Ultraviolet light bulb
IMAGE CREDIT: Wikimedia Commons / Artist: Anakin101

Ah, chipmaking: Unlike the early days of lithography, it’s no longer as simple as just shining some light onto a mask and creating a pattern. Now, the actual photons have to be created through torturous means. That means starting from scratch, more or less, in terms of light sources, machine configuration, photoresist and mask design, among other things (oh, so many other things). I’ve collated the ten most interesting new things I learned about extreme ultraviolet lithography at the EUV Litho workshop last week in Honolulu, Hawaii.

10. The end of Moore’s Law is always seven years away.
Lithography guru Chris Mack explained this tenet in a tutorial on microchip lithography, his tongue firmly planted in his cheek. Why seven?

The next generation (say the chips with 32-nm feature sizes Intel plans to ship this fall) is already in development; the following generation is in full R & D phase (22 nm in 2011), and the one right after that is already a gleam in the eye of the semiconductor researcher (15nm in 2013). The one right after (11 nm) that is the one no one’s really thought about yet (2015?), and they see only problems without solutions. And since that adds up to about seven years, it’s the ever-moving horizon beyond which all progress breaks down: human sacrifice, cats and dogs living together, transistors obstinately refusing to shed another nanometer.

Mack, as I have mentioned, has a bet with workshop organizer Vivek Bakshi that EUV will go the way of the dodo, or at least the way of 157-nm lithography: “Everyone will talk about it and talk about it for two more years,” he predicted, “and then after two years no one will ever talk about it again at all.”

The stakes? Mack’s Lotus. Bakshi says he already has a set of “EUVL” license plates—in fact he has two, because he “accidentally” ordered an extra set. Next year, he says, he’s going to give the annual EUV Litho workshop award in the form of a laminated EUVL license plate.

9. You can make 15-nm features with 193-nm light.
At least, that’s what Intel’s Sam Sivakumar said in his keynote on Wednesday. I’m talking about the 15-nanometer process node, which as we know from Bill Arnold’s article, is kind of an invalid concept these days. But for purposes of comparison, today’s cutting edge chips are at the 45-nm node, meaning the smallest features (usually transistors) have 45-nm dimensions. It’s more complicated than that, so if you like, check out Jeff Hecht’s tutorial.

Sivakumar speculated that Intel would be hedging its bets for the 15-nm node (coming to a laptop near you in 2013, which is a lot closer than you think). Though Intel is on track for development and production of those chips using EUV lithography, they also have a side bet on “extreme double patterning,” which is a term I just made up. It’s double-patterning lithography, which Intel is already going to start using for the 32-nm chips it will ship this fall, but twice as much of it, or let’s call it quadruple patterning. That should slow down their wafer throughout (now at 150, and then divide that by four) sufficiently to motivate them to get EUV on the roadmap.

8. There’s more than one way to skin an EUV photon

Apparently there is a big difference between the light source that will produce the 13.5-nm photons necessary for lithography, and the light source that will produce 13.5-nm photons for inspecting the masks. The key is finding defects that would interfere with the patterns and make them useless, and researchers don’t know yet how big a defect has to be to get in the way, or how many of them will cause a problem. It’s still very much a work in progress.

The masks are multilayered, and every layer has to be perfect (all sorts of numbers for maximum defects per square centimeter populated the week’s presentations, and they were all incredibly low). The only way to probe down into the mask layer cake to check for those defects is with 13.5-nm photons.

My first thought was that a scanning electron microscope or an atomic force microscope could tackle that job, but no, says Debbie Gustafson, sales VP at Energetiq, a company that produces light sources. “AFM would be much too slow,” she says, “and even if it weren’t, both SEM and AFM only image the surface layer.”

And the light source has to be brighter than the EUV photons that imprint the pattern—900 kilowatts per square millimeter, according to one presenter from Ritsumiken University.

7. Tin vs. Xenon: This time it’s personal.
Such light is created not by exploding tin droplets with a carbon dioxide laser (the so called laser produced plasma, or LPP, that creates the images), but with Xenon gas and electrodes. An electrode releases an electrical discharge into a cloud of gas, which heats it, magnetically constrains it into a small space, and then EUV photons are released. This is called discharge produced plasma or DPP, and Energetiq uses it for mask inspection sources).

Chris Mack says tin will be the light source of the EUV generation. “So far the industry consensus is that the tin and carbon dioxide method will be the winner,” he says. But he’s talking about LPP, not DPP. Xenon arc lamps, which are used in today’s 193-nm lithography, have new life in the defect inspection field.

6. Litho nerds have dreams, too
Q: Why would you want your own laser if you work at Lawrence Berkeley National Labs?
A: Because just like astronomers who have to sign up to get time on the telescope, researchers working on EUV lithography have to sign up to get time on the synchrotron or whatever other crazy light source they have at the national labs. So, LBNL lead scientist Patrick Naulleau would have preferred to have his own light source.

NanoUV manufactures next-generation EUV light sources. The company claims that their source is as bright as a synchrotron but without the oppressive footprint of, say, the Large Hadron Collider. The instrument is 45 cm long and 14 cm wide and, utilizing a controversial “plasma lens technology,” according to NanoUV presenter Sergey Zakharov, is capable of producing 120 Watts. Naulleau got a little unhappy in the Q & A session. “How is this possible?” he asked, more than once.

But Zakharov skirted the answer and put Naulleau off, telling him to wait until February, when the tool will be revealed to the world.

Unfortunately, it was postulated by attendees who prefer to remain anonymous that the NanoUV light sources appear to transcend the laws of physics.

Eric B. Szarmes from the University of Hawaii’s department of physics and astronomy presented another cutting edge high-brightness lithography source base on Compton backscattering from electron pulses. If I had to bet on eyebrow-raising numbers, I’d bet on his, because he works with John Madey, who invented the free electron laser at Stanford in 1975.

5. Photoresist: Meet the new boss…
However, unlike the newfangled ways of generating photons, one surprising variable probably won’t create a headache. The photoresist that you can use to expose a wafer using 13.5-nm photons is the same chemical structure that worked for the 248-nm lithography of 16 years ago (they’ve been stuck at 193 nm since 2001).

4. I have to learn French!
Pop quiz time. Define “étendue:”

A.     A treacherous postmodernist ballet move that combines a plié with an accent aigu.
B.     What your French teacher used to yell across the room when you weren’t listening.
C.    A measure of attenuation of a beam of light.
D.     Area of the entrance pupil times the solid angle the source subtends as seen from the pupil. These definitions are for infinitesimally small "elements" of area and solid angle however, and have to be summed over both the source and the diaphragm.

I cycled through A, B, and C (C out loud unfortunately, and in the form of a question) before it was explained to me that étendue in fact is defined as D. The equation is the area times the divergence angle. The larger the beam the smaller the divergence. If EUV goes mainstream, you can expect to hear this word as often as I did last week, and you’ll be well-equipped with a definition. So, you’re welcome.

3. EUV Needs A Vacuum

Contamination is a big problem for EUV lithography tools. You have a machine inside which a pulsed laser explodes a series of falling 20- to 40-nanometer tin droplets, creating a super-hot plasma, and that plasma beams out EUV photons.

But that’s just the easy part. Now you have to get them to the mask over the wafer—but remember you have to do it without letting the EUV photons touch anything.  It’s like that game from the 1980s, Operation, where you had to pick up the creepy little plastic liver without touching the metallic sides of the patient’s “cavity.”

EUV photons, which are just on the boundary of where they should be called x-rays, are absorbed by absolutely everything; EUVs go in, but they don’t come out. The list of what is the death of an EUV photon includes: water molecules, air, glass, and metal. In other words, everything you’d expect to find in your average lithography machine.

In order to minimize that absorption, researchers have “child-proofed” their machines almost beyond all reason. Every step—from EUV generation to plasma to the beam concentrators to the beam collectors—takes place in a vacuum. And now, because you can’t use glass or plastic or saran wrap or any kind of barrier, you can’t separate the vacuum chambers. With pretty much any other light, you could just the beam through windows that separate one vacuum chamber from the next. Not with EUV. The area has to be completely open. Now you’re stuck making a huge, long, labyrinthine vacuum chamber in your litho tool.
And everything that interacts with the beam (steerers, light collectors, lenses, and so on) needs to be made of super reflective mirrors. These are based on Bragg reflection, a concept used for optical fibers. A Bragg reflector is created by coating a smooth substrate with several dozen alternating layers of molybdenum and silicon.

And it gets worse!

2. Contamination is bad

University of Illinois at Urbana-Champaign plasma physicist David Ruzic uses the very user-friendly term “splats” to describe the small balls of tin that pancake onto the equipment.

The ions and tin splats exploding out of the plasma have no travel restrictions. They can migrate everywhere and anywhere in your machine. The splats are no fun but you can clean them up, more or less, with RF coils and fancy lasers.

What’s harder to do is guard against the devastation rained on your super sensitive mirrors by the emitted ions from the plsma. These, according to one presentation, cut short the life of the average satellite-grade Bragg reflector mirror from 30,000 hours to, uh, 2 to 4. HOURS. You think double patterning is expensive. Wait till you’re asking ASML for a new litho tool six times a day (though I’m sure they would find some way to get through the pain…).

1. No technology is immune from the Star Wars metaphor!

Ruzic tells me the preferred method of dealing with line edge roughness–-which according to many attendees is the limiter for lithography-- is shooting a tiny laser (pew! pew! pew!) down the canyon walls of those tiny features to smooth out the rough spots. “It’s like that part where they’re shooting lasers down the canyons in the Death Star,” he explained, warming the nerdiest regions of my heart.

Stay on target!
The Conversation (0)

The State of the Transistor in 3 Charts

In 75 years, it’s become tiny, mighty, ubiquitous, and just plain weird

3 min read
A photo of 3 different transistors.

The most obvious change in transistor technology in the last 75 years has been just how many we can make. Reducing the size of the device has been a titanic effort and a fantastically successful one, as these charts show. But size isn’t the only feature engineers have been improving.

Keep Reading ↓Show less