Semiconductor Immersion Lithography
Вставка
- Опубліковано 12 жов 2022
- I get it. Everyone wants to talk about EUV. It's the sexiest lithography around with all the mirrors and the purple UV light.
But I think we shouldn't discount 193 nanometer immersion lithography. Coming about in the early 2000s, it has taken the industry further than anyone could have ever expected.
193i as it is sometimes called is still a workhorse and at the core of many leading edge process nodes. For this video, I want to look at how it works and the challenges overcome in developing it.
Links:
- The Asianometry Newsletter: asianometry.com
- Patreon: / asianometry
- Twitter: / asianometry
People often say that motivation doesn’t last. Well, neither does bathing. That’s why we recommend it daily
Brilliant
It's not like if u just buy the motivation like u do with water
TSMC N5 and N4 are not even fully EUV. Almost only the fins in the FinFETs are made with EUV. All the TSVs and most the routing layers are made with 193i. And the reason is simple: The wafer throughput is much much higher on 193i. So if a chip has 80 layers and only 5 of them are EUV and the rest are 193i you need, say, 1 EUV machine and 8 193i machines for the same throughput as 25 EUV machines would give you. Considering just how brutally more expensive the EUV machines are this is the difference of having a break-even after 3 years at full capacity or 30 years at full capacity.
I thought the fins were quad duv? The big issue intel had with 10nm quad was using it for M0 - at least this is what I've heard around - (quad fin patterning was industry standard for a while I believe). I wouldn't be shocked if they changed the litho tech for the N5 fins though for throughput reasons.
If euv Is only used sparringly couldn't it be replaced with greater use of múltiple layering duv?
@@diegoantoniorosariopalomin9979 you could up until some point. But at quad EUV is just cheaper and higher throughput. Multi patterning also increases design rule complexity.
@@ab-lymphocite5464 does it matter too much when EUV Is only used sparringly like you said?
EUV may make comercial Sense for most nations, but china wont have access to it for several years, however what you said may mean that restriction wont be a killing blow to their chip industry
@@diegoantoniorosariopalomin9979 You want EUV anywhere where you would be quad patterning because it kills yields and makes your nodes much more expensive/slower. And quad patterning can’t really carry you beyond n5 maybe n3. With this in mind and the fact that smic’s n16 copy isn’t in hvm at any real scale, I am highly skeptical of the prc being able to cover it’s needs anytime soon. To say nothing of being cut off from the metrology, ald, and many more tools. The prc also has to develop their own native dry and then wet 193nm litho tools to even be able to get the scale to make any modern nodes.
This is great - many people think that EUV is the only industry tech out there since it makes the headlines. You're doing important educational work, thanks for putting these together!
Something modern like a Zigbee controller may very well be produced on 193i, just like so many other kinds of chips.
That's because it actually is at present. 153nm hasn't really been resolved and yields make it cost prohibitive. Wafers are not cheap! Bad yields and disposal of dirty water (or treatment) isn't cheap or easy either. The result is that EUV is the only tech that currently makes sense from cost and quality due to the combinations of problems and low yields.
It’s important to remember even in the n5 and smaller nodes especially in the back end of the line an EUV machine is overkill. Most process levels have CDs well In the range of 193i. Nearly all modern foundries operate these tools and will for many more years
The exception is the lower metals which are getting to the point of low NA EUV double or DUV quad patterning. But yeah it is frustrating how much EUV can sometime be overemphasized.
As an D&E Engineer at ASML for the DUV (193 nm) machines. I would like to thank you for this video. Too bad I can not share any technical details with you.
whoa you're a big boss
Can't wait for you to release the video explaining the new US restrictions and their implications.
what restrictions?
Unless US gives an exemption, otherwise ASML, Nikon and Canon can't sell any new or old lithography to China anymore. I am use all 3 uses some US tech one way or the other. Same for other machines that use in fab.
@@NoNameAtAll2 US ban many technologies export to China, which is good for the world
@@Theoryofcatsndogs I guess the machines that made it to China will be running 24/7.
@@NoNameAtAll2 Forbid US tech firms or those using US patents to sell advanced chips and equipment to China and the latest move included US persons to service equipment sold to China.
Great video as always. Thought you sounded a little down while narrating this. Hope you're doing well. Take care!
I realized that as well. Hope he's doing well
You sound very tired. We are very grateful for the effort you put into it.
I work in lithography for a big company. It's fun seeing these tools and the graphics you have. It really helps!
Love Asianometry! Your rock, dude! Keep on trucking.
U hit these outta the park every time in my opinion. I really, REALLY enjoy your videos so thank u very much! I do wish, however, that they could’ve been a little longer but I’m assuming the UA-cam algorithm perhaps doesn’t favor those longer videos as much. Well besides the fact that there’s already a lot of work going into a 16 minute video of course.
🙏✌️🖖🙌
I work for Carl Zeiss SMT and the immersion optical system is one of our most difficult systems with a lot of challenges in the production.
Why is that?
For you, maybe.
Again, excellent research into the history of lithography. You should look into reticle purge and how that also differentiated ASML and Nikon scanners.
Been waiting for this one, great as always!
Always learning something in this channel, thanks!
Our legacy fab is still using it!
Our cutting edge one is as well :D
10:52. That is an amazingly illustrative flaw! The system projected the bubble unto the waver, showing that the air does not shrink the image as the surrounding water, as the refractive index in air/gas is much closer to one.
When I read about the US export ban on semiconductor production equipment to China in Peter Zeihan's newsletter, I immediately thought of this channel and tried to find an existing video on the topic, but there is only one about China's equivalent to ASML, which is many years behind the cutting edge. I would love for you to cover the implications of the ban for Chinese domenstic semiconductor production. Zeihan says that they are screwed, because they are completely dependent on foreign imports to produce anything remotely marketable, but I was not able to confirm this assessment.
Believing that guy is humanity's greatest mistake. He talks of US isolation and how world will go to chaos without them. Absolutely astonishing.
On one hand, you've got the people who think China will catch up in 5 years. Then you've got people like Zeihan. The truth is in the middle if you ask me.
9:27 I love the note on the corner. I really needed help to figure out if they were using 10 cm shower heads on 197 nm wafers. Great mix of excellent education and dry (wet?) humor as always.
Way to go asianometry!!! Please be motivated please!!
I loved this. But also all the other subjects you approach. Thank you!
Excellent episode, thank you for enlightening this subject!
Hey Asianometry, love your content! Have you every considered doing a video on space-based microfabrication? It's something I've seen a lot of hype around recently, but have been unable to *any* papers or otherwise outlining the proposed fabrication processes - would love to see you tackle it!
Long overdue. Excellent. I learned a lot.
Awesome explanation, thank you!
your channel is fucking awesome, I love your passion on it
you doing better research than any techtuber
Another great video! Thanks!
Great video, thank you.
I believe you have a video on resists, but what about anti-reflective and surface prep treatments?
You're humor is so dry- I love it
Jheeze man, incredible work.
You keep it up man!!
You said they dont do well, one for the algorithm. 80k + people learned something, you did fkn amazing
Thanks for the video
Immersion lithography introduces added complexity and expense over dry lithography. Ensuring a clean environment for high-purity water and managing environmental issues related to water disposal are critical. Moving wafers into and out of the fluid can be intricate. It is vital to control the movement precisely while preserving the fluid layer. It is also crucial to minimize or eliminate bubbles. The immersion fluids must be transparent at the 153 nm wavelength (utilized in EUV lithography), and achieving such transparency presents a continuous challenge since the 153 nm wavelength is in the deep ultraviolet (DUV) spectrum, making it difficult to find materials that are transparent and compatible with semiconductor lithography processes.
I have a video suggestion:
With EUV reaching limits because of the wavelength theres the question "is there another way to produce even smaller features?". I heard that using electrons archives that, but it is a lot slower and can't be used for high volume production.
Physics is a "beach", bro. There is always a tradeoff
Immersion lithography works in large part because you literally CHANGE the wavelength of the light because of the high index of refraction of water for the corresponding light. The index of refraction depends on the frequency of the light -- that doesn't change. The wavelength (lambda) and frequency (f) satisfy
lambda * f = v
where v = speed of light in the medium. The speed of light in the medium satisfies v = c/N, where "c" is the speed of light in vacuum, "N" is the index of refraction of the medium at the given frequency, and N = 1 for vacuum. So in vacuum or in air, as the index of refraction of air is almost one, then you have lambda(air) * f = c, and for water lambda(watter) * f = c/N, which after a little bit of algebra gives you
lambda(water) = lambda(air)/N.
Since at the frequency when you have lambda(air) = 193 nanometers, N(water) is fairly large, you get that lambda(water) is smaller than in air, which allows (as you know) the manufacture of smaller features.
EDIT: I guess that what you have called NA in your formula is pretty much what I am simply calling N --- but I just had to correct the fact that you are actually changing the wavelength, it doesn't stay the same, because the medium where the light travels changes from air to water; rather, what is always the same is the frequency of light. Sorry, but I am a physicist, and I just could not let it go --- I still like your videos very much. I am not trying to disparage you in any way.
Doing the lords work sir
Cannon just announced that they're getting back in the lithography game.
Any thoughts on how difficult that'll be for them or guesses on how successful they will be?
42
Excellent vid🙏🙏🙏
I am always interested in alternate paths that might end up being the "next big thing".
X-rays, E-beam, etc.
a video on electron beam lithography/deposition/direct etching may be interesting as an alternative process
Electron beam lithography is much, much slower, like 1000s of times slower.
@@mx2000 less demanding on hardware by roughly the same amount though, and an interesting topic nonetheless
Thanks!
you are talking about technology very well and concise, maybe u should try to begin to review electronic products too
It's funny that the brick wall was hit over ten years ago with the legacy technology and leading edge technology lines are now blurred. The cutting edge is still in focus. Is the space race over?
All that sexy purple EUV light!!! Lolol
Never in a million years would I have expected the Nas Ether reference 🤣
Beautiful. The history of technology just behind the bleeding edge.
This is so cool, I don't understand but it's still cool
How did you manage to bring Nas into a video about this? That was cool! Thanks!!
The American banned the Chinese from buying the Lithography machines. Can the Chinese make it by themselves even taking 10 years to build?
They can. And some of the patents are starting to expire. But they lack the practical knowhow (a patent is practically useless other than telling you where to start looking), so yeah, 10 years sounds about right.
@@andersjjensen This is made easier by having working machines.
Given they don't have dry DUV tools being produced at any scale almost two decades after release of immersion DUV, and that there are no EUV tools in the PRC (to my knowledge) I would assume that it would be way longer than 10 years.
@@andersjjensen When has IP stopped mainland Chinese companies. Totally second you on how patents are not always super useful.
Sometimes with optics there is no substitute for experience, and one just has to work on it really hard for a rather long time to learn all the nuances -- it takes a lot of skill and sometimes luck too. That's one of the reasons why Zeiss in Germany is supplying the optics for the top of the line machines. They have almost two centuries of experience and had world's most renowned opticians developing their methods. It is not trivial to catch up to them.
I remember I first heard of the immersion technique just before I got laid off in 2013.
Every fab that was using DUV at the time I bet are now using the immersion steppers. Steppers when they stop being cutting edge find their way into commodity parts. Which means they get smaller as well.
Man, there are a lot of lens elements in this machine. Just to make a projected image smaller. I had a Nikonos under water camera, a 15mm lens for macro photography. It had many lenses nested on top of each other , convex, concave to be able to be an inch away from the subject and have a wide field of view with no fish eye effect. The no fish eye effect was the reason for all the lenses in a wide field. The lens itself cost $4,000.00 in the year 2000. Very pricey.
Amazing tech, this.
What did they think was easier to do extreme ultraviolet lithography over 157nm Immersion lithography?
I bet Burn Lin's dis tracks are epic.
has Burn Lin ever been to Berlin?
;)
I'd be interested to hear what types of devices use these processors. Thanks!
Your phone, laptop or desktop PC
Great video. We still need a LOT of not state-of-the-art lithography for all the workhorse chips industry needs. It would be nice to not have to depend on shipping these all from Asia.
Looking for an honest opinion. I am working as a semiconductor test engineer for 1.5 years I got no complains about that but I kinda always had some interest on chip designing but right now I am confused. Should I try to switch or stay with testing for some more time?
What happens when the paths of light cross? Does interference happen during travel as well?
Okay, okay, dad, I'll never call immersion lithography old again.
With more US restrictions coming on Chinese tech, it looks like a decoupling is on its way. Wonder what the Chinese will come up with. Will they work on 157nm light that western lithography abandoned, or work its own version of EUV, like SSMB EUV, or the known working method of vaporizing a tin droplet with a CO2 laser.
Any plan to cover ASL KAUST?
Didn't get your phrase "Mirrors split beams into two - the original beam and it's reflected twin - and you want to make sure that the two do not interfere with one another". Could you explain?
Ok, so China looks locked out of EUV with aminimum of 15 years to catch up. If they took a 157nm with immersion approach as a intermediate step, could we calculate how far that could take them? Any chance of getting past 5nm ?
It's still used so widely after all.
Just wondering: is it not possible to use electron beams instead of light? Like with electron microscopes, where the energy can be increased at will, effectively giving a variable "wavelength". 🤔
you are the coolest the best information giving channel in youtube about hardwares
Paying him is 💯 worth it.
Please make a video about canon's possible NanoImprint lithography that canon is working on. They are building a new factory to make lithography machines at a cost of ~$350m and if they succeed with getting nanoimprint working well they might make nanoimprint lithography machines there.
I wonder what would happen if you used heavy water instead of regular water in it.
Magic is just technology you don't understand yet, indeed.
"Any sufficiently analyzed magic is indistinguishable from science!" -- Agatha, from Girl Genius comic. 5th of Dec 2008.
Sick diss track ..hahaha.
true man of the culture for the Nas reference
>Modern immersion lithographies can take you all the way to the N7 node
Intel: Hold my N14+++++++++++++ node
I'd be curious to know what's the actual current limit with E UV in size for a wafer.
Like 2nm transistors?
Also not sure if I understood that correctly but when TSMC speaks bout 5nm it's not the actual transistor size but rather the light wavelength?
If i remember correctly, it's neither. "Nanometers" are just indication of how fast processor will be. N10 is better than N22, and worse than N5. Something like that.
First question, on limits of EUV for a perfect diffractive grating image (a series of lines or similar) assuming a k1 of the near impossible 0.25 with no multipatterning
“Old” EUV tool, 0.33NA = ~10nm
“New” EUV tool, 0.55NA = ~6nm
193i DUV, ~1.35NA, = ~35nm
Realistically multiply those numbers by 2 for practical limitations. Features can get smaller through various forms of multipatterning but that gets expensive.
Regarding your second question on “5nm”, Node sizes are a marketing term and don’t really correspond to a critical dimension of anything in particular. You will find that most critical dimensions are larger than 20nm regardless of the “Xnm” node size label. The main changes that are occurring are different transistor architectures are being used to reduce transistor efficiency losses by employing more advanced forms of GAA (Gate All Around) with atomic layer deposition tools and the like.
@@alexanderbrown6077 Worth noting is that DUV double patterning is still way cheaper than EUV single (probably lower throughput though). Otherwise good answer, you saved me a good bit of writing :D. If memory serves from the IMEC nodes we can get into like "18 or 16A" with DUV quad (not that you would do that besides some front end layers), with EUV double and High NA single carrying you a good bit farther than that.
I recommend the video titled "Why Every 3nm Process Node is Different" to get this answer.
@@alexanderbrown6077 6nm transistor size. that's truely impressive
Will you eventually do a video on X-Ray Lithography? From what I understand, Russia and China are attempting to develop the technology, and it would be nice to understand the hurdles ahead of them
In the next 5 to 10 years, I think China's chip industry will be very well paid and have very good working conditions and hours. This is because of China's growing chip worker shortage. And the Chinese government are making chips and technology their first priority. Unlike the US, increasing wages and standards of working won't be much of a hassle due to the fact that in China, rich businessman and corrupt CEOs don't have nearly as much power as in the US.
Yup, China will throw unlimited money and human resources at this chip war. And they are going to win
Bro deal between Vedanta and Foxconn is done ✅.
Make a updated video how this is going to help India in future and what kind of semiconductor will be manufactured here.
I am going to spam this massage if you read this pls reply.
You should just make a textbook and sell it on wafer tech. Your content is education quality! Learn so much here and in the comments section.
How is it possible to create features smaller than the light's wavelength?
With EUV.
First: to get a straight line, you will need to make a diffraction pattern on the template instead of a line.
Second: water compresses the wavelength by about 1.35 times, this is the essence of immersive technology
Third: overuse multipatterning as hell.
First, the response of the photo-sensitive material is non-linear, and the edges of the printed lines can be very sharp comparing to the width of the lines.
Second, the projection machine aligns the pattern with far greater accuracy than the wavelength.
So you can print one set of relatively wide lines, and then another set of relatively wide lines with a carefully chosen offset, and then the gaps between the two sets can be made much smaller than the wavelength and very accurately controlled.
Many other tricks along the same lines are possible.
@@cogoid sounds a little like a diffraction grating.
@@brodriguez11000 Conceptually, "double pattering" is much simpler than that -- you can think of it as parking cars closer together than the size of a car.
Playing with diffraction is the basis of a whole other bag of tricks for increasing the resolution in lithography -- phase shift masks, optical proximity correction and all that. In these methods, one tries to predict the smearing of the image due to diffraction and to undo this smearing by tweaking the mask. You cleverly modify the mask, so that it does not really look exactly like the features which you are trying to print. But when the light from different points on this mask adds up coherently on the resist, this creates the pattern that you want to get.
What about nano impress
Can anyone here point me to an analysis of the recent chip tech restrictions placed on China? I would really like to hear Asianometry's thoughts on this.
Please talk about the chip ban on china ?
Me: *listening to a video essay about technology that I'm only marginally capable of understanding*
Asianometry: "It was the Ether of lithography, sick ass diss track"
Me: O_O
About how many years/decades will it probably take for China to produce its own cutting-edge microprocessors with native cutting-edge lithography (EUV, etc.), including the ability to make all the specialized parts involved? In other words, how long until China can make everything itself? Is it even possible? Thanks in advance, everyone.
You should read EqualOcean as gives prefect insight about Chinese semiconductor industry. According to them it will take 10 to 12 years.
@@ASK-ko9qx thank you!
China is so hard pressed right now, that they are going to use their spies to copy asml and zies tech
When is 'good enough'?
When people stop paying for more
Good enough is changing all the time, and probably never reached. Its malleable with current world situations. I think people would rather get something more energy efficient at the cost of a bit of horsepower at this moment in time.
Is the factor k1 in DUV and EUV machines the same?
No. EUV is 0.33NA and High NA EUV is 0.55NA.
They are both manipulating the same number, but using different phenomenon to achieve it. Also EUV has much lower NAs due to the higher energy photons.
I'm sorry for using the EUV machine as a black light.
Hello @Asianometry, could we perhaps get a video about the Russian Baikal BE-S1000 that has been crippled due to their imperialist invasion of Ukraine?
Hope it dorsn't take fotever for xray litography.
I had an idea. Etch the image, then shrink it. That will achieve 1nm with todays tech.
thats... not how it works
You need to look up the sizes of the transistors and the sizes of wavelength, and wonder how that would work :)
honey, I shrunk the ... transistors
Look up "double patterning": print, etch, atomic layer deposition (ALD), and after a few more processing steps you can keep the pattern that stuck to the walls of your original pattern...
Could you do a video about Biden's new China chip restrictions? How will they effect the world and is China lost behind forever because of it?
But the other one's purple!
Konigsberg for Taiwanese people , please ! Love from Poland !
What would be the oldest lithography technology that could be condensed into a desktop machine. The machine can consume the entire desktop if need be.
What do you even mean by that? How small you can make some old tool? If yes then about desktop size for an old Hg arc lamp contact litho machine from the 80s is the norm.
There is a UA-cam channel "Sam Zeloof" -- the guy makes working chips at home, using home-made maskless photolithography stepper based on a DLP projector. I think his setup is capable of features 20-30 um wide.
@@cogoid I remember watching a couple of his videos for a class! I also joked that he was the guy global foundries was looking for when they wanted a new college hire with 8 years of semiconductor manufacturing experience.
@@ab-lymphocite5464 The guy is both brilliant and fortunate. He grew up in a technically oriented family which owns a manufacturing business.
I know of his existence, he has been working on the homemade tech for years I was just think that with his process known and classified what other technologies like @A B-Lymphocite had provided would fit within a span of a desktop.
Is this guy a ASML shareholder?
State of the art that China won’t be able to procure.
Are you sick? Your voice doesn't sound quite right
sounds suicidal.