Author Topic: Moore's Law Continues  (Read 2696 times)

0 Members and 1 Guest are viewing this topic.

Offline mawyattTopic starter

  • Super Contributor
  • ***
  • Posts: 4120
  • Country: us
Moore's Law Continues
« on: October 30, 2023, 08:49:05 pm »
Interesting article, surprised they convey this much information, but gives an idea of where they really are  ;)

https://spectrum.ieee.org/high-na-euv

Best,
Curiosity killed the cat, also depleted my wallet!
~Wyatt Labs by Mike~
 
The following users thanked this post: MK14, magic

Offline TimFox

  • Super Contributor
  • ***
  • Posts: 9003
  • Country: us
  • Retired, now restoring antique test equipment
Re: Moore's Law Continues
« Reply #1 on: October 30, 2023, 09:27:34 pm »
When I was still in grad school, there was a lot of interest in charged-particle beams (electrons or ions) for high-resolution lithography, since the diffraction limit was much better than for photons.
Instead of physical masks, they would use beam deflection, as in a CRT, to expose the pattern on a resist layer, or deposit metal ions directly.
(There are lots of other practical problems with charged-particle optics, especially aberrations, but electron microscopes had achieved spatial resolution better than 1 nm more than 50 years ago.)
The shorter and shorter wavelength UV clearly won out since then, but I wonder when massive particles will re-appear?
 

Offline magic

  • Super Contributor
  • ***
  • Posts: 7454
  • Country: pl
Re: Moore's Law Continues
« Reply #2 on: October 31, 2023, 05:14:39 am »
The optics required to cover 16×26mm field in a single exposure at 10nm resolution must be insane. It's like APS-C with more than million pixels in each dimension, and of course tack sharp from corner to corner.

Quote
Consequently, at the mask, the incoming and outgoing cones of light become larger and must be angled away from each other to avoid overlapping. Overlapping cones of light produce an asymmetric diffraction pattern, resulting in unpleasant imaging effects.
Do you know something about this effect and what it means for photograhpy?


When I was still in grad school, there was a lot of interest in charged-particle beams (electrons or ions) for high-resolution lithography, since the diffraction limit was much better than for photons.
Instead of physical masks, they would use beam deflection, as in a CRT, to expose the pattern on a resist layer, or deposit metal ions directly
I think achieving enough throughput for large chips would be very hard if not impossible. How much time would you like to spend exposing a single pixel of the photomask, is one nanosecond enough? At this rate, for each second you are progressing 10 linear meters and printing 50 lines across a wafer of average 20cm width, covering a strip 0.5μ wide. It would take hours days to finish one wafer and the technology would need to be much cheaper to compete with photolitography systems processing hundreds of wafers per hour (as per the article). Otherwise, it will be limited to manufacturing very small and very expensive things.
« Last Edit: October 31, 2023, 05:26:15 am by magic »
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: Moore's Law Continues
« Reply #3 on: October 31, 2023, 05:36:52 am »
When I was still in grad school, there was a lot of interest in charged-particle beams (electrons or ions) for high-resolution lithography, since the diffraction limit was much better than for photons.
Instead of physical masks, they would use beam deflection, as in a CRT, to expose the pattern on a resist layer, or deposit metal ions directly.
(There are lots of other practical problems with charged-particle optics, especially aberrations, but electron microscopes had achieved spatial resolution better than 1 nm more than 50 years ago.)
The shorter and shorter wavelength UV clearly won out since then, but I wonder when massive particles will re-appear?

Most likely never: e-beam has too poor imaging after all (the deposited charge ends up in an explosion of scattering ionization, i.e. the exposure spot is much larger than the incident beam, also secondary emission that rains back down as a halo around the spot), and similar effects play with ions (plus physical damage to the resist, sputtering and whatnot); it would be great if a beam could be imaged but there's probably no way to do that.  Even e-beams are difficult to project an image with; an example being a patterned cathode*, but imagine trying to maintain optical precision on a heated thermionic cathode, plus contamination due to ion bombardment, plus smearing/divergence due to space charge?!  There's just no such thing for ions, and you can't like mask or reflect them.  I guess it would be really interesting to have, like, a huge drift chamber where ions could move towards an oversized electrostatic or magnetic image of the mask, and kinda make an electromagnetic DLP, and somehow manage to refocus the beam into some meaningful image, but....... right?  And I don't think it'd be something you can really reduce all that much, to get any kind of resolution out of naturally-blurry EM fields you're going to need a huge area, even if intimately patterned.

*Supposedly, the pattern you see, say on an analog scope with deflection stopped (NORM mode pre-trigger, or X-Y mode) and the focus cranked way out, is characteristic of the cathode, i.e. imaging the emission itself, give or take the optical path of the gun.  So if it's rough, flakey, salt-and-pepper, whatever, that's indication of a poor or failing cathode.  I'm not actually sure how true all of this is; I'm far from an expert on electron optics...


Do you know something about this effect and what it means for photograhpy?

I'm guessing it's to say the PSF varies with coordinate, and not in a way that's easy to compensate for (and for all the compensation they're already doing, that must be saying something), but I'd also like to know more :)

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline Someone

  • Super Contributor
  • ***
  • Posts: 5156
  • Country: au
    • send complaints here
Re: Moore's Law Continues
« Reply #4 on: October 31, 2023, 06:39:35 am »
There are lots of other practical problems with charged-particle optics, especially aberrations, but electron microscopes had achieved spatial resolution better than 1 nm more than 50 years ago.
Imaging resolution /= machining resolution. 10nm feature size is pretty tough on a FIB and as above, veeeeerrrryyyyy sssssllllllooooowwwwwwwwww.
 

Offline magic

  • Super Contributor
  • ***
  • Posts: 7454
  • Country: pl
Re: Moore's Law Continues
« Reply #5 on: October 31, 2023, 09:03:38 am »
Do you know something about this effect and what it means for photograhpy?
OK, there is one damn obvious problem with the imaging cone overlapping the illuminating cone: part of it runs into illumination optics where it's diverted back to the light source. Or alternatively, part of the illuminating cone is obstructed by the imaging optics before reaching the mask. Either way, effective system pupil has "partial eclipse" shape and is narrower in X dimension than in Y, resulting in similarly distorted diffraction blur and loss of X resolution.

Perhaps this is all they meant to say.

BTW, this issue can be worked around by means of beam splitters aka half-transparent mirrors, a technique routinely used in microscopy for at least a century. But if nothing is transparent enough to make EUV transparencies then chances are that beam splitters weren't an option either.

FWIW, in reflected light microscopy fully overlapped cones normal to specimen surface are in widespread use and I haven't heard of this causing unexpected issues.
« Last Edit: October 31, 2023, 04:23:08 pm by magic »
 

Offline tom66

  • Super Contributor
  • ***
  • Posts: 7336
  • Country: gb
  • Electronics Hobbyist & FPGA/Embedded Systems EE
Re: Moore's Law Continues
« Reply #6 on: October 31, 2023, 10:54:18 am »
The optics required to cover 16×26mm field in a single exposure at 10nm resolution must be insane. It's like APS-C with more than million pixels in each dimension, and of course tack sharp from corner to corner.

Do they actually use 10nm resolution?  It was my understanding they use lower resolution combined with computational lithography (using interference patterns to reconstruct the design.)  The EUV wavelength of 13.5nm can still image details on a chip around 2nm in pitch now, with these techniques.  Apple has a SoC with 3nm technology, though the actual nanometer scale is distorted by marketing.

Edited to correct error.
« Last Edit: October 31, 2023, 11:03:03 am by tom66 »
 
The following users thanked this post: Someone

Offline coppice

  • Super Contributor
  • ***
  • Posts: 10035
  • Country: gb
Re: Moore's Law Continues
« Reply #7 on: October 31, 2023, 11:08:55 am »
When I was still in grad school, there was a lot of interest in charged-particle beams (electrons or ions) for high-resolution lithography, since the diffraction limit was much better than for photons.
Instead of physical masks, they would use beam deflection, as in a CRT, to expose the pattern on a resist layer, or deposit metal ions directly.
(There are lots of other practical problems with charged-particle optics, especially aberrations, but electron microscopes had achieved spatial resolution better than 1 nm more than 50 years ago.)
The shorter and shorter wavelength UV clearly won out since then, but I wonder when massive particles will re-appear?
There used to be e-beam lithography machines at around the 1 micron level, used for rapid prototyping. I don't think they were ever considered for mass production, because they take too long to scan. However, for prototyping skipping the slow and costly mask creation step is a huge plus point. Now masks costs have risen to eye watering levels, so you'd think that skipping them would be an even better deal for prototyping now than 35 years ago. I understand they can make the e-beam machines work down to the resolution of current ICs - 1 or 2nm. There must be other factors stopping their greater adoption. I'm not clear how much work was needed between the e-beam generated prototype and a mask generated production part, to allow for differing edge effects. Clearly the nature of how an e-beam fuzzes up the printed image is going to differ from how an optical printing system fuzz them up. Perhaps issues like this made the e-beam prototypes too weak a representation of how a production part would perform. Good for checking out the logic, but poor for predicting other performance issues, like speed or noise.
 

Online ejeffrey

  • Super Contributor
  • ***
  • Posts: 4034
  • Country: us
Re: Moore's Law Continues
« Reply #8 on: October 31, 2023, 02:25:34 pm »

Quote
Consequently, at the mask, the incoming and outgoing cones of light become larger and must be angled away from each other to avoid overlapping. Overlapping cones of light produce an asymmetric diffraction pattern, resulting in unpleasant imaging effects.
Do you know something about this effect and what it means for photograhpy?

This is just because they have to use reflective masks but need to separate the incoming and outgoing beams.  I'm pretty sure what they mean by "overlapping beams produce an asymmetric diffraction pattern" is that if you truncate one side of the cone into a "D" shape you get a weird diffraction pattern so you need to make sure they are completely clear of each other.
 

Offline TimFox

  • Super Contributor
  • ***
  • Posts: 9003
  • Country: us
  • Retired, now restoring antique test equipment
Re: Moore's Law Continues
« Reply #9 on: October 31, 2023, 04:11:20 pm »
There are lots of other practical problems with charged-particle optics, especially aberrations, but electron microscopes had achieved spatial resolution better than 1 nm more than 50 years ago.
Imaging resolution /= machining resolution. 10nm feature size is pretty tough on a FIB and as above, veeeeerrrryyyyy sssssllllllooooowwwwwwwwww.

< 1 nm resolution was obtained in scanning transmission electron microscopes (STEM) around 1970, which is a similar problem to machining resolution.
 

Offline mawyattTopic starter

  • Super Contributor
  • ***
  • Posts: 4120
  • Country: us
Re: Moore's Law Continues
« Reply #10 on: October 31, 2023, 04:36:10 pm »
MIT Lincoln Labs with USG sponsorship were doing nanometer level Direct Write E-Beam Lithography over 20 years ago. They created a special Class 1 Area within a Class 10 Room for E-Beam processing the special wafers, which was painfully slow!! To help with the throughput, conventional lithography was used for non-critical areas of the chip, while Direct Write utilized where meaningful. This helped with overall wafer processing speed, but still much too slow (expensive) for commercial use, and thus relegated to specialized chip development where actual chip cost didn't matter, but performance and chip size did!!

Direct Write can achieve remarkable resolutions but is hampered by throughput speed since it's basically a serial write process, whereas UV and EUV are essentially parallel exposure lithography processes, and thus much more economical, altho as mentioned the mask costs are far from economical!!!

Best,
Curiosity killed the cat, also depleted my wallet!
~Wyatt Labs by Mike~
 

Offline mawyattTopic starter

  • Super Contributor
  • ***
  • Posts: 4120
  • Country: us
Re: Moore's Law Continues
« Reply #11 on: October 31, 2023, 04:49:49 pm »
The optics required to cover 16×26mm field in a single exposure at 10nm resolution must be insane. It's like APS-C with more than million pixels in each dimension, and of course tack sharp from corner to corner.

Quote
Consequently, at the mask, the incoming and outgoing cones of light become larger and must be angled away from each other to avoid overlapping. Overlapping cones of light produce an asymmetric diffraction pattern, resulting in unpleasant imaging effects.
Do you know something about this effect and what it means for photograhpy?


The optics are by Carl Zeiss, and likely the most advanced optics ever conceived and well beyond our limited understanding. These folks at ASML and Zeiss have been doing what was thought quite impossible for some time now, and continue to do so in spite of what physics indicates are the limits.

Quite impressive indeed  :clap:

And this is what they are releasing, you can wager they are well beyond what's indicated in this article ;)

Edit: As you mentioned the optics to achieve those resolutions over a 16X26mm field are quite insane!! One of our most prized optics is the famous Printing Nikkor 105mm F2.8, this was created and used to copy high resolution movie film over the standard 35mm frame, but at resolutions that could be acceptable for movie theater use, so use as film reproduction. Nikon spend years developing this lens for this use, and thus the Printing Nikkor name. The Carl Zeiss lenses are doing something similar but at nanometer resolutions over a similar size field!!

https://www.savazzi.net/photography/printing-nikkor-105mm.html

https://www.closeuphotography.com/lens-tests

Guess where the idea of using processed silicon wafers for lens evaluation across the field came from  ;D

Best,
« Last Edit: October 31, 2023, 05:27:30 pm by mawyatt »
Curiosity killed the cat, also depleted my wallet!
~Wyatt Labs by Mike~
 

Online ejeffrey

  • Super Contributor
  • ***
  • Posts: 4034
  • Country: us
Re: Moore's Law Continues
« Reply #12 on: October 31, 2023, 05:33:02 pm »
When I was still in grad school, there was a lot of interest in charged-particle beams (electrons or ions) for high-resolution lithography, since the diffraction limit was much better than for photons.
Instead of physical masks, they would use beam deflection, as in a CRT, to expose the pattern on a resist layer, or deposit metal ions directly.
(There are lots of other practical problems with charged-particle optics, especially aberrations, but electron microscopes had achieved spatial resolution better than 1 nm more than 50 years ago.)
The shorter and shorter wavelength UV clearly won out since then, but I wonder when massive particles will re-appear?
There used to be e-beam lithography machines at around the 1 micron level, used for rapid prototyping. I don't think they were ever considered for mass production, because they take too long to scan. However, for prototyping skipping the slow and costly mask creation step is a huge plus point. Now masks costs have risen to eye watering levels, so you'd think that skipping them would be an even better deal for prototyping now than 35 years ago. I understand they can make the e-beam machines work down to the resolution of current ICs - 1 or 2nm.

Well, you can make masks with e-beam writers although laser scanners are more common, so the costs aren't independent.  The issue is that as you move to smaller feature sizes the e-beam writing process also gets slower and more expensive, so it's not clear that the tradeoff point actually changes.
 

Offline Someone

  • Super Contributor
  • ***
  • Posts: 5156
  • Country: au
    • send complaints here
Re: Moore's Law Continues
« Reply #13 on: October 31, 2023, 10:53:37 pm »
There are lots of other practical problems with charged-particle optics, especially aberrations, but electron microscopes had achieved spatial resolution better than 1 nm more than 50 years ago.
Imaging resolution /= machining resolution. 10nm feature size is pretty tough on a FIB and as above, veeeeerrrryyyyy sssssllllllooooowwwwwwwwww.
< 1 nm resolution was obtained in scanning transmission electron microscopes (STEM) around 1970, which is a similar problem to machining resolution.
Restating what you already said doesnt add much if anything. You might think they are similar, but care to share references to single digit nm machining ? Like with lithography the actual resolution achievable is variable depending on the design geometry and not able to be summed up as a single readily comparable number.
 

Offline TimFox

  • Super Contributor
  • ***
  • Posts: 9003
  • Country: us
  • Retired, now restoring antique test equipment
Re: Moore's Law Continues
« Reply #14 on: October 31, 2023, 10:55:53 pm »
There are lots of other practical problems with charged-particle optics, especially aberrations, but electron microscopes had achieved spatial resolution better than 1 nm more than 50 years ago.
Imaging resolution /= machining resolution. 10nm feature size is pretty tough on a FIB and as above, veeeeerrrryyyyy sssssllllllooooowwwwwwwwww.
< 1 nm resolution was obtained in scanning transmission electron microscopes (STEM) around 1970, which is a similar problem to machining resolution.
Restating what you already said doesnt add much if anything. You might think they are similar, but care to share references to single digit nm machining ? Like with lithography the actual resolution achievable is variable depending on the design geometry and not able to be summed up as a single readily comparable number.

Don't be a jerk.  I was clarifying my earlier statement.  I was near the lab that achieved that resolution with an STEM, but I am unfamiliar with nm machining.
 

Offline mawyattTopic starter

  • Super Contributor
  • ***
  • Posts: 4120
  • Country: us
Re: Moore's Law Continues
« Reply #15 on: October 31, 2023, 11:49:10 pm »
There are lots of other practical problems with charged-particle optics, especially aberrations, but electron microscopes had achieved spatial resolution better than 1 nm more than 50 years ago.
Imaging resolution /= machining resolution. 10nm feature size is pretty tough on a FIB and as above, veeeeerrrryyyyy sssssllllllooooowwwwwwwwww.

< 1 nm resolution was obtained in scanning transmission electron microscopes (STEM) around 1970, which is a similar problem to machining resolution.

Isn't that "Tunneling" rather than Transmission?

Recall IBM used a similar scanning "probe" technique to spell IBM with individual Xeon atoms!!

https://cen.acs.org/analytical-chemistry/imaging/30-years-moving-atoms-scanning/97/i44

Best,
Curiosity killed the cat, also depleted my wallet!
~Wyatt Labs by Mike~
 

Offline coppice

  • Super Contributor
  • ***
  • Posts: 10035
  • Country: gb
Re: Moore's Law Continues
« Reply #16 on: October 31, 2023, 11:50:52 pm »
There are lots of other practical problems with charged-particle optics, especially aberrations, but electron microscopes had achieved spatial resolution better than 1 nm more than 50 years ago.
Imaging resolution /= machining resolution. 10nm feature size is pretty tough on a FIB and as above, veeeeerrrryyyyy sssssllllllooooowwwwwwwwww.

< 1 nm resolution was obtained in scanning transmission electron microscopes (STEM) around 1970, which is a similar problem to machining resolution.

Isn't that "Tunneling" rather than Transmission?

Recall IBM used a similar scanning "probe" technique to spell IBM with individual Xeon atoms!!

https://cen.acs.org/analytical-chemistry/imaging/30-years-moving-atoms-scanning/97/i44

Best,
STEM - scanning tunnelling electron microscope. Maybe he mistyped.
 

Offline Someone

  • Super Contributor
  • ***
  • Posts: 5156
  • Country: au
    • send complaints here
Re: Moore's Law Continues
« Reply #17 on: November 01, 2023, 12:42:20 am »
There are lots of other practical problems with charged-particle optics, especially aberrations, but electron microscopes had achieved spatial resolution better than 1 nm more than 50 years ago.
Imaging resolution /= machining resolution. 10nm feature size is pretty tough on a FIB and as above, veeeeerrrryyyyy sssssllllllooooowwwwwwwwww.
< 1 nm resolution was obtained in scanning transmission electron microscopes (STEM) around 1970, which is a similar problem to machining resolution.
Restating what you already said doesnt add much if anything. You might think they are similar, but care to share references to single digit nm machining ? Like with lithography the actual resolution achievable is variable depending on the design geometry and not able to be summed up as a single readily comparable number.
Don't be a jerk.  I was clarifying my earlier statement.  I was near the lab that achieved that resolution with an STEM, but I am unfamiliar with nm machining.
So what did you add?
50 years ago vs 1970's
better than 1 nm vs < 1 nm

At least we now know you are unfamiliar with machining.
 

Offline TimFox

  • Super Contributor
  • ***
  • Posts: 9003
  • Country: us
  • Retired, now restoring antique test equipment
Re: Moore's Law Continues
« Reply #18 on: November 01, 2023, 01:22:13 am »
As I stated in my post, I was referring to the Scanning Transmission Electron Microscope, which pre-dates the tunneling microscope (developed in 1981).
It is an example of how fine a focus could be achieved with a 100 keV electron beam over 50 years ago.
« Last Edit: November 01, 2023, 01:35:52 am by TimFox »
 

Offline mawyattTopic starter

  • Super Contributor
  • ***
  • Posts: 4120
  • Country: us
Re: Moore's Law Continues
« Reply #19 on: November 01, 2023, 01:44:56 am »
Interesting was only aware of the Tunneling version!!

Best,
Curiosity killed the cat, also depleted my wallet!
~Wyatt Labs by Mike~
 

Offline coppice

  • Super Contributor
  • ***
  • Posts: 10035
  • Country: gb
Re: Moore's Law Continues
« Reply #20 on: November 01, 2023, 01:21:04 pm »
As I stated in my post, I was referring to the Scanning Transmission Electron Microscope, which pre-dates the tunneling microscope (developed in 1981).
It is an example of how fine a focus could be achieved with a 100 keV electron beam over 50 years ago.
The scanning transmission microscopes are usually called SEMs - just scanning electron microscope, because transmission came first - and the tunnelling ones add the T, so STEM.
 

Offline EPAIII

  • Super Contributor
  • ***
  • Posts: 1165
  • Country: us
Re: Moore's Law Continues
« Reply #21 on: November 01, 2023, 01:57:42 pm »
Probably already being done or I am just naive, but couldn't multiple beams write to different areas at the same time? Just keep them far apart enough so there are no bad effects. A 10 x 10 array of beams could decrease the time needed by a factor of 100.
Paul A.  -   SE Texas
And if you look REAL close at an analog signal,
You will find that it has discrete steps.
 

Offline TimFox

  • Super Contributor
  • ***
  • Posts: 9003
  • Country: us
  • Retired, now restoring antique test equipment
Re: Moore's Law Continues
« Reply #22 on: November 01, 2023, 02:00:02 pm »
As I stated in my post, I was referring to the Scanning Transmission Electron Microscope, which pre-dates the tunneling microscope (developed in 1981).
It is an example of how fine a focus could be achieved with a 100 keV electron beam over 50 years ago.
The scanning transmission microscopes are usually called SEMs - just scanning electron microscope, because transmission came first - and the tunnelling ones add the T, so STEM.

No.  "SEMs" are scanning electron microscopes, which usually generate the image of a surface by secondary electron emission.
The first electron microscopes were "TEMs" for "transmission electron microscope", where electrons passing through the target were focused by an objective lens onto a photographic film. 
Scanning Transmission Electron Microscopes were developed thereafter, but achieved good results around 1970.  https://en.wikipedia.org/wiki/Scanning_transmission_electron_microscopy  for a comparison and history.
The resolution with the scanned beam was well below 1 nm.
The scanning tunneling microscope (STM), which is really cool (Nobel Prize 1986), is completely different:  see  https://www.testandmeasurementtips.com/basics-of-the-scanning-electron-microscope/
 
The following users thanked this post: MK14

Offline mawyattTopic starter

  • Super Contributor
  • ***
  • Posts: 4120
  • Country: us
« Last Edit: November 11, 2023, 01:00:47 am by mawyatt »
Curiosity killed the cat, also depleted my wallet!
~Wyatt Labs by Mike~
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf