[ExI] Nanotechnology

[John Clark]

On Fri, Oct 17, 2025 at 1:56 AM Keith Henson via extropy-chat <
extropy-chat at lists.extropy.org> wrote:

 Adrian Tymes  wrote:
>
> *>>> So now that they're out, how long until we start
>> seeing reverse-engineered ripoffs, legal or not?*
>
>
> >I don't think so.  These things are too complicated to reverse engineer
> without a vast amount of effort.
>

*I agree, in my previous post I only hinted at the massive complexity of
this beast of a machine, I didn't mention how it moves things around. And
it needs to because at any one time the light beam only illuminates about
1% of the 300 mm (approximately 12 inch) wafers that these machines print
on, therefore it has to move either the wafer or the reticle (projected
chip pattern) about 100 times to complete the entire wafer; ASML decided to
do both because to be economically viable the machine needs to produce a
LOT of chips so it needs to do everything fast.  The magnetically levitated
wafer tables accelerates at 7g in one direction (only 5g in some earlier
models) while the reticle accelerates at 15g in the opposite direction, and
it needs to do this without producing any vibration or heating.  *

*And each of the 100 exposures needs to be positioned within a quarter of a
nanometer, to achieve this level of precision, all the moving components of
the machine need to be measured 20,000 times per second with an accuracy of
60 picometers. By comparison the diameter of a silicon atom is about 210
picometers. *

*The machine is only making money when it is operating so ASML would like
it to be working 24-7 although they haven't quite reached that lofty goal,
but they have managed to keep it operating 75% of the time, the other 25%
is needed for maintenance, and I think that's pretty good for something
this monstrously complicated. *

*I would not have thought a machine like this was possible but amazingly it
is. It took ASML 17 years to develop the technology before they could make
something that they could sell and the development was so expensive it
nearly bankrupted the company, but they stuck with it, and now they're
reaping the benefits. *


*How an ASML Lithography Machine Moves a Wafer
<https://www.youtube.com/watch?v=1fOA85xtYxs>*

*John K Clark*




>
> Keith
> > On Thu, Oct 16, 2025 at 9:31 PM John Clark via extropy-chat
> > <extropy-chat at lists.extropy.org> wrote:
> > >
> > > There are two different paths that I think could bring on the
> Singularity, AI and Nanotechnology, it looks like AI will hit first but
> nevertheless I want to talk about Nanotechnology, particularly the top down
> variety used by the semiconductor chip industry rather than the bottom up
> path proposed by Eric Drexler.  Chipmakers use light to edge patterns onto
> silicon, the smaller the wavelength of light the smaller the pattern they
> can etch. Until recently  the smallest wavelength of light used was 193
> nanometers in the Deep Ultraviolet range made with a Fluorine Argon laser
> (visible light ranges from red at 750 nm to violet at 380 nm). Then the
> Dutch company ASML started to sell Extreme Ultraviolet Lithography machines
> to chip makers that use 13.5 nm light from vaporized Tin. ASML has a
> complete monopoly on these super advanced and super complex machines
> because they were the only company willing to take the risk of spending
> vast amounts of money for well over a decade to develop them, it's so
> expensive nobody else is even trying to make something comparable; and the
> way their machines work is amazing.
> > >
> > > The first thing you need to do is make a bright point source of
> Extreme Ultraviolet light, nobody has been able to make a laser with a
> wavelength that short and synchrotron radiation is not a point source, but
> when Tin is vaporized it has a bright spectral line at 13.5 nm, so they
> decided to use that.  50 thousand times a second a tiny droplet of molten
> tin is fired into the machine at 80 meters a second. Each droplet is then
> hit twice by two different lasers, the first laser pulse is low power and
> comes from a YAG laser of the type used in eye surgery, it reshapes the tin
> droplet from a sphere to a concave sheet, to do this the laser must hit it
> at exactly the right instant or the droplet will be at the wrong
> orientation. The second laser pulse is far more powerful and is in the 10
> mm infrared region and comes from a 25 kW carbon dioxide laser which is 5
> times as powerful as lasers that are commonly used to cut steel,  this
> vaporizes the Tin droplet and produces a flash of 250 watts of 13.5
> nanometer light. Without that first laser pulse that reshapes the Tin
> droplet the EUV output would be less than 10 watts.
> > >
> > > One significant problem they had to overcome was that the vaporized
> Tin tended to land on the delicate optical surfaces and degrade their
> performance, they eventually solved that problem by introducing a very
> small amount of hydrogen into the laser reaction chamber instead of having
> a hard vacuum.
> > >
> > > You can't use any lenses if you're dealing with EUV light (they're
> really soft x-rays) because glass, sapphire, diamond or any known substance
> would just absorb the light, so your only alternative is to use mirrors
> instead and even those are not easy to make. Each machine needs 6 mirrors a
> meter wide made with an accuracy of less than a 10^-12 meters ( a
> thousandth of a billionth of a meter),  they're the most perfect mirrors
> human beings have ever made, and they must be coated with 50 layers of
> molybdenum and silicon to reflect light with such a small wavelength, even
> then each of the six mirrors only reflect about 70% if the light that falls
> on them, and that severely reduces the amount of light you can actually use
> to etch things, that's why the point source has to be so bright.
> > >
> > > Until recently the only EUV machine in the world was ASML's NXD360D.
> It costs $160 million, is the size of a school bus weighs about 200 tons
> and has a Numerical Aperture of 0.33 (a dimension list number that
> determines how much light a lens or mirror can bring to a focus) it can
> produce a 26 nm pitch (the minimum center-to-center distance between lines
> that connect circuit elements), it can make about 160 wafers an hour, after
> manufacture each wafer is worth about $17,000. A typical chip fabrication
> plant will need several dozen of these machines.
> > >
> > > The resolution a lithography machine can produce is proportional to
> (light wavelength)/ 2* ( Numerical Aperture) so a mirror (or lens) with a
> larger numerical aperture will be able to print finer details onto silicon
> than a mirror with a smaller NA because it can focus more of the EUV
> light.  ASML's next generation machine is the EXE5000, it uses the same
> wavelength of light but increases the NA from 0.33 to 0.55 and can write a
> line on silicon with 16 nanometer pitch not 26 with the older machine and
> process 220 wafers an hour not 160 like the older NXD360D, it even uses 45%
> less electricity. The drawbacks to the new machine are it's even larger,
> about twice the size of a railroad locomotive, and it costs 380 million
> dollars, and yet everybody is desperate to get one. Apparently Intel wants
> one more than anybody else because they had dibs on the very first one that
> rolled out of ASML's plant.
> > >
> > > ASLM is not allowed to sell its EUV Machines to Russia or China, but
> that hasn't hurt the company because even without those markets they're
> selling their machines as fast as they can make them, and ASML now has a
> net worth  two and a half times that of one of its much more famous
> customers, Intel.
> > >
> > > John K Clark
> >
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20251017/3c7ec8ac/attachment.htm>