IEEE Spectrum November, 2016 - 32

T h e M o lT e n T i n S o l u T i o n

Curves and Corners
EUV promises to create sharper
shapes [right] than those
that can be created through
multiple patterning with today's
193-nanometer light [left]. The
lines in these micrographs have
a minimum width of 24 nm.

The less light that reaches a wafer, the longer a wafer must remain in
the scanner to be exposed. And in a fab, time means money. For EUV
to make it into commercial use, it needs to be able to compete with the
cost of existing lithographic methods. So the losses among the mirrors
have to be compensated by a radiation source that is extremely bright.
And that proved to be really, really hard to engineer.

nov 2016

north AmericAn

SPectrUm.ieee.orG

HEADED EAST: A portion of an ASML NXE:3350B
EUV scanner is loaded onto a 747-400 extendedrange freighter at the Amsterdam Airport
Schiphol in the Netherlands, for shipment to
a customer. The scanner is split into nine parts
(some larger than others) for shipping.

technique the Cymer team began exploring before
being acquired by ASML. They found that if they
fired a "prepulse" before the main laser, they could
flatten each tin droplet into a pancake, creating more
surface area for the main laser to hit and increasing how much of the tin droplet was converted to
plasma. The change has boosted the laser-to-EUV
conversion efficiency from a meager 1 percent to
some 5 percent. Earlier this year, thanks to the
prepulse and other optimizations, ASML reported
that it had reached 200 W in the lab. Another lightsource developer, Gigaphoton, has also reported
great progress. The long-awaited production target of 250 W no longer seems far off. But the true
test of whether EUV is ready to go into production
will happen in the labs and fabs-and spreadsheets-
of ASML's chipmaking customers.
nobody doubts that EUV machines can make
fine features. Go to a semiconductor conference
and you'll probably come across a presentation
with sharp-looking micrographs of EUV-made pat-

asML (5)

In the early days, EUV researchers used just about everything they
could think of to generate X-rays, including lasers and particle accelerators. But the method that won out, which seemed to offer a practical and economical way to get sufficient brightness, employs plasma.
Zap the right material with a powerful enough laser or electric current
and you can separate electrons from the atoms they're attached to. The
resulting plasma will radiate EUV as the superhot blob cools back to
its prezapped state.
The oft-repeated target for such plasma-based light sources has
been 250 watts at the intermediate focus, the location where the EUV
light exits the source and enters the scanner. That level of light would
enable the machine to produce in the neighborhood of 125 wafers per
hour, a mass production target that is about half the rate of today's
advanced 193-nm machines.
But for a number of years, progress was slow, and brightness improvements consistently lagged behind predictions. By 2011, some five years
after ASML delivered its first test scanners to two customers, one of the
leading light-source developers, San Diego-based Cymer, had succeeded
in developing a source that could consistently deliver 11 W. "We probably underestimated how difficult it would be," concedes ASML's Hans
Meiling, who is responsible for EUV product marketing. In the end, in
an effort to accelerate development, ASML put in a bid for Cymer, formally completing the reported €3.1 billion acquisition in 2013.
To create EUV light, Cymer uses an approach called laser-produced
plasma, which fires 50,000 microscopic droplets a second of ultrapure molten tin across a vacuum chamber, hitting each with powerful
CO2-laser light generated by a series of amplifiers derived from a design
originally used for metal cutting. When a laser pulse hits a molten tin
droplet, it heats it up into an EUV-emitting plasma. A collector mirror
reflects light created in this process and casts it into the scanner. Because
the approach generates EUV light as well as tin debris, hydrogen gas
constantly flows across the collector mirror to keep it from being rapidly covered with a layer of tin.
"The first time I heard about it, I thought it was insane," admits ASML's
Alberto Pirati, who joined the company's EUV light-source program in
early 2013. But little by little, the team achieved the seemingly impossible. One of the biggest breakthroughs came with the introduction of a

http://SPectrUm.ieee.orG

Table of Contents for the Digital Edition of IEEE Spectrum November, 2016