Greetings from the Center for Free Electron Lasers at DESY the home of the european xray free electron laser ! It's really cool to see FELs on youtube and crazy to have the video actually talk about SASE. I personally mostly work with undulator light from the petra synchrotron, so take what I say with a grain of salt. Some issues with XFELs: - you need really nice beam parameters for sase to work and seeded fel isn't feasible for xray photons - they are fucking huge, european xfel is a >3km long line! compare that to high brilliance synchrotron with undulator beamlines like petra III at 730m diameter - repetition rate is hard some of the really really cool shit you probably couldn't talk about because of length: - the light is in phase, so you can do some really fancy optics stuff with it and get intensity proportional to N² instead of N (where N is electron number) - there are two different regimes for the self organisation. instead of injecting electrons that are monoenergetic but spread out in phase space you can also inject them with a wide energy spread in an ultra short pulse. That's one of the things people are trying to do with laser based acceleration systems. Ultimately I'd guess that FELs won't do lithography in the super near future, just because time averaged intensity matters more than peak intensity. European XFEL SASE1 has 27kHz time average repetition rate with 4mJ energy per pulse. That's only slightly above 100W! From a billion €, kilometers long device that only went online a few years ago. Petra's P61 beamline alone puts out more power than that (I've seen samples with holes melted through them by that beamline), admittedly that one has a wider spectrum, but you get what I mean. The size / capital investment to power ratio is pretty bad for FELs still. source: www.xfel.eu/facility/instruments/spb_sfx/instrument_design/index_eng.html#:~:text=These%20pulse%20trains%20can%20be,repetition%20rate%20of%204.5%20MHz.
The MuCLS is an even smaller sychotron-like X-ray source, just 5m x 3m, but I have no idea if it's suited for Lithographie. Still an interesting look at the other side of the extreme for monochromatic X-ray light sources.
You mastered presenting technical topics. Rise and fall of inflections, jokes, etc etc. night and day difference from your dry early videos. Hell ya buddy, thank you for teaching us so much.
Yeah current EUV is super duper power hungry because the reflective optics they use to bounce the EUV around for photolitho only each reflect ~70% IIRC, and after bouncing around a dozen or so times that means that only a super tiny amount of the EUV generated actually makes it to the wafer :(
Hearing about the insanity behind previous generations of fabs was crazy when I first found this channel but now imagine saying to a random person “yeah that phone chip was made with a particle accelerator”. Crazy tech
Synchrotron radiation has been around for what 6 or 7 decades? Sure, the machines are big but there's really nothing stopping anyone from making them smaller and more efficient. Synchrotron radiation is smaller than 13nm and has properties wholly different than it's neighbor EUV.
Free electron lasers are a really cool diy project. Deadly high voltages, vacuums, strong magnets, blinding laser radiation, and extremely dense beams of beta particles. Don’t know why they’re not more popular. Also the beam dump for an electron beam will just be some conductive surface that’s connected in a circuit back to the other end of the high-voltage source that accelerated the electrons out of the electron gun. Charges need to flow in a circuit, though when you’re dealing with free charges it can get kinda strange with net charge buildup.
I guess it depends on the beam energy: You may require *a lot* of shielding around your nice conductive beam dump :) Actually, I wonder about the return circuit. E.g. for SwissFEL, the electron bunches are 200pC at 100Hz repetition rate, that's 0.02µA. You don't need a special return path for that, but grounding is of course essential.
This looks quite good: High-power EUV free-electron laser for future lithography by Nakamura et al (2023); here's the abstract: The development of a high-power EUV light source is very important in EUV lithography to overcome the stochastic effects for higher throughput and higher numerical aperture (NA) in the future. We have designed and studied a high-power EUV free-electron laser (FEL) based on energy-recovery linac (ERL) for future lithography. We show that the EUV-FEL light source has many advantages, such as extremely high EUV power without tin debris, upgradability to a Beyond EUV (BEUV) FEL, polarization controllability for high-NA lithography, low electricity consumption, and low construction and running costs per scanner, as compared to the laser-produced plasma source used for the present EUV lithography exposure tool. Furthermore, the demonstration of proof of concept (PoC) of the EUV-FEL is in progress using the IR-FEL in the Compact ERL (cERL) at the High Energy Accelerator Research Organization. In this paper, we present the EUV-FEL light source for future lithography and progress in the PoC of the EUV-FEL.
I suggest you looked at TsingHua University's SSMB (Steady State Micro Bunching) to generate DUV, EUV and soft X ray.... all in one. China is already building an EUV factory at Xiong'an New Area in Heibei. 😁
Yep and per another comment here that likely has more time averaged power than FEL anyway as that seems to be main reason for study of SSMB... increases time averaged output.
Talking about getting the "Semiconductor Avengers" together, this at this point start to look even more complex. On the level of fictitious Black Mesa research complex. This will cause a large overhaul on floor planning too when trying to protect equipment from disaster scenarios (floods and earthquakes) and floor costs when putting in laser redundancy. I can almost here the quote: "I never thought I'd see a resonance cascade, let alone create one." 🙃
6:00 When I was a teenager back in the late 70's, I wrote out by hand, from the Encyclopedia Britannica Jr. the section on Laser. I remember Laser: Light Amplification by Stimulated Emission of Radiation. Please don't omit the "Radiation" part.
They had one of these at the uni I worked at, or at least they did research on it back in the 90s to 2000s. It was able to emit a few mW at a 5 mm wavelength. They had some mathematical modeling for reaching kW powers but I'm not sure if they got that far before the building was demolished. Removing the room it was in reportedly took a week because it was absolutely radiation hardened and had very thick, reinforced walls.
Electrons are not attracted towards the magnetic poles :) their path is just curved perpendicular to the magnetic field (the oscillations are actually in the plane perpendicular to your undulator picture at 7:55).
I like the idea of using particle accelerators to create light at the energy needed for future lithograph. Some particle physist are working on very small particle accelerators that are the size of your hand. Perhaps this avenue can enhance our understanding of the quantum world.
@@Cristopherdreamer _(Note: Given the small size I'm assuming these are cyclotrons, synchrotrons, or the like.)_ The typical limiting factor in how fast one can accelerate particles is the magnetic field strength. The faster the particle moves the more inertia it has and the more force that is needed to nudge it into a curve. The tighter the curve the stronger the magnetic field needs to be. The first cyclotron could only achieve low energies (80 keV) because it used magnets recycled from telephone switch relays. With modern superconducting magnets one could achieve much higher energies. At some point though the limiting factor becomes the material strength of the magnets; too high a field strength and the magnets will tear themselves apart. The only solution then, barring advances in material science, is making the ring diameter larger. I've glossed over/skipped a lot of details but that's the gist of it. Maybe the OP could provide a pointer (DOI) to some papers covering current research in this area?
You might want to take a look at new particle acceleration techniques based on lasers. It has the potential to do in a few meters what a linear accelerator takes a kilometer to do. It would not help with the undulator side of things, but it would help reduce the overall size of the FEL machine.
Great explanations, as always. Yet, in the undulator the magnets don't attract or push off electrons rather that deflecting them othogonally (in the picture to or from the viewer of the picture) to the magnetic field and orthogonally to the electron's trajectory. The resulting acceleration effect would be the same.
I learned the physiks behind a stimulated emission differently then your explanation. That it doesn't release a photon in a random direction but rather that if an atom gets hit by a photon while having one of its electrons exited at the exact same energy state then the incoming photon, it releases a second photon with exactly the same frequency and in the same direction then the first one.
Most important difference between plasma 13.5 nm light source and 13.5 nm laser (any type, but we can do it only on free electron laser actually) that is the first one is non-coherent. In properties way it is odrinary light source. Laser light is coherent. Most important laser part is resonator. With resonator light is producing in entire volume of gain medium. Photons emmited with act of spontaneus emmision have random vetor, they goes into random direction. Photons emmited durign stimulated emmision have this same (all) properties - also direction of propagation. Light moving parallel to resonator axis are amplified (because it bouncing many times form one mirror to another, and in each pass it is amplified by gain medium) - and all this photons can have this same properties - frequency, phase, direction of propagation, etc. This means this light can be focusing into very small point - smaller than cross section of gain medium and actually it can reach two diameter of light wave lentgh. Parallel beam also mean we can reach high NA.
A magnet doesn't attract electrons. Electrons are deflected by the magnetic field perpenticularly to the magnetic field, not towards it or away from it. So in your picture they would swerve in and out of the picture plane.
An electron travelling straight down the centre wouldn't be accelerated at all because it's parallel to the field. Off-center electrons would oscillate like an airplane changing the direction of rotation mid-barrel roll
I was wondering if “perpenticularly” was a word I’d never heard of. I tried looking it up and didn’t find such a word, so far. This has me thinking that you perhaps you meant to type “perpendicularly”?
@@shanent5793 I don't think you have the correct magnetic layout in mind. In the layout shown the magnetic field lines will go up and down alternately, and the electrons will in turn curve in and out of the screen plane. This is also true for electrons moving straight down the center of the undulator.
I whould love to see a video about this new ssmb technique China is supposedly investing. Building a particle accelator for lithography, really? Can it be cost effective, compared to standard euv?
Could be... a single EUV scanner cost $300-400M dollars, whereas a Chinese synchrotron particle accelerator can cost $200M but produce dozens of EUV beamlines.
Except China was not given a choice between cheap advanced chips made by foreign tin EUV and expensive advanced chips made by domestic particle accelerator EUV. China was given a choice between expensive advanced chips made by domestic particle accelerator EUV or having no advanced chips at all.
@@MrSpiritmonger they run about $50 million, the vacuum control system I've had to design for the AMSL EUV is around another $5 million, and then there's other support tools that are also required, and I couldn't tell you what they are since I've had no involvement on that side of the tool yet.
Odd how chip sanctions pushed China to go all out in the development on this technology. I remember how Russia sending rockets into space got America's space technology development go all out as well.@@Stroporez
The US Navy spent a lot of money on FEL (500M?). Insurmountable engineering problems scaling it up. Of course, that was for directed energy. Perhaps for this specific application it will work. FEL beams in the Navy effort were particularly nasty in terms of power density, simply ate the optics alive.
Are you sure you're not talking about the US navy's electromagnetic rail gun? Electron beams for directed energy weapons don't work. The electrons will interact with the atmosphere. In space they are bent by the Earth's magnetic field.
The SSMB (Synchrotron based Steady State microbunching) approach to generating EUV being pursued by Tsinghua University deserved a mention. Unlike a Free electron laser which has high peak power with low repitition rate or a normal storage ring synchrotron which has low peak power with high repitition rate, a SSMB based synchrotron has both high power and high repitition rate. If successful, SSMB EUV radiation would be far more coherent, tunable, powerful and cleaner than anything LPP can hope for. Tsinghua has promised a prototype by 2026-27.
@@msimon6808 SSMB is optically and mechanically much less complex than LPP. ASML didn't take this approach as it was discovered much later and remains kinda experimental plus the logistical problems that comes when you have to force chipmakers rebuild their multi-billion dollar fabs around your synchrotron facility rather than just shipping a machine to wherever their production lies. But there are lots of advantages to the SSMB approach: -High average power: the power aimed is > 1 kW per tool, each facility should be able to incorporate multiple tools. -Narrow-banded and collimated: the radiation spectrum bandwidth is < 1% and has a well collimated angular spread ≤ 0.1 mrad, which should help to reduce the number of reflection mirrors and thus increase the EUV power transport efficiency -Truly continuous wave: the temporal structure of the radiation is truly CW, this minimizes the chip damage problem; -Clean radiation: the radiation is clean and carries no debris, so that mirrors do not get contaminated and do not require frequent replacements during operation; -High stability: storage ring source has the advantage of a good stability, which helps to increase the machine availability and qualified production rate; -High efficiency: higher wall-plug electricity to EUV conversion efficiency than LPP source expected; -Reasonable price: storage ring technology is mature and the average price for each tool is expected to be reasonable; -Good scalability: easy to scale to shorter wavelength If Tsinghua can pull it off, it would be absolutely revolutionary.
@@xBlackWind Thank you very much. My son works (EE) at a company that employs a Belgian accelerator to produce isotopes. I sent him this article. And since electronics interests you - I did something about switch debounce. Hardware and software. I have designed a SPST SR Latch debouncer. It tells you - like the SPDT SR Latch circuit - as soon as the switch starts changing state (plus deglitch time). Either opening or closing. spacetimepro.blogspot.com/2023/09/switch-debouncer.html
@@fromfareast3070 what calculation are you talking about? SSMB light source power is almost an order of magnitude greater than that of LPP EUV. Modern LPP EUVs output an average power of 350W while SSMB can easily reach an average power >1kW with far lesser optical complexity.
Thanks for the analogies used in this video. I will use and credit your video in a part of training that I do on how coherent optics are used in high-speed networking above 100Gbps Ethernet.
Considering the size of the hole that they have dug up over there in Hsinchu... god knows what they can put down there... I saw it first hand back in May and I don't know what they were digging but from my what brother in-law told me, before the hole, there was a literal mountain in its place. It's insane!!!
Lol... Ask ChatGPT the stuff you don't get, or find a source that explains the parts that feel like you came late to your university lecture and you know not what they are talking about. If you don't like math, just find a source that doesn't mathematically encrypt its content. Trust me, nowadays if you can't understand something, it is by choice. Though I don't claim certainty, for I don't know your situation... But the human brain has too much computation power for anyone to excuse themselves as too dumb. It could be a software issue, (I mean, someone that can't use their brain to its potential can't expect to harness its full capabilities), but statistically speaking, I think no one is hopeless, for as they say, it is never too late to learn.
Contamination of the incredibly sophisticated optics are more than enough reason to move on to free electron lasers. No one wants a dirty neighbor. Improved energy efficiency and a more robust upgrade path are icing on the cake.
Love this one. Used to work with synchrotrons, and while your physics explanations have minor issues, the business history of this is fascinating. Superb work.
As a vacuum-electron device worker (CPI Microwave) I find this baffling and yet delightful. After years of telling ol' vacuum tube guys that our industry was not only shrinking but shrinking towards zero and would die, now solid state electronics engineers think they may need a linear-beam vacuum-electron device to keep their industry moving forward. This even as GaAs and GaN devices have largely finished moving the power-frequency curve up and to the right, and we still have lots of applications-- replacing your microwave oven's magnetron (a type of vacuum tube) with a GaN device would more or less double its power consumption, for example. I feel now like there will be a need for tubes forever. If you need some LINAC klystrons for a giant $400m FEL we'd be delighted to provide.
6:38, a slight correction on the meaning of "stimulated emission." Stimulated emission refers to a process where an excited electron can become sort of a photocopier for photons. When the electron is excited (technically I think this is called the Avatar state), and an external photon interacts with it, then the electron is prone to emit its energy as a second photon traveling in exactly the same direction with exactly the same wavelength, phase and polarization as the incident photon. This is why the laser emits all of its photons in a narrow beam at effectively a single wavelength.
Yeah, really important thing which makes laser not just fancy lamp. We get over this so much on the lectures I usually don't even pay attention towards explanation in videos. Also he missed "Radiation" then explaining acronim/word "Laser", which little weird.
Slight correction to the slight correction: The wavelengths must already match anyways, they are a consequence of the energy level difference. But yes, the phase e.t.c. make it special.
1.) Nd lasers are not commonly pumped by LEDs, in fact this has only been demonstrated in academic settings. 2.) Stimulated emission does not refer to any emission from an excited state as you explain, but is instead the result of an incoming photon stimulating an electronic transition and subsequent release of an additional photon with identical phase. What you effectively described is spontaneous emission.
I received notification of this video via the Asianometry Newsletter. Momentarily, I thought I had the opportunity to win a *free* electron laser. Disappointment.
It will never be disappointing to stop using the LPP method. Tunability of the wavelength and higher power ceiling for the light source are groundbreaking changes. I hope we see the FELs in production before the end of the decade. That would be amazing.
The mirrors have to be specially constructed for the specific wavelength if I understand it correctly. Normal mirrors don't work for that short wavelength and that is one of the problems. So the mirrors are made with thin layers making the light reflect with a big loss. The layers have to be tuned to have the exact thickness depending on wavelength.
@@lubricustheslippery5028 Yes, but you could still have a single light source for multiple machines operating at different wavelengths, if you partition in time. Even if it was just for development purposes, being able to operate and tune a newer generation machine without having to build a whole new light source would be a big deal. Simply being able to experiment with different wavelengths cheaply would be a massive deal in it's own right. The different layer thicknesses for the mirrors would not be something manufacturers would need to develop new technology for, and while not optimal, the resists that work for 13nm should do ok for other high energy wavelengths. You could say 'lets try 40nm', adjust the light source, and switch out the mirrors, and it might just work. The possibility to introduce a cheaper alternative to EUV is also made available my this type of light source. Nothing is stopping us developing machines around something like 40nm light, for example. These would have far less stringent requirements in terms of mirrors and masks than 13nm, while also reducing the energy doses required 3x. They may not quite hit the same quality* as 13nm, but potentially could be much cheaper to operate. If a fab were able to shift 60% of steps to a cheaper process that would be massive in terms of output cost. Fabs do miracles with 193nm, so an intermediate step might be warranted, if it could be developed cheaply. * It is possible that they would, or better. Photoresists for EUV are activated as much by secondary effects as the light itself, so reducing the energy of those secondary effects might allow the lithography to be more accurate. The reduced requirements from mirrors would also make getting higher NA with the same manufacturing precision possible, clawing back some of the accuracy almost for free.
@@lubricustheslippery5028 Yes, but you could still have a single light source for multiple machines operating at different wavelengths, if you partition in time. Even if it was just for development purposes, being able to operate and tune a newer generation machine without having to build a whole new light source would be a big deal. Simply being able to experiment with different wavelengths cheaply would be a massive deal in it's own right. The different layer thicknesses for the mirrors would not be something manufacturers would need to develop new technology for, and while not optimal, the resists that work for 13nm should do ok for other high energy wavelengths. You could say 'lets try 40nm', adjust the light source, and switch out the mirrors, and it might just work. The possibility to introduce a cheaper alternative to EUV is also made available my this type of light source. Nothing is stopping us developing machines around something like 40nm light, for example. These would have far less stringent requirements in terms of mirrors and masks than 13nm, while also reducing the energy doses required 3x. They may not quite hit the same quality* as 13nm, but potentially could be much cheaper to operate. If a fab were able to shift 60% of steps to a cheaper process that would be massive in terms of output cost. Fabs do miracles with 193nm, so an intermediate step might be warranted, if it could be developed cheaply. * It is possible that they would, or better. Photoresists for EUV are activated as much by secondary effects as the light itself, so reducing the energy of those secondary effects might allow the lithography to be more accurate. The reduced requirements from mirrors would also make getting higher NA with the same manufacturing precision possible, clawing back some of the accuracy almost for free.
I have been wondering about Samsung struggling with their 3nm manufacturing and even TSMC seems to have problems with 3nm such as low yield, which could be because of early technology but anyway. It seems that EUV lithography is running out of steam, and Ray Kurzweil says that when that happens, historically a new paradigm replaces it and continues the accelerating progress.
It doesn't matter if the stockastic outcome is random or chaotic - applying the binary distributiton it will work 50% of the time. ... because either it works or it doesn't ... 😁
I'm so used to people zoning out when i get nerdy,..... its nice to come here and feel like a layman and just barely being able to digest some of these processes...it gives me that feeling of awe and wonder again..... thank you
Synchrotrons use that exact feature to their advantage for a wide variety of applications. In fact, most Synchrotrons are capable of sweeping across ranges of energy for materials studies, and this capability let's you get molecular structures and all sorts of other information about a sample. In fact, Synchrotrons are better at doing this energy sweeping than free electron lasers (at the cost of maximum intensity). For example, x-ray absorption fine structure microscopy (EXAFS or XAFS) and x-ray absorption near-edge structure (XANES) can be used to find the complete molecular structure of a material in a matter of minutes, and you can automate beam-lines that do this. You basically sweep across a spectral line of an atom (one in the x-ray range) and take an absorption spectrum. These spectra have wiggles in them due to atomic bonds and such, so you can do some math (a fancy fourier transform) and find the distribution of atoms around that specific element in a sample. The advanced photon source (APS) at Argonne National Lab does this and a whole host of other things. Most FEL and SLS (Synchrotrons light sources) work in the GeV x-ray regime (1000x more energy than the MeV range discussed here) because that energy range is more useful for research purposes. Also, there is a program at Arizona state University to make a compact FEL by using an intense optical/IR laser as the undulator instead of a magnet bank. I imagine the future of high-energy lithography machines will use plasma Wakefield accelerator cavities and those compact laser undulators (if they can get them working, of course). Synchrotrons and FELs are absolutely incredible machines. Sad that they're so expensive
You probably know the most about chip fab and lithography as well as specific model chip fab equipment than anyone bot specifically employed in the industry.
There is nothing mind blowingly new here, not to disparage Jon, he is excellent at bringing down the big concepts to us noobs. But if you start to read the papers and interact with people in fabs you learn that a lot of this is just industry happenings. 🤣
Honestly I am interest in these topics and I love what you are talking about but perhaps the most important thing is that I like how serous and rigorous you are about all those topics the part which is the most important and to which you have always seriously adhered in a rigorous manner is the humours part of the presentation... I can not explain in words how impressive this is as it is like the seriousness brings us to a higher level then the fun part, the release or the acronym *R* (meaning the release) is when we loose the energetic state and get to the enlightened state acronym *G* (meaning ground state)... thanks for every an all videos and all the efforts and dedications...
@@guill90 My guess would actually be SMIC or Global Foundries, but there are a fair number of other obscure options so I would give it less than a 30% chance that one of my guesses is right.
Haha no one is going to give China's efforts a fair shake in YT channels like this. But I am glad that a lot of people are calling out for China's FEL efforts and they are probably at the leading edge for this application. Shows that the western propaganda about China is not as tight as they will like it to be.
@@gelinrefira Yeah, I don't know if or why you'd try to control information about China's progress if there is interest in handicapping that progress. But silly efforts to delete Chinese business from western helpers is starting to look like it could totally backfire and leave China at the leading edge without even needing to try to take over Taiwan to get there. Not sure it helps or hurts risks for Taiwan in Grand scheme, but seems it mostly hurts if China is made more self sufficient by outside pressures that force it.
@@stormsj That's the thing that westerners are chronically ignorant or deliberately refused to understand on how China looks at the Taiwan issue. I think it's deliberate because they are terminally unable to accept that China is doing great and is innovative, creative, dynamic and powerful. So of course they are going to control information flow. If those Americans who like to denigrate, insult and look down on China will to visit Shenzhen one day, they will either be shell-shocked, or gibbering in terror. "The white man is not on the top anymore!!!" Th thing is China doesn't NEED Taiwan to keep growing and progressing. What Taiwan represent is a wound that have not been healed. It's a political issue, not an economic one. But the Chinese are in no hurry to reunify with Taiwan and they ultimately don't even need TSMC. All China really wants is peaceful reunification and they can wait 10, 20, 50, even a 100 years to do that. But everyone and their dogs in the west talk like China can never have advanced chip technology without Taiwan or the west. Taiwan is a small little province. If you really know China, then Taiwan really is more like a 2nd rate, chronically mismanaged provincial backwater. All US sanctions do is solidify China's conviction and determine, and accelerate their progress because they can centrally planned to focused on project deemed as strategically important. Now you just force China to push their best and brightest to come up with innovative solutions to solve intractable problems. And everyone who bet against China from succeeding on their national project always lose.
@@stormsj At least my opinion, as someone who isn't pro China non particularly pro US. Having tight competion between superpowers in this case US and China reduces the risk of direct military conflict because stakes are to high and risks aren't worth it. But if one side falls behind or significantly threatened in some other way, risk of war increases to dangerous levels because either side might see it as a final option to regain their position.
@@vitaliibraslavets this is probably true as far desperation of one side falling, but I think risk of war is already heightened even if efforts to complicate China's competitiveness fail and it becomes more self sufficient as it has disconnected them from a need to cooperate with the west generally in this effort to isolate them. But yeah, even if efforts tend to have intended effects there may be some assessment there is a last chance to take Taiwan or the like as they lose capabilities. It's all been generally ignorant way of going about things ever since Xi took over. I totally get why the rest of the world has reacted to China post Xi as they have... but this reaction has had the effect of strengthening rather than weakening him. They needed to be more savvy and directed annoyance more clearly toward the top of the CCP and been more receptive to corporations and local governments in China willing to stand up to what has been happening under Xi.
If only the Buddha had known there was an easier way to get to that higher state of being. Geezzz. There's a Nobel Prize waiting for someone who can determine the maximum tensile strength of a metaphor.
How are the mirrors and masks going to stand up to coherent illumination? Is it necessarey to diffuse the light to prevent speckle, or does it average out over many pulses?
Any chance of using wakefield style accelerators to reduce the giant size needed and maybe put it into the lithography machines? Or do they not allow the needed preciscion?
Having included particle accelerator in semiconductor chip fabrication has opened gate for new technologies and thought process. One of such could be use of space technology which provides better vaccum and cleanliness apart from twenty four hours of solar energy for uv source (?) and energy. How is the question that needs to be sorted by experts but to explore possibilities I personally can try to invest around 100 million dollars in near future.
Could you make a video on wafer steppers? I feel like there is also a lot of RnD and Manufacturing detail hidden there, where else would you need to move something that fast but also so precisely at the same time.
@@dantesk1836 Part of the magic is clever use masks and etching that lets one use one feature to set the bounds of another feature (thus obviating the need for extreme precision on mask alignment). See for example self-aligned gates and multipatterning. Especially useful for the latter is atomic layer deposition which allows surfaces to be uniformly coated with atomic-level precision on the thickness. You still need to precise align the masks just not quite as accurately or as many times.
I remember reading an article years ago when EUV was still in early development about using particle accelerators. The author of the paper argued that it would ultimately be cheaper and more reliable than the crazy liquid tin Rube Goldberg laser.
A laser generally will be more efficient and reliable than a thermal approach to generating light, but the thermal approach is often the easy way that needs relatively low R&D investment.
author of the video didn't get the idea of the stimulated emission of radiation ( from 6.00 to 7.00) if radiation is emitted in all directions it is called spontaneous. and yes it exist all the time the medium is excited. for LaSER you have to redirect some part of this spontaneous emission with the mirrors to the laser mode. and then, when you have accumulated radiation in one direction, it starts to stimulate atoms to actually emit radiation in this particular direction ( mode). thats a laser. If you just collect spontaneous emission by means of light collector, like EUV source does, you just get incoherent luminescence, not laser. LASER actually stimulates to emit into the direction of the existing dense bunch of photons (and bunch of fast electrons actually give you this condition, no need for mirrors :-))
Cheeky video... sure still quantum tunneling still applies. Im curious if the higher resolution with smoother lithograph would mitigate it and allow much smaller nm gates.
That's HEPS Beijing Huairou Synchrotron.... but there is a Synchrotron SSMB EUV in Xiongan New Area by Tsinghua University that will be ready by 2026-27.
Greetings from the Center for Free Electron Lasers at DESY the home of the european xray free electron laser ! It's really cool to see FELs on youtube and crazy to have the video actually talk about SASE. I personally mostly work with undulator light from the petra synchrotron, so take what I say with a grain of salt.
Some issues with XFELs:
- you need really nice beam parameters for sase to work and seeded fel isn't feasible for xray photons
- they are fucking huge, european xfel is a >3km long line! compare that to high brilliance synchrotron with undulator beamlines like petra III at 730m diameter
- repetition rate is hard
some of the really really cool shit you probably couldn't talk about because of length:
- the light is in phase, so you can do some really fancy optics stuff with it and get intensity proportional to N² instead of N (where N is electron number)
- there are two different regimes for the self organisation. instead of injecting electrons that are monoenergetic but spread out in phase space you can also inject them with a wide energy spread in an ultra short pulse. That's one of the things people are trying to do with laser based acceleration systems.
Ultimately I'd guess that FELs won't do lithography in the super near future, just because time averaged intensity matters more than peak intensity. European XFEL SASE1 has 27kHz time average repetition rate with 4mJ energy per pulse. That's only slightly above 100W! From a billion €, kilometers long device that only went online a few years ago. Petra's P61 beamline alone puts out more power than that (I've seen samples with holes melted through them by that beamline), admittedly that one has a wider spectrum, but you get what I mean. The size / capital investment to power ratio is pretty bad for FELs still.
source: www.xfel.eu/facility/instruments/spb_sfx/instrument_design/index_eng.html#:~:text=These%20pulse%20trains%20can%20be,repetition%20rate%20of%204.5%20MHz.
The MuCLS is an even smaller sychotron-like X-ray source, just 5m x 3m, but I have no idea if it's suited for Lithographie. Still an interesting look at the other side of the extreme for monochromatic X-ray light sources.
Normally, coherence is undesirable for a lithography light source
Very cool!!!
@@stormsj SSMB?
@@ArawnOfAnnwn Steady-state micro bunching, which to put simply is a way to make synchrotron generated light useful for lithography.
You mastered presenting technical topics. Rise and fall of inflections, jokes, etc etc. night and day difference from your dry early videos.
Hell ya buddy, thank you for teaching us so much.
@14:03 “the current LPP approach uses 1.1 megawatts for 250 watts of EUV power” really put things back into perspective
Yeah current EUV is super duper power hungry because the reflective optics they use to bounce the EUV around for photolitho only each reflect ~70% IIRC, and after bouncing around a dozen or so times that means that only a super tiny amount of the EUV generated actually makes it to the wafer :(
Repurposed hadron colliders for purposes of manufacturing semiconductors is dope🔥
Hearing about the insanity behind previous generations of fabs was crazy when I first found this channel but now imagine saying to a random person “yeah that phone chip was made with a particle accelerator”. Crazy tech
Look up Ion implantation. Its a particle accelerator firing toxic chemicals to manufacture semiconductors.
OMG I'm gonna come my nigga one😮😮😮
Synchrotron radiation has been around for what 6 or 7 decades? Sure, the machines are big but there's really nothing stopping anyone from making them smaller and more efficient.
Synchrotron radiation is smaller than 13nm and has properties wholly different than it's neighbor EUV.
Electrons are much less massive than hadrons so simpler accelerators are better
Free electron lasers are a really cool diy project. Deadly high voltages, vacuums, strong magnets, blinding laser radiation, and extremely dense beams of beta particles. Don’t know why they’re not more popular.
Also the beam dump for an electron beam will just be some conductive surface that’s connected in a circuit back to the other end of the high-voltage source that accelerated the electrons out of the electron gun. Charges need to flow in a circuit, though when you’re dealing with free charges it can get kinda strange with net charge buildup.
Net charge build up? Get a bigger net. So to speak.
I guess it depends on the beam energy: You may require *a lot* of shielding around your nice conductive beam dump :)
Actually, I wonder about the return circuit. E.g. for SwissFEL, the electron bunches are 200pC at 100Hz repetition rate, that's 0.02µA. You don't need a special return path for that, but grounding is of course essential.
They don't fail in spectacular enough ways for UA-cam.
I'd love to find some resources to make a diy fel
This looks quite good: High-power EUV free-electron laser for future lithography by Nakamura et al (2023); here's the abstract:
The development of a high-power EUV light source is very important in EUV lithography to overcome the stochastic effects for higher throughput and higher numerical aperture (NA) in the future. We have designed and studied a high-power EUV free-electron laser (FEL) based on energy-recovery linac (ERL) for future lithography. We show that the EUV-FEL light source has many advantages, such as extremely high EUV power without tin debris, upgradability to a Beyond EUV (BEUV) FEL, polarization controllability for high-NA lithography, low electricity consumption, and low construction and running costs per scanner, as compared to the laser-produced plasma source used for the present EUV lithography exposure tool. Furthermore, the demonstration of proof of concept (PoC) of the EUV-FEL is in progress using the IR-FEL in the Compact ERL (cERL) at the High Energy Accelerator Research Organization. In this paper, we present the EUV-FEL light source for future lithography and progress in the PoC of the EUV-FEL.
"REDACTED" 😂 I laughed so hard and abruptly I scared my 🐈
I suggest you looked at TsingHua University's SSMB (Steady State Micro Bunching) to generate DUV, EUV and soft X ray.... all in one. China is already building an EUV factory at Xiong'an New Area in Heibei. 😁
Yep and per another comment here that likely has more time averaged power than FEL anyway as that seems to be main reason for study of SSMB... increases time averaged output.
Definitely..but he is Too "Reducted" to tell Us...😂😂😂
Talking about getting the "Semiconductor Avengers" together, this at this point start to look even more complex. On the level of fictitious Black Mesa research complex. This will cause a large overhaul on floor planning too when trying to protect equipment from disaster scenarios (floods and earthquakes) and floor costs when putting in laser redundancy. I can almost here the quote: "I never thought I'd see a resonance cascade, let alone create one." 🙃
"Gordon doesn't need to hear all this, he's a highly trained professional!"
6:00 When I was a teenager back in the late 70's, I wrote out by hand, from the Encyclopedia Britannica Jr. the section on Laser. I remember Laser: Light Amplification by Stimulated Emission of Radiation. Please don't omit the "Radiation" part.
I feel like this will be super popular when we develop space based semi conductor fabs.
They had one of these at the uni I worked at, or at least they did research on it back in the 90s to 2000s. It was able to emit a few mW at a 5 mm wavelength. They had some mathematical modeling for reaching kW powers but I'm not sure if they got that far before the building was demolished. Removing the room it was in reportedly took a week because it was absolutely radiation hardened and had very thick, reinforced walls.
@mz4637 the transmission needs more respect.
Electrons are not attracted towards the magnetic poles :) their path is just curved perpendicular to the magnetic field (the oscillations are actually in the plane perpendicular to your undulator picture at 7:55).
I like the idea of using particle accelerators to create light at the energy needed for future lithograph. Some particle physist are working on very small particle accelerators that are the size of your hand. Perhaps this avenue can enhance our understanding of the quantum world.
I remember seeing the first cyclotron ever made (on exhibit at Berkeley). It's just 11 cm across and accelerated protons to 80 keV.
More on this pocket accelerator stuff?
@@Cristopherdreamer _(Note: Given the small size I'm assuming these are cyclotrons, synchrotrons, or the like.)_
The typical limiting factor in how fast one can accelerate particles is the magnetic field strength. The faster the particle moves the more inertia it has and the more force that is needed to nudge it into a curve. The tighter the curve the stronger the magnetic field needs to be. The first cyclotron could only achieve low energies (80 keV) because it used magnets recycled from telephone switch relays. With modern superconducting magnets one could achieve much higher energies. At some point though the limiting factor becomes the material strength of the magnets; too high a field strength and the magnets will tear themselves apart. The only solution then, barring advances in material science, is making the ring diameter larger.
I've glossed over/skipped a lot of details but that's the gist of it. Maybe the OP could provide a pointer (DOI) to some papers covering current research in this area?
So not long until pulse rayguns
Sadly, when physics is involved, bigger is almost always better
You might want to take a look at new particle acceleration techniques based on lasers. It has the potential to do in a few meters what a linear accelerator takes a kilometer to do. It would not help with the undulator side of things, but it would help reduce the overall size of the FEL machine.
plasma wakefield accelerator? sounds super exciting
Great explanations, as always. Yet, in the undulator the magnets don't attract or push off electrons rather that deflecting them othogonally (in the picture to or from the viewer of the picture) to the magnetic field and orthogonally to the electron's trajectory. The resulting acceleration effect would be the same.
starts with a banger “recently I’ve been thinking about TSMC’s N3 process”
I learned the physiks behind a stimulated emission differently then your explanation. That it doesn't release a photon in a random direction but rather that if an atom gets hit by a photon while having one of its electrons exited at the exact same energy state then the incoming photon, it releases a second photon with exactly the same frequency and in the same direction then the first one.
This is crazy cool! Probably the future of photolithography!
Can you imagine that once free electron laser is down, you will also bring down a whole fab
I first heard about free electrón lasers in the context of light sources that china may use for euv machines, so i am very interested about the topic
Sinchrotrons and free electron laser are both simpler than tin láser produced plasma, so Many are speculating China Will use such a technology
Most important difference between plasma 13.5 nm light source and 13.5 nm laser (any type, but we can do it only on free electron laser actually) that is the first one is non-coherent. In properties way it is odrinary light source. Laser light is coherent.
Most important laser part is resonator. With resonator light is producing in entire volume of gain medium. Photons emmited with act of spontaneus emmision have random vetor, they goes into random direction. Photons emmited durign stimulated emmision have this same (all) properties - also direction of propagation. Light moving parallel to resonator axis are amplified (because it bouncing many times form one mirror to another, and in each pass it is amplified by gain medium) - and all this photons can have this same properties - frequency, phase, direction of propagation, etc. This means this light can be focusing into very small point - smaller than cross section of gain medium and actually it can reach two diameter of light wave lentgh.
Parallel beam also mean we can reach high NA.
"Sir, we need a new machine for our production line."
And how much would that be?
"Well it's a free ele-"
Perfect! Just get it done.
Impressive to say the least. I hope they build one in NY. Worth the money if you ask me 4 to 5 fabs running off of one laser is a game changer.
A magnet doesn't attract electrons. Electrons are deflected by the magnetic field perpenticularly to the magnetic field, not towards it or away from it.
So in your picture they would swerve in and out of the picture plane.
An electron travelling straight down the centre wouldn't be accelerated at all because it's parallel to the field. Off-center electrons would oscillate like an airplane changing the direction of rotation mid-barrel roll
Well you might be right, but Sulla>Marius so ur opinion is invalid.
@@Danji_Coppersmoke you almost made my day
I was wondering if “perpenticularly” was a word I’d never heard of. I tried looking it up and didn’t find such a word, so far.
This has me thinking that you perhaps you meant to type “perpendicularly”?
@@shanent5793 I don't think you have the correct magnetic layout in mind. In the layout shown the magnetic field lines will go up and down alternately, and the electrons will in turn curve in and out of the screen plane. This is also true for electrons moving straight down the center of the undulator.
A couple of days ago China announced its building a cycotron as big as two basket ball courts for chip manufacturing.
Synchrotron SSMB EUV in Xiongan by Tsinghua University to be ready by 2026-2027. Epic.
What in the USA doing? Biden pushed through billions of dollars to chip companies. How many Synchrotrons are in the plans?@@MrSpiritmonger
Wonder how big gonna look.
I whould love to see a video about this new ssmb technique China is supposedly investing. Building a particle accelator for lithography, really? Can it be cost effective, compared to standard euv?
Could be... a single EUV scanner cost $300-400M dollars, whereas a Chinese synchrotron particle accelerator can cost $200M but produce dozens of EUV beamlines.
Except China was not given a choice between cheap advanced chips made by foreign tin EUV and expensive advanced chips made by domestic particle accelerator EUV. China was given a choice between expensive advanced chips made by domestic particle accelerator EUV or having no advanced chips at all.
@@Stroporez domestic LPP EUV prototype by CIOMP will be ready by end of the year or early next year, mass adopted within 3 years.
@@MrSpiritmonger they run about $50 million, the vacuum control system I've had to design for the AMSL EUV is around another $5 million, and then there's other support tools that are also required, and I couldn't tell you what they are since I've had no involvement on that side of the tool yet.
Odd how chip sanctions pushed China to go all out in the development on this technology. I remember how Russia sending rockets into space got America's space technology development go all out as well.@@Stroporez
The US Navy spent a lot of money on FEL (500M?). Insurmountable engineering problems scaling it up. Of course, that was for directed energy. Perhaps for this specific application it will work.
FEL beams in the Navy effort were particularly nasty in terms of power density, simply ate the optics alive.
Are you sure you're not talking about the US navy's electromagnetic rail gun? Electron beams for directed energy weapons don't work. The electrons will interact with the atmosphere. In space they are bent by the Earth's magnetic field.
The SSMB (Synchrotron based Steady State microbunching) approach to generating EUV being pursued by Tsinghua University deserved a mention.
Unlike a Free electron laser which has high peak power with low repitition rate or a normal storage ring synchrotron which has low peak power with high repitition rate, a SSMB based synchrotron has both high power and high repitition rate.
If successful, SSMB EUV radiation would be far more coherent, tunable, powerful and cleaner than anything LPP can hope for. Tsinghua has promised a prototype by 2026-27.
Any information on why the various approaches are favored?
@@msimon6808 SSMB is optically and mechanically much less complex than LPP. ASML didn't take this approach as it was discovered much later and remains kinda experimental plus the logistical problems that comes when you have to force chipmakers rebuild their multi-billion dollar fabs around your synchrotron facility rather than just shipping a machine to wherever their production lies. But there are lots of advantages to the SSMB approach:
-High average power: the power aimed is > 1 kW per tool, each facility should be
able to incorporate multiple tools.
-Narrow-banded and collimated: the radiation spectrum bandwidth is < 1% and has
a well collimated angular spread ≤ 0.1 mrad, which should help to reduce the number
of reflection mirrors and thus increase the EUV power transport efficiency
-Truly continuous wave: the temporal structure of the radiation is truly CW, this
minimizes the chip damage problem;
-Clean radiation: the radiation is clean and carries no debris, so that mirrors do not
get contaminated and do not require frequent replacements during operation;
-High stability: storage ring source has the advantage of a good stability, which helps
to increase the machine availability and qualified production rate;
-High efficiency: higher wall-plug electricity to EUV conversion efficiency than LPP
source expected;
-Reasonable price: storage ring technology is mature and the average price for each
tool is expected to be reasonable;
-Good scalability: easy to scale to shorter wavelength
If Tsinghua can pull it off, it would be absolutely revolutionary.
@@xBlackWind Thank you very much. My son works (EE) at a company that employs a Belgian accelerator to produce isotopes. I sent him this article. And since electronics interests you - I did something about switch debounce. Hardware and software.
I have designed a SPST SR Latch debouncer. It tells you - like the SPDT SR Latch circuit - as soon as the switch starts changing state (plus deglitch time). Either opening or closing.
spacetimepro.blogspot.com/2023/09/switch-debouncer.html
The calculation wasnt in favor of SSMB the efficency is just too low...
@@fromfareast3070 what calculation are you talking about? SSMB light source power is almost an order of magnitude greater than that of LPP EUV. Modern LPP EUVs output an average power of 350W while SSMB can easily reach an average power >1kW with far lesser optical complexity.
Btw the newly released iPhone A17 MLB uses TSMC’s 3N(B) process node.
Lyncean and Paul scherrer institute XIL-ii would be notable mentions for other attempts at this technology
Another awesome video.
Can’t believe you passed on the opportunity to drop a Dr. Evil meme when introducing the topic of “Lasers”.
Thanks for the analogies used in this video. I will use and credit your video in a part of training that I do on how coherent optics are used in high-speed networking above 100Gbps Ethernet.
This channel is brain growing! Thanks
Thumbnail at a glance made me think this video was about toy trains lol
You weren’t too far off, lol.
Same 😅😅 looked like a Christmas train set
the series of bunches of particles going through an accelerator is called a train :D
great video, it for sure that many company like asml tsmc gf samsung are probably research in FEL for decade.
Even longer
Considering the size of the hole that they have dug up over there in Hsinchu... god knows what they can put down there... I saw it first hand back in May and I don't know what they were digging but from my what brother in-law told me, before the hole, there was a literal mountain in its place. It's insane!!!
I have no idea what 90% of the videos in this channel is about, but I still watch them.
It's like starting a movie 30 minutes in.
Lol... Ask ChatGPT the stuff you don't get, or find a source that explains the parts that feel like you came late to your university lecture and you know not what they are talking about. If you don't like math, just find a source that doesn't mathematically encrypt its content. Trust me, nowadays if you can't understand something, it is by choice. Though I don't claim certainty, for I don't know your situation... But the human brain has too much computation power for anyone to excuse themselves as too dumb. It could be a software issue, (I mean, someone that can't use their brain to its potential can't expect to harness its full capabilities), but statistically speaking, I think no one is hopeless, for as they say, it is never too late to learn.
Contamination of the incredibly sophisticated optics are more than enough reason to move on to free electron lasers. No one wants a dirty neighbor. Improved energy efficiency and a more robust upgrade path are icing on the cake.
Your work is much appreciated! Your EUV and foundry tech are my absolute favorite! Keep them coming!
Thanks!
less reflection using less mirrors is the ongoing project now
Love this one. Used to work with synchrotrons, and while your physics explanations have minor issues, the business history of this is fascinating. Superb work.
Hi Asianometry,,,thanks for your analysis of modern Lithography!
As a vacuum-electron device worker (CPI Microwave) I find this baffling and yet delightful. After years of telling ol' vacuum tube guys that our industry was not only shrinking but shrinking towards zero and would die, now solid state electronics engineers think they may need a linear-beam vacuum-electron device to keep their industry moving forward. This even as GaAs and GaN devices have largely finished moving the power-frequency curve up and to the right, and we still have lots of applications-- replacing your microwave oven's magnetron (a type of vacuum tube) with a GaN device would more or less double its power consumption, for example. I feel now like there will be a need for tubes forever. If you need some LINAC klystrons for a giant $400m FEL we'd be delighted to provide.
I'm older than transistors. I can relate.
6:38, a slight correction on the meaning of "stimulated emission." Stimulated emission refers to a process where an excited electron can become sort of a photocopier for photons. When the electron is excited (technically I think this is called the Avatar state), and an external photon interacts with it, then the electron is prone to emit its energy as a second photon traveling in exactly the same direction with exactly the same wavelength, phase and polarization as the incident photon. This is why the laser emits all of its photons in a narrow beam at effectively a single wavelength.
This!
Yeah, really important thing which makes laser not just fancy lamp. We get over this so much on the lectures I usually don't even pay attention towards explanation in videos. Also he missed "Radiation" then explaining acronim/word "Laser", which little weird.
Slight correction to the slight correction: The wavelengths must already match anyways, they are a consequence of the energy level difference. But yes, the phase e.t.c. make it special.
@@entcraft44Thank you.
SASE is vaguely similar to the "multiple pendulums on a table self-synchronizing" effect
Exactly. They like humming in unison. It makes them phase coherent.
In physics it is called entrainment.
1.) Nd lasers are not commonly pumped by LEDs, in fact this has only been demonstrated in academic settings.
2.) Stimulated emission does not refer to any emission from an excited state as you explain, but is instead the result of an incoming photon stimulating an electronic transition and subsequent release of an additional photon with identical phase. What you effectively described is spontaneous emission.
A for the effort, D- for the correctness of the explanation...
damn, the future is exciting!
So what the semiconductor industry ultimately ended up with is, in essence, a particle accelerator... What the heck.
Particle accelerator doing the thug shaker
*its all coming together*
sooner or later machine learning is gonna end up there as well
@@Summanislol why do you have a picture of Indra as your pfp? 😂
@@LanaaAmor because he did cool things
They might _only_ be etching with electrons, but soon they'll be printing on wafers with boson inc.
I received notification of this video via the Asianometry Newsletter. Momentarily, I thought I had the opportunity to win a *free* electron laser. Disappointment.
7:43 Stanfurd University? Are you a fellow California Golden Bear Mr. Asianometry? Great video.
can't wait to hear about X-ray lithography
It will never be disappointing to stop using the LPP method. Tunability of the wavelength and higher power ceiling for the light source are groundbreaking changes. I hope we see the FELs in production before the end of the decade. That would be amazing.
The mirrors have to be specially constructed for the specific wavelength if I understand it correctly. Normal mirrors don't work for that short wavelength and that is one of the problems. So the mirrors are made with thin layers making the light reflect with a big loss. The layers have to be tuned to have the exact thickness depending on wavelength.
@@lubricustheslippery5028 The losses reduce as layers are added.
@@lubricustheslippery5028 Yes, but you could still have a single light source for multiple machines operating at different wavelengths, if you partition in time. Even if it was just for development purposes, being able to operate and tune a newer generation machine without having to build a whole new light source would be a big deal. Simply being able to experiment with different wavelengths cheaply would be a massive deal in it's own right. The different layer thicknesses for the mirrors would not be something manufacturers would need to develop new technology for, and while not optimal, the resists that work for 13nm should do ok for other high energy wavelengths. You could say 'lets try 40nm', adjust the light source, and switch out the mirrors, and it might just work.
The possibility to introduce a cheaper alternative to EUV is also made available my this type of light source. Nothing is stopping us developing machines around something like 40nm light, for example. These would have far less stringent requirements in terms of mirrors and masks than 13nm, while also reducing the energy doses required 3x. They may not quite hit the same quality* as 13nm, but potentially could be much cheaper to operate. If a fab were able to shift 60% of steps to a cheaper process that would be massive in terms of output cost. Fabs do miracles with 193nm, so an intermediate step might be warranted, if it could be developed cheaply.
* It is possible that they would, or better. Photoresists for EUV are activated as much by secondary effects as the light itself, so reducing the energy of those secondary effects might allow the lithography to be more accurate. The reduced requirements from mirrors would also make getting higher NA with the same manufacturing precision possible, clawing back some of the accuracy almost for free.
@@lubricustheslippery5028 Yes, but you could still have a single light source for multiple machines operating at different wavelengths, if you partition in time. Even if it was just for development purposes, being able to operate and tune a newer generation machine without having to build a whole new light source would be a big deal. Simply being able to experiment with different wavelengths cheaply would be a massive deal in it's own right. The different layer thicknesses for the mirrors would not be something manufacturers would need to develop new technology for, and while not optimal, the resists that work for 13nm should do ok for other high energy wavelengths. You could say 'lets try 40nm', adjust the light source, and switch out the mirrors, and it might just work.
The possibility to introduce a cheaper alternative to EUV is also made available my this type of light source. Nothing is stopping us developing machines around something like 40nm light, for example. These would have far less stringent requirements in terms of mirrors and masks than 13nm, while also reducing the energy doses required 3x. They may not quite hit the same quality* as 13nm, but potentially could be much cheaper to operate. If a fab were able to shift 60% of steps to a cheaper process that would be massive in terms of output cost. Fabs do miracles with 193nm, so an intermediate step might be warranted, if it could be developed cheaply.
* It is possible that they would, or better. Photoresists for EUV are activated as much by secondary effects as the light itself, so reducing the energy of those secondary effects might allow the lithography to be more accurate. The reduced requirements from mirrors would also make getting higher NA with the same manufacturing precision possible, clawing back some of the accuracy almost for free.
@@lubricustheslippery5028Carl Zeiss would probably figure it out
7:44 "Stanfurd" eh? I love it.
{^_-}
I like the Red Head In Harmony.
I guess the people naming it must have noticed the irony: the "free electron laser" will cost half a billion dollar :D
the electrons are free; the equipment on the other hand...
It was banned in China for years, and scientists jailed, because it had "free" in its name.
The electrons are free. If you want a bag to put them in it'll be an extra USD500M.
@@dpride00 GNU electrons.
Perhaps Red Hat would like to get involved?
Can't wait for 2024 intel to bring the heat on TSMC.
💥🔥🔥
That Intel wait has been awfully long in the past…. and didn’t always meet expectations…
Expect 2025 might be when the heat comes…
@@AC-jk8wq I can live with that. The industry desperately needs serious competition. Global foundries and Samsung are lagging way behind
I like how there's an organization named KEK, KEK!
Next video, bro will upload how to build your own EUV lithography machine 🗿🗿
I have been wondering about Samsung struggling with their 3nm manufacturing and even TSMC seems to have problems with 3nm such as low yield, which could be because of early technology but anyway. It seems that EUV lithography is running out of steam, and Ray Kurzweil says that when that happens, historically a new paradigm replaces it and continues the accelerating progress.
It doesn't matter if the stockastic outcome is random or chaotic - applying the binary distributiton it will work 50% of the time.
... because either it works or it doesn't ... 😁
lol
But then the yield would be 50%?
I'm so used to people zoning out when i get nerdy,..... its nice to come here and feel like a layman and just barely being able to digest some of these processes...it gives me that feeling of awe and wonder again..... thank you
you are talking to the wrong people
@@MegaChickenPunch That reminds me of the party scene from the first _Thrilling Adventures of Lovelace and Babbage_ comic.
Amazing stuff! 🎉😊
You should discuss plasma generators used in semi conductor manufacturing. Research the company Advanced Energy.
Channel Host: Taiwanese
Accent: American
Metal Pronunciation: Aluminium (correct)
I love it.
Heyo! Just wanted to mention that Apple's M3 SoC chip is produced using TSMC's 3nm process.
Wiggling electrons with EM field sounds like the ultimate method for generating EM radiation of any desired wavelength.
I was thinking that.. many potential applicanrtions
Synchrotrons use that exact feature to their advantage for a wide variety of applications. In fact, most Synchrotrons are capable of sweeping across ranges of energy for materials studies, and this capability let's you get molecular structures and all sorts of other information about a sample. In fact, Synchrotrons are better at doing this energy sweeping than free electron lasers (at the cost of maximum intensity).
For example, x-ray absorption fine structure microscopy (EXAFS or XAFS) and x-ray absorption near-edge structure (XANES) can be used to find the complete molecular structure of a material in a matter of minutes, and you can automate beam-lines that do this. You basically sweep across a spectral line of an atom (one in the x-ray range) and take an absorption spectrum. These spectra have wiggles in them due to atomic bonds and such, so you can do some math (a fancy fourier transform) and find the distribution of atoms around that specific element in a sample. The advanced photon source (APS) at Argonne National Lab does this and a whole host of other things.
Most FEL and SLS (Synchrotrons light sources) work in the GeV x-ray regime (1000x more energy than the MeV range discussed here) because that energy range is more useful for research purposes.
Also, there is a program at Arizona state University to make a compact FEL by using an intense optical/IR laser as the undulator instead of a magnet bank. I imagine the future of high-energy lithography machines will use plasma Wakefield accelerator cavities and those compact laser undulators (if they can get them working, of course).
Synchrotrons and FELs are absolutely incredible machines. Sad that they're so expensive
Gee, I don't know... I think you're making it sound way too easy.
early development of radar and microwave is exactly that
Sadly, no matter how hard you wiggle them, hard x-rays are still off the charts, although UV is possible.
You probably know the most about chip fab and lithography as well as specific model chip fab equipment than anyone bot specifically employed in the industry.
There is nothing mind blowingly new here, not to disparage Jon, he is excellent at bringing down the big concepts to us noobs. But if you start to read the papers and interact with people in fabs you learn that a lot of this is just industry happenings. 🤣
Circulate like a 7-11 Slushee machine? Too Damn technical.
I suddenly have the urge to go to 7 11 for a blue Slurpee.
Lots of UA-cam videos on China’s SSMB EUV in the last two weeks.
How are the buildings holding up?
@@msimon6808what do you mean?
Nothing like problematic swagger to set you back a few quarters
This channel is amazing.
I have learned so much.
Keeping it real. Thank you!
Honestly I am interest in these topics and I love what you are talking about but perhaps the most important thing is that I like how serous and rigorous you are about all those topics the part which is the most important and to which you have always seriously adhered in a rigorous manner is the humours part of the presentation... I can not explain in words how impressive this is as it is like the seriousness brings us to a higher level then the fun part, the release or the acronym *R* (meaning the release) is when we loose the energetic state and get to the enlightened state acronym *G* (meaning ground state)... thanks for every an all videos and all the efforts and dedications...
That "redacted" was amazingly well done.
What's that mean?
Which fab was it? SMIC or Intel?
@@guill90 My guess would actually be SMIC or Global Foundries, but there are a fair number of other obscure options so I would give it less than a 30% chance that one of my guesses is right.
My guess would be Intel they have a long history of always choosing the 'Betamax' option when the rest is the industry decides to choose 'VHS'
@@guill90probably Intel, dumbest of the dumb
Next you need to cover China's SSMB EUV source efforts...and that approaches details in general.
Haha no one is going to give China's efforts a fair shake in YT channels like this. But I am glad that a lot of people are calling out for China's FEL efforts and they are probably at the leading edge for this application. Shows that the western propaganda about China is not as tight as they will like it to be.
@@gelinrefira Yeah, I don't know if or why you'd try to control information about China's progress if there is interest in handicapping that progress. But silly efforts to delete Chinese business from western helpers is starting to look like it could totally backfire and leave China at the leading edge without even needing to try to take over Taiwan to get there. Not sure it helps or hurts risks for Taiwan in Grand scheme, but seems it mostly hurts if China is made more self sufficient by outside pressures that force it.
@@stormsj That's the thing that westerners are chronically ignorant or deliberately refused to understand on how China looks at the Taiwan issue. I think it's deliberate because they are terminally unable to accept that China is doing great and is innovative, creative, dynamic and powerful. So of course they are going to control information flow. If those Americans who like to denigrate, insult and look down on China will to visit Shenzhen one day, they will either be shell-shocked, or gibbering in terror. "The white man is not on the top anymore!!!"
Th thing is China doesn't NEED Taiwan to keep growing and progressing. What Taiwan represent is a wound that have not been healed. It's a political issue, not an economic one. But the Chinese are in no hurry to reunify with Taiwan and they ultimately don't even need TSMC. All China really wants is peaceful reunification and they can wait 10, 20, 50, even a 100 years to do that.
But everyone and their dogs in the west talk like China can never have advanced chip technology without Taiwan or the west. Taiwan is a small little province. If you really know China, then Taiwan really is more like a 2nd rate, chronically mismanaged provincial backwater.
All US sanctions do is solidify China's conviction and determine, and accelerate their progress because they can centrally planned to focused on project deemed as strategically important. Now you just force China to push their best and brightest to come up with innovative solutions to solve intractable problems. And everyone who bet against China from succeeding on their national project always lose.
@@stormsj At least my opinion, as someone who isn't pro China non particularly pro US. Having tight competion between superpowers in this case US and China reduces the risk of direct military conflict because stakes are to high and risks aren't worth it. But if one side falls behind or significantly threatened in some other way, risk of war increases to dangerous levels because either side might see it as a final option to regain their position.
@@vitaliibraslavets this is probably true as far desperation of one side falling, but I think risk of war is already heightened even if efforts to complicate China's competitiveness fail and it becomes more self sufficient as it has disconnected them from a need to cooperate with the west generally in this effort to isolate them. But yeah, even if efforts tend to have intended effects there may be some assessment there is a last chance to take Taiwan or the like as they lose capabilities. It's all been generally ignorant way of going about things ever since Xi took over. I totally get why the rest of the world has reacted to China post Xi as they have... but this reaction has had the effect of strengthening rather than weakening him. They needed to be more savvy and directed annoyance more clearly toward the top of the CCP and been more receptive to corporations and local governments in China willing to stand up to what has been happening under Xi.
If only the Buddha had known there was an easier way to get to that higher state of being.
Geezzz.
There's a Nobel Prize waiting for someone who can determine the maximum tensile strength of a metaphor.
thanks for that redaction!
N3 BASED FR
How are the mirrors and masks going to stand up to coherent illumination? Is it necessarey to diffuse the light to prevent speckle, or does it average out over many pulses?
Perhaps rotation of the mirrors, but that would be incredibly complex and possibly a nightmare to keep calibrated
Any chance of using wakefield style accelerators to reduce the giant size needed and maybe put it into the lithography machines? Or do they not allow the needed preciscion?
Wakefield is still experimental
Great work. Thoroughly enjoyable!
Wow, a free electron laser? Getting more sci Fi every year....cheers!
Heeey, I caught that REDACTED. ;*[}
Good stuff. With your videos, I really feel close to the bleeding edge of technology.
Based Video.
Using a particle accelerator to do lithography is metal as f_ck.
Wow, these machines are crazy inefficient...
Having included particle accelerator in semiconductor chip fabrication has opened gate for new technologies and thought process. One of such could be use of space technology which provides better vaccum and cleanliness apart from twenty four hours of solar energy for uv source (?) and energy. How is the question that needs to be sorted by experts but to explore possibilities I personally can try to invest around 100 million dollars in near future.
7:42 "Stanfurd"! = Stanford university
Free electron laser with every purchase! Ba-da-DOMP-tssss!
Free Electron Laser EUV sounds like a clone of SSMB-EUV being developed by Tsinghua University.
Could you make a video on wafer steppers? I feel like there is also a lot of RnD and Manufacturing detail hidden there, where else would you need to move something that fast but also so precisely at the same time.
Also wafer aligners. Seems fascinating to me how you can put something so precisely, to a fraction of a nanometer, on top of each other.
@@dantesk1836 Part of the magic is clever use masks and etching that lets one use one feature to set the bounds of another feature (thus obviating the need for extreme precision on mask alignment). See for example self-aligned gates and multipatterning. Especially useful for the latter is atomic layer deposition which allows surfaces to be uniformly coated with atomic-level precision on the thickness. You still need to precise align the masks just not quite as accurately or as many times.
I remember reading an article years ago when EUV was still in early development about using particle accelerators. The author of the paper argued that it would ultimately be cheaper and more reliable than the crazy liquid tin Rube Goldberg laser.
A laser generally will be more efficient and reliable than a thermal approach to generating light, but the thermal approach is often the easy way that needs relatively low R&D investment.
Yea , but things have changed a lot over past 20 years.
The smartest man on youtube. I will click like.
Most clever deer*
author of the video didn't get the idea of the stimulated emission of radiation ( from 6.00 to 7.00) if radiation is emitted in all directions it is called spontaneous. and yes it exist all the time the medium is excited. for LaSER you have to redirect some part of this spontaneous emission with the mirrors to the laser mode. and then, when you have accumulated radiation in one direction, it starts to stimulate atoms to actually emit radiation in this particular direction ( mode). thats a laser. If you just collect spontaneous emission by means of light collector, like EUV source does, you just get incoherent luminescence, not laser. LASER actually stimulates to emit into the direction of the existing dense bunch of photons (and bunch of fast electrons actually give you this condition, no need for mirrors :-))
you should do a video on NIL and xray lithography. nobody talks about them but they compete with EUV.
Pretty sure he did a video comparing the various options and why EUV was chosen
0:21 Yep it's based
Cheeky video... sure still quantum tunneling still applies. Im curious if the higher resolution with smoother lithograph would mitigate it and allow much smaller nm gates.
China already has a working prototype, allegedly in Huairou 😮
That's HEPS Beijing Huairou Synchrotron.... but there is a Synchrotron SSMB EUV in Xiongan New Area by Tsinghua University that will be ready by 2026-27.