Atoms and Sporks - 2019-01-23
In this video we'll learn about how quantum mechanics and Stefan Boltzmann are conspiring to kill viral cat videos!... or something like that. Technical Caveat #1: I'm intentionally ignoring an entire side of the story. In reality, there are also electrons in the drain, and because it is a fundamental aspect of quantum mechanics that two electrons can't share the same state, these electrons “block” electrons from crossing the channel if no voltage is applied. In general, the role of this “blocking” is very important to understand to get a complete picture of what's happening here, but I chose to ignore it to make the specific points I was trying to emphasize conceptually clearer. Technical Caveat #2: In general, these dates are quite soft. I'm not trying to be a historian, and am being pretty generous in some rounding and such. Technical Caveat #3: In this whole section of the video, I'm trying to emphasize the direct parallel between classical frustrated total internal reflection (and evanescent fields) and quantum tunneling. However, there are things – like polarization, most notably – that are different between these two, and that difference is playing a role in the images I'm showing. So I'm aware that it may not be entirely fair to say the “bleeding” light shown in the photographs is tunnelled light. Technical Caveat #4: Not all of the years I mention here were full meetings, some were just “updates”. Furthermore, I intentionally neglected to mention that there IS a new rebooted roadmap that is trying to recover from the ashes of the ITRS, so industry collaboration in this regard isn't fully 100% dead as I imply. But my point still stands. Technical Caveat #5: People who know quite a bit about emerging solid-state technology might have heard that a 1:1 relationship between gate voltage and barrier lowering actually ISN'T the theoretical limit. There's a very good chance I might talk about so-called “negative capacitance” at some later date. Technical Caveat #6: Throughout I only talk about the Boltzmann distribution, even though it should more accurately be the Fermi-Dirac distribution. I do this: because the two are equivalent in semiconductors in technological scenarios; because industry exclusively thinks in terms of Boltzmann; and because, well, that's where the NAME (Boltzmann's Tyranny) comes from! Also, the Density of States I draw, which I call “perfect”, that actually isn't what would be ideal for switching. But, I think it reflects the idea without dragging things down into a discussion of what would actually be an ideal switching DoS. Technical Caveat #7: For those with a deeper knowledge of transistor physics, you might think I'm being misleading here as I'm blurring supply voltage vs. overdrive voltage, as well as supply voltage vs. gate voltage and am also avoiding a discussion of threshold voltage in general. I think the specific wording I've chosen conveys the correct ideas, without having to bog things down in delineating the specific differences and how each effect affects each differently. But others are free to disagree.
Excellent Job. You explained a very complex issue, very simply.
+1
HOW DOES THIS VIDEO HAVE ONLY 3000 VIEWS???
Wow, how is it possible that you have so little views and suscribers! if you want we cant talk about search engine optimization :)
I will totally reference this video in my next video about transistors!
Thanks. Glad you enjoyed it!
Well you should be grateful that so few people are watching this channel but you are watching it. To put it directly, for tech-based channels, the number of views and subs is inversely proportional to the value of its content.
Hi! Thank you for this fantastic video! I have a little question: how the Boltzmann' Tiranny is connect with the limit of 60 mV/dec (called also the Subthreshold Slope in MOSFET device)? Thanks!
Your voice very expressive. Immediately captured my attention.
Thanks a lot 😊
Does quantum tunneling play a role in the on current as well? What about BJTs?
yes!
Great video, not enough people talking about this issue. Ill be the next Y2K except it will actually be a real problem.
👍
I thought quantum tunneling was the way we explain why insulators tendency to approximate resistors as they strink.
Yep, you're right! You're talking about Schottky contacts and junctions. When a Schottky junction is not very abrupt the main way flow happens is by high energy electrons over the top (thermionic emission as I discuss) but if it's very abrupt (this is done by having high doping) then it's thin enough that tunneling is dominant. When you're deep in that regime transport through the junction will be Ohmic-like and you get an "Ohmic contact".
@Atoms and Sporks The ohmic analogy help you intuit why quantum tunneling is a problem. The understanding that electrons can move though a barrier does explain the resultant heating that is impacting the efficiency of microchips.
Loved the video. Why not just do both videos? :)
So the answer to keep Moore's law going is ultimately to solve the energy crisis, so we continuously can run more and more expensive cooling systems.
I suppose that would sort it. The current strategy it to somewhat consider it a loss cause but to abandon the notion of a "general processing unit" for most applications and more and more consider bespoke "Application Specific Integrated Circuits" (ASICs) which achieve greater performance for one special task through specialization in design.
Why don't they just make the processors bigger?
✰✰✰✰✰⠀EXCELLENT WORK⠀⠀♦⠀BEST CONTENT⠀⠀🎆⠀BEST PRESENTATION
So much thankful!
Glad you like it! Thanks.
Valleytronics, baby!
What impact would this problem have on recent CE grads? For example, if someone recently graduates having only learnt about transistors etc. would their degree and knowledge become useless after the limit of transistors is reached?
Hi, great question! It wouldn't. Honestly, I feel it's the opposite and it's a very exciting time to be a new CE grad.
What my video is about is basically the (very quickly) approaching physical limits of Moore's Law. Moore's Law exists not because it's a physical law of nature, but because it was a self-fulfilling economic prophecy on the part of the semiconductor industry. Chip manufacturers would produce a brand new fast chip using new fabrication technologies produced by the manufactures of fabrication equipment (which is a different industry). As a result programmers and software companies would produce new programs and applications that demanded even more computational power than ever before. Furthermore, new software application areas open up with computing "horse power" and computers leave the lab (scientific and military codes) and enter the work place (spreadsheet, business programs, e-mail) and then the home (computer games, cat videos) and then the lap and then the pocket ("there's an app for that"). With these new softwares on the market more and more people would start buying computers and there'd be a knock-on effect back down the line: more people demanding computers to use new software, more software sold, more chips sold, more fabrication equipment demanded. So then the fabrication equipment industry gets out their new line of equipment (lithography, etch, etc.) and the whole thing ratchets forward a generation.
It's a self-reinforcing cycle. And it's a cycle that only had two metrics for its looping: switch speed (a.k.a. clock speed) and number of transistors (well, on the part of the CPU at least).
However, totally irrespective of these physical limits, this cycle has started to change. What is happening is a diversification of the nature of demand on both the software and hardware end. The "wants" are becoming more varied and complicated. The biggest example of this is mobile phones. Mobile phone manufacturers don't actually demand faster chips, it's not their first priority, they want LOWER POWER chips, and the also want chips that interface very natively with peripherals like radio circuitry (i.e. the talking and wireless data functionality), with touch screens and cameras and so on. Their primary "want" isn't more transistors per chip, it's a more customized chip at all levels that is specialized to a different set of motivators than the programmers of old ( who just thought in terms of "More computing power! Nothing else matters!"). Internet of Things applications (to the extent that that isn't just an over-used buzz-word) demand even more specialized chips than that with a very different set of criteria for "better than last year's version" than Moore's Law suggests.
The chip industry is being pulled in a direction of specialized or so-called Application Specific Integrated Circuits (ASIC) that are requiring from-the-ground-up redesigns of all parts of chip making: the device level (the FETs), the architecture level, the front-end and the back-end. And although the idea of an ASIC has existed forever, it has taken on whole new meaning in the last 20 years as the breadth of applications is exploding. People want dedicated chips that are: Lab on Chip (LoC), that are System on Chip (SoC), that are passive and can harvest their own energy, they want "smart dust", they want thin-film transistor networks for LCD screens, and so on. These applications require changing everything, they are in a sense about forgetting "number of transistors" and "switching" speed and starting a whole new journey.
So, to bring this rather long comment to a close (sorry for that!), in some sense it's a great time to be a new CE grad, because the industry is changing things up at the architecture level in ways it never has before. And to do that they need young blood and new ideas. And if you're a research physicist like me, many of these things like spintronics, flextronics, valleytronics, photonics, polaritonics, graphene, carbon-nanotubes, 2D materials, quantum computers, and the kitchen sink, have ZERO chance of saving Moore's Law, but we're still interested because the industry isn't actually asking them too.
Keep in mind that there's still a huge market for "traditional" transistors. In fact a lot of innovation is being driven by mature technology nodes. Also, knowing the fundamentals is still extremely valuable.
I'd be interested to see videos on negative capacitance. Do you per chance attend stuff like IEDM and VLSI conferences?
you can still fill infinite small space with emotional sized particles
some spots orbs of light can heat up
now the spin of electron makes others time zones chaos
like in the brain
use intuition
theres Not a limit
We can't make smaller transistors, but what prevents us from making bigger dies?
Nick Hoffmann - 2019-01-23
Well, I would rather advise you to stick with physics-related video's, especially because you've already promised to cover so many other topics in your previous video's. I would, for example, love to see a continuation of the atoms and sporks journal club series.
Great video, keep the good work up.
Atoms and Sporks - 2019-01-31
Apologies for that, I'm trying to cycle through things so I don't linger on one topic for too long. Next one should be "Why does light have momentum" that I originally promised way long ago (I'd actually finished it awhile ago but I was worried it was too technical and boring, so I'm reworking it). Aiming to get a Journal Club in every 4th or 5th video.
radhenitsri - 2020-06-08
@Atoms and Sporks you are doing such a wonderful job, my best wishes.