Maskless momentum - Mark Allen
Transcript of Maskless momentum - Mark Allen
As semiconductor fabrication
technology progresses, one of
the growing problems is
lithography. Semiconductors are
created using optical techniques
to expose photoresists. However,
the wavelength of the light used
to expose the photoresists is
now much larger than the size
of the features being
created. If this problem
isn’t solved, then progress
to ever smaller nodes will
be stalled.
Needless to say,
work is underway to
solve the problem
and – in some
cases – this work
has been going
on for some time.
The industry is
keen to use current technology for as long
as possible. Today, most lithography
systems are based on 193nm laser light.
But the time is rapidly approaching when
the systems will no longer be of use – even
with such tricks as immersion, which takes
advantage of higher indices of refraction.
Two schools of thought dominate the
research. One is to use extreme ultraviolet
(euv) light. Instead of the current 193nm
wavelength, euv uses a light source with a
wavelength of 13.5nm. At first sight, this
would appear to solve the problem. But it
hasn’t been as easy to develop the process
as was originally believed. Now,
companies like Intel
and IBM, who had been
looking to use euv in the
near future, are putting
developments on hold.
But optical lithography also needs
masks and, as feature size decreases,
mask costs increase dramatically. The
other approach being pursued is to do
away with masks altogether. Instead of
exposing photoresists to light, maskless
lithography uses an electron beam to
‘write’ the circuitry directly onto a wafer
coated with an appropriate resist. It’s a
development of a process currently used to
create low volume asics under the guise of
direct writing.
Belgian research institution IMEC has
been working on euv for some time but,
interestingly, is now looking at maskless
lithography as one of its future options.
Kurt Ronse, director of IMEC’s
lithography department, noted the
techniques are very different. “With euv,
everything has to happen in a vacuum. It’s
broadly similar to existing lithography,
but it has to use mirrors instead of lenses
because there are no lenses which are
transparent to euv.”
EUV also uses masks, but these are also
reflective. “It’s a logical step beyond
193nm,” Ronse claimed. But it is a big
step. “We have tried 157nm,” he
admitted, “but it wasn’t successful.”
Meanwhile, a European programme
called Rimana (Radical Innovation
Maskless Nanolithography) has just
concluded initial research into maskless
nanolithography technology. Dr Hans
Loeschner, project administrator and chief
technology officer of Vienna based IMS
Nanofabrication (www.ims.co.at) noted:
“The objective was to show proof of
concept for maskless lithography.”
Rimana was working on projection
maskless lithography, or PML2, regarded
as a potentially cost effective solution for
32nm devices and beyond, offering
resolutions of less than 10nm.
The approach is based on a system
which creates thousands of electron
beamlets and focuses these on the wafer.
According to Dr Loeschner: “The approach
19w w w . n e w e l e c t r o n i c s . c o . u k 2 7 J a n u a r y 2 0 0 9
MasklessmomentumMaskless lithography is being considered as a viable alternative to
extreme ultraviolet. By Graham Pitcher.
S P E C I A L R E P O R T
Fabrication Technology
Mirrors have to be
with euv lithography
because lens are not
transparent to the
13.5nm wavelength
light.
fab.qxp:Tech Temp 21/1/09 17:16 Page 19
wasn’t proved when Rimana started. It
was a good idea, but could it work? We
think we have shown the main questions
have been answered.”
Dr Loeschner believes maskless
lithography is, in principle, simple. “It’s a
combination of known things,” he
claimed. “You start with an electron
source in a vacuum and extract an
electron beam.” Once that’s done, a set of
condenser optics shapes the electrons
into a 20mm beam. These then
pass through an aperture
plate. “This is a silicon plate
about 20µm thick,” Dr Loeschner
explained. “It has many 4µm holes, which
allows the creation of beamlets.”
The beamlets then pass through a
blanking plate and pass to a deflection
grating. This device, powered by cmos
electronics, starts organising the beams
into a form that allows a chip to be created.
“At the right moment,” Dr Loeschner
explained, “a small voltage is applied to
form an electronic filter with sufficient
ability to deflect the beamlet. In this way,
some of the beams are filtered out.”
Unlike the direct writing approach, the
wafer moves across the scanning field at a
rate of around 50mm/s. The beamlets
created by the equipment allow the chip
to be exposed. “Thousands of beams are
working in parallel to create the chip,”
said Dr Loeschner. Currently, the process
is such that a throughput of five wafers an
hour can be achieved.
Wafers need to be coated with a
suitable resist. Each beamlet then exposes
a spot on the resist several times until the
correct dose has been applied. A further
benefit is that the control electronics can
also provide features equivalent to the
optical proximity correction used in
current lithographic systems.
But production ready equipment is
some way off. “A prototype tool needs to be
built, then a full production prototype,” Dr
Loeschner explained. The tool is targeted
at the 22nm half pitch node, which is
expected to come into use in 2016. That
means the production prototype would
need to be ready by 2014 at the latest.
A production tool is likely to generate
millions of beams, speeding throughput.
By running machines in parallel,
throughputs rivalling those of current
lithographic systems may be possible.
“IMS is a small company and we knew
from the beginning that we wouldn’t be
able to take the idea to production. But
we’ve been talking to equipment
companies and there has been strong
interest.”
Ronse believes it won’t be easy to take
maskless lithography to production.
“Several groups have been trying to build
a prototype and it doesn’t seem to be easy
to do. All the beams need to be
calibrated,” he noted, “and there are
carbon contamination problems. But the
system will still need to match the optical
throughput of more than 100 wafers/hr.”
Ronse also sees maskless lithography
as a back up, should euv development
efforts fail. “IMEC is seeing progress in
maskless lithography from quarter to
quarter and member companies are
recommending that we now focus on this
technology.”
For the moment, IMEC will continue to
watch developments. “There are no
prototype tools available,” Ronse noted,
“and that’s typically where IMEC gets
involved. “We have to see which tool ships
first and its specifications. That can take a
while and it may be the end of this year
before we can make a decision.”
However, euv work is continuing.
“We’re finding more promising resists,”
Ronse claimed, “and feeding back
information to resist manufacturers.
We’re also monitoring the alpha tool to
see whether it’s stable and which
components are failing.”
But one reason why scepticism exists
around euv is the optical performance.
“The optics on the alpha tool have
limitations which makes it difficult to go
beyond 32nm,” Ronse admitted.
“However, tools coming next year should
have less flare in the optics and that will
enable us to address 22nm,” he
concluded.
20 w w w . n e w e l e c t r o n i c s . c o . u k 2 7 J a n u a r y 2 0 0 9
S P E C I A L R E P O R T
Fabrication Technology
Figure 2: The principles of projection maskless lithography
electron source
condenser optics
programmableaperture platesystem
aperture plate
blanking plate
deflecting electrodes
first lens
stopping plate atbeam crossover
second lens
substrate/stage
“We’re finding more promising resists
and feeding back information to resist
manufacturers.” Kurt Ronse, IMEC
fab.qxp:Tech Temp 21/1/09 17:16 Page 20