IBM advanced lithography facility: the first five years
07/01/1998
IBM advanced lithography facility: The first five years
L. Grant Lesoine,* Jeffrey A. Leavey, IBM Microelectronics Division, East Fishkill, New York
In five years of industrial R&D operation, the advanced lithography facility (ALF) has proved the viability of synchrotron-radiation-based x-ray lithography for IC manufacturing. The Oxford Helios electron storage ring (ESR) has been shown to be reliable and easily controlled. Specially adapted commercial equipment deals successfully with such unique x-ray lithography issues as keeping large particles out of small mask/wafer gaps.
As feature sizes of ICs shrink, there will come a time when the paradigm must shift and the semiconductor industry must adopt new technology to move forward. Optical lithography has taken us far, but with exponentially increasing costs and technology challenges at each generation. The industry must soon decide what comes next. We believe that x-ray lithography -which we are using today for sub-150-nm development -is the best option for future manufacturing. The IBM ALF has accumulated five years of operational experience as a center for synchrotron-radiation-based x-ray lithography development. This paper will review the operational history and accomplishments of ALF and describe what we have learned about the intricacies of proximity x-ray lithography in an industrial environment.
The table shows some notable milestones in the development of ALF, which became fully operational in early 1992 as an addition to the IBM Advanced Semiconductor Technology Center (ASTC) in East Fishkill, NY. Initially, ALF was equipped with a limited tool set, since it was only expected to provide x-ray exposures, with the ASTC performing the rest of the processing (i.e., resist coating, development, etc.). However, once operational, ALF`s work load quickly increased as IBM and other users began to explore the advantages of x-ray lithography. Today, ALF is preparing to become a complete ASTC lithography sector with 300-mm wafer capability.
 |
ALF program
ALF is a unique facility for proximity x-ray lithography development. The key tool is the ESR synchrotron-radiation source, which was delivered in March 1991. Nine months later, we produced the first significant 0.3-?m ground-rule lithography test exposures (Fig. 1). The ESR provides exposure radiation from as many as 22 beamlines, emanating from superconducting bending magnets on two sides of the racetrack-shaped ring. As the magnetic force changes the direction of the relativistic electrons circulating around the ring, those electrons emit synchrotron radiation at x-ray wavelengths. One RF cavity restores the lost energy to the electron beam. The desire to demonstrate that such a seemingly exotic system could be operated as an industrial facility provided much of the motivation for the development of ALF.
 |
Figure 1. First exposures in ALF: 0.3-?m lines/spaces, 1991.
In 1994, IBM, Lockheed Martin Federal Systems, AT&T, and Motorola formed the Proximity X-ray Lithography Association (PXrL) to explore x-ray lithography for three years, until October 1997. Also in 1994, ALF developed a higher-energy x-ray beamline to perform micromachining experiments using the LIGA process in which very thick resists (200-500 ?m) were exposed to make micromolds, lenses, microelectronic mechanical systems, and other structures [1]. Currently ALF`s customers are focused on early 1-Gbit-generation development. We are also actively seeking partners to join us in other efforts.
The ALF facility overview
The 50,000-sq.-ft ALF facility shown in Fig. 2 is part of the IBM ASTC and contains 3000 sq. ft of Class 1000 cleanroom space. Because beamlines can be located on both ends of the electron storage ring, the ESR and all potential lithography floor space is located on a monolithic slab with 20 ?-in of differential-motion vibration isolation. At present, the beam ports of one magnet feed the cleanroom, while the other magnet provides radiation for a nonclean experimental area [2-4].
 |
Figure 2. ALF schematic building layout. The source area is shown in blue, the cleanroom in pink, and the nonclean area in green.
The ALF can be roughly divided into three areas (Fig. 2): the source and supporting equipment space, the cleanroom area, and the nonclean exposure area. The source area houses Oxford Instruments` Helios-1 ESR, the electron linear accelerator (Linac) injector, power supplies, and the liquid helium cryogenic systems needed for the superconducting magnets.
The ESR source. ESR availability, tracked since 1991, reached greater than 95% in 1993 and continues at about 99% today (Fig. 3), proving that it is the most reliable tool in ALF. The lower availability in September 1997 is attributable to the klystrons (high-powered RF power tubes in the Linac injector), which have reached the end of their anticipated life span and are being replaced as part of normal operation. The cryogenic system captures and reliquifies gaseous helium that has cooled the ESR magnets, thus avoiding the high ($5000/day) cost of purchasing liquid helium. The most serious cryogenics problem was a failure in a liquid helium heat exchanger in 1994. This was repaired and the system modified to prevent a reoccurrence. Should the helium recycling system fail, the ESR can be run for three days using liquid helium stored in a 3000-liter dewar, which can be refilled from a cryogenic tanker truck.
 |
Figure 3. Helios-1 availability.
There have also been two planned shutdowns for ESR upgrades. The first, after six months of operation in 1991, was to switch from the first superconducting dipole magnet to the third (or spare), which had significantly better magnetic field properties. The second, in 1996, was to repair and upgrade the magnets for higher beam-current (and x-ray-flux) operation. These two changes nearly doubled the x-ray output over the initial value. Because of the experience obtained with ALF`s Helios-1 system, Oxford`s Helios-2 design will store 300 mA at initial operation and needs neither of these upgrades. Today, beam currents of 350 mA (far above the original ESR design specification of 200 mA) are routinely stored at full energy and the excellent beam lifetime provides nearly 24 hr of usable x-rays. In fact, the ESR can store approximately 50% more current than we are able to use for x-rays, due to limitations of the present beamline window. New window designs are being considered, however, that will permit ALF to take full advantage of the maximum ESR current capacity.
From very early in ALF`s operational history, Oxford instituted preventative maintenance and detailed tracking of system operations and faults. Each problem has been investigated and appropriate changes made to prevent future occurrences. During the first three years of operation, one day/month and two week-long periods each year were dedicated to maintenance and ESR technical studies to learn the operations envelope of the Helios-1 system. Today, we operate ALF with one major maintenance week/year, but the timing is flexible to meet facility demands. In a production mode, maintenance activities are limited to 6-8 hr/week distributed over a full 24-hr/day, 7 day/week operating schedule.
One criticism of ESR x-ray technology has been that typical technicians cannot operate the source, because synchrotrons are too complicated to use in an industrial setting. ALF has refuted this criticism by training many technicians and contractors with diverse backgrounds to operate the ESR system successfully. Our newest ESR technician, with no synchrotron experience, routinely stored beam in the synchrotron without assistance after about two weeks of on-the-job training (more in-depth training requires several months). In fact, with the continuing improvements in system monitoring and machine protection, our automated control system has achieved unattended overnight and over-weekend operation.
Laboratory environment. The ALF cleanroom has Class 1 or better minienvironments at critical tools (with SMIF handlers on such key tools as the SVG coater and x-ray stepper) and Class 1000 performance elsewhere. X-ray-mask shipping and storage employs a new SMIF-design cassette. ALF can now provide full litho-sector services: coating a variety of customer-requested resists, litho exposures, resist development, CD and overlay metrology, and wafer and mask cleaning. Other tools in Class 1 environments include KLA-Tencor and Inspex wafer inspection systems, a specialized x-ray-mask inspection tool, and a unique 193-nm laser ablation cleaning tool for masks and wafers. After struggling with customer demand for four different wafer sizes (75, 100, 125, and 200 mm), ALF has converted all customers to 200-mm wafers and has begun planning for the change to 300-mm wafers as part of IBM`s recently announced 300-mm strategy.
Initially, two Suss XRS200 steppers scanned the mask and wafer (with their mounting assemblies) through fixed x-ray beams with throughput limited to 4 wafers/hr [5]. One of these steppers produced defect-free 64-Mbit DRAMs in 1994, and is still in operation for resist development work and for wafers other than 200 mm. The newest exposure tool, a 200-mm SVGL aligner, scans the incident x-ray beam to expose the entire chip [6]. This design greatly increases the potential throughput of the tool compared to earlier systems.
Finally, in the nonclean area, one beamline and one exposure chamber are now used for automated full-field exposures (doses up to 5 kJ/day) in projects not requiring cleanliness or a stepper. The plan is to convert this area into a cleanroom, nearly doubling the space available for future steppers.
Recent ALF accomplishments
The Suss steppers provided all of our exposures until the SVGL aligner, developed under a DALP contract, was installed in August 1995 [6]. This stepper is based on and compatible with the commercial 200-mm SVGL Micrascan. The mirror-scanned beamline [7], which rapidly exposes the static wafer and mask in this system, maintains an exposure uniformity of approximately ?2% over a 50-mm wide beam by controlling the scan rate. The present stepper overlay performance is about 35-40 nm (3 s) in x and y; an upgrade is planned for later this year to <30 nm (3 s).
The SVGL stepper is currently the primary exposure tool in ALF, and most recently was used for four levels of 1-Gbit memory test-site exposures. Figure 4 shows two of the four x-ray test-site levels: 175-nm capacitor and 175-nm isolation features. Bit yields for the x-ray-exposed wafers were comparable to optical lots run in parallel: about 99% for the 1-Mbit test arrays. A Power PC gate-level with 150-nm ground rules was printed in ALF during 1997 as a logic chip demonstration. For 1998, demonstrations are planned for memory at 150 nm, logic chips below 150 nm, and sub-100-nm exposures for R&D.
 |
Figure 4. 175-nm x-ray capacitor and isolation levels in 1-Gbit test site.
To demonstrate the extendibility of x-ray lithography, we have printed 70- and 100-nm features at ALF, including the 100-nm lines shown in Fig. 5. Smaller linewidths require smaller mask-wafer gaps, so that a 15-?m gap is needed for 100-nm lines and a 10-?m gap for 70-nm lines [8].We regularly run at 20-?m gaps and will be at 10-?m gaps by next year. Preventing wafer and mask damage by particles in these narrow gaps is a continuing issue.
 |
Figure 5. 100-nm lines/spaces produced by x-ray lithography at ALF.
Controlling particulates and FM
Particulates or foreign matter (FM) on the surfaces of the mask and the wafer are not acceptable for any next-generation lithography. X-rays will print through low-density FM with a thickness below 5 ?m [9]. In practice, the number of printed defects found on x-ray-printed wafers has been about 10-50% of the number reported on the mask itself [10]. This immunity to small defects is a distinct advantage for x-ray lithography, and can help contain cleanroom costs by reducing the need to control very small particulates on the mask and on top of resist-coated wafers.
Of more immediate importance to x-ray lithography is the control of large FM (also called "boulders") that exceeds a specified height limit. Because proximity x-ray requires small gaps between the mask and wafer (typically 20 ?m or less), FM with a height exceeding the gap could cause mask damage. In ALF, we currently require FM height to be less than 1/2 the smallest gap used, or a limit of 5 ?m. This safety factor is not fixed; it just works well for ALF. Any other combination might work, such as 1/4 gap width on the mask and 3/4 on the wafer; the idea is that any two particles will not touch if they happen to pass each other.
Large FM is not ordinarily a concern in a production environment, but boulders can appear on wafers due to mishandling or when processes and tools are under development or out of specification. This FM often ends up buried under process layers or strongly attached to the surface. ALF`s experience is that the number of buried boulders were typically fewer than 5/wafer, but they can be as high as 30 ?m above the wafer surface. The presence of just one of these large surface anomalies requires the wafer to be rejected and not exposed or to be cleaned. Large FM has been observed on masks in ALF, but this more than likely results from the fact that ALF is a development facility and the SVGL aligner is a prototype tool. A production-quality tool is expected to meet much tighter contamination standards.
While a single boulder on a wafer may cause it to be rejected, surface FM on wafers can be entirely removed with a variety of existing techniques. We prefer wafers to receive brush scrubbing, where possible, before x-ray exposure. Once brush scrubbed, we know that any remaining FM is either buried or strongly attached and will require special cleaning. To handle these infrequent special cases, ALF developed a 193-nm excimer laser ablation-cleaning tool called ASAT (ALF Spot Ablation Tool). With the ASAT, we can find and remove single pieces of strongly adhering FM from masks, wafers, or other surfaces. For example, the ASAT was recently used to clean the SVGL stepper x-ray Image Sensor (XrIS), which is used for wafer alignment. Because the XrIS operates at the ~10-?m mask-wafer gap to set alignment, it requires cleaning so that no feature or FM is more than 5-?m tall.
The goal of the ASAT system is to save the wafer for exposure by sacrificing a few of the chips on the wafer. To date, the ASAT tool has been extremely successful, having saved a number of masks and wafers, including some of our recent 1-Gbit test-site wafers. Figure 6 shows the removal of a large boulder from the 1-Gbit gate-level test-site mask recently used in ALF. Other advanced x-ray mask-cleaning methods are in use or being developed at IBM`s Advanced Mask Facility (AMF) in Burlington, VT. The change to TaSi absorber and SiC membrane for x-ray masks has greatly increased the mask`s robustness for piranha / HuangA and laser-ablation cleaning.
 |
Figure 6. a) Removal of a FM boulder from an x-ray mask using the ASAT tool. Part b) shows the partially ablated FM and part c) the fully cleaned mask.
We have developed specialized particle-characterization tooling based on commercially available platforms for mask and wafer inspection to address x-ray lithography issues. That equipment has shown that most FM has a low aspect ratio. Damage-warning systems that assume a spherical FM shape issue a high percentage of false alarms. Incorporating automatic or manual height measurement into the inspection process for large FM solves the problem. ALF`s inspection tooling uses built-in review microscopes to determine FM heights by DOF measurements. Only FM found to exceed the height cutoff are flagged for cleaning. We are also investigating interferometric techniques to automate the height-measurement step, as would be required for high production throughput. In the AMF, masks are inspected with both specially modified wafer tools and a KLA-Tencor SEMSpec. The SEMSpec allows defect detection down to about 70 nm in the pattern area, but must be supplemented with other tooling to inspect the remainder of the mask for large FM.
As a backup to boulder checks before exposing wafers, SVGL is developing a Particle Detection System (PDS) that is located in the stepper itself. After wafer leveling is complete, the wafer is passed in front of the PDS, which is set to detect FM exceeding the gap. If a particle is found that does exceed the gap, the wafer is rejected, preventing mask damage.
FM control for x-ray lithography is like that for optical lithography except that it requires FM height awareness. We have developed the inspection and cleaning techniques that meet today`s needs and are extendible to future ground-rule requirements.
Conclusion
Today, the focus in ALF has changed from such R&D questions as: "Can it be done? Will it work? Is it reliable?" to concerns about the technical details and manufacturing infrastructure of a working technology. We have progressed over the years from 350-nm devices to 180 nm and to sub-100 nm for litho development. We have received defect-free masks with 0.18-?m ground rules, and the IBM AMF has completed the transition from gold to refractory metal. Newly established shipping and handling processes maintain the cleanliness and low defect levels of these masks.
Minienvironments and robot-automated processes planned for the 300-mm program will also benefit x-ray lithography. Since the SVGL x-ray aligner uses the same technology and alignment marks as the Micrascan, x-ray to optical mix-and-match is easier to perform. The recent storage ring upgrade allows the source to run at more than two times the previous flux levels, giving higher throughput with the promise of even more to come. The synchrotron vendor has developed a new model of ESR that has better automation and control features plus simplified maintenance and significantly higher flux output [11]. Faster resists that will further improve throughput are now under test.
Over more than five years of operation, the x-ray source and beamlines in ALF have exceeded the original expectations for reliability, usability, and function. The first-generation exposure and processing tools have kept up with the demands placed on them, and are improving. The cleanroom has been augmented with minienvironments and automated handlers to meet the ever-more-stringent wafer- and mask-handling requirements. A manufacturing-capable stepper and an advanced e-beam mask-writer are in the planning stages, both of which are necessary for full production using x-ray lithography.
However, ALF is capable of providing samples of very small features not readily available otherwise. This work is necessary to substantiate device models and for early learning in etch and other semiconductor processes at IBM and our partners. ALF plans to make integrated devices with =100-nm ground rules, and will be defining the processes that can be used in manufacturing these devices.
One major Japanese semiconductor company [12] has recently reaffirmed its intention to use x-ray lithography to manufacture semiconductor devices in a new facility to be completed around the year 2000. Within IBM, ALF recently completed three lots with four levels of exposure (capacitor, isolation, gate, and metal) of a 180-nm, 1-Gbit test site in partnership with the ASTC. In November 1997, as part of a SEMATECH contract to explore mask lifetime, ALF completed 3286 wafer passes with a single mask over 12 days of 24 hr/day running. This was the first test of its kind under conditions similar to those in manufacturing. The SVGL aligner and ESR performed extremely well. With its advantages in DOF, image quality, and extendibility to 70-nm ground rules, x-ray lithography stands as the most likely successor to conventional optical lithography.
Acknowledgment
SMIF is a trademark of Asyst Technologies Inc. and Micrascan is a trademark of SVG Lithography Inc.
References
1. D. Morris et al., "Micromachining Using Helios," SPIE, 2437, pp. 134-138, 1995.
2. L.G. Lesoine et al., "ALF: A Facility for X-ray Lithography," 1263, Proc. SPIE, pp. 131-139, 1990.
3. L.G. Lesoine, K. Kukkonen, J. Leavey, "ALF: A Facility for X-ray Lithography II: A Progress Report," Proc. SPIE, 1671, 299-311, 1992.
4. J. Leavey, L.G. Lesoine, "Design Considerations for the IBM X-ray Lithography Facility," IBM J. of Research and Development, Vol. 37, No. 3, 1993.
5. A. Chen et al., "First X-ray Stepper in IBM Advanced Lithography Facility," J. Vac. Sci. Technol. B, Vol. 10, No. 6, p. 2628, 1992.
6. A. Chen et al., "Evaluation of the Defense Advanced Lithography Program (DALP) X-ray Lithography Aligner," Proc. SPIE, 3048, p. 200, 1997.
7. Private communication, R. Rippstein, IBM Microelectronics Division, 1997.
8. H.I. Smith, F. Cerrina, "X-ray Lithography for ULSI Manufacturing," Microlithography World, Vol. 6, No. 1, p. 10, 1997.
9. B. Bollepalli et al., "Simulation of X-ray Mask Defect Printibility," Proc. SPIE, 3048, p. 155, 1997.
10. ALF internal review of mask verification prints by J. Leavey, 1997.
11. Helios-2 literature, Oxford Instruments.
12. Dempa Newspaper, Mitsubishi, March 12, 1997.
*On assignment to Lockheed Martin Federal Systems under DARPA contract N00019-94-C-0035, which partially supported this project.
L. GRANT LESOINE works in IBM`s x-ray lithography program office. He coordinated the construction of ALF and is responsible for its operation. IBM Advanced Lithography Facility, 1580 Route 52, Hopewell Junction, NY 12533; ph 914/892-3833, fax 914/892-6043, e-mail [email protected].
JEFFREY A. LEAVEY joined the ALF team in 1992. He is an advisory engineer in radiation safety, contamination control, and tool availability.