The @MATEC Archives
|Volume 1, Number 1||Deep UV
by: Mike Lesiecki, Ph.D.
In their study of the photolithography process, our learners often record this statement in their notebooks, "illumination wavelength drives feature size." They come away with a general feeling that smaller is better. Take a look at the technology trend that shows the march toward ever-shrinking feature sizes.
In this article lets discuss aspects of light sources, discover advantages and limitations of shorter wavelengths and learn how excimer laser sources can be used in the production of next generation devices.
Mercury lamps, used to generate ultraviolet light near 356 nm (the so-called i-line), produce feature sizes down to 0.3m m. For features in the range 0.25-0.18m m, deep UV (DUV) light from a krypton fluoride (KrF) excimer lasers are employed. For the 0.15m m and lower generation the industry is turning to 193nm lithography using argon fluoride (ArF) excimer lasers as sources.
We tend to teach our students the following relationships:
minimum feature size » l / na
depth of focus » l / na2
You can see that small wavelengths (l ) and large numerical apertures (na) are best for small feature size. However, a large na really hurts the depth of focus. In addition, large na optics have larger chromatic aberrations. This causes different wavelengths of light to be focused at different distances from the lens resulting in a degraded image. Therefore, the current trend is to emphasize shorter l s.
The key message for our learners is that life is not as simple as a convenient shift to ever-shorter wavelengths. One of the biggest challenges at 193nm is the optical materials for the projection systems that transfer the mask image to the wafer surface. A refractive system uses light transmitted through optical elements, typically made from fused silica (quartz). At 193nm most materials including quartz are absorptive, thus creating losses. Quartz lenses also suffer from compaction phenomena which cause light induced density variations in the material. This significantly degrades the optical system. An alternate material (calcium fluoride) is highly transmissive and immune to compaction but is very difficult to fabricate. Successful designs combine the elements of both materials but force restrictions on the laser source itself.
The Linewidth Problem
We usually think of laser light sources as monochromatic. However, its all a matter of degree. The figure below shows the output of a krypton fluoride (KrF) excimer laser near 248nm. The linewidth (the width at half the peak height) is 100pm (picometers). The laser output ranges from about 248.3 to 248.5nm.
This relatively broad output is part of the nature of exicimer lasers (excited
state dimer). In our chemistry courses, we are told that compounds like ArF or KrF
do not exist in their normal states. True, however in the high voltage discharge section
of the laser, the ArF or KrF molecules are created in an excited electronic state and
immediately dissociate after lasing. The lack of a well-defined lower state means that
lasing can occur over a range of different energies (wavelengths). Many DUV systems need
linewidths 100 times smaller for optimum performance. Laser manufacturers like Cymer
An alternate solution is to use a combination of refractive (lenses) and reflective (mirrors) objectives, a so-called catadioptric system.
As a learning activity send your students to the web and search the term catadioptric and look for examples of lithography systems that employ that design.
These systems have a relaxed laser linewidth specification (100-300 pm) which makes life a bit easier. In the FAB you will find whats known as a "mix and match" strategy where high precision (DUV) systems are used to produce critical structures and more conventional i-line systems are used to generate less critical areas (see http://www.cymer.com/PressRoom/techarticles1.html ).
Wrap up your discussion with reminders of the need and drive for smaller feature sizes and highlight some of the complications.
Beyond Laser Technology
Electron beam (e-beam) lithography is advancing to demonstration stages under DARPA, SEMATECH and other funding. A team from Lucent Technologies recently announced feature sizes of 0.08m m ( http://www.lucent.com/press/0796/960708.blb.html ).
A Final Note
"Beyond 0.18m m, life is going to get more difficult."
Mike Lesiecki is the director of MATEC. His degree is in physical chemistry and his research interests include lasers and laser material interactions.