157nm optical lithography: The accomplishments and the challenges
06/01/1999
 |
In a relatively short period, 157nm lithography has gone from "not possible" to the forefront of optical lithography development efforts. While significant challenges remain, particularly for mask and pellicle technologies, recent but quiet development work on calcium fluoride, fluorine-laser sources, and alternative mask substrates has brought the industry back to a "can do" attitude.
From our view today, it is not unlikely that, as 193nm exposure tools evolve from current 0.6 NA to 0.75 NA and with the use of strong phase-shift masks, 193nm lithography may offer partial solutions into the 100nm technology node. But what are the lithography solutions after that?
The Semiconductor Industry Association (SIA) Roadmap clearly indicates a need for a solid solution for the 100nm node by 2005, supporting pilot chip sets. But many semiconductor manufacturers are working well in advance of the SIA Roadmap and want a 100nm-node lithography solution as early as 2003. Their concerns are twofold:
- The performance of 193nm lithography, even with strong phase masks, is a questionable total solution to 100nm-node needs.
- Projected availabilities - estimates vary dramatically from 2005-2006 and beyond - for so-called next-generation lithography (NGL) methods, such as extreme ultraviolet (EUV) and scattering with angular limitation projection electron-beam lithography (SCALPEL), are too far out to meet anticipated needs.
Indeed, the industry seems to be facing a technology gap at the 100nm-technology node (see Fig. 1 below) that needs to be filled with 157nm optical lithography.
Why 157nm now?
 |
As recently as one year ago, very few semiconductor manufacturers or semiconductor capital equipment companies gave any consideration to using 157nm lithography because the basic enabling technology was not available. One of the few known optical materials capable of transmitting 157nm laser energy - calcium fluoride (CaF2) - was not considered to be of sufficient optical quality and could not be fabricated in large enough size to support the approach. Other materials such as magnesium fluoride (MgF2), which has unacceptable intrinsic optical birefringence and lithium fluoride (LiF), which is relatively soft and hydroscopic, were also discounted. In addition, 157nm fluorine-laser source technology (F2) was immature, and there were serious questions about its stability, repetition rate, overall power, and bandwidth. The conclusion then was that 157nm lithography was not feasible.
So why is it attractive today? Significant advances have been made in enabling technologies. Both F2 laser and CaF2 material development have been quietly worked on to ensure credible results before declaration that 157nm lithography is possible.
Lambda Physik has significant experience with an installed base of >40 commercial 157nm F2 lasers in nonlithographic applications. They extended this application and executed a 157nm development program for lithography, demonstrating extremely encouraging results that show a 600Hz, 6W laser is possible.
Specifically, its data showed these lasers had a relatively wide 2pm bandwidth. Because all refractive designs would require very narrow bandwidths (<0.1pm at 157nm ), the only exposure tool solution may possibly be a beamsplitter catadioptric design (i.e., combination of lenses and mirrors with a beamsplitter cube that folds the beam over itself, see Fig. 2). This design is similar to SVG Lithography`s (SVGL`s) optical designs used in both 248nm Micrascan II and III and Micrascan 193nm systems.
Our analysis has shown that the optical design, while following a similar form as previous Micrascan systems, will require all transmitting components to be made of CaF2 material. Figure 2 illustrates that the 157nm-optics form is only 25% larger than the Micrascan III design. This is driven principally by an increase in the 157nm numerical aperture (NA), namely 0.75 to support <100nm resolution.
Making the 157nm cube beamsplitter out of extremely high-quality CaF2 represents a significant challenge above and beyond that of making a transmitting lens. Thus, while lenses only need transmission quality in one direction, the cube, which is as big or bigger than its corresponding lenses, needs optical transmission quality in two directions. (Figure 2 shows how the optical path uses two directions, 90? apart, of the beamsplitter cube.)
To demonstrate that fabricating this key component was possible, SVGL joined forces with Schott Glaswerke of Germany to develop a 1/5th (volume) scale version of the cube material (Fig. 3). This collaboration took over 1.5 years, but resulted in the successful growth and characterization of a single-crystal CaF2 cube >100mm on a side with good optical quality in all three axes, successfully demonstrating that this material is feasible.
Continuing this development work, SVGL and Schott Glaswerke are now fabricating a full-scale 175mm CaF2 cube. In addition, we have begun similar CaF2 development efforts with Bicron (Cincinnati, OH) and Optovac (Sturbridge, MA) to ensure an adequate supply and second sourcing.
These advances with F2 lasers and CaF2 material have clearly changed the conventional wisdom about 157nm lithography from less than a year ago. While 157nm optical lithography still requires full development of a resist and processing infrastructure, the basic approach is continued evolution (not a revolution as some contend) of reduction optical imaging: an optical "engine" uniformly illuminated by a laser light source that illuminates and reduces a transmission reticle pattern onto a wafer. Our approach will be to couple this engine within scanning to produce a 157nm step-and-scan system that provides the benefits of a large 26mm X 34mm scanned field and excellent across-the-chip linewidth control and uniformity.
Yet we cannot dismiss the challenges that are still in front of us before this technology can be considered viable.
Addressing 157nm challenges
Participants at last year`s Next Generation Workshop (December 1998) voted 157nm as the leading lithography contender of choice for the 100nm technology node. In parallel, International Sematech has begun a 157nm feasibility program - a corporate research and development agreement (CRADA) with MIT Lincoln Labs. Industry participants have been invited to participate in this CRADA by submitting development samples that demonstrate feasibility for materials, optical coatings, mask material and fabrication, and pellicle material and fabrication - the infrastructure needed to bring about successful 157nm technology. This program is well underway; for example, Lambda Physik has installed F2 lasers at MIT Lincoln Labs and initial test results demonstrating material feasibility are expected about July 1999. There have already been some good results reported from initial laser testing of fused silica materials, indicating that the materials may be suitable for 157nm masks.
Also, International SEMATECH sponsored a 157nm workshop (February 1999, Phoenix, AZ) that was attended by more than 145 managers and engineers from 54 companies and organizations. Attendees heard 42 presentations about 157nm lithography materials, exposure tools, lasers, reticles, pellicles, metrology, and resists. This workshop provided a common forum for the international community and gave the supplier infrastructure a chance to review existing data, plans, strategies, and the related challenges pertinent to 157nm optical lithography solutions.
CaF2 optics
The size of the single crystal needed for a 157nm catadioptric optical design, as we outlined above, is only part of the CaF2 challenge. This material must be manufactured from extremely high-purity synthetic crystalline material. Basically, purified raw starting material, in a powder form, is put into a high-temperature chamber and melted. A single-seed crystal contacts the molten liquid and slowly pulls out a large single crystal. When finished, the crystal must exhibit superior optical quality (i.e., low strain and homogeneous to <0.5ppm) in two dimensions and be free from impurities that can lead to absorption problems or color-center formations that are susceptible to laser irradiation and resultant laser damage.
The crystal goes through a week-long growing process and is cooled in another week-long controlled process that minimizes strain-induced birefringence. SVGL is working closely with its partners to ensure not only the development of the CaF2 cube and its quality, but also to ensure capacity for lens material. This is important as needs for 193nm lithography and 157nm lithography increase.
Once large enough CaF2 crystals are available, these must be worked into finished optical elements. CaF2 is much different than glass and fused silica and requires special processing and precautions during fabrication. For example, the CaF2 required for 157nm lithography must be single crystalline to satisfy the stringent optical qualities, but single-crystal material presents some manufacturing challenges not normally associated with working fused silica:
- Like all crystalline materials, CaF2 has a natural cleavage plane that is susceptible to temperature and shock effects.
- CaF2 has a thermal expansion coefficient 36X greater than fused silica (see table), so heating and cooling must be carefully controlled when it is fabricated into an optical component. Interestingly, CaF2 has a 7X greater thermal conductivity than fused silica. This negates some of the thermal expansion, which also means CaF2 will do a significantly better job of dumping absorbed energy during a 157nm-exposure cycle.
- CaF2 is somewhat softer than fused silica (i.e., ~0.5 knoop value or measure of hardness) and is prone to sleeks and scratches during polishing and handling. Special polishing and handling procedures are required.
Certainly optical engineers are not afraid of CaF2; it has been used as an optical material for decades. They are aware, however, that they need to understand and respect its properties fully, because they are taking it to a higher level of sophistication.
Optical coatings
There are few optical coating materials available that exhibit low absorption (specific requirements are proprietary) at 157nm and also provide the range of index of refraction (i.e., 1.40-1.80 ) needed to provide both antireflective coatings and the required beamsplitter coatings. Interestingly, 193nm lithography faced this same problem. Fortunately, it appears that some of the types of materials, basically fluorides, that were found and developed for 193nm optical-coating solutions also have good transmission at 157nm and will become the initial basis for coating efforts.
SVGL has initiated a substantial in-house coating development effort and has already started deposition of initial coatings. We are also participating with MIT`s testing, where the coatings will undergo characterization with laser irradiation, specifically looking for laser-induced damage mechanisms.
F2 laser source
The natural F2 laser spectrum provides a two-line emission -157.523nm and 157.629nm. The 157.629nm is the strongest line (Fig. 4), ~5X the spectral brightness of the smaller 157.523nm line.
Lambda Physik`s development activities and the early availability of a 157.6299nm line on its NovaLine F2 laser operating at 600Hz and 6W of power helped enable the SVGL 157nm lithography development approach. It demonstrated that a light source was feasible and this encouraged SVGL to continue to pursue CaF2 material development. Lambda Physik`s patented resonator design allows up to 10W of power in single-line operation [1]. The company plans to deliver 1000Hz 10W systems by the end of 1999 and 2000Hz 20W lasers for SVGL full-field systems in 2001. Dose energy stability performance is <0.6% in a 50-pulse moving average.
SVGL is also forming a relationship with Cymer to develop a second source for 157nm lasers. Cymer, well known for its 248nm lasers, has also been working on an F2 system and has shown initial data similar to the Lambda results [2].
Both companies have measured their respective lasers and have verified the laser bandwidth to be <2pm (Fig. 4), which fits well within the catadioptric design bandwidth requirements, as explained above.
Masks
Maskmaking for 157nm lithography is perhaps the biggest concern among users. Of course, feature sizes <100nm put tougher demands on mask writing, but the limited materials available for a mask substrate add a significant complication. CaF2 and MgF2 are candidates because they transmit 157nm energy. Unfortunately, the substantial inherent birefringence of the latter possibly precludes its use as an optical reticle material.
The large thermal expansion coefficient of CaF2 is problematic during mask writing, because as the mask is being written, energy is also absorbed, causing the substrate to expand and distort feature placement. Mask-writing experiments on CaF2 have been conducted demonstrating that it is possible to write a CaF2 mask, but with some degree of feature placement error. Much more experimentation needs to be conducted to see if the writing errors can be compensated.
Today, a new version of a familiar material in lithography -fused silica -shows significant promise as a mask substrate for 157nm lithography. Silica used for lithography projection optics typically contains a high amount of structural OH groups. This species absorbs at short wavelengths and therefore limits the utility of this particular blend of fused silica for 157nm transmission. However, high transmission in the deep ultraviolet (DUV) at 157nm is achieved by minimizing OH content in silica. Corning is taking this approach and the concept has already been confirmed [3]; such materials have already been manufactured, tested, and demonstrated to have 70-80% transmission at 157nm. Nikon`s presentation at the 157nm workshop suggested that fluorine doping of dry silica can also be an approach to further increase the 157nm transmission.
Both of these fused silica materials show significant promise as a 157nm mask substrate and offer the advantages of reduced thermal expansion during mask writing and exposure, thus minimizing distortion effects. Four companies have development efforts in process for fused silica mask material solutions: Corning, Nikon, Heraeus, and Asahi Glass.
The susceptibility of these newer materials to 157nm laser-induced damage, either in the form of material compaction or induced absorption, is being investigated in the Sematech-MIT CRADA initiative.
Pellicles
At 157nm exposure, pellicles present two problems:
- finding a pellicle membrane material that transmits at 157nm and
- purging or effectively evacuating and maintaining an N2 environment in the 5mm pellicle-to-mask surface gap without the presence of dead air spaces that could cause significant 157nm energy loss, and doing so without a detrimental effect on system reticle change time.
The nitrogen purge environment is important because an air environment provides an absorption medium for 157nm energy. In fact, the entire optical path including the reticle and the wafer must be a purge nitrogen environment. Here again, the Sematech- sponsored MIT CRADA is pursuing an effective solution to the pellicle material issue.
Resists
Users` concerns about 157nm resist issues are similar to those about masks. At the recent 157nm workshop, MIT Lincoln Labs reported on an earlier DARPA-funded effort that yielded 80nm features printed with a simple 157nm "tool" microscope using a silylation resist process [4]. Comments during the workshop were that 157nm resists had more in common with 193nm resists than with 13nm extreme ultraviolet (EUV) processing. There is some consensus that existing commercial resists, if limited to 50-90nm thickness, might be usable for initial tool testing. However, this is not expected to be the answer for full tool development where an estimated 10mJ/cm2 exposure dose will be needed and thinner resists are prone to more defects. In fact it appears that new resist chemistries need to be developed and hard-mask processing might be required.
The saving speculation is, considering the competitive nature of semiconductor manufacturers, we would not be surprised if some are already developing the needed resists or at least giving the problem serious consideration.
The platform
Even if all the optics problems and other challenges noted above are taken care of, there is still the need for an integrated platform capable of supporting 100nm and 70nm geometries. This platform also has to provide a solution that improves both 200mm and 300mm wafer throughputs, improves overlay driven by the smaller feature size, improves focus setting and control to improve customer process latitude, and provides compensation techniques to correct for thermal-induced errors.
Obviously, acceptable error budget allocations that accommodate CD control for 150nm and 130nm features are no longer permitted when the exposure system must support 70nm features with single-digit linewidth control. In almost all cases, the smaller feature sizes drive an error budget allocation that in turn drives significant improvements to all the exposure subsystems.
At SVGL we have begun development of a new platform capable of handling 200mm and 300mm wafers, and incorporating improved module designs that support substantially more stringent linewidth control requirements. We are addressing enhanced wafer throughput and improved alignment and overlay, targeting 25nm within a system and 35nm system-to-system. In addition, we are working on enhanced process latitude by improving focus control and reducing optical and stage contributions to the overall depth-of-focus budget.
Our approach will be incorporated in a new body basis for a VHNA (very high NA), large-field 193nm and 157nm exposure system (Fig. 5). Both the VHNA 193nm and the subsequent VHNA 157nm system will provide a 0.4-0.75 variable NA coupled with variable 0.3-0.8s illumination. In addition, both will provide a full array of illumination fills: standard, annula, and quadrapole.
The 157nm optical train puts a significant burden on the platform because the entire optical path, including the reticle and wafer space, must be purged free of oxygen. This must be done while allowing for movement of both reticles and wafers as scanning takes place without impact on system throughput and performance. Fortunately, SVGL has developed substantial integrated purging and stage technology, fabricated and proven under our previous x-ray lithography development program; this work has a direct correlation to the work at hand.
Conclusion
A significant amount of work is still needed to position the remaining infrastructure and provide 157nm optical lithography as the solution for the 100nm "lithography gap" that exists for the semiconductor industry. However, the industry in general is rising to the challenge and has already brought forth amazing results in a relatively short time. The momentum that this effort has already gathered demonstrates the industry`s interest in finding an optical solution for the 100nm technology node.
SVGL is developing an evolutionary solution that makes use of almost three decades of optical and scanning experience as the body basis for its 157nm approach. We will have a 4mm X 22mm 157nm miniscanner in 4Q 00, a vehicle for resist and process development. To accomplish this, SVGL is joining forces with Tropel, incorporating its small-field optical train into a modified SVGL Micrascan III system. This miniscanner will also provide overlay capability for users who want to do early chip development. Our 26mm X 34mm full-field system is also under development at SVGL and will be available in 1Q 02. n
Acknowledgment
We thank the following individuals for written and verbal discussions in preparation for this article: Bruce Tirri and Tom Fahey, SVGL; Charlene Smith, Corning; Cliff Martin, Odyssey; Robert Willard, Lambda Physik; Ewald Morsen, Schott Glaswerke; Morty Rothschild, MIT. We commend International Sematech for sponsoring the 157nm workshop and the CRADA at MIT Lincoln Labs, which has put 157nm on a fast track.
Micrascan is a registered trademark of SVGL. NovaLine is a registered trademark of Lambda Physik. Micrascan is a registered trademark of SVGL.
References
- U. Stamm et al., "Excimer Laser for 157nm Lithography," Proceedings of SPIE, Volume 3676, paper 84, pending publication.
- T. Hoffman et al., "Revisiting F2 Laser for DUV Microlithography," Proceedings of SPIE, Volume 3679, paper 49, pending publication.
- C. Smith, L.A. Moore, "Fused Silica for 157nm Transmittance," Proceedings of SPIE, Volume 3676, paper 95, pending publication.
- T.M. Bloomstein et al., "Lithography with 157nm Lasers," JVST B, Vol. 15, No. 6, Microelectronics and Nanometer Structures, MIT, pp. 2112-2116, Nov./Dec.1997.
Authors
 |
James A. McClay received his ASEE at Salem Tech and BSEE at Bridgeport Engineering Institute. He has more than 18 years of experience in lithography development and manufacturing. McClay is vice president of the 157nm program at SVG Lithography, 77 Danbury Rd., Wilton, CT 06897-0877; ph 203/761-4422, fax 203/761-6388, e-mail [email protected].
 |
Angela S.L. McIntyre received her MBA from Sloan School of Management at MIT and her MS in material science at MIT. She has worked at TI as an R&D collaborations manager. McIntyre is currently project manager for lithography capital equipment at Intel Corp., 2200 Mission College Blvd., Santa Clara, CA 95052-8119; ph 408/653-7499, fax 408/765-2162, e-mail [email protected].