Plasma treatment for restoration ofdielectric multilayer mirrors in short-wavelengthfree-electron lasers

Kawakatsu Yamada, Tetsuo Yamazaki, Takashi Shimizu, Norihiro Sei,and Tomohisa Mikado

Dielectric multilayer mirrors, degraded through irradiation by high-energy undulator radiation, weresuccessfully restored by surface treatment with RF-induced O2 plasma. Themirror loss, which had beenincreased up to,1000 parts in 106 1ppm2 through themirror degradation, was drastically reduced to,100ppm during the treatment. Such a mirror-restoration technique has been desired especially inshort-wavelength free-electron lasers 1FEL’s2, because the laser gain is so small that even a mirror loss assmall as ,1000 ppm interferes with the FEL oscillation. The mirror degradation is most likely causedby the deposition and doping of carbon atoms onto the dielectric surface. The surface analysis by thex-ray photoelectron spectroscopy revealed that the plasma treatment effectively removed the carboncontamination covering the mirror surfaces without serious surface damage by high-energy particlesfrom the plasma.Key words: Free-electron lasers, plasma treatment, mirror restoration.

1. Introduction

At the Electrotechnical Laboratory 1ETL2, we havebeen conducting short-wavelength free-electron laser1FEL2 experiments with two electron storage rings,the Tsukuba Electron Ring for Accelerating and Stor-age 1TERAS2 and the NIJI-IV. The first lasing in thevisible wavelength was obtained in 19911,2 on TERAS,which was designed for general users of synchrotronradiation. The new ring, NIJI-IV,3 dedicated to aFEL, has been under construction since 1990 torealize a FEL in the ultraviolet wavelength. Nowwecan obtain laser oscillation over the wavelength from595 to 350 nm with the NIJI-IV FEL system.4,5However, some problems remain to be solved. Themirror degradation during the laser operation is oneof the most critical issues.In storage-ring FEL’s, the laser cavity is usually

composed of high-reflectance and low-optical-loss di-electric multilayer mirrors, because of their smallgain. However, it is well known that the optical lossof the dielectric multilayer cavity mirrors rapidlyincreases when the mirror surface is irradiated by a

The authors are with Electrotechnical Laboratory, 1-1-4 Ume-zono, Tsukuba City, Ibaraki 305, Japan.Received 3 November 1994.0003-6935@95@214261-05$06.00@0.

r 1995 Optical Society of America.

higher harmonic component in the undulator radia-tion.6–9 The situation is particularly serious at wave-lengths shorter than the visible, because the lasergain becomes so small that the total cavity loss,including the transmittance for output coupling, mustbe of the order of 10–100 parts in 106 1ppm2 to achievelaser oscillation. It was suggested that such a mir-ror degradation was caused by the deposition anddoping of carbon atoms onto the dielectric surfaces.6–8To eliminate the carbon contamination and reduce themirror-degradation problem, we tried surface treat-ment with RF-induced O2 plasma for the first time insuch a very low-loss region, to our knowledge.Although similar methods have been used to cleanoptical elements for space vehicles10,11 or synchrotronradiation beam lines,12,13 in those cases the loss wasas much as several percent to several tens of percent,which is on a much larger level than that of themirrors usually used in FEL experiments. The mir-ror surfaces were analyzed by the use of x-ray photo-electron spectroscopy 1XPS2 before and after oxygen–plasma treatment.

2. Experimental

The sample mirrors were from the Ojai Research Co.and had been used in the TERAS FEL experi-ments.1,2,14 These mirrors have alternating TiO2and SiO2 layers of 1@4-l thickness, with a 1@2-l-thick

20 July 1995 @ Vol. 34, No. 21 @ APPLIED OPTICS 4261

SiO2 layer on the top for surface protection, where lrepresents the light wavelength to which the mirrorsare optimized. Although their original loss was ,40ppm around 585 nm, including 30-ppm transmit-tance, it increased rapidly up to several hundred ppmwith irradiation by the undulator radiation. Figure1 shows the wavelength dependence of mirror lossesfor three degraded samples, d, f, and h. Mirror losswas measured with the cavity-decay-time method bythe use of a cw dye laser.7,9 The mirror exposure,defined as a product of the ring current and theexposure time, was of the order of 10–100 mA hr forthese samples. The open circles show the originallosses for a new mirror. It is found that there aredifferent types of degradation. For samples d and f,the bandwidths of the mirrors are found to be nar-rowed around different optimum wavelengths, whichcan be explained as the degradation of the mirrorsurface.7 On the other hand, in sample h the losscurve does not indicate a strong dependence on thewavelength, although the loss is relatively large.This may be caused by the volumetric degradationinside the dielectrics. Here we concentrate on thesurface degradation, because it is much more seriousthan the volumetric degradation in the visible FEL’s.Treatment with oxygen plasma was tried to clean

and repair the degraded mirror surfaces. Figure 2shows the reaction chamber for the plasma treatment.The chamber was evacuated down to ,2 3 1025 Torrwith a turbo molecular pump, TMP, and a rotarypump, RP. After the evacuation, oxygen gas wasintroduced from above the sample along a quartz tubeat a flow rate of 50 standard cubic centimeters perminute. The oxygen pressure inside the chamberwas set at 2.0 Torr during the plasma treatment,except in a couple of experiments in which it was setat 0.3 Torr. The plasma was generated inside thequartz tube by feeding a RF at 13.56 MHz with a netinput power of 450 W through a water-cooled two-turn coil. The distance between the center of the

Fig. 1. Wavelength dependence of mirror losses for degradedsamples d, f, and h. Dashed curve, original-loss curve for a newmirror.

4262 APPLIED OPTICS @ Vol. 34, No. 21 @ 20 July 1995

plasma and the mirror surface was approximately20 cm, which was sufficiently long to avoid surfacedamage by high-energy particles from the plasma.The temperature of the mirror substrate was ,30 °Cduring the treatment.Various kinds of species, such as atomic oxygen

radicals and atomic and molecular oxygen ions, areexpected to be generated in the plasma. To identifythese species, the emitted light was observed with amonochromator 1Jarrell-Ash Monospec 182 equippedwith a photodiode array 1Princeton Instruments SMAIRY-700G2. The total wavelength resolution of thesystem was ,1.5 nm. Typical spectra from theplasma are shown in Fig. 3. The upper and lowertraces are those spectra observed near the plasmaand near the sample surface, respectively. It isfound that the majority of the spectral lines are thoseemitted from atomic oxygen radicals, and only twovery intense lines of this type are observed near thesample surface. This suggests that the atomic oxy-gen radicals play an important role in the treatingprocess.Mirror surfaces were analyzed with XPS to investi-

gate the mirror-degradation mechanism, where thephotoelectrons excited by Mg Ka of 1253.7 eV wereanalyzed with a hemispherical energy analyzer 1VSWScientific Instruments HA1502with the energy resolu-tion of ,0.2 eV. To calibrate the binding energy, thesilicon 2p peak from the SiO2 was scaled at 103.4 eV.15

3. Results and Discussion

Figure 4 shows the successful restoration of thedegraded mirrors by the use of the plasma treatment.The ordinate and the abscissa indicate the loss permirror and the treatment time, respectively. Themeasurement was made at the wavelength of 611 nm.The O2 pressure during the treatment was 2 Torr for

Fig. 2. Schematic drawing of the reaction chamber for the plasmatreatment.

Fig. 3. Typical spectra from various species generated in the O2 plasma. The upper and lower traces were observed near the plasma andnear the sample surface, respectively.

samples a, b, and c and 0.3 Torr for d. It is found insamples b and c that the mirror loss was decreaseddown to ,1 3 1024 1100 ppm2 after 60-min treat-ment, which is a usable level for the laser oscillationexperiment, even though it is larger than their origi-nal loss 1,40 ppm2. On the other hand, the mirrorreflectivity was not restored so effectively in sample a.This is because only sample a was placed upstream forthe electron motion in the laser cavity and was notexposed to the intense undulator radiation directly.The mirror loss in this case should have been causednot by surface degradation by carbon atoms but byinner defects inside the dielectric coatings as a resultof, for example, an exposure to the high-energyradiation scattered from the ring components duringbeam injection, because the sample a was placed inthe FEL system for approximately 10 times as long aperiod as other samples. It is clear that such volu-metric degradation cannot be repaired with the sur-face treatment by the plasma. This type of degrada-

Fig. 4. Dependencies of the mirror losses on treatment time.The O2 pressure during the treatment was 2 Torr for samples a, b,and c and 0.3 Torr for d. Only sample a was placed upstream forthe electron motion in the laser cavity.

20 July 1995 @ Vol. 34, No. 21 @ APPLIED OPTICS 4263

tion can also occur in samples b and c, although itseffect must be much smaller than in case a. Thedifference between the final loss 1,100 ppm2 of plasma-treatedmirrors b and c and their original loss 140 ppm2may be caused by such a volumetric degradation.Sample d was treated in low-pressure O2 10.3 Torr2,where the bright region in the plasmawas observed tocome close to the sample surface. As a result, theloss was reduced in a much shorter time but tended toincrease again. This is probably due to the onset ofthe surface damage by high-energy particles from theplasma. This implies that the plasma conditionmustbe finely controlled for the surface treatment in verylow-loss regions. The RF plasma is found to besuitable for this purpose, because the plasma param-eters, such as the input RF power and the distancebetween the plasma and the sample, are easily con-trolled.Figure 5 shows the typical photoelectron spectra

before and after the plasma treatment. In the uppertrace of Fig. 5, an intense carbon peak is observed inaddition to the expected silicon and oxygen peaksfrom the SiO2 top layer. This indicates that carbonwas deposited onto the dielectric surface in the case ofdegraded mirrors. On the other hand, it is shown inthe lower trace that the plasma treatment makes thecarbon peak weaker and the silicon peak more in-tense, which infers that the carbon contaminationcovering the mirror surface has been effectively re-moved. More detailed information on the carbonpeak is given in Fig. 6. The shape of the carbon peakis asymmetric with respect to the binding energy

Fig. 5. Typical x-ray photoelectron spectra before and afterplasma treatment.

4264 APPLIED OPTICS @ Vol. 34, No. 21 @ 20 July 1995

before the plasma treatment, as shown in Fig. 6.This suggests that the carbon peak is formed by acertain convolution of several chemical bonds, includ-ing Si–C 1283.4 eV2, C–C 1284.7 eV216 and some others,such as C–O and hydrocarbonic bonds, which arewidely distributed at 285–291 eV. Moreover, it shouldbe noted that the binding energy at the maximum ofthe envelope lies closer to that for Si–C than that forC–C, which means that a large portion of the carbonatoms was doped into the SiO2 layer to form SiC.A long tail toward the higher-binding-energy sideimplies that some infinitesimal molecules remainingin the vacuum, such as CO and hydrocarbons, areadsorbing on the mirror surface, which will be thesource of carbon atoms. The reduction of the peakheight after the plasma treatment indicates the effec-tive removal of the doped carbon as well as thedeposited carbon. This means that the carbon-dopedlayer is so thin that the doped carbon is easilyremoved by the surface treatment only. According tothe results from XPS measurement, it was confirmedthat the mirror degradation is mainly caused by thedeposition and doping of carbon atoms onto the dielec-tric surface in the visible FEL’s.

4. Conclusion

In the dielectric multilayer mirrors used in short-wavelength FEL’s, different types of degradation areobserved, that is, surface degradation by carbon andvolumetric degradation by inner defects. Surfacetreatment with RF-induced O2 plasma was applied tothe degraded mirrors to mitigate the surface degrada-tion and was found to be very effective in restoringthe mirror reflectivity even at a loss level of the orderof 10–100 ppm. XPS analysis indicated that thecarbon peak, which is a certain convolution of Si–C,C–C, and some other bonds, including CO and hydro-carbons, was drastically reduced by the plasma treat-

Fig. 6. Detailed observation of the carbon 1-s peak in the x-rayphotoelectron spectra before and after plasma treatment. Thepeak positions expected for Si–C and C–C bonds are indicated witharrows. The C–O and hydrocarbonic bonds are distributed at285–291 eV.

ment. This fact means that the mirror degradationin the visible FEL’s was mainly caused by surfacedegradation as a result of deposition and doping ofcarbon atoms onto the dielectric surface. RF plasmais suitable for surface cleaning at a very low-loss level,because the surface damage by the high-energy par-ticles from the plasma can be avoided by control ofplasma parameters, such as the distance between theplasma and the sample mirror and the RF powerinput to the plasma.Now we are investigating the mirror-degradation

mechanisms in the UV wavelength, where the volu-metric degradation will be more important. In thiscase some other mirror-restoration techniques will benecessitated. Mirror-restoration experiments withboth RF-induced O2 plasma and thermal annealingare in progress.

References and Notes1. T. Yamazaki, K. Yamada, S. Sugiyama, H. Ohgaki, T. Tomi-

masu, T. Noguchi, T. Mikado, M. Chiwaki, and R. Suzuki,‘‘Lasing in visible of a storage-ring free electron laser at ETL,’’Nucl. Instrum. Methods 309, 343–347 119912.

2. K. Yamada, T. Yamazaki, S. Sugiyama, T. Tomimasu, H.Ohgaki, T. Noguchi, T. Mikado, M. Chiwaki, and R. Suzuki,‘‘Visible oscillation of storage-ring free electron laser onTERAS,’’ Nucl. Instrum. Methods 318, 33–37 119922.

3. M. Kawai, K. Aizawa, S. Kamiya, M. Yokoyama, Y. Oku, K.Owaki, H. Miura, A. Iwata, M. Yoshiwa, T. Tomimasu, S.Sugiyama, H. Ohgaki, T. Yamazaki, K. Yamada, T. Mikado,and T. Noguchi, ‘‘Present status of a 500MeV compact electronstorage ring for UV FEL experiments,’’ Nucl. Instrum. Meth-ods 318, 135–141 119922.

4. T. Yamazaki, K. Yamada, S. Sugiyama, H. Ohgaki, N. Sei, T.Mikado, T. Noguchi, M. Chiwaki, and R. Suzuki, ‘‘First lasingof the NIJI-IV storage-ring free-electron laser,’’ Nucl. Instrum.Methods 331, 27–33 119932.

5. T. Yamazaki, K. Yamada, N. Sei, H. Ohgaki, M. Kawai, M.

Yokoyama, S. Hamada, S. Sugiyama, T. Mikado, R. Suzuki, T.Noguchi, M. Chiwaki, and A. Iwata, ‘‘Lasing at 352 nm of theNIJI-IV storage-ring free-electron laser,’’ Jpn. J. Appl. Phys.33, 1224–1227 119942.

6. M. Billardon, P. Elleume, J. M. Ortega, C. Bazin, M. Bergher,M. Velghe, D. A. G. Deacon, and Y. Petroff, ‘‘Free-electron laserexperiment at Orsay: a review,’’ IEEE J. Quantum Electron.QE-21, 805–823 119852.

7. P. Elleaume, M. Velghe, M. Billardon, and J. M. Ortega,‘‘Diagnostic techniques and UV-induced degradation of themirrors used in the Orsay storage ring free-electron laser,’’Appl. Opt. 24, 2762–2770 119852.

8. D. A. G. Deacon, ‘‘Optical coating damage and performancerequirements in free electron lasers,’’ Nucl. Instrum. Methods250, 283–288 119862.

9. K. Yamada, T. Yamazaki, S. Sugiyama, T. Tomimasu, T.Mikado, M. Chiwaki, R. Suzuki, and H. Ohgaki, ‘‘Free electronlaser experiment at ETL,’’ Nucl. Instrum. Methods 304, 86–92119912.

10. R. B. Gillette and B. A. Kenyon, ‘‘Photon-induced contaminantfilm effects on ultraviolet reflecting mirrors,’’ Appl. Opt. 10,545–551 119712.

11. W. R. Hunter, G. N. Steele, and R. B. Gillette, ‘‘Increase intransmittance of unbacked aluminum filters exposed to rf or dcdischarges in oxygen,’’ Appl. Opt. 12, 2800–2802 119732.

12. W. R. McKinney and P. Z. Takacs, ‘‘Plasma discharge cleaningof replica gratings contaminated by synchrotron radiation,’’Nucl. Instrum. Methods 195, 371–374 119822.

13. T. Koide, S. Sato, T. Shidara, M. Niwano, M. Yanagihara, A.Yamada, A. Fujimori, A. Mikuni, H. Kato, and T. Miyahara,‘‘Investigation of carbon contamination of synchrotron radia-tion mirrors,’’ Nucl. Instrum. Methods 246, 215–218 119862.

14. K. Yamada, T. Yamazaki, S. Sugiyama, H. Ohgaki, T. Noguchi,T. Mikado, M. Chiwaki, R. Suzuki, and T. Tomimasu, ‘‘FELoscillation on TERAS,’’ Nucl. Instrum. Methods 331, 103–106119932.

15. See, for example, C. D. Wagner, W. M. Riggs, L. E. Davis, J. F.Moulder, and G. E. Muilenberg, eds., Handbook of X-RayPhotoelectron Spectroscopy 1Parkin-Elmer Corp., Eden Prairie,Minn., 19792, p. 52.

16. W. Zhu, H.-S. Kong, and J. T. Glass, ‘‘Characterization ofdiamond films,’’ in Diamond Films and Coatings, R. F. Davis,ed., 1Noyes, Park Ridge, N.J., 19932, Chap. 6, p. 260.

20 July 1995 @ Vol. 34, No. 21 @ APPLIED OPTICS 4265