Excimer Laser Sources For Mask Inspection
Since their first technical realisation in 1976, discharge pumped excimer lasers have found numerous industrial, medical and scientific applications. The fast development of excimer lasers in recent years has succeeded in designing very compact, turn-key systems delivering up to 10W of radiation at 248nm (5W at 193nm and 1W at 157nm) with repetition rates up to 1000Hz. Excimer laser based 193nm (ArF) and 248nm (KrF) photoresist exposure have become established in semiconductor production lines. In the near future mask inspection systems operating at the lithography wavelength will become more and more important in increasing yields in fabs. Another potential application is in maskwriting.
In searching for mask defects, there is no need for the large line-narrowed excimer lasers used in the stepper/scanner exposure systems. Rather, the emphasis is on cost-driven miniaturisation of the laser. The throughput requirement of mask inspection is much lower than for wafer processing. At the same time these smaller systems have to meet the same temporal and spatial properties as their 'bigger brothers', especially concerning pulse-to-pulse fluctuations, pointing stability or gas lifetime. It is therefore useful to report on the output characteristics of current compact ArF and KrF excimer lasers and compare them to continuous wave (CW) light sources, such as mercury vapour lamps and frequency doubled argon ion lasers.
Spectral, temporal and spatial beam properties
Figure 1 shows the spectral distribution of the KrF laser emission without the special line-narrowing techniques used in creating sources for exposure systems. The spectrum was created by means of a 75cm spectrograph in a Czerny-Turner configuration using a 600grooves/mm grating. The instrumental response function had a gaussian shape with a full width at half maximum (FWHM) of approximately 160nm. The instrumental response function was used to deconvolute the measured laser emission spectrum. The corrected bandwidth has been quantified as 320.10pm (FWHM). The ArF free running spectrum is reported to be approximately 500pm.
The beam profile was measured using a fluorescent screen to transform the UV laser radiation into visible light. The visible image on the screen was recorded by a CCD camera and converted into a greyscale intensity plot. Figure 2 gives an example of a typical near field beam profile of an ArF excimer laser, with greyscale values towards the centre indicating higher intensity. The FWHM beam dimensions are 3mm by 6mm. By comparing these values to the diameter of the fundamental transversal mode TEM00, the number of transversal modes can be estimated in the range of several hundreds, with the assumption of a rectangular symmetry. This high number of modes has a severe impact on the coherence properties of the excimer laser beam.
The output parameters of an excimer laser vary with time in three different component life cycles: laser gas, optics and laser tube. Usually the life time of a component is given as the number of laser pulses required to reduce the output to 50% of the maximum value. In contrast to the transient behaviour of excimer lasers, most applications need constant output parameters. A common way to assure e.g. constant pulse energy is to adapt the discharge voltage as the laser gas or the components are aging, to operate the laser in an energy-stabilised mode.
Gas life time in energy-stabilised mode is defined as the number of laser pulses a control loop is able to maintain an energy of 50% of the maximum (starting) energy by increasing the discharge voltage (Figure 3). Pulse-to-pulse fluctuations for this measurement were distributed with a standard deviation of approximately 3%. Gas life time in discharge-voltage-constant mode is defined as the number of laser pulses until the energy dropped to 50% of the maximum (starting) energy, with the laser running at maximum discharge voltage (Figure 4 ). Output power was measured with a thermopile power meter exhibiting a long integration time in the order of several seconds. Therefore pulse-to-pulse fluctuations of the laser were not resolved with this measurement. Typical values for the gas life time of an ArF (193nm) excimer lasers using Corona preionisation are 4-6 million pulses. The values for KrF (248nm) excimer lasers are 9-25 million pulses. The output power at a specific working point strongly depends on laser tube and the age of the optics. The factors involved include the inevitable dust build-up within the laser tube. During a tube life pulse-to-pulse fluctuations tend to increase as power decreases.
The beam position was measured as the centre of mass of the laser beam profile at a distance of 1m from the output aperture. To indicate the pulse-to-pulse behaviour, 100 subsequent shots were recorded and the standard deviation . was calculated for the x- and y-directions (Figure 5). Pulse-to-pulse fluctuation is mainly caused by local inhomogeneities in the discharge.
Unlike beam position stability, beam direction stability is measured in the focal plane of a lens and is not dependent on measuring distance. It reveals the angular intensity distribution of the laser beam. The positional information was measured with the beam profiler and then translated into angular terms.
Figure 7 shows the angular deviation of 100 consecutive laser pulses of an ArF excimer laser distributed around their mean value. The 2. values of the beam direction stability vary from 30-40µrad. Angular fluctuations are also caused by local discharge inhomogeneities. The long term behavior of the beam direction is correlated with beam position and shows a walk-off in the order of 200µrad, as well.
A typical pulse shape of an ArF laser is given in Figure 8. The full width at half maximum (FWHM) of the transient is ....=12ns. 95% of the total pulse intensity is lying in a width of ...%=27ns. Neglecting side effects, ....-values can vary in a range of 10-30ns for both ArF and KrF lasers by using output couplers with different reflectivity or tuning the discharge pulse. For the given data, maximum pulse peak power is approximately 1MW for a typical energy of 8mJ. The KrF excimer laser can produce peak powers of up to 3MW.
In Figure 9, output energy and pulse-to-pulse fluctuations are presented as a function of consecutive laser pulses in discharge-voltage-constant mode for an ArF laser running at 500Hz. The standard deviation values were calculated from samples of 1000 shots each. Generally, fluctuations tend to increase towards the end of the gas lifetime, approximately by a factor of two for both the ArF and KrF lasers. For an excimer laser running at maximum voltage, a typical interval for pulse-to-pulse fluctuations is 1%<Û<3%
Results for spatial coherence showed this parameter to be tunable just with different resonator set-ups between 400µm<Îc<1mm. The relation of high divergence and low coherence length could be confirmed.
Comparison
Energy/power fluctuations
Some beam parameters such as the energy/power fluctuation of frequency doubled argon ion lasers and lamps are technologically superior to those of compact excimer lasers _ their disadvantageous properties have to be overcome in the case of pulse-to-pulse fluctuations by different algorithms or averaging techniques. Long term energy/power fluctuations are present with all sources because of heat dissipation and component life cycles. But all sources can be stabilised by controlling the discharge current or voltage.
Bandwidth
Reticle inspection systems are designed to provide similar resolution to stepper/scanner systems, but with much lower throughput. Throughput needs determine the field size of the imaging system. Reticle inspection systems therefore have much smaller field sizes, allowing higher bandwidth for the illumination sources and simpler imaging optics. The higher bandwidth means that line narrowing is not needed on excimer laser based inspection applications.
Beam direction
Beam direction turns out to be a serious obstacle for reticle inspection. Beam direction stability cannot be reduced by homogenisers. As a consequence it has to be actively stabilised, as is necessary for all types of laser source.
Peak intensity
The short pulse width of an excimer laser forces high peak intensities. For standard optical materials in the beam line this results in increased degradation by colour centre formation via two photon absorption. Recent improvements however, have succeeded in the development of high purity materials such as CaF2 with extraordinary lifetimes under UV radiation. Considering the low fluence required for inspection applications, system operation of several years with insignificant degradation of optical components can be expected.
Temporal structure
The defined temporal structure of an excimer laser pulse can turn into an advantage concerning signal-to-noise ratio, when time gated measurement technologies are used to record the image. Considering an 1kHz repetition rate excimer laser with a pulse duration of 10ns, the signal to noise ratio could be decreased by up to a factor of 100,000.
Coherence
Addressing the speckle problem in imaging systems, it must be pointed out, that excimer lasers in general are highly multimode lasers and have poor coherence, without the application of expensive measures. The effort to lower the coherence of excimer lasers is therefore limited.
Power consumption/UV output power
In general the overall UV output power of excimer lasers is unmatched by other light sources - especially when inspection has to match microlithography at the 193nm and 157nm wavelengths. Frequency doubled argon ion lasers can provide reasonable output power down to a lower limit of 244nm. Comparing efficiency, they are well below both lamp sources and excimer lasers.
Acknowledgements
The authors wish to thank Maik Hohmann and Dennis Fischer for measurements of coherence properties and Michael Muencker and Stefan Geiger for providing their analysis and expertise. All measurements were carried out using ExciStar S-1000 KrF or ArF lasers developed and manufactured by TuiLaser.
