- Email: [email protected]
Cleaning of optical surfaces by excimer laser radiation
applied
surface science ELSEVIER
Applied Surface Science 96-98 (1996) 463-468
Cleaning of optical surfaces by excimer laser radiation K. Mann Laser-Laboratorium
*,
B. Wolff-Rottke, F. Miller
Gijttingen, Hans-Adolf-Krebs- Weg I, D-37077 Gbttingen, Germany Received 22 May 1995
Abstract The effect of particle removal from Al mirror surfaces by the influence of pulsed UV laser radiation has been studied. The investigations are closely related to the demands of astronomers, who are looking for a more effective way to clean future very large telescope (VLT) mirrors [l]. A systematic parameter study has been performed in order to determine the
irradiation conditions which yield the highest dust removal efficiency (i.e. reflectivity increase) on contaminated samples. The particle removal rate increases with increasing laser fluence, being limited however by the damage threshold of the coating. Data indicate that on Al coated BK7 and Zerodur samples KrF laser radiation yields the optimum result, with cleaning efficiencies comparable to polymer film stripping. The initial reflectivity of the clean coating can nearly be restored, in particular when an additional solvent film on the sample surface is applied.
1. Introduction The effect of particle removal from solid state surfaces by pulsed UV laser irradiation was demonstrated several years ago. It has also been used for restoration of icones [2], cleaning of Si wafers [3], as well as cleaning optical components [5]. However, the mechanisms responsible for the laser induced detachment of particles are still under discussion. It is conceivable that contaminants are mechanically ejected from the surface due to the instantaneous thermal expansion of the pulsed laser irradiated substrate [3]. On the other hand, the importance of thin liquid layers on the substrate for the cleaning effect has been pointed out 131. UV photon induced bond breaking and/or explosive evaporation of the liquid film beneath the dust particles might also be respon-
* Corresponding 503599.
author. Tel.: +49-551-503541;
fax: +49-551-
0169-4332/96/$15.00 0 1996 Elsevier Science B.V. All rights reserved SSDI 0169-4332(95)00459-9
sible for the detachment. In order to achieve optimal cleaning efficiency, a comprehensive study of the removal power in terms of the relevant parameters is needed. Particle contamination is known to be a major problem for astronomical mirror surfaces, causing losses in reflectivity and high image background. The work presented here was initiated by demands of astronomers from ES0 (European Southern Observatory/Garching, FRG). ES0 plans to build ‘Very Large Telescope (VLT)’ systems with 8.2 m diameter primary mirrors on Mount Paranal (Chile). For the cleaning of such large mirror surfaces innovative procedures have to be developed [6]. In this paper we report on results of a feasibility study, in which the influence of the various laser parameters on the particle removal and corresponding reflectivity enhancement has been investigated systematically in order to define a process window for optimum cleaning performance.
464
K. Mann et al./Applied
Surface Science 96-98
energy Homogenizer
variable
(1996) 463-468
He-Ne laser ,,,
xy-translation
attenuator video microscopy Fig. 1. Experimental set-up used for on-line reflectivity measurements
2. Experimental The investigations were performed on Al coated BK7 and Zerodur substrates (film thickness 80-100 nm) provided by ESO. Samples had been exposed to dust contamination at Mount Paranal (Chile), the projected site of the VLT, for various times from 3 to 12 months, The main contamination is quartz sand with particle sizes from several up to a few hundred pm. Water marks are also present on some of the samples.
In order to guarantee a safe operation of a future laser cleaning system, laser induced damage thresholds (LIDT) have to be determined accurately both for the Al coatings and the underlying Zerodur substrate. Corresponding ‘l-on-l’ as well as ‘S-on-l’ measurements have been performed within the UV laser damage testing facility of Laser-Laboratorium Gijttingen [7]. In accordance with respective standards damage is defined as any alteration of the sample surface morphology detectable by microscopic means (Nomarski as well as on-line video
Rg. 2. Spahal intensity profile of homogenized KrF laser (4 mm X 4 mm).
K. Mann er al/Applied
465
Surface Science 96-98 (1996) 463-468
microscopy [7]). The laser cleaning efficiency is monitored by reflectivity measurements, which are performed both on-line with a He-Ne laser and off-line with a spectrometer. For the on-line measurements (cf. Fig. 1) a laser beam homogenizer [4] is employed in order to obtain a flat-topped intensity profile of 4 x 4 mm [2]. A corresponding 2D energy density distribution is displayed in Fig. 2. A He-Ne laser (633 nm) is directed onto the excimer laser irradiated sample site at an angle of incidence (Y= 62”. The intensity of the reflected beam is measured with a photodiode, yielding the relative change in reflectivity AR upon irradiation. Spectrally resolved absolute reflectivities are measured both before and after excimer laser irradiation with a UV-VIS-NIR spectrometer with reflectivity accessory (PerkinElmer Lambda 19, angle of incidence (Y= 6”). Obviously, due to the shadow cone cast by the particles under a shallow angle, the reflectivity increase observed upon particle detachment during the on-line He-Ne laser measurement (i.e. the sensitivity) is much higher than for the measurement at normal incidence, scaling with the factor (1 + tan (Y) given by the surface area obscured by a particle [l]. The cleaning efficiency, however, which can be defined as the reduction in particle covered substrate area at normal incidence, is overestimated by the same factor.
3. Results The influence of the following laser parameters on the cleaning effect was investigated: wavelength (A = 193, 248, 308, 351 nm (excimer laser), 355 nm (tripled Nd:YAG laser)), pulse width (T= 30 ns, 50...200 ns), pulse number per site (n = l-1000), jluence (H = 50- 1000 ml/cm’) and pulse repetition rate (f, = l-100 Hz). For all utilized UV wavelengths dust particles were at least partly removed from the mirror surfaces. In order to find the optimum wavelength for cleaning both the reflectivity enhancement AR and probability of laser induced damage of the coating have to be taken into account. The measured damage thresholds are found to increase monotonically with increasing laser wavelength (cf. Table l), which obviously reflects the spectral absorptivity data of
Table 1 ‘IO-on-l’ damage thresholds Ho of Al coated BK7 and cleaning fluences H,, needed for a certain reflectivity enhancement AR at different laser wavelengths (10 laser pulses per sample site, AR measured with He-Ne laser under 62”) A hm)
Ho
193 248 351
115 220 490
hJ/cm*)
H,, (mJ/cm* )
AR(%)
90 130 200
13 20 20
Al. Although this behaviour would speak in favour of a longer cleaning wavelength, the efficiency of the cleaning process is lower, since also higher fluence values are necessary and a smaller area can be cleaned per pulse (cf. data for 248 and 351 nm in Table 1). On the other hand, at 193 nm a maximum reflectivity enhancement of only about 13% is achieved with 90 mJ/cm*, which is already very close to the damage threshold. Due to the low efficiency this wavelength can be excluded for the process. Cleaning with a Nd:YAG laser at 355 nm can be ruled out for another reason: due to the nonlinear frequency tripling the beam profile of this laser is composed of strong intensity spikes which lead to a rapid thermal breakdown of the coating. In contrast, for 248 nm the reflectivity enhancement can reach 30% at a fluence of about 160 mJ/cm’; since the multiple pulse (‘S-on- 1’) damage threshold is considerably higher and tends to a saturation value for high pulse numbers (H, > 200 mJ/cm’ for S = 1000 on moderately contaminated samples [l]), a large number of consecutive cleaning cycles is possible without occurrence of thermal defects. The influence of the pulse duration was investigated for the wavelength 248 nm. The initial pulse duration of the applied excimer laser (30 ns) can be extended up to 200 ns by an optical delay line. According to the thermal breakdown model the damage threshold of Al coatings should show a square root dependence of the pulse duration [8]. But again, similar to the longer wavelengths, also the fluence required for efficient cleaning increases with increasing pulse length. Obviously, longer pulses do not improve the cleaning efficiency. The pulse width of 20-30 ns provided by standard excimer lasers seems to be well suited for cleaning.
K. Mann et al/Applied
466
Surface Science96-98 11996)463-468
In Fig. 3 the reflectivity increase AR at 633 nm is shown as a function of the number of laser pulses at 248 nm, indicating that the major enhancement of the reflectivity occurs during the first few pulses: already 43% of the maximum increase is achieved after the first pulse at 160 mJ/cm2 and 80% after 5 pulses. An equivalent saturation behaviour is also obtained from a statistical evaluation of the particle densities on the sample sites under test. Thus, only 3-5 cleaning pulses per site appear to be sufficient for an effective and rapid cleaning process. The laser fluence on the target is one of the most sensitive parameters for cleaning. In Fig. 4 the reflectivity increase is shown as a function of the fluence for an Al coated BK7 glass with a very high particle density. Below 50 mJ/cm’ no significant increase of the reflectivity is observed. For larger fluences the reflectivity strongly increases before it reaches a maximum, the position of which depends on the number of applied laser pulses (230 mJ/cm2 for 5 pulses, 160 mJ/cm2 for 15 pulses). For even higher fluences the reflectivity decreases due to the onset of laser damage to the coating. Thus, for 5 cleaning pulses, the optimum fluence without risk of damage can be determined to be 160 f 30 mJ/cm2 at 248 nm. For up-scaling of the cleaning process it is important to find the maximum pulse repetition rate tolerable with respect to damage resistance and cleaning
reflectivity
increase
6
[%]
8
# of laser
10 pulses
12
14
16
18
on 1 site
* H=136mJ/sqcm * H=162mJ/sqem
Fig. 3. Relative reflectivity increase on contaminated Al sample vs. number of pulses for 248 nm radiation (pulse length 30 ns).
_30
_ 248nm 30ns
e a226I F’OE ._
4x4mm2
.g15 .z gio5 L 5m
o;j
/.
50
100
150
laser fluence
200
250
3 10
[mJ/sqcm]
Fig. 4. Relative increase in reflectivity of a contaminated Al sample as a function of KrF laser fluence, using 5 and 15 laser pulses per site (He-Ne laser under 62”).
efficiency. Damage threshold measurements indicate no influence of the repetition rate on the durability of the Al coating. However, the cleaning efficiency was found to decrease with higher repetition rates; especially, a large amount of smaller particles is observed which were originally not present. These originate from a radiation induced decomposition of already detached particles or larger particle clusters, which are still in the region of the laser beam during subsequent pulses. Hence, the process speed can not simply be enhanced by increasing the repetition rate.
However, there is a way to overcome this problem, i.e. the application of an array of parallel streak profiles at a high pulse repetition rate using adjacent cylindrical lenslets. By scanning this profile over the sample surface at a speed such that each following pulse irradiates the area adjacent to the previously irradiated zone without overlap, the actual repetition rate per site is reduced without reducing the process speed 111. Additionally, an auxiliary gas flow or suction should be applied in order to remove already detached particles out of the irradiated zone. Summarizing the above results, the optimum cleaning effect on contaminated Al coatings is obtained by 3-5 KrF laser pulses (Q-= 30 ns) with a homogeneous spatial profile and a fluence H = 160 f 30 mJ/cm2. The effective pulse repetition rate fP on each site should be less than 5 Hz. Using this ‘process window’, cleaning is possible for a variety of different contaminants (dust, hydrocarbon films,
K. Mann et al. /Applied Surface Science 96-98 (1996) 463-468
467
water mark found on the samples provided by ESO. The residual bright dots on the cleaned areas of both micrographs can be attributed to already existing coating defects as well as to readsorbed particles.
4. Discussion With respect to the mechanisms responsible for particle desorption the results indicate that the driving force is the absorption of UV radiation by the Al coating and not by the particles themselves, since the process works even for quartz sand as contaminant with a very low UV absorptivity. Moreover, it is observed that humidity is of great importance for cleaning. This is documented by the fact, that artificial contamination with dry quartz sand in a dry atmosphere as well as readsorbed particles are very hard to detach with the laser. It is conceivable that moisture builds a very thin liquid film beneath the dust particles due to capillary forces. As discussed by Tam et al., this film is strongly superheated at the interface upon absorption of the laser pulse by the substrate. The resulting extremely high liquid pressure can cause an explosive ejection of the dust particles [3,9]. For particles which readsorb on already cleaned areas detachment becomes very inefficient, since moisture cannot build up again beneath the particles in the time interval between successive laser pulses. For this reason an auxiliary air gas flow has to be applied during laser cleaning in order to remove the detached particles and to avoid their fragmentation and redeposition. The important role
Fig. 5. Dark field micrographs of laser cleaned areas on strongly contaminated Al mirrors, showing the removal of dust particles (above) and of water marks (below); image size about 3 mm (her )
water marks, finger prints etc.). E.g., Fig. 5 (above) shows in its upper half the laser induced removal of quartz sand from a strongly contaminated Al coating; Fig. 5 (below) displays a laser cleaned track on a
76.888
---.-+----
---_ ---L-/-e
Fig. 6. Absolute reflectivity as a function of wavelength for Al coating before (lower curve, c) and after cleaning, using polymer film stripping (a) or laser process assisted by ethanol film (b).
468
K. Mann et &./Applied &face
of moisture is also demonstrated by the observation that a thin liquid film sprayed onto the sample surface just in time to be completely evaporated by the next laser pulse leads to a strong further improvement of the cleaning efficiency [1,3]. A corresponding result is displayed in Fig. 6, showing the spectrally resolved absolute reflectivity of a laser cleaned mirror in comparison with data obtained from a mirror site cleaned by polymer film stripping. In both cases a 12% increase in reflectivity can be achieved.
5. Summary We have found that 248 mn/30 ns pulses are well suited for desorption of quartz particles from Al mirror surfaces. The laser induced reflectivity enhancement increases with higher laser fluence, the upper fluence limit being given by the damage threshold of the coating. At fluences slightly below the damage threshold up to 80% of the lost reflectivity can be reinstalled, which requires, however, a homogeneous laser intensity profile in order to avoid damage of the coating. The reflectivity also increases with increasing pulse number per site with a maximum increase within the first few pulses. There is strong evidence from our data that humidity plays a dominant role in the laser desorption process.
Science 96-98 (19%) 463-468
Acknowledgements The work was sponsored by the European Southern Observatory, Garching (FRG). The authors especially like to thank P. Giordano (ES01 for stimulating discussions as well as for providing the Al mirror samples.
References [l] K. Mann, F. Miller and B. Wolff-Rot&e, UV Laser Cleaning of Very Large Telescope (VLT) Mirrors (European Southern Observatory (ESO), May 1994). [2] E. Hontzopoulos, C. Fotakis and M. Doulgetidis, SPIE 1810 (1992) 784. [3] A.C. Tam, W.P. Leung, W. Zapka and W. Ziemlich, J. Appl. Phys. 71 (1992) 3515. [4] K. Mann, A. Hopfmliller, H.Gerhardt, P. Gorzellik, R. Schild, W. Stijffler and H. Wagner, Proc. of OE’Technol., Boston (USA), SPIE 1834 (1992) 184. [5] J.D. Kelley, in: Laser Induced Damage in Optical Materials, 1991, SPIE 1624 (1991) 153. [6] P. Giordano and A. Tormjon, in: The ES0 Messenger 75 (199419. [7] K. Mann and H. Gerhardt, Proc. of EC0 IV, Int. Conf. on Opt. Sci. and Eng., Den Haag, SPIE 1503 (1991) 176. [8] K. Mann, H. Gerhardt, G. Pfeifer and R. Wolf, in: Laser Induced Damage in Optical Materials, 1991, SPIE 1624 (1991) 436. [9] W. Zapka, W. Ziemlich, W.P. Leung and AC. Tam, Microelectron. Eng. 20 (1993) 171.
surface science ELSEVIER
Applied Surface Science 96-98 (1996) 463-468
Cleaning of optical surfaces by excimer laser radiation K. Mann Laser-Laboratorium
*,
B. Wolff-Rottke, F. Miller
Gijttingen, Hans-Adolf-Krebs- Weg I, D-37077 Gbttingen, Germany Received 22 May 1995
Abstract The effect of particle removal from Al mirror surfaces by the influence of pulsed UV laser radiation has been studied. The investigations are closely related to the demands of astronomers, who are looking for a more effective way to clean future very large telescope (VLT) mirrors [l]. A systematic parameter study has been performed in order to determine the
irradiation conditions which yield the highest dust removal efficiency (i.e. reflectivity increase) on contaminated samples. The particle removal rate increases with increasing laser fluence, being limited however by the damage threshold of the coating. Data indicate that on Al coated BK7 and Zerodur samples KrF laser radiation yields the optimum result, with cleaning efficiencies comparable to polymer film stripping. The initial reflectivity of the clean coating can nearly be restored, in particular when an additional solvent film on the sample surface is applied.
1. Introduction The effect of particle removal from solid state surfaces by pulsed UV laser irradiation was demonstrated several years ago. It has also been used for restoration of icones [2], cleaning of Si wafers [3], as well as cleaning optical components [5]. However, the mechanisms responsible for the laser induced detachment of particles are still under discussion. It is conceivable that contaminants are mechanically ejected from the surface due to the instantaneous thermal expansion of the pulsed laser irradiated substrate [3]. On the other hand, the importance of thin liquid layers on the substrate for the cleaning effect has been pointed out 131. UV photon induced bond breaking and/or explosive evaporation of the liquid film beneath the dust particles might also be respon-
* Corresponding 503599.
author. Tel.: +49-551-503541;
fax: +49-551-
0169-4332/96/$15.00 0 1996 Elsevier Science B.V. All rights reserved SSDI 0169-4332(95)00459-9
sible for the detachment. In order to achieve optimal cleaning efficiency, a comprehensive study of the removal power in terms of the relevant parameters is needed. Particle contamination is known to be a major problem for astronomical mirror surfaces, causing losses in reflectivity and high image background. The work presented here was initiated by demands of astronomers from ES0 (European Southern Observatory/Garching, FRG). ES0 plans to build ‘Very Large Telescope (VLT)’ systems with 8.2 m diameter primary mirrors on Mount Paranal (Chile). For the cleaning of such large mirror surfaces innovative procedures have to be developed [6]. In this paper we report on results of a feasibility study, in which the influence of the various laser parameters on the particle removal and corresponding reflectivity enhancement has been investigated systematically in order to define a process window for optimum cleaning performance.
464
K. Mann et al./Applied
Surface Science 96-98
energy Homogenizer
variable
(1996) 463-468
He-Ne laser ,,,
xy-translation
attenuator video microscopy Fig. 1. Experimental set-up used for on-line reflectivity measurements
2. Experimental The investigations were performed on Al coated BK7 and Zerodur substrates (film thickness 80-100 nm) provided by ESO. Samples had been exposed to dust contamination at Mount Paranal (Chile), the projected site of the VLT, for various times from 3 to 12 months, The main contamination is quartz sand with particle sizes from several up to a few hundred pm. Water marks are also present on some of the samples.
In order to guarantee a safe operation of a future laser cleaning system, laser induced damage thresholds (LIDT) have to be determined accurately both for the Al coatings and the underlying Zerodur substrate. Corresponding ‘l-on-l’ as well as ‘S-on-l’ measurements have been performed within the UV laser damage testing facility of Laser-Laboratorium Gijttingen [7]. In accordance with respective standards damage is defined as any alteration of the sample surface morphology detectable by microscopic means (Nomarski as well as on-line video
Rg. 2. Spahal intensity profile of homogenized KrF laser (4 mm X 4 mm).
K. Mann er al/Applied
465
Surface Science 96-98 (1996) 463-468
microscopy [7]). The laser cleaning efficiency is monitored by reflectivity measurements, which are performed both on-line with a He-Ne laser and off-line with a spectrometer. For the on-line measurements (cf. Fig. 1) a laser beam homogenizer [4] is employed in order to obtain a flat-topped intensity profile of 4 x 4 mm [2]. A corresponding 2D energy density distribution is displayed in Fig. 2. A He-Ne laser (633 nm) is directed onto the excimer laser irradiated sample site at an angle of incidence (Y= 62”. The intensity of the reflected beam is measured with a photodiode, yielding the relative change in reflectivity AR upon irradiation. Spectrally resolved absolute reflectivities are measured both before and after excimer laser irradiation with a UV-VIS-NIR spectrometer with reflectivity accessory (PerkinElmer Lambda 19, angle of incidence (Y= 6”). Obviously, due to the shadow cone cast by the particles under a shallow angle, the reflectivity increase observed upon particle detachment during the on-line He-Ne laser measurement (i.e. the sensitivity) is much higher than for the measurement at normal incidence, scaling with the factor (1 + tan (Y) given by the surface area obscured by a particle [l]. The cleaning efficiency, however, which can be defined as the reduction in particle covered substrate area at normal incidence, is overestimated by the same factor.
3. Results The influence of the following laser parameters on the cleaning effect was investigated: wavelength (A = 193, 248, 308, 351 nm (excimer laser), 355 nm (tripled Nd:YAG laser)), pulse width (T= 30 ns, 50...200 ns), pulse number per site (n = l-1000), jluence (H = 50- 1000 ml/cm’) and pulse repetition rate (f, = l-100 Hz). For all utilized UV wavelengths dust particles were at least partly removed from the mirror surfaces. In order to find the optimum wavelength for cleaning both the reflectivity enhancement AR and probability of laser induced damage of the coating have to be taken into account. The measured damage thresholds are found to increase monotonically with increasing laser wavelength (cf. Table l), which obviously reflects the spectral absorptivity data of
Table 1 ‘IO-on-l’ damage thresholds Ho of Al coated BK7 and cleaning fluences H,, needed for a certain reflectivity enhancement AR at different laser wavelengths (10 laser pulses per sample site, AR measured with He-Ne laser under 62”) A hm)
Ho
193 248 351
115 220 490
hJ/cm*)
H,, (mJ/cm* )
AR(%)
90 130 200
13 20 20
Al. Although this behaviour would speak in favour of a longer cleaning wavelength, the efficiency of the cleaning process is lower, since also higher fluence values are necessary and a smaller area can be cleaned per pulse (cf. data for 248 and 351 nm in Table 1). On the other hand, at 193 nm a maximum reflectivity enhancement of only about 13% is achieved with 90 mJ/cm*, which is already very close to the damage threshold. Due to the low efficiency this wavelength can be excluded for the process. Cleaning with a Nd:YAG laser at 355 nm can be ruled out for another reason: due to the nonlinear frequency tripling the beam profile of this laser is composed of strong intensity spikes which lead to a rapid thermal breakdown of the coating. In contrast, for 248 nm the reflectivity enhancement can reach 30% at a fluence of about 160 mJ/cm’; since the multiple pulse (‘S-on- 1’) damage threshold is considerably higher and tends to a saturation value for high pulse numbers (H, > 200 mJ/cm’ for S = 1000 on moderately contaminated samples [l]), a large number of consecutive cleaning cycles is possible without occurrence of thermal defects. The influence of the pulse duration was investigated for the wavelength 248 nm. The initial pulse duration of the applied excimer laser (30 ns) can be extended up to 200 ns by an optical delay line. According to the thermal breakdown model the damage threshold of Al coatings should show a square root dependence of the pulse duration [8]. But again, similar to the longer wavelengths, also the fluence required for efficient cleaning increases with increasing pulse length. Obviously, longer pulses do not improve the cleaning efficiency. The pulse width of 20-30 ns provided by standard excimer lasers seems to be well suited for cleaning.
K. Mann et al/Applied
466
Surface Science96-98 11996)463-468
In Fig. 3 the reflectivity increase AR at 633 nm is shown as a function of the number of laser pulses at 248 nm, indicating that the major enhancement of the reflectivity occurs during the first few pulses: already 43% of the maximum increase is achieved after the first pulse at 160 mJ/cm2 and 80% after 5 pulses. An equivalent saturation behaviour is also obtained from a statistical evaluation of the particle densities on the sample sites under test. Thus, only 3-5 cleaning pulses per site appear to be sufficient for an effective and rapid cleaning process. The laser fluence on the target is one of the most sensitive parameters for cleaning. In Fig. 4 the reflectivity increase is shown as a function of the fluence for an Al coated BK7 glass with a very high particle density. Below 50 mJ/cm’ no significant increase of the reflectivity is observed. For larger fluences the reflectivity strongly increases before it reaches a maximum, the position of which depends on the number of applied laser pulses (230 mJ/cm2 for 5 pulses, 160 mJ/cm2 for 15 pulses). For even higher fluences the reflectivity decreases due to the onset of laser damage to the coating. Thus, for 5 cleaning pulses, the optimum fluence without risk of damage can be determined to be 160 f 30 mJ/cm2 at 248 nm. For up-scaling of the cleaning process it is important to find the maximum pulse repetition rate tolerable with respect to damage resistance and cleaning
reflectivity
increase
6
[%]
8
# of laser
10 pulses
12
14
16
18
on 1 site
* H=136mJ/sqcm * H=162mJ/sqem
Fig. 3. Relative reflectivity increase on contaminated Al sample vs. number of pulses for 248 nm radiation (pulse length 30 ns).
_30
_ 248nm 30ns
e a226I F’OE ._
4x4mm2
.g15 .z gio5 L 5m
o;j
/.
50
100
150
laser fluence
200
250
3 10
[mJ/sqcm]
Fig. 4. Relative increase in reflectivity of a contaminated Al sample as a function of KrF laser fluence, using 5 and 15 laser pulses per site (He-Ne laser under 62”).
efficiency. Damage threshold measurements indicate no influence of the repetition rate on the durability of the Al coating. However, the cleaning efficiency was found to decrease with higher repetition rates; especially, a large amount of smaller particles is observed which were originally not present. These originate from a radiation induced decomposition of already detached particles or larger particle clusters, which are still in the region of the laser beam during subsequent pulses. Hence, the process speed can not simply be enhanced by increasing the repetition rate.
However, there is a way to overcome this problem, i.e. the application of an array of parallel streak profiles at a high pulse repetition rate using adjacent cylindrical lenslets. By scanning this profile over the sample surface at a speed such that each following pulse irradiates the area adjacent to the previously irradiated zone without overlap, the actual repetition rate per site is reduced without reducing the process speed 111. Additionally, an auxiliary gas flow or suction should be applied in order to remove already detached particles out of the irradiated zone. Summarizing the above results, the optimum cleaning effect on contaminated Al coatings is obtained by 3-5 KrF laser pulses (Q-= 30 ns) with a homogeneous spatial profile and a fluence H = 160 f 30 mJ/cm2. The effective pulse repetition rate fP on each site should be less than 5 Hz. Using this ‘process window’, cleaning is possible for a variety of different contaminants (dust, hydrocarbon films,
K. Mann et al. /Applied Surface Science 96-98 (1996) 463-468
467
water mark found on the samples provided by ESO. The residual bright dots on the cleaned areas of both micrographs can be attributed to already existing coating defects as well as to readsorbed particles.
4. Discussion With respect to the mechanisms responsible for particle desorption the results indicate that the driving force is the absorption of UV radiation by the Al coating and not by the particles themselves, since the process works even for quartz sand as contaminant with a very low UV absorptivity. Moreover, it is observed that humidity is of great importance for cleaning. This is documented by the fact, that artificial contamination with dry quartz sand in a dry atmosphere as well as readsorbed particles are very hard to detach with the laser. It is conceivable that moisture builds a very thin liquid film beneath the dust particles due to capillary forces. As discussed by Tam et al., this film is strongly superheated at the interface upon absorption of the laser pulse by the substrate. The resulting extremely high liquid pressure can cause an explosive ejection of the dust particles [3,9]. For particles which readsorb on already cleaned areas detachment becomes very inefficient, since moisture cannot build up again beneath the particles in the time interval between successive laser pulses. For this reason an auxiliary air gas flow has to be applied during laser cleaning in order to remove the detached particles and to avoid their fragmentation and redeposition. The important role
Fig. 5. Dark field micrographs of laser cleaned areas on strongly contaminated Al mirrors, showing the removal of dust particles (above) and of water marks (below); image size about 3 mm (her )
water marks, finger prints etc.). E.g., Fig. 5 (above) shows in its upper half the laser induced removal of quartz sand from a strongly contaminated Al coating; Fig. 5 (below) displays a laser cleaned track on a
76.888
---.-+----
---_ ---L-/-e
Fig. 6. Absolute reflectivity as a function of wavelength for Al coating before (lower curve, c) and after cleaning, using polymer film stripping (a) or laser process assisted by ethanol film (b).
468
K. Mann et &./Applied &face
of moisture is also demonstrated by the observation that a thin liquid film sprayed onto the sample surface just in time to be completely evaporated by the next laser pulse leads to a strong further improvement of the cleaning efficiency [1,3]. A corresponding result is displayed in Fig. 6, showing the spectrally resolved absolute reflectivity of a laser cleaned mirror in comparison with data obtained from a mirror site cleaned by polymer film stripping. In both cases a 12% increase in reflectivity can be achieved.
5. Summary We have found that 248 mn/30 ns pulses are well suited for desorption of quartz particles from Al mirror surfaces. The laser induced reflectivity enhancement increases with higher laser fluence, the upper fluence limit being given by the damage threshold of the coating. At fluences slightly below the damage threshold up to 80% of the lost reflectivity can be reinstalled, which requires, however, a homogeneous laser intensity profile in order to avoid damage of the coating. The reflectivity also increases with increasing pulse number per site with a maximum increase within the first few pulses. There is strong evidence from our data that humidity plays a dominant role in the laser desorption process.
Science 96-98 (19%) 463-468
Acknowledgements The work was sponsored by the European Southern Observatory, Garching (FRG). The authors especially like to thank P. Giordano (ES01 for stimulating discussions as well as for providing the Al mirror samples.
References [l] K. Mann, F. Miller and B. Wolff-Rot&e, UV Laser Cleaning of Very Large Telescope (VLT) Mirrors (European Southern Observatory (ESO), May 1994). [2] E. Hontzopoulos, C. Fotakis and M. Doulgetidis, SPIE 1810 (1992) 784. [3] A.C. Tam, W.P. Leung, W. Zapka and W. Ziemlich, J. Appl. Phys. 71 (1992) 3515. [4] K. Mann, A. Hopfmliller, H.Gerhardt, P. Gorzellik, R. Schild, W. Stijffler and H. Wagner, Proc. of OE’Technol., Boston (USA), SPIE 1834 (1992) 184. [5] J.D. Kelley, in: Laser Induced Damage in Optical Materials, 1991, SPIE 1624 (1991) 153. [6] P. Giordano and A. Tormjon, in: The ES0 Messenger 75 (199419. [7] K. Mann and H. Gerhardt, Proc. of EC0 IV, Int. Conf. on Opt. Sci. and Eng., Den Haag, SPIE 1503 (1991) 176. [8] K. Mann, H. Gerhardt, G. Pfeifer and R. Wolf, in: Laser Induced Damage in Optical Materials, 1991, SPIE 1624 (1991) 436. [9] W. Zapka, W. Ziemlich, W.P. Leung and AC. Tam, Microelectron. Eng. 20 (1993) 171.