and Methods in Physics Research B 113 (1996)
Beam Interactions with Materials & Atoms ELSEVIER
ERDA of solar selective absorbers W.Assmann
a**, Th. Reichelt a, T. Eisenhammerb, H. Huber a, A. Mahr b, H. Schellingerb, R. Wohlgemuth b a Sektion
Physik, Universitdf Miinchen. D-85748 Garching, Germany b Sektion Physik, Universitdt Miinchen. D-80539 Miinchen, Germany
Abstract ERDA (elastic recoil detection analysis) with 200 MeV I beams was used during the development of new solar selective absorber materials (TiN,O, and AlCu,Fe,) . Depth profiles of the light and heavy components were measured simultaneously using a detector system with particle identification. Two oxidation processes were found in TiN,O, absorbers at elevated temperatures which result in absorber degradation. Coating of the absorber with SiO, improved both the efficiency and the life time. For the production of the quasi-crystalline phase of AlCu,Fe, an over-stoichiometric sputter target was necessary due to additional Al& fOrtK?tiOIL
1. Introduction Conversion of solar radiation to thermal energy at temperatures below 1OO’Cis a highly developed technique. However, application of solar thermal energy in industry or thermal solar power plants requires much higher temperature levels of up to 400°C and more [ 11. At these temperatures radiation losses of the absorber become important due to the T4 dependence. Selective absorbers have been developed for this purpose, which absorb a large fraction of the solar radiation at wavelengths shorter than about 2 ,um and have very low emittance for thermal radiation [ 21. For example, coatings of TiN,O, on Cu have shown emittance of 0.04 at 3OO’C allowing 50% efficiency at 28O’C [ 31. These coatings are stable in high vacuum, but degrade quickly above 230°C in the presence of oxygen. Thermal stability is a general problem of absorber coatings. Therefore one point in the present study was to understand the mechanisms of degradation and to improve the long-term stability of the otherwise promising TiN,O, solar selective absorber. Recently, quasi-crystalline thin films of AlCu,Fe, were shown to have large stability and peculiar optical properties, which make AlCu,Fe, a possible candidate for use as solar selective absorber . The second part of this paper deals with the development of this rather new absorber material. Both studies required an analysis technique which could deliver the concentration depth profiles of light and heavy components as well, Because of the large number of samples
* Corresponding author. Bewhleunigerlabor der Universitit Miinchen, D85748 Garching, Germany; tel.: f49 89 3209 4283; fax: +49 89 3209 4280; e-mail: [email protected]
0168-583X/96/$15.00 Copyright SSDIO168-583X(95)01303-2
to be studied, the method had to be fast and easy to normalize. Ion beam based techniques have been used for similar problems in the past, mainly RBS and NRA [ 5-71. NRA has the disadvantage of needing a different reaction for every sample component, and RBS cannot separate the light components from the background of the heavier substrate. Heavy ion ERDA has shown to be well suited to measure depth profiles of light and heavy elements simultaneously, and if a particle identification detector system is used, for each element separately [ 81. In this report we will briefly describe our detector system and the ERDA method itself. Then we will present examples of ERDA on solar absorbers and demonstrate the improvements in our understanding of these coatings.
2. Heavy ion ERDA ERDA is routinely used for thin film analysis at the Munich 15 MV Tandem accelerator with different setups which have been described in detail in [ 9,101. Briefly, high energy heavy ions (typically 200 MeV 12’1ions) are directed onto the sample surface and recoil atoms from the sample are detected. The use of heavy ions for ERDA has the advantage of needing very low beam doses. When different ions are compared at constant velocity, the recoil cross section increases with Z&j. Thus typically lOi* iodine ions are sufficient for the analysis. However, heavy ion induced irradiation effects have to be considered. In principle all effects proportional to the electronic or nuclear energy loss should be independent of the projectile mass [ 10,111. But it is known that energy loss thresholds exist for different materials, where nonlinear
@ 1996 Elsevier Science B.V. All rights reserved
W! Assmann et al./Nucl.
Instr. and Meth. in Php.
Res. B 113 (1996) 303-307
I. A&,!&, coincidence spectrum of liNxOy solar absorber with SO2 cover and Al substrate. Absorber components and C contamination are indicated.
radiation induced effects have been measured [ 121. Thus in ion beam analysis any new material has to be checked for radiation induced compositional changes. Thick samples have to be measured with ERDA in reflection geometry. We use typically 19Oincidence and 18.5” exit angle to get a total scattering angle of 37.5”. Up to twelve samples can be mounted simultaneously on a 2-axes goniometer through a load lock system, thus the vacuum in the scattering chamber of about 5 x lo-’ mbar does not need to be broken. Recoils are detected by a AE-Exs, ionization chamber with a solid angle of 6.2 msr. The grid supported polypropylene entrance foil has a thickness of about 0.5 pm. Typical detector gas pressures ate 30 to 70 mbar isobutane. Because of the high projectile energies recoils at least up to mass 65, i.e. Cu, can be identified by their specific energy loss. The large acceptanceangle of 85 mrad causes a kinematic energy shift of the recoil energies of about 13% which severely limits the possible depth resolution of the detector. To restrict the kinematic shift to values comparable to the detector energy resolution of about l%, a position sensitive cathode is used in the ionization chamber. With this device the scattering angle of every detected recoil can be measured with better than 3.5 mrad resolution and the kinematic energy shift can be calculated. With this correction a depth resolution below 10 nm has been achieved. The necessary data evaluation is an integral part of the
data recording program which generates a kinematically corrected energy spectrum for every sample component online. This allows us to accumulate spectra normalized to a substrate component and to test for dose dependence for certain elements during the irradiation. Raw data are also written on tape for further data processing. To convert the measured energy spectra we use the program KONZERD [ 131 which gives us the final concentration depth profile of every measured component.
3. Analysis of TiN,O, absorbers 3.1. Sample preparation and characterization The coatings were prepared by activated reactive evaporation (ARE) of titanium [ 141. Oxygen and nitrogen were activated by means of a plasma discharge. Process parameters are substrate temperature, plasma current density, deposition rate and partial pressures of oxygen and nitrogen. During deposition, thickness and deposition rate were monitored by a quartz crystal monitor. Film thickness and density were measured afterwards by grazing incidence X-ray reflection (GXR). Different substrates were used during these studies: copper and aluminum sheets as possible substrates for applications, and also glass to see or avoid surface roughness effects. Chemical composition of deposited films was mea-
Instr. and Meth. in Phys. Rex B II3 (1996) 303-307
sured by Auger electron spectroscopy (AES) and energy dispersive X-ray analysis (EDX). The crystalline phases were determined by X-ray diffraction (XRD). The resulting picture of the as prepared coatings can be summarized as a system grown in columnar structure with different coexisting chemical and crystalline phases. Depending on the deposition parameters, amorphous and crystalline phases of TiN,, TiO, and TiO2 have been identified [ 151. The absorber properties have been determined by measuring the spectral reflectance and the directional thermal emittance. To study absorber degradation thermal treatments at different temperatures, gas pressures and Ox/Nz-ratios have been performed. The degradation can be seen as a characteristic change in the emittance from metallic to dielectric behavior which could be measured in situ. After similar temperature treatments samples have also been characterized by the techniques mentioned above. 3.2. ERDA results Because of the phase mixture in the TiN,O, films, no reasonable stoichiometric information could be determined by ERDA. Therefore concerning the composition of the absorber ERDA was restricted to measure the depth dependent total amount of Ti, 0 and N, and the contaminants as C or heavy impurities. The essential information delivered by ERDA was a characteristic change in the depth profiles of certain components after different temperature treatments. A typical 2-dimensional spectrum of a 95 nm thick TiN,O, film on an Al substrate, covered by a 65 nm thick SiO2 coating, is shown in Fig. 1. Different components of the multilayer system are well separated in the AE-E,,-matrix and allow the extraction of depth profiles without any background. This and all the following examples were measured at 37.5” scattering angle using a 200 MeV ‘“I beam. Typical beam dose was less than 10’” ions/cm2, where no radiation induced change of the depth profiles was observed. The first question we tried to answer with our analysis, was the physical reason of absorber degradation on Cu substrates. TiN,O, films of 65 nm thickness were deposited on a 150 nm thick Cu layer, which was evaporated onto a glass substrate. Depth profiles of Cu are shown in Fig. 2 for samples heated in air at a temperature of 2OO’C for 0.5, 1.0 and 2.0 hours. The spectra were normalized to the simultaneously measured Si depth profile of the glass substrate. A strong Cu diffusion through the TiN,O, layer to the surface can be easily seen. Even in the as prepared sample Cu has diffused into the TiN,O, film probably during the deposition, where the substrate is at a temperature of 17O’C. A strong CuO component has been found in XRD at the last stage of degradation which can be identified by ERDA as oxidation of the out-diffused Cu at the absorber surface. This changes the measured emittance to more dielectric behavior and explains the degradation of the selective solar absorber [ 15 1. The columnar structure of TiN,O, possibly enhances the Cu diffusion due to the adsorbed surface oxygen which
0.35 0.30 2
OS prepared 0.5h 200°C
Channel Fig. 2. Depthprofilesof Cu after different annealing times. Cu was covered with 6.5 nm TiNxOy.
is known to increase the Cu diffusion. Identical samples heated at 270°C for 44 h in vacuum ( lOA mbar) showed no significant Cu diffusion and only out-gassing of adsorbed 0 and N. Thus to avoid this kind of absorber degradation, Cu diffusion with successive surface formation of CuO has to be prevented. Two possibilities were tried in this study: change to a different substrate and/or sealing of the TiN,O, structure by a top layer. Recently, it was found that SiO2 layers on TiN,O, absorbers act as antireflection coatings that improve the optical properties [ 161. Therefore, to test the sealing properties of SiO2 coatings, 65 nm thick TiN,Oy films were deposited on Cu sheets and covered with 100 nm SiO,. These coatings which were deposited by e-gun evaporation of SiO:! had a stoichiometry of typically SiO2.2. Cu depth profiles were measured on samples both as prepared and after temperature
85 so I I , I I I 1 , I
(MeV) 95 100 105 , , 1 I . , r , , , , , ,
Cu profile 0.4 9 'F -u
-OS prepared 20h 250°C
Channel Fig. 3. Depth profiles of Cu after long-time annealing covered with 65 nm liNxOy with 100 nm SiO2 sealing.
on air. Cu was
M! Assmann et nl./Nucl.
Instr. and Meth. in Phys. Res. B 113 (1996) 303-307
treatments. Taking into account the higher temperature and longer annealing time compared to the samples of Fig. 2, the benefit of SiOx sealing can be clearly seen in Fig. 3. There is still some Cu diffusion into the absorber, but no Cu buildup at the surface. The additional roughness of the substrate influences the depth resolution, but apparently not the Cu diffusion. A substantial slow down of degradation in air was also seen in reflectance measurements [ 161. However, evaporation deposited SiO, has only limited sealing properties, after 400 hours at 2OO’C in air the oxygen content of TiN,ZO, increased by 63%. With a SiO, thickness of 130 nm instead of 100 nm this change was only 4%. Another suitable substrate for solar absorbers is Al. This also allows testing for more absorber specific oxidation effects which cannot be measured in the presence of copper. Samples have been prepared by ARE of TiNIO, on Al substrates. Some of the samples were covered afterwards with about 100 nm thick layers SiOx (Fig. 1). The time dependence of oxidation during tempering on air at 200°C has been measured for identical samples without SiO, cover. The oxidation shows a passivation-like dependence (Fig. 4) which is connected to the formation of a TiOz phase as seen in XRD. This oxidation also changes the emittance characteristic of the absorber to insulator behavior [ 151. Tempering at 27O’C for 52 hours in vacuum of 10m5 mbar decreased the oxygen as well as the nitrogen content of the TiN,O, film by about 10%. This treatment seems, according to infrared reflection measurements, to improve the absorber properties [ 181. Comparable temperature treatments with SiO, covered samples resulted in similar, but delayed behavior. Thus without copper substrate a second degradation process was seen for TiN,O, absorbers.
4. Analysis of AlCu,Fe,
Because of the degradation problems, TiNJOy absorber seems to be limited to operating temperatures below 3OO’C. A promising candidate for absorber materials, which could withstand higher temperatures without degradation, is quasi-crystalline AlCu,Fe,. Resulting from a numerical optimization of the optical properties a sandwich structure of dielectric/quasi-crystalline/dielectric thin films on copper substrate should have highly selective properties. Solar absorptance of about 0.9 and hemispherical emittance at 4OO’C well below 0.1 were calculated [ 41. However, the calculations demand thicknesses of the quasi-crystalline film of only about 12 nm. Such thicknesses of thermodynamically stable quasi-crystalline films have not yet been produced. We have tried to make AlCuFe films with a sputter technique which we had used to deposit Si3NJ/Ag/Si3Nd multilayers of comparable thickness [ 191. A rotating sputter target of appropriate composition should ensure the desired film stoichiometry. Films deposited on Si wafers at substrate temperatures around 270°C were found by XRD to be crystalline. Deposition at higher substrate temperatures of 46049O’C formed the quasi-crystalline phase. Film density and thickness was measured with GXR. During development of the deposition technique ERDA was used to check the film stoichiometty and test for contaminations. ERDA depth profiles of first deposited films showed Si diffusion into the AICu,FeY layer after annealing at 5 IO”. To avoid further Si diffusion at elevated temperatures, alumina buffer layers of 40-50 nm thickness were sputtered onto the substrate. To separate the Al of this Al203 layer from the AlCu,Fe, aluminum we used the oxygen depth profile. The depth resolution of our setup was adequate to distinguish Al203 oxygen from a small 0 contamination (about 3%) in the 180 nm AlCu,Fe, film (see Fig. 2 in Ref. [ 111). The oxygen content of the A1203 layer allowed us to de-
-‘* I 0
@ I 50
Fig. 4. Time dependence of 0 and N to Ti ratio in TiNxOy during annealing at 2WC on air.
annealing time at 200°C [h] Fig. 5. Concentration KONZERD [ 131.
depth profiles of AICuxFey absorber calculated with
W. Assmann et al./Nucl.
Instr. and Meth. in Phys. Res. B 113 (1996)
termine the corresponding amount of Al and hence the correct stoichiometry of the Ah&Fe, film. With smaller incidence angles of about 3” in ERDA and the corresponding improved depth resolution this procedure could be used for the required 15 nm thick AlCu,Fe, layers, too. In the concentration depth profiles of the main components (Fig. 5) a clear surface peak for oxygen can be seen. The additional increase of the Al concentration is probably due to A1203 formation at the surface. After annealing of AlCu,Fe, films on air at 400’C for 70 hours an Al203 passivation layer of about 10 nm thickness was formed. Oxidation of AlCu,Fe, and formation of Al203 seems also to occur during deposition within the film. The oxygencontamination of about 3% is an indication of this process. A stoichiometry for the quasi-crystalline phase of Al62.3Cu24.sFeI2.shad been reported in [ 201, while for our quasi-crystalline samples a stoichiometry of Al6s.oCu22.sFeI2.5was measured. The slightly higher Al content is again a hint at this oxidation process. Thus to create the quasi-crystalline phase a sputter target had to be used with excess Al, which was determined from the ERDA spectrum of oxygen.
5. Conclusion Structural and compositional analysis plays an essential role in developing new materials. Heavy ion ERDA has proven to deliver substantial information on the composition and depth distribution for development of solar absorbers. Two degradation mechanisms could be identified in temperature treatments of TiN,Oy absorbers: Cu diffusion with CuO buildup and oxidation of Ti compounds. These effects could be slowed down by a SiO, coating. ERDA depth profiles helped also to understand the oxidation of AlCu,Fe, during film deposition. After corresponding corrections of the deposition conditions the desired quasi-crystalline phase could be produced. Thus ERDA seems to be a very effective analysis method for optical or mechanical coatings development.
C.-J. Winter, R.L. Sizmann and L.L.Vant-Hull (Eds.), Solar Power Plants (Springer, Berlin,1991) [Z] GA. Niklasson and C.G. Granqvist, I. Mater. Sci. 18 (1983) 3475. [ 31 C. Seiffert, T. Eisenhammer, M. Lazarov, R. Sizmann and R. Blessing, Proc. of the ISES Solar World Congress, Budapest, Hungary, August 1993, Vol. 2, p. 321.  T. Eisenhammer, A. Mahr, A. Haugeneder. T. Reichelt and W. Assmann, Proc. 5th Int. Conf. on Quasicrystals, Avignon, France. 22-26 May 1995. [.5] K.P. Lieb. W. Bolse, Th. Carts, W. Miiller, Tb. Osipowicz and Th. Weher, Nucl. Instr. and Meth. B 50 (1990) 10.  K. Tanaka, Y Kubo, N.B. Chilton and M. Kumagai, Nucl. Instr. and Meth. B 83 (1993) 525.  W. Wagner, F. Rauch and K. Bange. Nucl. Insw. and Meth. B 89 (1994) 104. [ 81W. Assmann, H. Huber, Ch. Steinhausen, M. Dobler. H. Gliickler and A. Weidinger, Nucl. Instr. and Meth. B 89 ( 1994) 131.  W. Assmann. P. Hartung, H. Huber, P. Staat, H. Steffens and Ch. Steinhausen, Nucl. Instr and Meth. B 85 (1994) 726. [IO] G. Dollinger, T. Faestermann and P. M&r-Komor, Nucl. Instr. and Meth. B 64 ( 1992) 422. [II] W. Assmann, J.A. Davies, G. Dollinger, J.S. Forster, H. Huber, Th. Reichelt and R. Siegele, Proc. 12th Int. Conf. on Ion Beam Analysis, Tempe, AZ, USA, 22-26 May 1995, to be published in Nucl. Instr. and Meth. B. [ 121 Z.G. Wang, Ch. Dufour, E. Paumier and M. Toulemonde, J. Phys.: Cond. Matter 6 (1994) 6733. [ 131 A. Bergmaier, G. Dollinger and C.M. Frey, Nucl. In&r. and Meth. B 99 (1995) 488. [ 141 M. Lazarov, PhD Thesis, Univ. Munich (VDI Verlag, Dtisseldorf, 1993). [ 151 H. Schellinger, M. Lararov, H. Klank and R. Sizmann, in: Optical Materials Technology for Energy Efficiency and Solar Energy Conversion XII. Proc. SPIE 2017 (1993) p, 345. 1161 M. Lazarov, R. Sizmann and U. Frei. ibid. p. 366. [ 171 T. Eisenhammer, F. Muggenthaler and R. Sizmann, ibid. p. 46. [ 181 H. Schellinger, M. Lazarov, W. Assmann, B. Bauer and Ch. Steinhausen, in Optical Materials Technology for Energy Efficiency and Solar Energy Conversion XIII,Proc.SPIE2255( 1994) p. 172. [ 191 T. Eisenhammer, Proc. Int. Conf. on Metallurgical Coatings and Thin Films, San Diego, CA, USA, 24-28 April 1995, Thin Solid Films, in press. [ 201 D. Gratias. Y. Calvayrac, J. Devaud-Rzepski. F. Faudot, M. Harmelin, A. Quivy and P.A. Bancel, J. Non-Cryst. Solids 153/4 (1993) 482.