Excimer lasers for future lithography light sources
07/01/2000
Olivier Semprez, Cymer Inc., San Diego, California
OVERVIEW
To sustain optical lithography for the next five years, the industry must systematically address the technical challenges that will be encountered in developing and using the progression of excimer laser light sources needed, specifically KrF, ArF, and F2 lasers. While the challenges are formidable, there are no fundamental limits to the physics behind excimer laser performance.
Goals of the 1999 ITRS through 2005 put optical lithography on the semiconductor industry's critical technology path. Hence, continuous enhancements to lithographic ultraviolet light source technologies (i.e, excimer KrF for 248nm, ArF for 193nm, and F2 for 157nm) will, of necessity, be one of the enabling prerequisites. Deep UV laser sources will dominate what many experts believe will be optical lithography's final arena before an industry jump to the post-optical era.
The KrF laser has become the light source of choice for semiconductor production lithography at 250nm. Enhanced by optical techniques such as phase shifting and off-axis illumination, KrF is in place for ramp-up of 180nm IC production. ArF lasers will displace KrF between 130nm and 100nm, albeit not until ArF maturity is reached in production, sometime around 2002. ArF has the potential to be the dominant source down to the 100nm to 70nm technology nodes. Below 70nm, the F2 laser now appears as a possibility.
Laser characteristics
Excimer lasers are basically high-voltage pulse discharge devices operating in a gas-filled optical cavity. Their emission characteristics, including intensity and spectral bandwidth, vary significantly with the lasing medium and with the specific optical, mechanical, and electrical characteristics of the laser system design. The discharge chamber and associated optical modules are critical components, but so are the solid-state, pulse-power modules that trigger the light output. System phenomena that affect ultimate lithographic performance are related to electrical and thermal conditions arising during gas excitation by the pulse discharge, and to very critical discharge-related aerodynamic events that can adversely affect laser stability and that must be controlled to achieve necessary lithographic tolerances.
Figure 1. Improved KrF laser pulse-energy stability vs. repetition rate. |
Photonic energy output is a critical laser parameter for lithography. For example, recently improved output characteristics of KrF systems are such that the relationship between output power and laser repetition rate is essentially linear and the <2.5% (3s) deviation of stability over the total pulse repetition range to 2 kHz (Fig. 1) is significant for consistent production lithography [1].
Figure 2. a) FWHM and b) 95% integral spectrum pulse bandwidth definitions. |
Two important parameters define the shape and energy content of a laser pulse. "Full width at half maximum" (DlFWHM) is the traditional measure. Recent experience and analysis has shown, however, that a significant amount of energy can reside in the tails of the pulse, energy that can have a significant effect on resist exposure. Hence, the industry has adopted a second measure of spectral pulse width called dl95%; it defines the spectral range that includes 95% of the energy of the pulse (Fig. 2).
KrF emission is centered at 248nm, but the natural emitted pulse is 300pm wide. For lithography, this creates a problem that requires line narrowing. The dispersion (the change of optical index with wavelength) of optical components in the DUV region is high. Thus, spectral bandwidths, even in the picometer range, can produce significant chromatic aberration in systems with refractive optical components, making it difficult or even impossible to focus and image lithographic patterns properly. Fortunately, means exist to narrow sharply the spectral output of the laser beam so that lithography within acceptable dimensional tolerances can be accomplished.
UV excimer laser engineering is presently focused on optimizing pulse energy content and bandwidth, increasing available optical power for enhancing lithographic throughput, stabilizing pulse energy output to meet the increasingly tight exposure-dependent dimensional specifications of lithography at advancing technology nodes, and decreasing net semiconductor manufacturing cost. Therefore, it is important to address how this is being done for production KrF sources and how these developments will affect next-generation ArF and F2 sources.
Production advances for KrF lasers
KrF lasers with 1kHz pulse repetition rate are firmly established for 248nm lithography and 0.25mm IC production [1]. Recently, it has been shown that doubling pulse repetition rate to 2kHz will deliver twice the exposure energy within the same unit of time and therefore cut the exposure time approximately in half. In production situations where exposure time limits microlithographic throughput, a higher pulse rate represents a direct path to increased productivity.
Figure 3. Line-narrowing Etalon or grating in Littrow configuration. |
To illustrate this point, Hitachi recently modeled a 300mm scanner with a 250mm/sec stage. Results showed that, with a 30mJ/cms resist, a throughput enhancement of 30% is achievable simply by shifting the pulse repetition rate from 1kHz to 2kHz [2]. A cost analysis comparing the 1kHz system with the 2kHz system showed a 13% decrease in the operating cost/pulse of the 2kHz laser. Further, a cost of ownership assessment demonstrated a 10% to 25% overall decrease for the higher repetition rate when the laser was operated with a scanner in a 10,000-wafer/month operation. The Hitachi conclusion was that higher pulse repetition rate is extremely effective in boosting throughput. Not only can an exposure dose be delivered in less time at the 2kHz pulse rate, the higher repetition rate improves the averaged dose variations, therefore improving critical dimension (CD) control, which in turn affords gains in yield.
Advances in line narrowing
As noted above, achieving acceptable resolution for 180nm manufacturing with a KrF light source is feasible with optical enhancement technologies such as phase shifting and off-axis illumination. But another fundamental problem, namely chromatic aberration in refractive systems with optics, must be corrected by laser line narrowing. Until recently, state-of-the-art line narrowing methods could yield a KrF beam having a spectral width of <0.8pm FWHM and a dl95% <3.0pm at an average power of 10W and a pulse repetition rate of 1000Hz. Now, newer 20W, 2kHz KrF lasers can bring the KrF bandwidth below 0.6pm and a dl95% <2.0pm.
Figure 4. Pulse energy and stability vs. repetition rate for a KrF laser running in the constant charge voltage mode. |
One approach to line narrowing uses Etalon (reflecting surface) optics or a reflecting grating operating in a high order mode in what is described as the Littrow configuration (Fig. 3). In Cymer's KrF laser, for example, the size of the resonator allows the round trip of a light beam in 6-8nsec. The whole energy is gained within three round-trips. This constraint requires a spectrum narrowing method that brings the bandwidth to within ~1pm in just a single round-trip to achieve an average multiple-pass narrowing of <1pm.
Beyond 180nm lithography, even a DlFWHM of 0.6pm and a dl95% <2.0pm are inadequate. At the 180nm-technology node, resolution requirements demand that the laser spectrum be narrowed to a DlFWHM <0.4pm and a dl95% of <1pm. Recently [3], using a double pass Etalon with a 10W, 2kHz laser, laser pulses were produced that had an actual spectral bandwidth centered at 248.3nm, a DlFWHM of 0.37pm, and a dl95% of 0.93pm.
Pulse stability, which has a direct bearing on exposure dose uniformity, is an issue of considerable importance for lithography. Experiments have demonstrated that high repetition rate improves stability. Figure 4 shows pulse energy deviation at repetition rates up to 2kHz. The 3s deviation of <4% is achieved using a faster electrical excitation circuit. Since discharge instabilities require a certain amount of time to develop, the faster the excitation system, the less the time available for development of instabilities. Combining rapid pulse excitation circuitry with hardware and software means that to adjust total dose on a pulse-by-pulse basis allows the energy dose to remain within 0.4% of target. Dose stability better than ±0.5% over a 16msec exposure window has been demonstrated.
Next-generation ArF sources
Between the 130nm and 100nm technology nodes, semiconductor-proven efficiency considerations dictate a shift from 248nm KrF illumination to shorter-wavelength 193nm ArF. Considerable engineering effort has gone into ArF technology development during the last several years, but ArF demands are stringent. Challenges include increasing the power output, increasing repetition rate, further line narrowing, improving pulse uniformity for accurate exposure control, and addressing an entire set of issues that deal with the accelerated degradation of materials and system components caused by 193nm radiation.
Figure 5. Relative 3s energy distribution in a sequence of 2000 ArF bursts with 125 pulses/burst. |
KrF results have been extrapolated for ArF systems. One recent analysis suggests that with off-axis illumination that leads to a process factor of 0.38 (k1), and with a 0.7NA lens, a basic line-space resolution of <105nm will be achievable with ArF illumination [4]. Selete, the Japanese Semiconductor Leading Edge Technology effort, demonstrated 100nm and 70nm linewidth feasibilities using a Cymer ArF laser. It was observed, as with KrF, that use of an ArF system operating at a 2kHz repetition rate doubles the available laser power at the wafer surface compared to 1kHz operation. Experiments with a system operating at 2kHz and 10W (with an intensity of 100 mW/cms at the wafer surface) showed an average wafer throughput enhancement up to almost 50% over 1kHz operation for a DRAM process.
It is quite important for lithography that a laser source exhibit good dose stability. Figure 5, which shows an analysis of 2000 ArF bursts from a 2 kHz prototype system, illustrates progress in this area prototype system [5]. The 3s energy stability deviation was <16% for all cases, and for continuous operation it was <8%. This does not yet match the stability now achievable with KrF, but control is moving in the right direction. The longer the pulse shape, the lower the degradation of the optics, the longer the lifetime of those same optics. To the chipmaker, it results in a lower cost of operation for the overall system thanks to the laser.
The temporal shape of the laser pulse is another major concern for 193nm lithography because optical damage to transmissive elements in the beam line occurs at high peak power. Materials degradation due to compaction (for fused silica) and color center formation via two-photon absorption (in both fused silica and CaF2) have a nonlinear dependence on peak power.
Figure 6. ArF pulse shape, peak power vs. time. |
In terms of the temporal evolution of a 193nm pulse, a value described as the integral-squared pulse width is a better measure of the pulse's capacity for optical damage than a temporal FWHM value [6]. The former tends to suppress the effects of extremely narrow, high-amplitude features appearing in the temporal profile. Such features have high power values, but, unless sustained over time, contain little total energy to effect optical damage. In Fig. 6, both the FWHM value and the integral-squared value are more than doubled over other recently reported values. This points to a reduction in damaging peak-intensity power while increasing average delivered photonic power needed for resist exposure.
The dispersion of fused silica is greater at 193nm than at 248nm, thus exacerbating chromatic aberration. This forces exposure tool manufacturers to design expensive CaF2 chromatic correction elements into the optics. Methods for ArF line narrowing are thus especially important, since they could ease the need for this critical high-cost material.
ArF line narrowing is demanding. In a typical ArF cavity, the short pulse duration allows no more than approximately three round-trips. Even more than with KrF systems, first pass narrowing to 0.5pm is essential. In this regard, a new Cymer ArF optical cavity design has been reported that uses a new output coupler and achieves a DlFWHM of 0.3pm and a dl95% of 0.8pm [7]. The reduction in spectral pulse width greatly lowers chromatic aberrations. It does, however, allow lens designs with significantly diminished quantity and quality requirements for CaF2. This contributes to lowering the total running costs of future ArF systems.
Some success has been achieved in extending temporal pulse widths to maintain average power while reducing peak levels to lower optical damage. Current work includes efforts to extend module life, lower cost of ownership, and increase average power and repetition rates for improved lithography processing. Recent development in these areas suggests that when 193nm lithography becomes a production process, ArF lasers will exhibit performance comparable to that of KrF systems.
157nm the final optical frontier?
It is likely that 70nm will be the last technology node for optical lithography, before post-optical lithography. Even here, though, there are extremely challenging hurdles. These include 3s CD control requirements of 5nm, 3s overlay of 25nm, and 25 x 44mm field size (1100 mms). Difficulties in developing a molecular F2 laser source may seem quite tractable by comparison.
At the 157nm exposure wavelength, only a limited number of materials are suitable for optical components (e.g., CaF2, MgF2) and those materials are known to undergo severe deterioration in performance when exposed to short wavelength radiation over extended periods, affecting system cost of ownership.
As with the ArF laser, reducing peak power, but increasing average power, will be important. F2 lasers operating at multi-kHz pulse rates will be needed. Required power and energy levels are expected to be similar to those for recently improved KrF and ArF lasers (i.e., 10-40 W). Spectral linewidth demands at 157nm will depend on lens designs. It seems likely though, that a DlFWHM of 0.2pm will be needed.
F2 lasers emit two bands of light at 157.63nm and 157.52nm, but 87% of the laser energy is located at the longer wavelength. A prototype molecular 2000Hz F2 laser, based on a current ArF design, was recently disclosed in which single-line operation at 157.63nm was achieved [8]. Convolved linewidths of 1.14pm DlFWHM and dl95% of 2.35pm were measured.
The output power of this F2 laser prototype scaled linearly with repetition rate. Its energy stability was better than 6% (3s) over 100 bursts and better than 4% over more than 80% of the burst range. Gas lifetimes appear comparable to current ArF lifetimes. From previous experience with ArF lasers, one can estimate a gas lifetime of about 25 million shots.
Conclusion
Excimer laser light sources will be essential enabling elements for lithography for the next three critical dimension nodes of industry's roadmap. Major challenges to be overcome in developing KrF, ArF, and molecular F2 lasers include:
- Boosting average power not only to match but to maximize production wafer throughput with current and evolving photoresist chemistries;
- Stabilizing pulse power for optimum exposure dose control a critical yield issue; and
- Maximizing line narrowing for each laser's pulse spectrum to reduce chromatic aberration, achieve optimal CD resolution, and potentially reduce dependence on expensive color-correcting CaF2 optics.
These goals will be achieved through advances in electronic power supplies that boost total available laser power, increase pulse repetition rate, and improve the linearity of the output power vs. pulse frequency response; excitation circuit improvements that increase output pulse temporal stability; reduction of system factors (e.g., aerodynamic and thermal phenomena) that destabilize pulse temporal and spectral shape; and improvement in optical designs.
Despite significant challenges, there appear to be no fundamental physics issues impeding realization of excimer lasers for DUV and vacuum UV lithography. What is required is a solid, sustained engineering effort.
References
- D. Meyers, et al., "Production Ready 2 kHz KrF Excimer Laser for DUV Lithography," Proceeding of SPIE 1999 Intl. Symposium on Microlithography.
- K. Takehisa, "Improved Productivity at Higher Power," presented at the 6th Annual Cymer DUV Symposium, Tokyo, Dec. 1998.
- A. Ershov, et al., "Feasibility Studies of Operating KrF Lasers at Ultra-Narrow Spectral Bandwidths for 0.18µm Linewidths," Proc. SPIE, Vol. 3334, Feb. 1998.
- Y-M. Ham, "DUV Source Requirement for Next Generation Optical Lithography," presented at the 6th Annual Cymer DUV Symposium, Tokyo, Dec. 1998.
- T.P. Duffey, et al., ArF Lasers for Production of Semiconductor Devices with CD <0.15mm, Proc. SPIE, Vol. 3334, Feb. 1998.
- R. Sandstrom, "The development of line-narrowed ArF lasers for 193nm optical lithography," presented at the Semicon/Kansai '96 ULSI Technology Seminar, Osaka, Japan, May 1996.
- A. Ershov, H. Besaucele, P. Das, "Performance characteristics of ultra-narrow ArF laser for DUV lithography," Proceedings of SPIE 1999 Intl. Symposium on Microlithography.
- T. Hofmann, et al., "Revisiting F2 Laser for DUV Microlithography," Proceeding of SPIE 1999 Intl. Symposium on Microlithography.
Olivier Semprez is an EE master graduate from the Institut Superieur d'Electronique de Paris with a masters degree in marketing and corporate finance from the University Paris, Dauphine. He has researched optoelectronics at the Hitachi Central Research Laboratory in Kokubunji, Tokyo, and spent two years at the Japanese Optoelectronic Industry and Trade Development Association in Tokyo. Semprez is a marketing engineer at Cymer, 16750 Via Del Campo Ct., San Diego, CA 92127; ph 858/385-7372, fax 858/385-6035, e-mail [email protected].