How Carl Zeiss Crafted a House of Mirrors for EUV Light
Merry Christmas, everyone
Author’s Note: If you want to watch the video, you can below
I have nothing to add to this video except to complain that I had a huge headache writing it. Several people have asked for sources on this if they wanted to read more on it. Here are a selected few:
If you are interested in helping to support the channel, I’d appreciate if you check out the Patreon. Check out the Early Access Tier, where you can watch videos before they are released. Every little bit helps.
Extreme Ultraviolet Lithography is the next step on the semiconductor fabrication roadmap. It is a disruptive technology using a new ultraviolet wavelength.
Dutch company ASML is currently the only company shipping these million dollar machines. But ASML in turn relies on Carl Zeiss in Germany for the all-important optics system. The two are intimate strategic partners.
The key issue that ASML and Carl Zeiss engineers had to overcome is that high-energy UV light gets absorbed by virtually all materials, which means that lenses are not feasible. The lens material will literally just eat up those rays. Thus, Zeiss crafted an optics system entirely out of multi-layer mirrors.
It is super cool. In my last video I gave a brief overview of Carl Zeiss the company and a few of their products. In this follow up, we are going to go deeper and look at how Zeiss made a EUV optics system.
## The EUV Optics System, Outrageously Simplified
Let me start with this. I am not a vision scientist and the technical details behind the EUV Starlith 3000 systems are some of the most mind-bending stuff I have ever read.
I collected the details for this video from over 30 conference presentations academic theses, and industry papers on the topic. Forgive me for any errors that I will inevitably make here.
Here are the basics. The Starlith 3000 series is a multi-component optics system consisting of the collector optics, an aperture, illumination optics, reticle, and projection optics. We will go over what those individual components actually do later.
The light source is where it all begins. I discussed how such EUV light is created in my previous video about ASML's supply chain. So let's skip the heavy details. For now, all you need to know is that they use lasers and tin plasma to generate high energy EUV light.
Once the light is generated, it needs to be collected. And that's what the collector does.
## The Collector
The Collector Optics system is set up to collect as much EUV light as possible from the source and transfer those photons to the next module in the Starlith system.
In order to do this, it is made in the shape of an ellipsoid, kind of like a car headlight lamp. The only gaps in the mirror would be for the laser and the tin droplets.
Figuring out the collector's shape was not the only challenge engineers had to overcome. They also had to make sure it stays clean and functional.
The collector is uniquely difficult in that it is exposed to molten tin plasma from the EUV light source. This leads to its inevitable deterioration. But in order for EUV to be economically viable, ASML customers like TSMC, Intel, or Samsung should only have to replace the collector once each year or every 3000 exposure hours.
To protect the mirror, Zeiss engineers took a two fold approach.
First, they attempted to prevent layers of tin from forming on the collector mirrors. They created a buffer zone of inert gas and employed magnetics to discourage ions away from the mirrors.
But no matter what, tin will get to the collector and form a layer on it. And a tin layer of just 1.2 nanometers thick will cause a 20% decline in collector efficiency and a 10% decline in tool output.
Thus, Zeiss engineers had to figure out a way to constantly clean it right inside the machine. As often as every 2 hours or less, as according to studies of the alpha demo tool.
The way they figured out how to do it was to send hydrogen radicals at the tin layers. Radicals, to jog your memory from chemistry class, are atoms or molecules with one unpaired electron.
Their unpaired electron makes these things quite reactive. Thus, these hydrogen radicals will react with the tin, floating away from the mirror in the form of a gas called tin hydride.
This cleaning does not stop the collector's deterioration. Merely delays it. It is inevitable. But it did help Zeiss get substantially closer to its one replacement a year goal and overcome one of EUV’s major obstacles.
## The Illuminator
After the UV light is collected, it is sent into the optics system through an aperture at intermediate focus. Then it enters the illuminator system.
The illuminator has two big jobs. Its first job is to evenly spread out the EUV light so that it remains uniform across the entire field of focus.
Second, it ensures that the image of the EUV light source does not affect the rest of the optics system.
Zeiss drew on their extensive experience building similar products for optical microscopes to deliver this very special component.
After that, the light hits the mask or reticle, which contains the IC chip's patterns. Uniquely, this reticle is itself a mirror.
When you get to these sizes, the information on the reticle is extremely dense. The ASML NXE:3400B, an EUV machine, has a field size of 33 millimeters by 26 millimeters. This is quite large.
To put it into perspective. Let us call a single 13 nanometer by 13 nanometer square a pixel. Now if that pixel was to be blown up to the size of a pixel in a regular full HD TV, then the television would be 17,622 meters high.
That would be 50% taller than Taipei 101. And that entire television screen would be filled edge to edge with the chip designs of something like a Intel Core i7.
Conceptually at a high level, this mirror reticle is made up of two things: The first is a highly reflective multi-layer mirror substrate called a reticle blank.
The second is an absorbing layer applied on top of it containing the chip design pattern. The light bounces off the mirror and makes an EUV shadow.
With the reticle, Carl Zeiss focused on two big fundamental challenges. The first is making sure that the reticle blank is as clean and reflective as possible. ASML asked that the blank should have only 0.003 defects per square centimeter.
The second major challenge was how to detect and repair the reticle when defects do arise. Blank or absorber-related defects will happen no matter what. These have to be detected and repaired.
In the optics system for the 193nm DUV lithography generation, defects are detected with a scanning electron microscope. Easy.
For EUV, they needed to do more as some types of reticle defects are invisible to an electron microscope. So Zeiss took that machine and added to it an atomic force microscope, a probe device that scans by sort of running its "finger probes" across the surface of a sample. That is the best way I can put it.
This new machine is now capable of dynamically switching between electron microscope and atomic force microscope to find defects the other cannot.
Once a defect is detected, repairs are done using electron beam processing. It is kind of like welding at an atomic level, where a precursor is laid on the sample and then struck with a beam of electrons from an electron gun. Depending on the type of defect, you can either add more absorber or etch away absorber.
And oh yeah, the repair has to be entirely done autonomously in a vacuum to prevent any contamination. The Zeiss MeRiT system is able to meet all of these requirements and helped overcome one of the more daunting challenges in EUV commercialization. That being said, the costs of entry for producing an EUV reticle are expected to be prohibitive for all but the richest of companies.
Don't these titles all sound like really poorly named professional wrestlers?
Anyhow. The Projector optic system is a multi-mirror system designed to expand out the reticle image and project it onto the wafer.
It is crucial that aberrations on the projector optics are reduced as much as possible. At first, the mirrors that Zeiss created for the ASML alpha demo tool's projector back in 2010 had very slight, but significant defects.
These defects are measured in a unit called surface-rms, or the root mean square average of the profile height deviations from the line. Too much of a surface-rms leads to what is essentially the microlithography version of lens flare.
Most of us just know lens flare as that thing that J.J. Abrams really likes to put into his movies. But what it actually is is an image artifact that comes as a result of scattered light in a lens system.
Unlike with the esteemed work of J.J. Abrams, flare causes defects on the wafer and thus, is unwanted.
The 2010 alpha demo tool had a surface rms of 0.25 nanometers, leading to a 25% flare, which was unacceptable. In order to get the flare down, the mirror aberrations had to be substantially reduced. Something they managed to achieve over the multiple generations of the Starlith EUV system.
So let us talk about the mirrors then.
As you might expect for a system made of mirrors, Carl Zeiss paid very close attention to how those mirrors were made. The mirrors are so important that they were the reason why the whole enterprise chose to work with the 13.5nm spectrum of EUV light.
It was a decision between the 11.3 to 11.6 nm wavelength using molybdenum and beryllium mirrors or the 13.5nm wavelength with molybdenum and silicon mirrors. In the end they chose the latter.
And as you might expect, these are not your ordinary bedroom mirrors. These are multi-layer mirrors. Sometimes also called Bragg Reflectors.
Your traditional mirror is usually made by applying a layer of silver or aluminum onto glass.
An EUV mirror on the other hand is made up of over 50 pairs of molybdenum and silicon layers deposited on top of a single silicon wafer. Each of those layers within the pair is about 6.75 to 7 nanometers thick. 2.7 nm of molybdenum and 4.1 nanometers of silicon.
You want to have all of these layers because the EUV light is still going to pass through some of them. So you have more layers, with the hope of reflecting as much EUV as possible.
The silicon and molybdenum atomic layers are applied using a technique called Direct Current Magnetron sputtering. This is a type of thin film physical vapor deposition, something I talked about in my video about Tokyo Electron.
So to make these EUV mirrors, each 2 to 4 nanometer layer of molybdenum and silicon has to be deposited one by one. Fifty times, without fatal defects.
After that, each mirror has to be painstakingly polished. The acceptable surface deviation metric is 50 picometers, or 50 trillionths of a meter. That is a staggering number.
To put that into perspective. 50 picometers on a mirror that is 450 millimeters wide. If that 450 millimeter mirror was blown up to the size of the United States, 4,500 kilometers or 2,800 miles wide, then 50 picometers would be just 0.5 mm tall.
In DC Magnetron sputtering, you want to layer a coating onto a surface. You bombard the coating with ionized gas molecules. Those molecules "sputter" off atoms which then condense as a thin film onto the surface.
Even with all that, each mirror can theoretically reflect just around 70% of the EUV light that hits it. The rest is absorbed by the multilayer. And it can only achieve that number at a very specific angle, requiring exquisite equipment to position it just right.
Considering there are so many mirrors across the entire optics system, you can appreciate the challenge that ASML engineers face in getting enough power through to the wafer.
The thing about this EUV optics system that I keep marveling at is just how radically different it is from its predecessor. It is much like the shift from internal combustion engines to electric vehicles.
The vapor deposition techniques behind multi-layer mirror alone took over 20 years of research. It required close collaboration between ASML, Carl Zeiss, Philips, and various universities and institutes across Europe.
To have been able to deliver something like this is a stunning achievement, the culmination of thousands of years of man-hours. Throughout this video, I hope that I have impressed upon you that notion, that understanding of just how difficult it was.
Feel free to share the newsletter if you are interested.
And follow the Twitter too