193-nm lithographic system lifetimes as limited by UV compaction

04/01/1997

Second in Series

193-nm lithographic system lifetime as limited by UV compaction

William G. Oldham, Richard E. Schenker, University of California, Berkeley

Next-generation lithography tools will operate with 193-nm UV light (ArF excimer laser) to print device features with resolutions of ?0.18 ?m. Throughput comparable to current lithography systems will require in excess of 2 million laser pulses/hour (20 billion/year) to pass through the half-meter or so of optical material in the system. Yet today, compaction can be observed in state-of-the art fused silicas after only about 1 million pulses of 193-nm radiation at an intensity level of 1 mJ/cm2. To avoid imaging degradation, we must minimize exposure of fused silica elements in real systems and search for improvements in radiation tolerance.

Improvements and innovations in optical lithography have been among the main driving forces in the production of denser and denser ICs. In order to continue this progress, lower-wavelength radiation sources must be used for printing of even smaller device patterns. Currently, 248-nm lithography (KrF excimer laser source) is used in advanced production to manufacture device features of 0.25 ?m. The introduction of 193-nm lithography (ArF excimer laser source) for future device "generations" seems almost inevitable.

The choice of lens materials is severely limited at such a short wavelength. While crystalline materials (such as quartz, calcium fluoride, lithium fluoride, magnesium fluoride, and a few others) should be free from compaction, other problems such as absorption, birefringence, poor thermal stability, and difficulty of surface figuring have so far limited their use in large diffraction-limited UV optics. Fused silica (synthetic amorphous SiO2) is the only optical material available at this time. The properties of different fused silicas, including transparency, homogeneity, and susceptibility to laser-induced damage, vary owing to subtle differences in fabrication and annealing.

At wavelengths around 200 nm, lamp sources have insufficient power for high-throughput lithography systems. While an ArF laser offers high power at a short wavelength, pulsed laser sources also raise the instantaneous light intensity significantly, greatly increasing the risk of photo-induced damage. Pulsed UV radiation induces both color center formation and densification in fused silica. The most important color center for UV applications, the E` center, has an absorption peak at about 210 nm with a full width at half maximum of about 40 nm. Absorption not only causes power loss in lithographic systems, but also induces temperature rises within the optics, leading to refractive index changes and hence imaging aberrations. In many new fused silicas [1], the induced absorption from color center formation saturates at low levels (<0.1% absorbed/cm) when per-pulse energy densities are <1 mJ/cm2. This important finding significantly improves the prospects for large fused silica optics at 193 nm, but places increasing importance on compaction and its consequences.

UV laser-induced compaction is less well understood than color center formation. The structural rearrangements or lattice imperfections leading to radiation-induced densification are still unknown. Densification directly changes the refractive index, which in turn induces imaging aberrations. Compaction rates increase substantially with lower-wavelength sources. Fused silica, for instance, compacts 20-30 times faster from 193-nm laser pulses than from 248-nm pulses of equal intensity and energy [2].

Unlike color center formation, compaction does not saturate over the range of damage important to lithographic optics. In fact, compaction increases uninterrupted even after over 40 million, 500-mJ/cm2, 248-nm pulses; long after all color center formation has saturated [3].

Figure 1. UV-induced densification in Corning 7940 vs. pulse count at three different pulse energy densities.

Compaction vs. pulse count

Only within the last year has UV-induced densification in fused silica been characterized sufficiently to allow accurate prediction of the effects of compaction on lithographic system lifetimes. A technique using birefringence measurements, developed at UC-Berkeley, enables real-time monitoring of irradiation-induced compaction even at very low fluences. Localized compaction owing to UV irradiation sets up stress gradients that can be measured by probing the resulting birefringence with polarized light. The magnitude of laser-induced densification can then be extracted from the stress distribution using the known mechanical properties of fused silica.

This method can measure compaction in the parts-per-billion range. For comparison, phase measuring interferometry (PMI) can only monitor the effect of compaction on index of refraction at damage levels of hundreds of ppb. Figure 1 shows the pulse count dependence of compaction for a single sample of Corning 7940 fused silica at three different pulse energy density levels (12-28 mJ/cm2). The dependence is nonlinear, with compaction rates highest initially. For a fixed pulse energy density, compaction* follows the pulse count raised to a power of about 0.65 to 0.7.

Compaction vs. pulse intensity

When radiation travels through a material, there is a non-zero probability that multiphoton absorption will occur. The second-order description of the radiation intensity in a sample can be written as:

where

I = the intensity, usually in W/cm2

a1 = the linear absorption coefficient (cm-1)

a2 = the two-photon absorption coefficient (cm/W)

z = length, increasing in the direction of propagation (cm)

If the two-photon absorption term were ignored, one would get the familiar equation for intensity vs. depth in an absorbing medium: I(z) = I0 exp (-az).

Though two-photon absorption is only a small portion of the total at typical intensity levels, it is apparently the source of UV damage in fused silica. This observation is consistent with an ionization model for damage initiation: the effective energy band gap of fused silica is about 8.3 eV, requiring two UV-photons to initiate an ionization event (hn = 6.4 eV at 193 nm).

Figure 2 shows total compaction in the same Corning 7940 fused silica sample for irradiations performed at five different energy densities (1-28 mJ/cm2). The data are plotted vs. a dose parameter, pulse count ? pulse energy density squared divided by the pulse length. This parameter, which is proportional to the number of two-photon absorption events, allows extraction of a universal relation describing UV-induced compaction vs. dose. The universal curve fitting through all the data shows that the density change equals a material-dependent constant ? the dose parameter raised to a power of about 0.7. In other words, the total UV-induced densification (Dr
) can be described by:

where

t = the pulse length (ns)

N = the number of pulses (millions)

k = a constant

I = the 193-nm energy density (mJ/cm2)

This general behavior - damage is proportional to the effective fluence to the 0.7 power - is consistent with past compaction studies [4] using electron beam and gamma radiation, suggesting like densification mechanisms. For a typical value of k of 0.2 ppm, Eqn. 2 predicts a density change of 0.2 ppm after 10 million, 1-mJ/cm2, 10-ns long pulses. This relationship allows appropriate scaling of higher energy density tests to predict damage rates at lower fluences.

Figure 2. UV-induced densification in a single sample of Corning 7940 for five different pulse energy densities. The x-axis, number of pulses ? the pulse energy density squared divided by pulse length, is proportional to the number of two-photon ionization events within the sample.

Compaction rates of fused silicas

Five 1990-1994 grade fused silica samples were evaluated: Suprasil 311, SV2G1, Suprasil 2, and Suprasil 300 from Heraeus-Amersil, and Corning Excimer Grade 7940. Suprasil 300 is a "dry" fused silica (<1ppm OH content) and is generally only used for IR applications. SV2G1 is a partially processed precursor of Suprasil 2. Suprasil 311, Suprasil 2, and Corning 7940 are all high-quality UV-grade fused silicas. Figure 3 shows the density changes for the five different materials vs. the number of pulses delivered ? the intensity squared. The x-axis is directly proportional to the total number of two-photon events within the sample because the two-photon absorption coefficients of all fused silica types are roughly equivalent [5]. The dehydrated fused silica sample has by far the worst compaction performance while the "precursor" sample is the most durable at higher fluence levels. The compaction rates among all samples tested vary by about a factor of three, but the UV-grade specimens have less than ?25% variation in damage rate.

Figure 3. UV-induced densification for five different samples vs. number of pulses ? the pulse energy density squared divided by pulse length. Suprasil 300 is a "low water content" fused silica, not intended for UV use.

Five experimental fused silicas were also tested as part of a SEMATECH study. The five fused silicas were versions of Corning 7940, Suprasil 311, Suprasil 1, Shin-Etsu X103, and Shin-Etsu X103A. Several batches of each material were evaluated. The samples are randomly labeled A, B, C, D, and E in accordance with the wishes of the suppliers and SEMATECH. Samples were exposed at both Berkeley and MIT Lincoln Laboratories. The pulse length used for the Lincoln tests was approximately 22 ns, twice that for the Berkeley tests. Experiments performed at Berkeley generally lasted for about 10 million pulses while the exposures performed at MIT lasted for 150 million pulses (5 days at 400 Hz). All the samples tested followed roughly the same k (N I2/t)0.7 dependence, differing only in the magnitude of the constant k. The pre-exponential coefficient (k) varied from 84-660 ppb among all the experimental fused silicas tested.

Figure 4 plots the measured compaction for one of the more promising experimental fused silicas. The line shown corresponds to Dr
= 0.12 ppm ? (N I2/t)0.7, illustrating an approximate factor-of-two improvement in damage durability over (1990-1994) grade fused silicas.

Figure 4. UV-induced densification in an experimental fused silica vs. number of pulses ? the pulse energy density squared divided by pulse length. MIT data points are for three individual samples measured after 100 and 150 million pulses.

The effects of compaction on optical properties

The true measure of damage to an optical element is imaging performance. Optical path differences (OPDs) are the best "unit" for quantifying imperfections in optical elements. The optical path length is simply the refractive index of the material ? the length, so a measure of changes in index and length is sufficient to calculate OPDs. The sample length change and the index change are a function of density changes and sample/damage geometry. For a uniformly compacted sample, the 193-nm index change is approximately:

When only a small area of a lens is exposed (compacted), the rest of the lens resists shrinkage of the damaged portion, so the net densification is less than if the sample had been uniformly exposed. Nonetheless, the refractive index can be calculated from the known strain optic coefficients. For given sample and damage geometries, both the index and length changes are directly proportional to the densification. Surface depressions formed by compaction partially counteract OPD increases, due to the larger refractive index. Finite element analysis (FEA) can characterize the OPD dependence on sample thickness and shape as well as the shape and size of the exposed area (Fig. 5). For typical elements used in a lithography system (diameter about 5? larger than thickness) and for typical intensity distributions within the optical elements (damaged area radius about one half of lens radius), simple relations approximate the path length and index change. The 193-nm refractive index change is given roughly by:

where (Dr
) unrestricted is the density change that would have been produced if the sample had been uniformly irradiated. The surface depths are approximately:

where L is the sample thickness. The OPD, given by [OPD = Dn L + (n-1)DL], for a 1-cm-thick sample with 1-ppm densification would thus be approximated by:

OPD = 0.47 ppm ? 1 cm + (1.56-1) ? (-1.9 nm ? 2 surfaces)= 2.6 nm(6)

This value corresponds to about 0.014 wavelengths at 193 nm. In this case, over 40% of the OPD increase from densification is canceled by path length decreases due to surface depression formation.

Figure 5. Finite element analysis calculation of distortions from a 1-ppm laser-induced densification in a fused silica lens (distortions magnified 600,000?). The lens is 10 cm dia., 1.5-cm thick, and has a compacted region with a 5-cm dia. A 2.8-nm trench depth is calculated.

How much compaction is allowed?

A model optical system [6], based on an extrapolation of the prototype 193-nm scanner at Lincoln Laboratories, was used to evaluate the effects of compaction (Fig. 6). The path length through several optical elements is compressed into three blocks (18-, 16-, and 6-cm long) of fused silica and the focal lengths (46, 12, 16, and 16 cm) of the four lenses are selected to emulate the rays` paths in the lithography tool being modeled. The model system has a 4? reduction, a 5 ? 26-mm field size, a 4-mm spacing between the wafer and the last optical element, and is capable of operation with a numerical aperture of 0.6. The compaction damage rate at any element depends strongly on the "filling" of the element by the optical energy passing through the system. From Eqn. 2, the magnitude of the damage is proportional to (intensity)1.4, i.e., [1/(beam diameter)]2.8, but the impact of the damage also depends on the optical path length between the element and the image plane. Elements near the pupil are very sensitive to damage, but because of lower intensity are damaged relatively little. Elements near the wafer, because of high intensity, are damaged more severely, but do not play as key a role in imaging.

A detailed analysis of the effects of compaction on the model

system has been presented [7]. Equation 2 is used for the calculations with an optimistic compaction rate constant k of 0.15 ppm. The net OPD is assumed to be 60% of that predicted by refractive index changes alone due to the offsetting effects of thickness shortening. The lifetime of a system is defined here as the time to induce an additional 0.05 l total RMS aberration, not including terms such as pure defocus that are directly correctable. "Diffraction-limited" lithography systems [8] are designed for maximum intrinsic wavefront aberrations much less than 0.05 l; simulations predict significant distortion and focal shifts at the edge of the image field for a 0.05 l total RMS compaction-induced wavefront aberration.

System lifetime depends strongly on illumination conditions, image field size, and wafer plane intensity. Lifetimes increase substantially for systems with less source coherence and lower wafer plane intensities. If a 70% clear field mask and 0.4 mJ/cm2 per pulse wafer energy density are used (50 pulses/field for a resist with a sensitivity of 20 mJ/cm2), we predict lifetimes of four (partial coherence parameter s = 0.5) and 10 months (s = 0.7) using the 0.05 l wavefront aberration criteria and assuming continuous operation (1000 Hz laser producing 86 million pulses/day). These lifetimes scale as (resist sensitivity)-2, (mask pattern transmission)-2, and (tool utilization)-1.

Figure 6. Model optical configuration used for evaluation of compaction effects on 193-nm lithographic system imaging. The path lengths of several optical elements are compressed to three blocks of fused silica. Rays shown are 0, +1, and -1 diffracted orders for minimum-sized lines and spaces at the edge of the image field. One of the diffracted orders travels through a large portion of the damaged region while the other travels through mostly undamaged material.

Generalizations about system lifetime

The lifetime of an actual lithographic system is a complex function of system design and operating conditions. Because of the high relative intensity at the last few elements in the system and the super-linear dependence of compaction rates on intensity, those elements are naturally the primary area of concern. If the last few lenses in the system were easily replaceable, some suggest the system lifetime could be indefinitely extended at the cost of periodic replacement of those elements. By the same reasoning, calcium fluoride elements might be appropriate for optics near the wafer plane since the system is less sensitive to material inhomogeneities in those last few elements.

These techniques would actually have limited usefulness, however, because optics in the system away from the wafer plane contribute significantly to the aberration budget with only modest compaction damage. In fact, in our model system [7] at s = 0.5 we found about a 50% contribution to the total aberration from elements near the pupil. The relative contribution to total system aberrations from elements away from the wafer plane will increase substantially for systems with larger image field sizes.

Lowering the peak optical intensity by increasing the pulse repetition rate and/or using longer laser pulses at lower peak power would increase lifetime. For example, doubling the pulse length and halving the intensity would double the system lifetime while keeping throughput constant. The availability of both positive and negative 193-nm resists would allow exclusive use of dark field masks in printing, and thus would lower the energy density within most of the optics without lowering the energy density at a bright area of the image. More sensitive resists would permit a reduction in optical intensities within all the optics, including those near the wafer, while maintaining a constant throughput.

Conclusion

For low-throughput, 193-nm lithography tools, UV-induced compaction should not limit the useful lifetime of the system. For high-throughput machines constructed with presently available fused silica optics, however, a high-sensitivity resist may be required to achieve acceptable system lifetime. If, on the other hand, recent improvements in the quality and durability of fused silica continue, the risk of compaction may disappear. Only a few years ago the transparency of fused silica was considered unacceptable for 193-nm lithographic applications. The ultimate determination of whether or not UV-induced damage will limit 193-nm system lifetime will be made on the production line.

References

1. M. Rothschild, et al., "Laser Induced Damage in Optical Materials," oral presentation at the Second Int. Symposium on 193-nm Lithography, 1996.

2. P. Schermerhorn, "Excimer laser damage testing of optical materials," SPIE, Vol. 1835, pp. 70-79, 1992.

3. D.J. Krajnovich, et al., "248-nm Lens Materials: Performance and Durability Issues in an Industrial Environment," Proc. SPIE 1848, 544-560, 1993.

4. W. Primak, R. Kampwirth, "The Radiation Compaction of Vitreous Silica," Journal of Applied Physics, Vol. 39, No. 12, pp. 5651-5657, Nov. 1968.

5. R. Schenker, et al., "Ultraviolet Damage Properties of Various Fused Silica Materials," Laser-Induced Damage in Optical Material: 1994, 26th Annual Boulder Damage Symposium Proceedings, pp. 458-468, SPIE proceedings 2428.

6. Model based on D. M. Williamson, US Patent # 5,212,593, and is similar to the experimental 193-nm system at MIT-Lincoln Labs. Because the beam path in such a system folds over itself within certain elements, a higher effective intensity was used in elements near the pupil plane.

7. R. Schenker, W. Oldham, "The Effects of Compaction on 193-nm Lithographic System Performance," presented at 1996 Symposium on Electron, Ion and Photon Beams, Journal of Vac. Science and Technol. B, Nov./Dec. 1996.

8. D. M. Williamson, et al., "Micrascan III, 0.25-?m resolution step and scan system," SPIE Vol. 2726, Optical Microlithography IX, 1996.

WILLIAM G. OLDHAM received his PhD in electrical engineering from Carnegie Mellon University in 1963. After a year on the research staff of Siemens AG in Erlangen, he joined the faculty of the University of California, Berkeley, where he is now a professor of electrical engineering. He is a fellow of the IEEE.

RICHARD SCHENKER graduated in February from UC-Berkeley with a PhD in electrical engineering. He has written several papers on UV-damage to optical materials and its impact on lithography. He is now a senior process engineer at Intel Corp.; ph 408/765-2431, e-mail [email protected].

* The density changes given are the computed "unrestrained" compaction levels that would be produced in the sample if it were irradiated uniformly. In most test geometries, only a small area of the sample is irradiated and the net density change is less than the unconstrained density change, owing to the influence of the undamaged portions of the sample.