Howdy, Stranger!
It looks like you're new here. If you want to get involved, click one of these buttons!
Quick Links
Samsung announces that 7nm LPP using EUV is in production
https://news.samsung.com/global/samsung-electronics-starts-production-of-euv-based-7nm-lpp-process
https://www.anandtech.com/show/13496/samsung-starts-mass-production-of-chips-using-its-7nm-euv-process-tech
That's quite a bundle of jargon, but this is a big deal, even if TSMC will probably also make a similar announcement soon. EUV lithography has been delayed by many years, and that has caused serious problems in trying to move to new process nodes.
To explain what that means, for a number of years now, foundries have been using argon fluoride lasers to etch patterns in silicon. You know how in chemistry, they said that the noble gases don't normally form compounds? Well, you can force them to by certain means, but the compound really tries to break apart, and releases photons with a rather short wavelength when it does. In the case of argon fluoride, it's 193 nm. That's far into the ultraviolet range, so much so that it quickly gets absorbed by the Earth's atmosphere. They do that because a shorter wavelength lets you cleanly etch smaller features in silicon.
But how can you make a "7 nm" process node that really has 50 nm or so features using a 193 nm laser? Well, you need to take multiple passes. If you carve a piece 193 nm across in one spot, and then another one 100 nm to its side, the overlap between them will have a deeper hole than the parts that you only hit once. They call this double patterning, and needing two passes for a single layer is more expensive than one. Then they needed triple patterning, and then quadruple, and then it was really getting out of hand.
The solution is to stop using 193 nm ArF lasers to try to carve ever smaller features. Get a 13 nm laser and you can carve your features in a single pass. That's much faster and potentially cheaper, and can mean fewer things that can go wrong. That's not quite X-rays, as the official boundary between X-rays and ultraviolet is 10 nm, but it's pretty close. Rather, it's extreme ultraviolet, and hence the name EUV.
So why didn't everyone start using EUV years ago? Well, they wanted to. They tried to. But making a light source that can fire over 100 W lasers of 13 nm photons turns out to be rather hard. One problem is that you'd better do it in a vacuum or else they'll get absorbed by the air. Another is that 13 nm photons will tear apart a whole lot of things very quickly, as that's far more energy than the ionization energy of literally every element in existence.
But today, after all of the years of delays, EUV lithography is finally here and Samsung is finally using it in production. I'd expect TSMC to follow shortly, and Intel whenever they can. If it works as well as people hope, this could be a major contributing factor to future die shrinks, and hence future performance and efficiency gains from computer hardware, for years to come.
https://www.anandtech.com/show/13496/samsung-starts-mass-production-of-chips-using-its-7nm-euv-process-tech
That's quite a bundle of jargon, but this is a big deal, even if TSMC will probably also make a similar announcement soon. EUV lithography has been delayed by many years, and that has caused serious problems in trying to move to new process nodes.
To explain what that means, for a number of years now, foundries have been using argon fluoride lasers to etch patterns in silicon. You know how in chemistry, they said that the noble gases don't normally form compounds? Well, you can force them to by certain means, but the compound really tries to break apart, and releases photons with a rather short wavelength when it does. In the case of argon fluoride, it's 193 nm. That's far into the ultraviolet range, so much so that it quickly gets absorbed by the Earth's atmosphere. They do that because a shorter wavelength lets you cleanly etch smaller features in silicon.
But how can you make a "7 nm" process node that really has 50 nm or so features using a 193 nm laser? Well, you need to take multiple passes. If you carve a piece 193 nm across in one spot, and then another one 100 nm to its side, the overlap between them will have a deeper hole than the parts that you only hit once. They call this double patterning, and needing two passes for a single layer is more expensive than one. Then they needed triple patterning, and then quadruple, and then it was really getting out of hand.
The solution is to stop using 193 nm ArF lasers to try to carve ever smaller features. Get a 13 nm laser and you can carve your features in a single pass. That's much faster and potentially cheaper, and can mean fewer things that can go wrong. That's not quite X-rays, as the official boundary between X-rays and ultraviolet is 10 nm, but it's pretty close. Rather, it's extreme ultraviolet, and hence the name EUV.
So why didn't everyone start using EUV years ago? Well, they wanted to. They tried to. But making a light source that can fire over 100 W lasers of 13 nm photons turns out to be rather hard. One problem is that you'd better do it in a vacuum or else they'll get absorbed by the air. Another is that 13 nm photons will tear apart a whole lot of things very quickly, as that's far more energy than the ionization energy of literally every element in existence.
But today, after all of the years of delays, EUV lithography is finally here and Samsung is finally using it in production. I'd expect TSMC to follow shortly, and Intel whenever they can. If it works as well as people hope, this could be a major contributing factor to future die shrinks, and hence future performance and efficiency gains from computer hardware, for years to come.
Comments