypothetically superconducted at room temperature to 20 or 30 degrees celsius below ambient while it sips 1-3 powers of ten less power of the traditional circuit power is going to be dozens of times easier than approaching 4K in cost, heat transfer performance and energy input. It's quite possible IBM's circuit size was to increase surface area for liquid helium contact to stay in superconducting range while operating. The biggest issue with superconductor logic right now is quite literally that we need to design them around cooling them to absolute zero. Getting a ceramic that's difficult to create but doesn't break at 100K is probably going to vastly change the designs around superconductor logic in ways no one can predict.
I don't disagree with all of your statements, but some of your points felt blatantly ignorant on the decades of material design that goes into semiconductors. In the short term, your points are 100% correct about the introduction of a superconductor, but in the long term it's doubtful any of the material choices for semiconductor design decisions we made for silicon will remain in Superconducting logic if we get a superconductor, meaning except for josephson junctions your leakage current and RC arguments are mearly moot if a material change happens with superconducting logic design.
Thanks Jon, for another very informative video. Your content just keeps getting better.
It was over in 1988 when room temperature super conductors were breaking records weekly...all Busted.😂
"At any time in the future"
So, not after a billion years of research? After a googol attempts to make one? You're in no position to write off the future. We simply don't know, that's why they science the hell out of it.
This is exactly what I tell people. I studied Computer Architecture as a CS student, and it's common knowledge that one of current challenges is heat/power, which is influenced by capacitance of circuits
"They" have been promising the ULTIMATE 3 TECHNOLOGIES since I was a kid in the 70s:
1) Jetsons hover-cars
2) Lightsabers, &
3) Tri-D TV...
We're still waiting, guys...c'mon! If I'm not watching ppl accidentally cutting their own heads off while hovering metres above the ground, all in glorious (spex-free) 3D, by the time I die I'm going to be very tetchy...
Asianometry is mistaken. The heat dissipated by a chip is proportional to resistance (not inductance or capacitance). So, no resistance (aka superconductivity) means no heat. Since by far most of the energy consumed by chips are released as heat, room-temperature superconducting computing will dramatically slash energy costs. All data centers will want to go superconducting!
Yes, I did laugh out loud when you gave the Rankine and Leiden temps. Nice work on this.
That’s so frustrating, imagine living through the 1940’s till the 1980’s… you would see a total different world… I barely see any change from 2010 till now, am I wrong?
If I give my opinion on this, the vast majority of "drastical" chances nos come in the realm of the digital, because the Internet os much more accesible than ever before.
If you were to consider the rise of ChatGPT as a paradigm shift, is still strictly prominent in the tech/academic world than anything.
The power shift from west to a multilateral world and the end of US hegemony. The anthropocen with all the climate changes. OLED screens and 4k120Hz. All theses are the event of our time, they are interesting I think. Less than enjoy life and finding meaning in simple things, but interesting nonetheless
Processing speeds have increased tenfold on high-end desktop chips, more so on the server side. So there's plenty of progress in that space. Tech in general has gone way more mobile - with mobile phones now essentially being everything many people need.
First of all the 1940s to 1980s span over 40 years. From 2010 to now is only 13 years. That's a big difference. We also haven't had a world war since 1940s. A lot of progress was made in that era because the war pushed the limits of technology. A lot of the electronics like the transistor started in that era.
Smartphones in 2008 vs smartphones in 2023 is magnitudes different. Cameras, bandwidth, video calls everywhere, processing speed, battery life, charging speed, wireless charging, peripheral devices (apple watch, vr headset, etc)
There's a lot that's improved but it's been so smooth and steady that it's hard to notice
I am starting to fall in love with your Content .
Thanks Brother. 🙏
In a way the only benefits that the industry if IC can obtain from a superconductor application is the lessen use of decoupling capacitors due to less loss of power and stronger signal integrity while using less power.
People invest too much into things that they cannot even participate in. Like even if LK99 was the perfect superconductor, what would most people do with it? They would wait years to buy products. So better just ignore it until a product comes out.
-73degC is a very easy temperature to reach for a industrial application of almost any scale. Forget nano-scale for now, you'd revolutionize a LOT stuff at or above the milimeter scale. The key question is what happens when it fully leaves the superconducting state at higher temps. Either way, it would be advantageous in so many applications.
The most successful applications of superconductors is for High Q RF filters, since loss reduces Q.
Similarly, graphenes were said to have non-resistive current conduction. Could you please make a video about it?
well, If superconductor cables are invented, I doubt that inductance and capacitance would be a real problem at this point, because we could lower the signal voltage even further, allowing us to transmit ultra-low voltage signals. This requires further research to find the optimal application. The bottom line is that we cannot predict our life beyond the discovery of superconductors.
Scammers should be composted!
the oven temperature superconductor, we can still dream
I need a superconductor induction stovetop to heat up my ramen.
Hey buddy, i love your vids. They have taught me so much. Surely if you have a superconductor that has zero resist and low heat loss it would make a difference! Even if it is hard to manufacture?
Keep up the fantastic work buddy, your vids are an inspiration.
The signal interconnect delays relate to the LC factor rather than the RC as you stated. The Velocity Factor is defined as the reciprocal of the square root of the K factor. The Velocity Factor results in a per unit ratio of the speed of light.
The K Factor is defined as the product of the relative permeability multiplied by the relative permeativity of the dielectric medium - in other words the K Factor is proportional to the inductance L and the capacitance C
Such a pleasant professional and enjoyable narrator voice. Not at all HARSH, clipped, hard on the ear or unprofessional, not at all unpleasant to listen to .
Why do I get the feeling you're trying to scare me off of joining the local pottery club and making my own lk99 killer app 😂
Super conductors for SoC is maginal benefit because the interconnects are not that long and the CMOS gate charge and discharge interconnect that are more capacitive in nature than resistive. This is parasitic capacitance from the metal interconnect to the substrate which is grounded and attached to another CMOS gate(s) which is a capacitor given its gate is one half of a capacitor, the other half being the substrate or well. Thus the dominant power dissipation comes from the clock lines which switch at high frequencies driving a lot of gates ( assuming a synchronous register design).
The only resistive long lines are power and ground and they do not dissipate a lot of power at steady state. The power lines need to provide peak power so the DC lines do not drop their rated voltage and slow down the switching of the logic processing info.
Current and long established solutions is to have large peaking capacitors under the power and ground lines. As per above this is just CMOS transistors with their drains and sources shorted together and tied to ground/power. The power and ground lines would connect to the gates of these large CMOS transistors.
Ps
I used to be a full custom memory and microcontroller design engineer. I worked in the EU and UK for many design houses for @10yrs.
The real application is in power and distribution lines and components such as electric motors of all types and spec. You will have scientific and space applications as well. Forget semiconductors. You will have Quantum computing applications as well, maybe.
Don't current models of quantum computers rely on supercundoctors? Even a room temperature supercunductor rather than an oven temperature supercunductor would be pretty useful for reducing the cooling needs, despite not eliminating them.
Same goes for MRI machines: most use a neodymium-titanium alloy supercunductor cooled with helium. There's a composite ceramic superconductor that could already replace that if it wasn't so difficult to fabricate as a long wire that would only require nitrogen for cooling, cutting operating costs by a factor in the range of 20x.
LK-99 is also a ceramic, which makes it even harder to manufacture, since you need 1 large piece since you cant join ends together. Cracks are also fatal, so any strain or expansion due to heat of it or surrounding material are a challenge.
There are other ceramic super conductors which have been discovered before which operate on liquid CO2 levels (200K), versus the liquid nitrogen (77K) or helium (4K). But if you cant spin a wire of it... is it really game changing?
Your last argument is based on BCS theory which describes only a subset of super conductor materials. So its a bit too early to be generalization that to a potentially new type of super conductor.
Thanks for the video. It's refreshing to see a realist's point of view on LK-99. This is quite different to other videos I've seen on this topic. Although I can understand the general message, several statements can be rephrased to improve the clarity for the viewers. I'm not an expert myself, so I may have some incorrect information as well, but here are my suggestions:
1. at 2:51, it's a little confusing what Mr. Sludds said about transmitting a zero or a one. Since superconductors dissipate zero energy, using superconductors for this purpose doesn't immediately paint a clear picture for me in how this would be done. Maybe I'm one of the only few who didn't immediately get it, but in case there are others like me, it can be elaborated. It was confusing to me, that the logic level used in the circuit will have a finite voltage level (say, 5V), and the leads used are made of superconductors, so this energy would break the superconductivity and defeat the purpose of using superconductors for this application.... (The I-V of a superconductor has a flat voltage up to the critical current, beyond which the material stops behaving like a superconductor). Also, using superconductors for ultrashort leads could be challenging. I'm not sure what the coherence length of a room temperature superconductor would be - probably ~1 nm, so ~5 nm leads may still work - but making "short lengths of wire" for transistors will not be the best application for the superconductor. Proximity effect can also complicate the behavior of the transistors. It may be better to use the superconductor to connect different chips together, or to supply current; as you said, in large interconnects.
2. 4:00 If a superconductor is used, there will still be a net gain in the efficiency. You swap out some of the R=finite Ohms with R=0 Ohms in an RLC circuit, the net impedance will be lower.
3. at 4:50 about the RC time, if you replace all the loads with superconductors, you would have RC=Tau=0. But I think, in reality, there will always be some R since, you'd want to keep the transistors to be made of silicon, but again, there would be a net gain due to the decrease in R.
4. at 6:15 regarding the current densities, superconductors can take quite a bit of a beating. if you take YBCO for example, the maximum supercurrent density is ~1e10A/m^2. This is more than enough for an interconnect. The cuprates never made it to the "IC arena" because there is a too small of a market for ICs targeted below 90K. No foundry is going to invest in multimillion dollar equipment for a small market. There are some talks about using cuprates for MRIs, and transmission lines, but I'm not sure if this was just an academic thought experiment...
5. Regarding the integration of LK-99 into CMOS foundry at 7:00, no foundry in the world is going to implement LK-99 unless it is indeed a superconductor. Materials like Pb and Hg are strictly banned, because of hazardous issues. LK-99 won't pass the RoHS criterium by the EU, which would make the companies lose one of its biggest markets. Not only that, Pb is a nasty material to include in your ultra-clean CMOS process. If it IS a superconductor, that's a different story. The EU is pushing strongly for quantum computers, and high temperature superconductors are needed in various applications (i.e. solenoids used to generate super high magnetic fields). If it does attract foundries, and EU eases the restrictions on Pb, then it's a matter of time, before material scientists and engineers to optimize the large scale synthesis of these sort of materials. Typically, for unconventional alloys which need high crystallinity, high temperature sputtering can be used to deposit high quality materials coming from a high quality target. Of course this would be done before the CMOS process. Or maybe plasma enhanced growth processes can reduce the growth rate, but this will probably not result in single crystals. However, single crystal superconductors may not be strictly necessary. It depends on the application. The superconducting wires used for MRI are not single crystals. Superconducting Josephson arrays demonstrated by NEC were not single crystals.
"we've been playing around with copper for 10000 years", yes. but only a fraction of that time was spent analysing and optimizing the growth of copper. Just because something is difficult now, doesn't mean that it shouldn't be explored. It's about the return on investment. If the return is the growth of a real, irrefutably proven room temperature superconductor, companies will invest. Also, we learned a lot about material growth and different synthesis over the past couple hundred years. We can apply what we learned towards a new material not only to learn about the material but also to optimize its properties.
6. 8:17 Josephson junctions can be constructed with other weak links, not just insulators. Josephson junction come in different varieties: Superconductor - Insulator - Superconductor (S-I-S), Superconductor - Normal metal - Superconductor (S-n-S), Superconductor - constriction - Superconductor (S-c-S), etc. Supercurrents are in superconductors even without the josephson junction. It's the superconducting current. What you are referring to, is probably the Josephson current. Josephson effect is one of MANY examples of quantum mechanics that we have observed.
7. 10:27 The issue on scalability is true but at the same time, no company really tried that hard. 1. Since there was a small market (if it existed at all) for low temperature junction arrays, companies had very little incentive to explore superconducting devices in large scale. 2. Even if you cool down the devices, the limitation is in the coherence length. This is the lengthscale that determines the smallest feature on your superconducting circuit. Since room temp superconductor would have an ultra small coherence length (if it abides by the same rules as other superconductors) the circuits made of the room temp superconductor could be miniaturized, fitting much more per area than what was done with Nb or Al. 3. The tools for fabrication are better nowadays. in 1990, node sizes were ~600 nm. Nowadays, we are pushing below 5 nm.
8. 12:24. Rule of thumb, is rule of thumb. This is a practical, but arbitrary blanket criterium which is very useful for 'regular' superconductors (especially type I superconductors which are more susceptible to disturbances like magnetic fields) used in applications at cryogenic temperatures but does not necessarily hold for high temperature superconductors. It all depends on what the application is. The critical current scales as (Tc-T)^0.5, so at around 300 K, the critical current density will already be 50% of what the maximum would be, assuming that the Tc is 400K.
9. 13:43 It would be helpful for viewers to understand why room temperature supercomputers are not likely even if we discover a room temperature superconductor. I will briefly mention here. This is unfortunately true, because the leading method of producing qubits is by using cQED devices e.g. transmon qubits. The typical resonances are at several GHz, and a lot of engineering go into the cryostats to filter out other noises above and below the few GHz range. However, a qubit at 300K, meaning all the circuits, electronics and filters will be at 300K will have blackbody radiation of 6THz. Meaning that the qubit will likely be destroyed by the thermal noise. BUT, this is only considering the cQED qubits. If you go the route of NV centers, then you don't need room temp superconductors but the qubit operations can still be performed at room temp. These are difficult to scale up though. Perhaps there is a way to make use of superconductors in supercomputers even at room temperature. But I think room temp superconductors will find other valuable uses in our civilization, e.g. affordable MRI, Maglev, Energy transportation without dissipation, etc.
10. It seems that this video is focused purely on state of the art computational processing units with high density transistors. But foundries also make other chips for other applications too.
Thanks for the video and good luck!
Cheers,
Lmaoooo you brought up the cursed unit of Rankine, i remember learning about that
Finally! I thought this video would never come.
I can hear your suppressed rage.
12:43
I got enlightened by all possible temperature scales 🤯
Maybe traditional supercomputers wouldn’t benefit from LK-99 but I could see it pushing the quantum computing industry to new heights.
When you think about it, ''room temperature'', depends on the ''room'' you're talking about. Perhaps this is something for out of earth manufacture and use.
Thank you for the nice video. I'd like to point out the currently leading quantum computers are all based on superconducting Josephson qubits (usually aluminum), though other implementations exist. So yes, there are real applications for high-temperature superconductors in computation aside from interconnects. Outside of computing, the first real-world application for high-T superconductors seem to be high-field superconducting magnets, like the operational 32 Tesla magnet (SCM-4) at MagLab in Tallahassee, and the ones developed for fusion tokamaks by Commonwealth Fusion.
they would be very useful in particle accelerators also
Cooling and throttling is a huge problem with modern processors: if the lack of resistance keeps the cpu/gpu from heating, the clock speed could be increased dramatically.
imagine having a processor that can just go super critical if it gets too hot lmao, gets too hot in one spot, resistance goes up then everything gets hot.
Don't semiconductors work by dynamically adjusting their resistance?
did you even watch the video lol
@@buzhichun Who knows?
One of the BEST overviews on the subject I have ever seen. Very insightful and really well done! LK-99 may be down, but it's FAR from out. Consider that the best minds in the WORLD will now be standing on the shoulders of the South Korean researchers that first broke this study to the world. What they didn't complete, perhaps OTHER researchers someday soon will! This is ALL just getting started - and it is FAR from being over.
High speed data signals are transmitted over transmission lines. Transmission line losses have multiple sources - resistance, skin effect, dielectric losses and radiative losses. Superconductors would eliminate resistive losses. Skin effect could possibly cause local current densities above the critical currents. It would not do anything to dielectric losses and radiative loss.
Technically speaking the energy that goes into creating electric and magnetic fields isn't lost, it's simply stored in a different form, the field itself. The energy is then returned to the source when the field is discharged.
You mention in your video that superconducting process unit density is an issue. It's worth noting that with CPUs at least (notably not graphics cards) the vast majority of transistors contribute only logarithmic value for execution speed at best. My current understand is that a ~50,000 transistors CPU designed in a modern TSMC fab would be about 50% as fast as the best CPUs, despite not having the billions of transistors that modern CPUs have. This is because the vast majority of transistors in a single-threaded application are used for branch prediction and other things which are very ineffective per unit transistor in improving performance. If you can push your processor to 100 ghz, but only fit 50,000 transistors, you could conceivably still improve single threaded CPU spead by ~12x, which would be an industry breakthrough. I think a proper, easy to manufacture, high purity room temp superconductor would definitely revolutionize at least single-threaded CPUs, if nothing else.
You're only looking at part of the problem. The biggest performance hit you see in the average program are cache misses, often because of missed branch prediction. Your cpu tries to load a bunch of instruction and data in advance to fill the pipeline, but if there's a branch it didn't predict correctly, it's going to have to load a bunch of stuff from RAM (most likely) and it will be stalled for quite a bit, for a duration that doesn't depend on its frequency since that's an interconnect limitation.
If you're making a "dumb" dsp or a gpu with no branching that works (and it's already what is being done more or less).
What stopped Intel from going further with netburst (P4) was that they could have raised the frequency (to a point, TDP was exploding already), but because of the latency when fetching data from RAM, it would require an even bigger pipeline and would just lead to no improved performance is the branch predictor wasn't working well. Which leads to putting two pipelines on one core so unless both stall you're still doing work.
Unless you program in a very specific way to avoid predicted branches as much as you can, just ramping up the frequency won't increase performance that much. It will in the tasks that are already better left to a gpu, but not for the average cpu task that is highly branching.
Ambient superconductors could be used to build a free-electron EUV/X-ray laser. Such a laser would be more efficient and durable than the current Sn plasma EUV sources. With a brighter and more monochromatic light source, more complicated optics can be used to improve image quality or enlarge the reticle, while reducing exposure time and increasing throughput
2:25 "80-90% of the power budget spent moving" - do you have a link for more information on chip power budgets?
What about semisuperconductors?
Unfortunately this video was my first entry to this channel. You come off like a self proclaimed genius, with interviews that are 'trust me bro' esq. This video reminds me of a college assignment
Solid B
What if the "wires" are superconducting at room temperature? Your answering question is: what about mass production?
Said differently, 0 resistance does not mean 0 impedance. Physically, the charge storage is still slowing things down. Photonics, optical interconnects, and optical computing are probably a better future platform.
"Current density" - Now I feel like Ohms law is a lie... I was imagining superconductors as a hair thin wires to run 230 volts with millions of amps across the country without need for high voltage lines and transformers.
It's more like a guideline. For most electrical work, it's good enough. When you get into microelectronics, the details matter.
7:25 fun fact, the "Su" in the references is Dr. Lisa Su, CEO of AMD.
The speed of electromagnetic waves of a transmission line is determined by its impedance given by its inductance and capacitance per unit of length. And this makes the transmission line dispersive so the speed depends of frequency distorting the signal and digital signals have a lot of high frequency components.
7:41 I'm still waiting for silicon photonics to really take off too...
At any time in the future is a bold statement - there are infinite combinations of material, material structures and manufacturing processes. At some point AI will make material simulation experiments happen at 100s then thousands then billions and so forth possible - learning each step of the way. It's unconceivable what kinds of materials humanity will discover if it doesn't first destroy itself. This is not possible at the moment as we don't have data for things we have not discovered, but I would be very suprised if the fundamental variables won't be discovered by the end of the next century. If it is at all possible to build up knowledge in the physics system, it will be done and won't take milleniums.
i just assumed the craze was about electromagnetism and maybe power transfer. Fusion, long range power lines etc. But i appreciate the video on semiconductors too.
5:53 What in the heck is the board in the picture?? Some kind of weird network thing, but what???
Um... inside chips is one of the least critical applications for superconductors. Power generation, solid state motor design, energy grid power distribution, space travel, medical equipment, electric cars, long distance telemetry... we really don't need superconductors in computers. We need them in everything else.
When you understand, it was FAKE!!!!!
The patent says it may have semiconductor properties, which means it could be a super semiconductor or semi superconductor.
Even if LK-99 didn't work out as a high-temperature superconductor, I wonder if some of the more traditional known low-temperature superconductors would work with quantum computers? Both quantum computers and superconductors work around the same temperature ranges, near Absolute Zero.
You left out that LK-99 is a ceramic, not a metallic alloy. It can't be etched in an IC in the same way either.
It's sad that LK-99 isn't a superconducting.
Bubble-mode popped, again.
When completely exposed, that person who has a new statistical theory of Superconduction will be interesting to follow-up.
Small gain from switching from metal to superconductor. Issue with keeping temperature low, low current capacity, and brittle ceramics.
IMO the best chance for superconducting in chips is if there is a surface/interface superconductor- a layer of material 1 touching material 2 (eg, many of the ideas surrounding graphene) - however we already do use complex alloys - eg: doped Gallium Arsenide, or taking advantage of 'strained silicon' forcing unusual atomic spacing ..
In recent decades the mentality of politicians has entered the Scientist. Not that the discovery and invention were gone bad but making a big issue in media has taken over publishing in scientific journals. Wonder why Korean are leading?
TBH whenever people ask me about this I'm like it fixes the helium shortage and that's maybe about it. People were talking about using these things for long-distance electrical transmission and it's just like, sure, let me know when you can get ceramic materials into a spool of wire without it shattering...this stuff isn't flexible like the glass in fiber optics, folks.
How many Chicken Wings does it have to cost to get up and go to work. Lunch and Dinner!
Thanks for so much information this is superb. Loving the appropriate wavy "quantum" background!! 🙂
10/10 stars.
Still waiting for my cold fusion powered graphene magic carpet.... Still waiting for my cold fusion powered graphene magic carpet....
A substantial portion of the computer's cycle time is spent for the hard drive to spin.
In all seriousness superconductors revolutionize transportation, not ICs. Any time you need magnetic fields and levitating stuff like trains and MRIs
Silver is kind of cheap in small quantities. Why isn't it used more in semiconductor products for iterconnects and such?