Silicon Photonics: The Next Silicon Revolution? 

AI chips! I think photonics is very useful for inference in particular. Lightmatter is an MIT photonics AI chips spin-off that everyone is talking about.

How this would increase SINAD in modern audio gear and how it would push the performance past 130dB SINAD for just about every consumer level audio device. If optical circuits are used? That takes care of magnetic fields on the traces used in common logic that we use today. It also would make it theoretically immune to Radiation to a point. Especially if you used something like a Quartz crystal or something of the sort of oscillator for the CPU clock timing so that the light could be powered by that so the Logic itself would be immune to much of the interference from space travel. Something like this would be perfect for noisy audio environments. I would also have to say that something like this would also make it so that you would not need to use expensive gold conductors or silver ones as you won't have to worry about metal breaking down over a very long lifetime. I know it's a little bit further than what you were talking about on the video but I thought it would be a good idea to run it by you.

There are several startups working on all-optical quantum computation using integrated photonics. Psiquantum and Xanadu are the two most prominent ones.

@@Extys I second this and also maybe go over the topic of what makes for efficient chips in a neutral net, like Tesla's new super computer, is he just bullshiting like normal elon or not.

I prefer Graphene and "Graphene - Photonics" over silicon. What's better than silicon integrated - Photonics? Graphene - Photonics / carbon nanotubes, and there have been new methods to scale up Graphene since 2020, you just need to research it, some ways are found if you go to my channel to check out all the Graphene playlists, and Graphene - Photonics on the other channel found under my channel called Technology Research, it's a green T icon / thumbnail, once you get there go right to the "created playlists" and start at the bottom with just the first 2 videos at the top of the Graphene playlist by clicking on the "view full playlists" below each playlist, scroll down to the bottom of the created playlists to find it. After those 2 videos at the top of the Graphene playlist, go to the Photonic computing playlist and so on, work your way to the top, watch all the Graphene playlists, and maybe even the STL playlist after those, you're going to see very big pictures here. Yeah I kind of get tired of the bias around silicon and nothing else like Graphene isn't mentioned ever with Photonics. Check out the Graphene computers for home and starships playlist on the carbon nanotubes scale up breakthrough in 2020, and Graphene aerogels made in space under the Graphene with starships playlist. Though please watch all videos from top to bottom in that order under each playlist, no cherry - picking, please force yourself to watch them in order, there is a very good reason I placed them that way. Again start at the bottom of the "created playlists", so click on the words created playlists and scroll to the bottom, the Graphene playlist should be on the bottom.

Now thats overclocking

🤣🤣🤣

LED is silicon, you need to revisit basics. Your solar panel is just LED in reverse mechanics. Oh sorry maybe germanium.

It did for a few fractions of a second, I bet.

It's sure give you smoke particle.

That's what happens when you check every 8 seconds.. 😆

@@markrix 😀

Now RU-vid algorithm controls your mind.

Yeah I was left somewhat confused after watching this video. A lot of information about how it works, but not why it's better or why we'd want to use this technology.

@@maybehuman2148 He said it at the outset, faster, lower heat.

there's a lot to be said about photon computing. Another really important reason is that in order to make quantum computing work, you need photon logic.

​@@megalonoobiacinc4863 excellent point on quantum computing

@fox 0ps, as was mentioned in the video light cannot pass through die features smaller than the wavelength of the light used itself. Which means that the chip will be very large and potentially very hot.

I wonder which technologies will be developed once we hit the end of the road.

Great 👍 photonics took pace by the invention of laser

@@mangeshgaikwad345 I was integrating lasers and waveguides detectors and modulators on the same IC 27 years ago. Challenge was doing this as a one person team. All the processing, simulation, testing. Really quite impossible in the basement of a university.

It will be quite specialized in the field as the knowledge to pull this off has a lot of factors.

no it is useless bubbles

Yes. Exactly my thoughts.

Hybrid analog/digital neuromorphic hardware using photonic components

Analog based computation is more efficient than digital, but its lack of accuracy makes the jump from digital to analog a huge one.

I came to comment about the refractive index slowing light in glass, and just had my mind blown. Hollow core fiber!? And it's been around since the '90s? Wow.

@@hikingpete it may have been around for awhile but hollow core fibers with low loss that comes close to traditional fiber is new.

That's why starlink can have lower ping in long transits than fiber on ground :p

Hollow core fiber research and developments are being accomplished in Australia with funding from the military specifically to test jet fuel integrity... fascinating developments

Yes, plus multimodal fiber has the beams bouncing off the cladding adding a lot of distance to the path…

Do you recommend any books for photonics?

Current prediction by social blade based on historic growth of the channel: 2024-04-18

What prevents us from using high energy ultraviolet lasers to make smalmer optical gates?

A gate isn't that useful if the horse is too big to walk through it.

@@Asianometry Hence my question why we can't use higher-energy lasers, whose wavelengths will fit through smaller gates. The high-energy end of ultraviolet light is in the nanometer range.

You need a wavelength that silicon is transparent. This limits your light to infrared

@@kazedcat I didn't know that, thanks!

It’s always cool watching these types of channels rise from a few thousand to hundreds of thousands and eventually millions of viewers.

Yeah I want to know his background. He's either deeply involved and knowledgeable in the economic, tech and semiconductor space or very very good at researching and writing scripts that are dense while being easy to follow paired with excellent visualizations. No other channel I know has been pumping out this much in depth content of complex topics. He's definitely has a rare talent.

@@Baulder13 I agree! Jon is in a league of his on in this field.

@@Baulder13 Agree, this guys breadth of knowledge and research ability, plus the fact he can present it to many of us non-PhD's is unique.

@@Baulder13 From what he said previously he doesn't know much about the stuff - he "just" researches topic very in depth. Those scientific papers that he shows - you can be certain the he reads all of them and more. I wish I could research like him. That takes a lot of practice and helluva brain to understand it all. Regardless - this is one of the best channels when it comes to both Semiconductor/Technology/Industry part of the world AND political/economy part of the world. He focuses mostly on pacific area (duh - Asianometry) and what is important there. But silicon industry is one of the most important parts of Asia's strength.

Alex was so helpful for me on this topic. I really give him all the props.

Interesting! I always thought the whole "information travelling at the speed of light" explanation for why photonics are good was a bunch of BS (since voltage waves do too).

This could be the key for phased array optical systems too. Light field based displays or something, if each emitter is synchronized. I have no idea though

@@yelectric1893 Very cool idea to incorporate photonics in coherent optical MIMO. In fact, it's recently been demonstrated that speeds of over 1 petabyte per second are possible in a fiber optic cable. That's a lot of data. They use Raman spectroscopy, which is another great topic for this channel since a lot of the papers coming out of Asian countries need to use raman spectroscopy or other techniques to optimize their new ideas for information technology. As you pointed out, this could be used for displays as well.

@@howwitty Hi, I’ve never heard of the Raman concept but is it like an EDA software? Seems very interesting.

@@yelectric1893 It's like a quantum ruler for molecules I guess!

almost as insane as the insides of electrical chips, pure black magic going on inside all those fancy procesors, even mine, I can't belive intel managed to run windows in a fucking potato

That makes integrating them into ics a bit worse of an idea eh?

@@xxportalxx. no, you just have to solve the problems first

@@JL-pc2eh it certainly gives even more credence to the separate chiplet strategy, however I wonder if you'd need like a separate heatspreader to decouple it from the ic running computations. If the physics itself is sensitive to temperature variation then thermally coupling it to an ic that's temperature is highly variable with load is a bad idea.

@@xxportalxx. what people do right now is add little resistances with feedback loops to heat the sensitive components and adjust their frequency response, which is not free by the way, but if you're operating around 50GHz the cost per bit goes down. In my lab there are some guys working on the thermal response of devices and trying to make more robust photonic circuits with this regard

@@raphaelcardoso7927 that seems rather difficult to maintain to me without significant thermal isolation (i.e. an isolated chiplet), as you're fighting both your active cooling loop, and the random heat spikes from the cpu. If it weren't isolated I would think you'd be limited to the range at or above the max temp of the cpu-cooler system, below that and the cpu could push your temps too high, unfortunately this also means you'd be fighting your cooling system constantly (and likely be degrading cpu performance).

100%. The other common myth is that "light is faster than electronics", which is not true if this is taken to mean light (in the visible/near IR) will travel faster than an electromagnetic wave through a copper wire. The wave in the copper wire is significantly faster and causes problems for the design of electro-optic devices, like modulators. The reason we say "optics is way faster than electronics" is because of the available bandwidth at ~193 THz (optics) vs at ~3 GHz (RF). We can use 100s GHz of available bandwidth at the high frequencies of optics to stuff lots of data, much more so than can be stuffed in the narrow bands near 3 GHz.

@@christianb1176 The speed of electricity through a wire or circuit in comparison to light through a fiber optic cable is a matter of much scientific debate. But, you are correct about the bandwidth. Light can carry significantly more data. Technically, electrons should move at a similar speed to photons, but there's a little thing called electrical resistance. That's why billions are being poured into research to find super conductors that can operate at room temperature.

It's in the market right now. Products already exist from companies like Intel and Luxtera.

I think the coolest technology should win.

Now you’re thinking like an engineer! Lockheed will be in touch.

That's what I wondered, if the silicon components are getting smaller than the wavelength of visible light, how to integrate them? Wow!

i bet they could eventually use shorter wave lengths too. surface plasmas are messy though

@@vevenaneathna the hard part with shorter wavelengths is finding good materials to make waveguides for them. Silicon right now is on 1550nm mostly

@@raphaelcardoso7927 ya thats over my head. idk, some kind of lead or gold alloy? i know there are quatnum mechanical effects that get a lot worse when u have "surface plasma" state, and I feel like heizenburg uncertanty might come into play if the channel was able to be super narrow or super flat/polarized, and im thinking that would limit the bandwidth/frequency range you could use if it wasnt just a binary 1,0 data

The problem is not propagation the problem is keeping the signal to noise ratio so that it is still possible to recover the signal in the other end.

For sure! However, charging the parasitic capacitor on the line at a few GHz gets expensive really quickly

@@raphaelcardoso7927 a reel of coax makes an excellent dummy load above a couple of 100 Mc. Shut down the transmitter when smoke starts to develop.

Major problem with the "optical computer" is a lack of optical memory. Optics is good as the interface between electronics-based computers.

@@christianb1176 You do have a good point. To simplify things early on, however, we could create a kind of "translator" that allows conventional memory to be used. This will cause bandwidth issues, but it would help to prove the concept until a better solution is found. For the record, though, I'm absolutely not an expert, so that's mainly baseless guessing on my part.

If you include organic compound then carbon will win but if you limit the scope to inorganic crystals then silicon wins.

Well they're still trying to find cost effective methods of industry scale production for those materials. Most of the benefits of graphene especially in electronics specifically requires uninterrupted sheets as well, which further complicates the production process.

I want to add, the problem of creating a photonic transistor of the same size of a current transistor with 7 nanometer gate, can only be solved with a deeper theoretical knowledge of what is common to a photon and electron, the quantum spin number. Quantum spin is not a real spin, rather it is a description of the invisible connections the particle has with Space and all surrounding the particle as it moves. This knowledge is still very lacking and it is at base of all the modern theories - trying to explain the very nature of things...

I forgot to say, thank you very much for this video and all your deep and informed journalism on current and future technologies...

@@rayoflight62 Go read up on some of the companies he listed. OP transisters may not be the standard across billions on a chip, however highly likely that key areas will be where they can do the most good initially, soon.

I didn't mean to diminish the achievements of the big names involved. I was trying to point out that the current OP technology is not the equivalent of doped silicon and electric fields technology, but OP is a component based on light path geometries and interference effects, because we don't have (yet) the theoretical foundations for a 1:1 equivalent...

That is actually correct. Electronic books never make that distinction for whatever reason. Well done😎👏👏👏👏

I've never quite understood why that is, sure it's a little more complicated but not by all that much compared to say complex board design et cetera and it gives rise to many dangerous myths on the nature of electrical dangers, most people think water is an excellent conductor (only moderately so) but are completely oblivious to the dangers of airborne dust and smoke being as good or even better conductors resulting in quite a number of electrocutions every year.

Laser socket? you don't need electricity for those chips?

​@@shadowshadow2724 Not for my hypothetical ones. As were mentioned in this video modern approaches (as in photonics logics chips or tranceivers) mainly use either another semiconducter material for the laser inside the chip or a laser beam guided from another source, but we still rely on electro-magnetical excitation of some crystal in order to make a laser beam. Not really sure about actual signal fade in my theoretically endless scalability and real world quality of full internal reflection inside modern silicon photonics and how often would you need a repeater (more added energy) in such scenario. The obvious elephant in the room is the price (or energy and materials) wasted on manufacturng such devices relative to the lifetime energy consumption plus manufacturing expenditures of alternative electro-magnetic ones. On the other hand it's really hard to define the efficiency of modern electric chips since it's always relative to the electricity and price of equipment to the number of operations calculated during certain period of time. That period of time is usually defined not by the lifetime of a chip but rather by monetary benefit of upgrading your system into some newer generation technology. At the same time there should be always done this software/hardware weigh in in order to know whether your logic circuits are even worth making to be "set in stone" forever. The real benefit of such photonic approach would be almost indefinite lifespan of such a system (with only the risc of physical destruction and maybe some lychens). It should not fear solar wind or any rapid shift of magnetic field of our planet or even electro-magnetic waves from nuclear blasts or any nuclear blasts in ionosphere for that matter. Unlike crypto where you just burn coal for nothing you can "burn" some logic circuit forever whenever you have some free wind or sunshine... Em but for the lasers you would still need to transfer energy for them in some manner, preferably electric, if only there were some method to make a laser from concentrated sunlight. OK, I'm not an expert neither in lasers nor in photonics but at one point I had a plan (among many others) to get my masters degree in Germany in order to get somehow to the GlobalFoundries since they have a lot of IPs in photonics, but that never happened unfortunately.

Yes! This is a research field known as optical networks on chip. It's been demonstrated that if the distance is above a few mm, photonics is more worth it than an electrical connection