Photolithography using a dual-chamber light source
04/01/2003
Future photolithography applications demand that light sources generate narrower and more stable spectral bandwidths to pattern finer chip features. A dual-chamber technology for excimer laser light sources will provide lithographers with narrow spectral bandwidths, high average output power, and extended maintenance cycles for all DUV exposure wavelengths.
Optical resolution as defined by the Rayleigh equation — resolution = k1∑l/NA — limits the rate at which design rules can be shrunk. In this equation, l is the wavelength of light emitted by the source in the lithography system, and NA is the numerical aperture of the lithographic projection lens. The k1 factor represents the process, including RETs, used to improve lens performance. Improving the efficacy of RETs is essential in the DUV era because lithographers want to optimize their capital equipment investment by extending the use of krypton fluoride (KrF–248nm) and argon fluoride (ArF–193nm) wavelength production as long as possible.
With ArF lithography systems only now coming into production, and KrF lithography systems actively in use as the workhorse tools for advanced design-rule applications, lithographers must implement more aggressive RETs (to further drive the k1 factor down below 0.4) and adopt lenses with higher NA values (≥0.85) to enable the production of feature sizes significantly smaller than the wavelength of the exposure source.
Subwavelength challenges
Lower k1 factors and higher NAs both pre-sent significant manufacturing challenges. For example, RETs—which include off-axis illumination, phase-shifting masks, and optical proximity correction—reduce the overall effective transmission of the light path and reduce the throughput of the lithography system. To maintain maximum throughput, a high-power light source is required to improve the economics of lithography and expose resists of different sensitivities. For a given exposure, increasing the output power increases the wafer throughput of the lithography tool until the stage speed reaches a plateau at max. velocity.
As lens designers increase NA above 0.8, the light source spectral bandwidth must be reduced to avoid chromatic aberration in the projection lens, which causes a loss of image contrast. Today's refractive DUV lenses can only be made of a few materials (i.e., fused silica and CaF2), limiting the designer's ability to correct for chromatic effects. The problematic availability of high-quality CaF2 also constrains lens design. Aberration-free imaging with such projection lenses requires a light source with a narrow spectral bandwidth and a stable center wavelength.
Until now, laser light sources for lithography have used a single-chamber design, in which one discharge chamber not only generates a narrow emission spectrum, but also produces high power with precisely controlled pulse energy. The chamber operates with optical components within the light source to produce tight spectral bandwidths by selecting a narrow part of the full spectrum generated by the chamber, and discarding the remaining light energy.
However, since the line-narrowing process results in reduced light energy, a single chamber must increase repetition rate to meet output power requirements. This forces the single-chamber source to make a trade-off between optimizing spectral bandwidth or pulse energy, since it has insufficient operational margins for both.
Dual-chamber light source design
A new light source architecture, master oscillator power amplifier (MOPA), has been developed in which the functions of spectral bandwidth and pulse energy generation are separated between two chambers, with each chamber optimized for one performance parameter, enabling improved spectral bandwidth performance and increased output power without increasing the repetition rate [1, 2].
 Figure 1. A spectral bandwidth of ≤0.25pm FWHM, much narrower than a free running laser, enables high-NA-imaging lenses and reduces the number of CaF2 lens elements. |
null
null
The first discharge chamber, the master oscillator (MO), is designed to produce output with a very narrow spectrum (typically ≤0.25pm full width half-maximum [FWHM]) at a very low pulse energy (typically 1mJ) after receiving short-duration, high-voltage electrical pulses from its attached electrical pulse compression head. The MO is situated within a resonator structure, which consists of an output mirror and a high-dispersion line-narrowing module (LNM).
The LNM is a multiple prism pre-expanded Littrow grating arrangement, with active control of the angle of incidence of light onto the grating (and hence resonating wavelength) via a piezoelectric transducer and a stepper-motor-controlled mirror. With high-speed wavelength control, the light source measures and adjusts the wavelength on a pulse-by-pulse basis—enabling the user to lock the light source wavelength to within ≤ ±0.025pm of the target wavelength over a given exposure window.
In order to generate sufficient output power, the line-narrowed light from the MO must be directed through the second chamber called the power amplifier (PA). Two wavefront engineering boxes direct and manipulate the laser beam between the chambers. The PA amplifies the line-narrowed, low-energy output from the MO to a higher pulse energy, to meet necessary power levels (typically 40–80W). The amplification process occurs with no change to the spectral properties of the light. The discharge occurring in the PA gain medium simply amplifies the line-narrowed light injected by the MO at the existing wavelength to a higher pulse energy. Synchronization between MO and PA chamber discharges is essential to the amplification process. The PA discharge must occur at the correct point of the MO signal to ensure spectral bandwidth properties are maintained.
An in situ metrology subsystem measures several performance characteristics, including the pulse energy of the MO, PA, and final output; the output wavelength; the spectral bandwidth of the output; and synchronization of both chambers. These performance characteristics are subject to feedback control based on sensor signals.
The PA output pulse energy is not directly released into the exposure tool. The light is directed through the optical pulse stretcher (OPuS) subsystem, which lengthens the pulse to reduce peak power and provides the capability for automated shuttering of the system's output beam, as well as an energy sensor for measuring the final pulse energy output.
It is desirable to incorporate an optical pulse stretcher into the light source to reduce the possibility of index-of-refraction changes in fused silica optical elements after long-term DUV exposure. Otherwise, the peak high-pulse energies produced might impact lens material lifetime via compaction and/or de-compaction processes [3]. Peak pulse energy intensities are reduced, and the pulses lengthened, by splitting off a portion of the output, sending it through multiple delay lines, and relay imaging the delayed (or "daughter") portion(s) back onto the "parent" beam. The spectrum remains unchanged.
Dual-chamber performance
The dual-chamber system also results in an improved light source dose stability. In a single-chamber design, several factors impact dose stability, including the voltage repeatability of the high-voltage power supply, the inherent discharge stability of the chamber, and the dynamics of DUV light generation. As light oscillating in the chamber is line-narrowed from its natural bandwidth to meet the energy and linewidth specs, significant energy in the gain medium is lost (see Fig. 1).
The onset of lasing is delayed by this initial high loss due to line-narrowing. The gain in excimer lasers decays quickly, which means a line-narrowing chamber usually achieves laser threshold only after the high-gain period has passed. The chamber operates at the tail end of the gain, which is more susceptible to slight fluctuations in discharge conditions and gas chemistry. These instabilities result in slight pulse-to-pulse energy variations, which result in overall dose fluctuation.
Since PA chamber operation does not impact spectral characteristics and is optimized to generate high pulse energy levels, the PA chamber can be driven to gain saturation. Discharge synchronization between the MO and PA enables the MO signal to reach the PA at the point of gain inception, which results in a more efficient extraction of stored energy and less susceptibility to gain fluctuations mentioned above.
The entire stored energy in the PA is used for each discharge (i.e., the gain is saturated), which means that the PA's output pulse energy does not fluctuate due to changes in the injected MO pulse energy. Figure 2 shows that the system can typically maintain twice the dose stability of a single-chamber light source. This is crucial for lithographers who are trying to conserve dose latitude while operating within shrinking process windows.
Conclusion
Next-generation photolithography requirements are driving a fundamental change to excimer light source architecture. To enable Moore's Law, allow for shrinking process windows, provide higher throughput, support cost reduction, and facilitate the industry's strategy of mix-and-match DUV lithography, light source manufacturers must transition from the single-chamber light source design that has served for the past 10 years.
In addition to avoiding design challenges encountered if single-chamber-based light sources are pushed to a higher performance level, a dual-chamber-designed light source allows spectral bandwidth generation to attain a production performance ≤0.25pm FWHM. Advanced, high-NA lens designs are enabled and CaF2 element requirements are reduced. Increased source power will improve wafer throughput, drive lithographers to develop faster wafer stage speeds and more robust resists, and enable optically inefficient RETs to extend KrF and ArF DUV lithography.
Daniel J. Colon III, Cymer Inc., San Diego, California
References
1. M.H.R. Hutchinson, Topics in Appl. Phys., V. 57, Springer Verlag, Munich, 1990.
2. C.B. Dane, T. Hoffmann, R. Sauerbrey, F. Tittel, IEEE Journal of Quantum Electronics, QE-27, 2465, 1991.
3. J. Moll, P. Dewa, Proc. SPIE, 4691, 1734–1741, 2002.
For more information, contact Daniel J. Colon III, Cymer Inc., 16750 Via Del Campo Court, San Diego, CA 92127; ph 858/385-7153, fax 858/635-6035, e-mail [email protected].