Huygens Optics - 2023-12-21
A visit to ASML with a deep dive into diffraction and imaging: 0:00 Intro 3:14 How big big are chip patterns nowadays? 5:00 Arriving at ASML Veldhoven 5:50 Interview Sander Blok part 1 7:40 About diffraction and image formation 9:36 Fraunhofer (far field) interference / diffraction explained 11:15 Diffraction on photolithography masks 15:11 About critical dimension 17:27 Example of computational photolithography 19:23 Interview Sander Blok part 2 21:46 EUV is difficult... by the way, the frequency of the tin droplets apparently is 50.000 not 15.000 per second. Animations were made using a Phyton script for Nvidia cuda, which was supplied to me by https://www.youtube.com/@DiffractionLimited The script with instructions to install it can be downloaded from github: https://github.com/0x23/WaveSimulator2D Third party images / video: Sleepy guy: https://www.youtube.com/watch?v=bBqLCrIwHbA Source mask optimization: https://opg.optica.org/directpdfaccess/837c889c-5b88-4dcb-8bfbdda02863dd9a_441841/oe-28-22-33371.pdf If you cannot access the article, search for the title in google: " Source mask optimization using the covariance matrix adaptation evolution strategy" Royalty free music used: Road Trip - Slynk Sharp Edges - half.cool Floating - Early Birds All imagery inside the ASML manufacturing facilities is stock video material shown Courtesy of ASML and IBM. ASML YouTube Channel: https://www.youtube.com/@ASMLcompany Want to work at ASML?: https://www.asml.com/en/careers/find-your-job Into viewing wave simulations? Nils Berglund has a ton of them: https://www.youtube.com/playlist?list=PLAZp3rbgWLo3VO2rqVKyL1T6DUmnDAaEN Did I forget anyone? Please let me know and we will work it out.
Alright these were the last trade secrets I needed to finish my DIY home EUV setup. I just have to find a nice value for k1 and I'm golden.
For EUV in your garage, probably you can be extremely happy with a k1 of 10000. :D
Awesome! What's your light source ?
@@Hippida a kid rubbing two sticks together really fast?
And of course it's got to be 3D printable so that all of us can make a copy and start a DIY wafer revolution ! 😂
14:17 As a telco engineer working with signal analysis I had an heureka moment. Basically thinking in fourier domain and radio signals, it is enough to recreate transmission with carrier and a single side-band and here we had something similar with light source positioned at angle to transmit 0th and half of 1st interference pattern through the imaging system. Thank You for the great content =) and merry Christmas
That's so true! All these fields are similar. The lens is simply a band pass filter with size of NA/lambda (in frequency space). By changing the incidence angle of the light onto the mask, you can change the position of the filter from minus half the lens size to plus half the lens size. As the diffraction pattern is simply a fourier transform of the electric field at the mask, it is easy to calculate the final image! You just have to iFFT(FFT(mask)*lens) and BAM! You're an imaging engineer!
@@DrBlokmeister🤯 it really is fourier sequences all the way down. Wheels within wheels...
Also didn’t think my basic understanding of “beamforming ” would apply, but shouldn’t be surprised that an array of light sources would not behave in a similar manner haha
@@DrBlokmeister Interesting! So I guess I can think of the lenses role in imaging as a convolution, since multiplication in k space is a convolution in r space.
I've never thought about optical imaging in this way, only XRay and MRI imaging (medical physicist by training), but I guess it makes sense that it's similar. If the lense is a rectangular band pass filter in k space, where would I find the corresponding sinc in real space?
@@chalkchalkson5639 I think you're right here. The lens has a point spread function and the final image is a convolution of the point spread function and the mask. I think it's a bit more complex in the case of off-axis illumination. This sinc function is the point spread function I mentioned. It's the response of the lens when you try to image an infinitely small dot. Since for an infinitely small dot, the diffraction pattern is infinitely wide, the entire lens is filled with light, the intensity being 1 over the entire lens. Therefore the image will be a fourier transform of this square.
This is a beautiful addition to the Asianometry content on Lithography. Really nice seeing visual representations of the things going on.
I have put Asianometry (read Chinaometry) on ignore.
@@3pan1Why? That channel provides useful info and makes some historical videos too.
@@ErnestasMage If you are able to read between the lines; something you are apparently not capable of, you would have intuitively felt and understood that channel is a quasi universal chinese propaganda instrument.
@@3pan1 Apparently just because the creator lives in Taiwan, he produces propaganda? I have watched numerous videos of his and found no inclination to believe that he supports that stuff, maybe it's just you, who only watched a little bit of his content, formed an inaccurate opinion?
@@3pan1 Watched alot of his video and i disagree, just because you cover a stories or a sector's progress doesnt mean he is doing propaganda. His video that tae place in China in the last 40 videos is 3(exclude those from Taiwan). Stop cherry picking to fit your delusional assumption, it only prove that you are a fool in disguise.
Worked for a year for a metrology system manufacturer, mostly phase shifting interferometry. Not only the production of these really small features is a challenge, but being able to visualize them and measuring their height was also incredibly difficult (at the required throughput)! The thin films mess up the interferometry signal, which threw off the existing algorithms to estimate the height. At a certain point the lateral dimensions of the features also started to become a challenge, as they diffracted the light used to visualize them. Everything's so cutting edge in this sector, it's awesome!
It is insane! When you work on the nanometer or picometer scale, suddenly everything matters! I never hear about the challenges that other companies have, so it's interesting to hear these things!
I work in metrology team at ASML. Indeed, being able to measue incredibly small features at high enough throughput is incredibly challenging.
@@irobot-ng6liyour lack of a command of basic language and grammar says otherwise.
Something interesting on our lens that I realized after the recording is the huge size of our field. If we take the half pitch resolution as a single pixel (10nm), our field size is roughly 2,6 million by around 0.1 million pixels, so if we make it into a camera, it would be a roughly 260 terapixel camera, give or take a factor of two.
so do you use any spherical optics at all?
and in terms of tolerance, what are we talking about?
ion polishing everywhere?
@@diegogmx2000 I don't know, and if I did, I'm not allowed to say. Zeiss makes the lenses, they are the lens designers and fabricators, so lens design is technically not an ASML thing. We only use the lens. Although in reality the difference more subtle.
@@DrBlokmeister damn, im still surprised how compartimentalized these things are, good ways to stifle innovation to keep monopolies
@@diegogmx2000 I think it's also a result of the complexity of these machines. I don't know a lot about the complexity of the wafer/reticle handling robots, I have a limited time and brain capacity. The machine is too complex to understand completely. I'm not an expert here, but I think this is also why ASML outsources for example this lens design and manufacture to the companies that are the experts in this field. We know roughly what kind of a lens we need, and let the experts decide how to make that lens. That way we get the best lens we can and can focus on our expertise.
Unfortunately, the field in the animations is quite a bit more limited.😁
While studying we went over some basics of euv lithography, but NEVER in such a high level of detail like you just provided.
Amazing to see behind the scenes how much this technology needs to account for.
This is the most informative channel I've come across on YouTube. Not only it's accessible, but is also rigorous enough and it doesn't shy away to show at least the easier part of the maths modelling the subject that you actually can learn something!
As a software developer, I am amazed that ASML is still pulling this off. About 25 years ago, there were projections that they would pretty soon need every SW developer in the country. That hasn't happened, but still ASML is turning out these ever more complex machines.
I have the greatest respect for ASML, it is without a doubt THE most technologically advanced company in the world, and a major driver in advancing the State of the Art for optics, mechanics, systems engineering, sensor & actuator tecnology, physics and SW. And electronics, obviously.
This video was a TREAT! ASML has always seemed like some ultra-secret black box of an organization. Very cool to see inside, meet people there, and get a deeper understanding of just how insane the tech they develop is.
Just an marvelous presentation. Listening to two dutch men speaking english, listining as an dutch man and understanding it all. MIND BLOWN. :)
Yay. I was hoping you would make a video like this. As a litho engineer this company doesn’t get enough credit as to its impact on our modern world.
Nice one, Jeroen. And thanks for advertising my channel! (featuring some optical fibers today).
Yes I saw. Nice high resolution. Did you modify the code in some way? I remember that in the past you had some trouble simulating high-frequencies.
@@HuygensOptics Not in any fundamental way. But I increased the resolution, so this may explain the difference.
@@NilsBerglundI knew I recognized those animations. I love your videos!
@@andrewcornelio6179 Thanks! In this video, Jeroen used another software to create the animations, though.
Ooh what a surprise! I've been a fan of you channel for a while now, but was surprise to see you make this video about where I work! I work on the team that makes the EUV light source. Suffice to say I have only a surface-level understanding of the scanner optics you spoke with Sander about, and I also usually leave talks about this subject agreeing "this is really hard stuff". But it is really hard in every part of the machine, which is why it is cool to work here.
Would love to see you make more videos like this occasionally when you get the chance to speak with engineers and scientists about their work. I really like the way you think about and present problems in optics!
Please send me (SBLS) a message and I'll arrange my famous lecture for your team if you're in VHV!
By the way, mad props for you source guys and girls. You guys do impossible stuff. Building steppers is something we've been doing for decades, but shooting steel-cutting lasers at supersoakered microscopic tin droplets flying through vacuum at 300 km/h twice fifty thousand times a second is completely insane. The fact that there isn't a conspiracy theory around this is a miracle. Hats off to you!
Thanks, I guess the work on the EUV source also has its difficult aspects but these would be more in the field of lasers, plasma generation and timing. Anyway, glad you liked the video!
Absolutely love this channel and was delighted to see ASML come up as a topic. From the little I know about it, it looks like one of the most important and most under-appreciated companies on the planet. The EUV light source seems like one of the zaniest components in the whole mind-boggling machine. Piece of trivia which I'm sure you know -- the holder of the patent on the liquid-metal-jet method of generating EUV is Hans Hertz, great grand-nephew of Heinrich, and professor at the Swedish Royal Institude of Technology. He's from a long line of physicists and I've wondered if his liquid-metal-jet technique is in any way connected to the fact that his father, Carl Hellmuth Hertz, held the patent on inkjet printing technology.
@@ps200306Supposedly they are indeed very similar problems, yes!
You need to control the volume very precisely, but also quickly, and produce a globule of the same shape consistently.
I forget which one, but one of Asianometry’s videos mentioned this.
For printers, MEMS is used. I wonder how it’s done for these machines! The working fluid is more difficult to work with, and its higher temperature, and continuous. Very difficult working parameters!
@@ps200306 I never heard of this. That is so cool!
You are by far the best educator of optical principles I have hear. I went to school for computer engineering and was able to make my own chips in our school's lab, and your videos are incredible.
What a beautiful video. Thank you so much. As a semiconductor wafer technology developer and optics enthusiast, I must say that you ticked all the tick marks. Thank you very much Mr. Vleggaar
this is great, definitely the best optics channel on youtube, and damn i would be sooo nice to take a look at that expecience center
really looking forward to the next videos on this
really to consider imaging as a consequence of the interference of the diffraction orders is just mindblowing yet makes things so much more intuitive once you incorporate it
Mind boggling is an understatement! I always knew the "shrink" created by EUV was extreme but man, your 1 micron to 12nm example really helped me
Another great video on an intersting subject. Like how you discuss topics at a high level of complexity in an stright forward and simple manner. The animations helped understanding a lot. Already looking forward to learn more super technical stuff!
Mind boggling precision, Thank you for this video!
Computational Photolithography. Something we couldn't even do without the GPUs from the previous generation
Is that a computation that parallelizes well?😂
@@salmiakki5638The light simulation step part is, iirc. But the steps before it that apply design rules so that it works well, are totally dependent on CPUs only as they’re too complicated for GPUs (way too much branching logic)
Itschain- cell-hainelopment process.
Chip calculates mask for even more powerfull chips ...
Skynet is coming ...
AI will prevail
I for one would welcome all the details from the ASML visit, speaking as a physicist who studied photolithography and quantum optics. Thanks and godspeed!
I would love to see a longer version of your visit. Thank you for your channel. It help me alot to understand light.
It's nice to see the optical process described in terms of waves instead of rays. That made the strange looking mask and light source a lot easier to understand.
I'd watch hours and hours of your videos. No matter how deep and technical because you are really good with didactic. Please, more!
The quality of your videos remains peerless. Not just on a streaming platform like YouTube, but in general, I think anyone would struggle to find such accessible yet detailed and insightful videos on this subject. Thank you.
Your comparison of a 1 micron feature to 12 nanometer features is great! I laugh out loud so amazing that it's funny.
Speak about finding borders of what is possible! The level of these guys and girls at ASML are unimaginable for us mere mortals. Moving around parts with several g-force and keeping it in distance within some nanometres is incredible, even for me as an mechanical engineer! Hats of for them!
Thanks! But even for us guys and girls, the stuff that we do is incredible. I focus only one tiny aspect, but when I speak to colleagues about other aspects, every single one comes with jaw-dropping things.
@@DrBlokmeister I think you meant “jaw dropping,” not “draw dropping.” Otherwise it sort-of sounds like you are talking about peoples’ underwear falling down.
While comical to imagine happening at ASML, I don’t think that was your intention.
@@WhatYouHaventSeen Ooops! 😅
A FANTASTIC video! These concepts are found in so many areas. Computational diffraction photolithography ~ holography ~ layered neural networks. Meso scale materials….so many connections!!❤
As you mentioned, what amazes me is that they can keep everything perfectly aligned across the full exposure area, such that everything ends up within, say, 5 nm of where it’s supposed to be. Mind-boggling!
Somehow that nm precision is never that amazing when looking at micro electronics (everythings is small so in relative terms it is not as impressive). But when you look at large physical systems moving at high speeds and distances for us humans, I am always amazed that it still as precise as some of the smallest things we can make
Most astounding! On numerous levels , , , thanks HO !
Your viewers are very fortunate that you are such an 'optics geek'. It is highly likely that your skills applied to the ASML motion system would also provide your viewers a fascinating peek into what I consider almost impossibly difficult mechanical repeatability. You hinted that the wafer experiences accelerations up to 30gs and yet each wafer has multiple exposures. And just imagine the thermal distortions affecting that 'brilliant' optical pattern generation system! That you deal with the esoterica in details that makes me excited when you post a new video. Congratulations on this one.
This is a magnificent video. The animations give a rare intuition for the optics and wave physics considerations that a scientist/engineer working in this field would use.
One of the very few truly high tech video’s on YouTube that doesn’t dumb down reality. So in fact all iPhones and Samsungs are made in The Netherlands…
Fascinating!
Is truly absolutely fascinating.
But have in mind that ASML has workers, engineers and brilliant minds for all over the planet and also the main components and technology comes from several other countries and without them this technology and machines will be impossible.
What this machines do and also the others thousand machines do to fabricated the logic and memory chips is absolutely the state of the art and are some kind of magic for an ordinary human being.
Funny, I just applied for a job at ASML. They're not far from me in CT, USA. Thanks for the overview! Just what I needed!
Good luck, hope you get invited for an interview!
@@HuygensOptics Thank you. Fingers crossed!
Good luck! If you get hired and are interested, reach out to SBLS. We can geek out over lens designs and imaging!
@@DrBlokmeister Awesome. Would be fun. I'm currently building a drift scanning camera for astronomy (my other hobby) so your thoughts would be much appreciated!
@vincei4252, ASML is on hire freeze right now, don’t get discouraged if you don’t get the position at this moment.
I loved everything about this video... simply amazing.
Thank you! The overlay of modern chip features on an i486 via blows my mind - 1000 times smaller linearly, 1M smaller by area?
I bet many of us would be happy to see a couple of hours of your un-edited chat with Sander, even if we won't understand most of it.
Thanks, people like you make the world a better place. Learned a lot since I discovered your channel.
Wow! I appreciate the simplification. So many things I had not considered go into wafer manufacture. Truly mind blowing.
This is the BEST youtube channel. High quality content of rare topics that have interested me all live but didn't pursue as career.
This technology is just unbelievable.
And the video to present all of that is insanely good and clear. Thank you sir.
I do not know how to thank you for all of these useful materials. I deeply appreciate you. Please keep going…
This video was top tier and a very much appreciated deep(ish) dive into modern nanoscale technology. I can appreciate the levels of complexity required to achieve this imaging and I can also appreciate the level to which you have simplified the discussion in order for someone like me to be able to usefully understand the important layers in such a dense complex process.
This was a very well explained and informative video. Hoping to see several more of these videos from this channel. Thank you for sharing.
I am so glad you got to visit, and to tell us about it! Astonishing stuff they do there.
Wow the engineering of the ASML machines to reproduce these features is beyond comprehension, to even fathom how to align and control this is just cutting edge. Excellent Video
Content is pure gold. Automatic like
Very interesting video! Also love the Everyday Astronaut RD-171 T-Shirt from your guide :)
Pre-patterning orientation check: masky end up, wafery end down! All systems norminal.
To my regret, this nerdy T-shirt is the only visible achievement of the Russians in ASML technology. Печально.
Fantastic video! What a show!
Boy do I miss my comp-litho days; you make me nostalgic
I love the level of physicality you are hitting. The simulations also help a lot.
@JorenVaes - 2023-12-21
As a chip designer, I never stop having my mind blown when we get silicon back from the foundry. You slave for months over these chips, and you are used to seeing them the size of a massive screen, having to zoom in like crazy doing the connections of small signals, when zoomed out not even being able to see whole functional blocks like complex operational amplifiers or so on, let alone the huge individual transistors (huge with respect to the tiny transistors we use in digital circuits). And then you get silicon back, and that huge chip is suddenly so small in that wafer gel pack that if you were to sneeze, you would lose them all, and never find them again.
@DrBlokmeister - 2023-12-21
As a lithographer, I'm always amazed by how chip designers optimize and design their chips so they can actually be manufactured on our machines, given all the conditions that we give you guys. Mad respect!
@ddopson - 2023-12-21
@@DrBlokmeister I'd be very eager to learn if both of you listed some of the more interesting constraints. Perhaps the chip designed and the lithographer might even have different takes about which constraints are the trickiest.
I have some vague ideas of the constraints, but am neither a chip designer nor a lithographer, so my understanding is a bit abstract and still rather hand-waived. For example, I recall some of the constraints deriving from the diffraction pattern on the mask such that if certain desired patterns are placed to near to each other, there's no way to construct a diffraction mask that generates that pattern due to various interference effects. Corners and ends of lines are particularly tricky. I recall that many chips now use multiple patterns per layer in oder to work around these imaging constraints by putting part of the pattern one mask and part on a second mask.
@DrBlokmeister - 2023-12-21
@@ddopson The constraints are I think more subtle. You are definitely correct. But in the end it's all about achieving a level of contrast and the correct feature width (CD) that is acceptable for the rest of the fabrication process. Depending on the topology of the structures, you need a certain depth of focus and might need to tune the machine to reach the desired values. What you're saying is also not wrong.
I think in the end our constraints are combined with all the other constraints in the manufacturing process, determined by the resist, material properties, etching, etcetera. Those final constraints are then communicated to the chip designers. At least, that's how I thought the process goes. If a chip designer tells you differently, believe that person!
@JorenVaes - 2023-12-21
@@ddopsonFrom what I understand, a lot of these details are ironed out by the foundry through their DRC (Design rule check) ruledeck. These have rules at their most basic do things like 'don't put lines closer than x nm to each other' or 'no line shorter than y nm', but quickly starts doing very complicated things like 'minimum line spacing is x nm unless you have more than y lines next to each other in parallel of the same width then you can decrease the line spacing to z nm except in the case where you need to contact the lines with a via in which case it changes again' blah blah blah, to the point that these rule decks are literally hundreds of pages (I think the one for the technology I worked in most recently is over 400 pages).
This is where a lot of the crazy stuff really is - there are rules that will allow you to only place lines on the finest structures in a certain pattern, with certain repetitiveness and on a grid, to make use of diffraction patters (or at least that is how I understand it - Sander might be able to correct me here).
Another example is that you might start using multiple patterning, where one metal layer is split up into multiple masks and soon to make more complex shapes possible.
The chipdesigners who really come close into contact with this are the people who design the standard cells---the basic logic gates---which really need to make as optimal use of all these tricks and features as possible, since they drive area of chips. In what I do (millimeter-wave chip design), the things that determine cost and area are not the transistors, but big passives like inductors and capacitors, and so we mostly look at performance of the transistor and not the area.
@DrBlokmeister - 2023-12-21
@@JorenVaes Very cool insights and context that you provide! All these rules you mention in the beginning make complete sense from an imaging point of view. Of course if all you want to do is place lines and spaces together, you can squeeze them in to the limits of your optics system. But as soon as you need something else, then that's gonna come at a cost. The more repetitive your structure, the easier it is to optimize for. If you want multiple structures at the same time, you have to optimize for all at the same time, and this means that all feature will perform worse than they could if you'd print them separately.
Also if you put things on a fixed grid, then your diffraction pattern will be discrete, as the Fourier transform of a grid is also a grid. This makes it easier to optimize the illumination settings (incidence angle of light). As soon as you want to place one feature off-grid, it will likely underperform and maybe not print well.