|
DAMAGE RESISTANT
OPTICS FOR EXCIMER BEAM DELIVERY
Long life,
damage resistant coatings are improving the economics of commercial
excimer applications.
By David Collier
and Wayne Pantley, Alpine Research Optics
David Collier is
President and Wayne Pantley is Sales Manager of Alpine Research Optics,
3180 Sterling Circle, Boulder, CO 80301, 303-444-3420, FAX 303-444-1686,
E-mail AROcorp@AROcorp.com, http://www.optics.org/arocorp/
Introduction
The excimer
laser's unique combination of output characteristics make it an ideal
choice for a wide variety of precision materials processing and biomedical
applications. Advances in excimer laser reliability and economy over the
past few years have now made it a practical source for many of these
applications; in fact, excimers represent one of the fastest growth areas
in the entire laser marketplace. However, until recently, optical damage
to the beam delivery optics, and to the coatings in particular, has
limited excimer based system reliability. In this article, we will examine
how optics manufacturers have addressed this problem, extending typical
coating lifetimes by over an order of magnitude. We will also see how new
research programs should lead to even further improvements in damage
resistance.
Excimer
Applications
To understand
the economic impact of optics failure, let's consider the two fastest
growing applications areas for excimer lasers - photorefractive keratotomy
(PRK) and microlithography.
In PRK, beam
uniformity and pulse energy must be tightly controlled to ensure a
successful surgical outcome. These parameters are directly affected by any
degradation in the multilayer dielectric coatings on the mirrors and
beamsplitters used in the beam delivery system. Until recently, many
systems in the field required several optics to be replaced during
scheduled service and maintenance, with a typical service interval of 4-6
months. Also, the total cost of the system optics (up to $10K) is quite
significant, compared for example, to the typical service contract costs
of around $40-50K/year.
George Caudle is
a Staff Laser Scientist at Visx, a leading supplier of PRK systems. He
explains, "At first, optics lifetimes definitely represented a "weak link"
in the system, that we felt could be improved upon by working with vendors
of these optics."
After years of
using successively shorter mercury arc lamp wavelengths, microlithography
is now turning to excimer lasers to provide the resolution for the next
generation of integrated circuits (IC's). The retooling of IC
manufacturers with 248nm krypton fluoride systems is now well under way
with sales of several hundred systems predicted for 1997. In addition,
prototype systems based on 193nm have already been developed and are
expected to arrive on the production floor within 3-5 years.
Microlithography
uses lower fluences (1-50mJ/cm2/pulse) than PRK (>100mJ/cm2/pulse). Yet
optics failure is much more of a problem for this application because of
the economics of continuous equipment operation. Simply stated, these
machines are expected to operate virtually around the clock, and even
routine (monthly) scheduled maintenance time is minimized. As a result,
the total number of laser pulses is astronomical; a 1kHz laser delivers
over 1 million pulses in less than 20 minutes.
Estimates for
total downtime costs for an IC manufacturer range up to several hundred
thousand dollars/hour for a 248nm system. These systems cost over $6
million each, whereas 193nm systems are expected to cost up to $15
million. Operational economics will be similarly scaled up, and downtime
is therefore projected to cost over $1 million/hour!
Defining Optics
Failure
In the early
1990's a typical 193nm turning mirror used in a PRK beam delivery system
had a useful lifetime of 0.5 million pulses or less. What do we mean by
useful lifetime? With high energy optics used for pulsed Nd:YAG lasers at
1064nm, optics failure has usually meant catastrophic failure - the
coating is pitted, burned or partially blown off the substrate. In the
case of short wavelength excimer photons, damage mechanisms are different,
and performance deterioration may occur a long time before catastrophic
failure. For a high reflector, failure has therefore been defined by a
decrease in reflectivity to some arbitrary level.
For both PRK and
microlithography, a 5% decrease in reflectivity often constitutes the
definition of failure. In the case of PRK, a decrease in coating
uniformity of more than a few percent would also constitute failure. The
5% value derives from the large number of optics which comprise a system.
A typical lithography system may use as many as 12 fold mirrors just for
the laser projection section - and this doesn't even include those used
for remote laser delivery. A 5% drop in reflectivity on 12 mirrors used
serially results in a 50% decrease in delivered laser energy. In
lithography, decreased pulse energy slows the exposure process, costing
the end user time and money.
In addition,
excimer laser damage is an accelerated process - a 5% drop in reflectivity
will quickly become a 10% decrease. End user economics dictate that the
optics be replaced as soon as they begin to show noticeable (-5%) signs of
deterioration - and certainly long before catastrophic failure is even a
possibility.
Improving
Coatings
In 1994, we
began a program at Alpine Research Optics to develop longer life coatings
for both 193nm and 248nm. The first obstacle we encountered was a lack of
information and basic understanding about both the mechanism(s) leading to
damage and failure, as well as reliable testing protocols to certify
batches of optics. It was soon apparent that what we knew about producing
long-life laser mirrors for the visible and near-IR didn't apply in the
case of the deep UV.
For example,
with pulsed Nd:YAG lasers at 1064nm and 532nm, it is well known that
coating damage is related to peak power. Indeed, lamp-pumped Nd:YAG lasers
with beam hot spots are notorious for quickly producing burns and pits in
both coatings and substrates. Thermomechanical stress, choice of
materials, and poor substrate preparation or cleaning techniques, all play
a role in this type of catastrophic damage.
In contrast,
excimer laser damage appears to correlate more closely with total
accumulated flux. The operating atmosphere also plays an important role.
Visx's Caudle notes that "We used to see significant optics lifetime
differences between individual medical facilities. We suspected this was
due to differences in air quality. This was confirmed by enclosing the
optics in a closed filtered environment, which significantly reduced these
site-specific performance anomalies."
Possible
mechanisms for decreased coating reflectivity include photooxidation
induced refractive index changes and/or absorption increases, problems
with the coating/substrate interface, and solarization of the coating
layers. Fortunately, research programs have now been funded to provide
definitive answers (see side-bar) to the damage question.
In the area of
testing, the qualifying specification for excimer optics used to be
resistance to several high energy pulses (2J/cm2). This is perfectly
acceptable for Nd:YAG optics but fairly useless for 248 and 193nm
excimers. Indeed, as our work progressed we found that coating designs
could be independently optimized for either high peak power resistance or
longer life.
With no rapid
testing method and no definitive information about failure mechanisms, we
had to partner with our OEM customers. Over a two year period, we
systematically refined coating designs and chamber conditions in an
iterative manner - relying on feedback from these end users about the
actual performance of each batch of coatings.
Improving the
248nm coatings proved to be a much simpler proposition than those for
193nm. The principal reason is that hafnium oxide is a well understood
coating material that can be used for the high index layers. But HfO2
begins absorbing at l<226nm, so this couldn't be used at 193nm.
Instead, we had to develop a unique, proprietary design based on a
combination of fluorides for both the high and low index layers. However,
we decided not to use thorium fluoride, a material sometimes used by
others because of its desirable optical properties. We avoided thorium
fluoride because it is a toxic, radioactive material which requires EPA
and OSHA approved handling and disposal.
We also found
that one of the most critical parameters in producing stable coatings is
the substrate temperature during deposition. If the temperature of the
optic is too low, the fluorides deposit poorly with unpredictable
refractive index and shortened lifetime. On the other hand, if the
temperature is too high, the coatings tend to craze as the optics are
cooled.
Finally, it may
surprise some readers to learn that we use a conventional e-beam to
evaporate the fluoride coating materials. Several techniques, most notably
ion-assisted deposition, have been developed in recent years, with the
goal of improving layer hardness and consistency. But in the case of the
fluorides used for deep UV mirrors, ion-assisted deposition typically
increases the absorption at these wavelengths, thereby actually reducing
the lifetime of the optic.
The other
advantage of e-beam evaporation is that good material distribution and
consistency can be obtained throughout the coating chamber. By ensuring
similar uniformity of substrate temperature, we are able to produce
excimer beam delivery optics in batches of 1-100 with diameters up to 8".
Results
Continuous
incremental improvements to our coating process have now increased 193nm
mirror lifetime from 0.5 million to 5 million pulses at 200mJ/cm2, as
certified by Spica Technologies. As Visx's Caudle notes, "We have been
shipping our latest model PRK system for about 18 months. In that time
we've seen a dramatic reduction in optics failure."
In addition to
the improvements summarized, we have found that these coatings can be
optimized for high peak power usage or extreme lifetime. This bodes well
for 193nm lithography, given that this economically important application
will use 10X lower pulse energies than PRK.
Of course,
longer life, more reliable optics will now benefit all excimer users, from
plasma physicists to atmospheric chemists to micromachining job shops.
After years of the commercial sector being serviced by trickle-down
products from research and aerospace applications, we believe that this
reverse process is representative of an important trend - indicating the
continuing maturation of the electro-optics industry.
Damage Research
and Testing
Producing
long-life high reflectors for excimer applications has been hampered by
the lack of both definitive information about failure mechanisms, and
lifetest data and models. Work now being conducted in these areas should
soon lead to further significant improvements in the reliability of these
optics.
Companies such
as Spica Technologies Inc. (Nashua, NH) now offer rigorous life testing of
optics as a service for both manufacturers and end users. According to
Mike Thomas, Chief Scientist and President of Spica, "At present, true
lifetesting of these optics is essential. In the past we have avoided this
type of testing because it is very time consuming and requires several
dedicated excimer lasers. A recent cooperative agreement with J.P. Sercel
Associates, Inc. (Hollis, NH) however, combines their excimer resources
and expertise with our knowledge and experience of testing optics. This
has enabled us to offer the first commercial life testing of excimer
optics from 157nm - 308nm.
"As we collect
additional data, we hope the results of these experiments will lead to the
development of a life test model. This may allow us to reliably
extrapolate shorter test data, which will be to everyone's advantage."
The testing and
manufacturing of these optics should be further simplified once the damage
mechanisms are completely understood. One of several research programs
aimed at this problem was recently funded by SEMATECH at MIT Lincoln
Laboratory. This group plans to use a variety of optics techniques to
monitor the changes in microscopic and macroscopic properties of single
and multilayer coatings. "Most of the existing research on excimer damage
of optics has focused on the bulk materials," according to Lincoln Lab's
Mordechai Rothschild. "With multielement lithography lenses for 193nm
costing several million dollars each, naturally attention has been
directed on finding long life substrates for the lenses. But now, the
importance of prolonging the life of their AR coatings and the reflective
optics has been fully recognized."
|