- Email: [email protected]
Characterization of ZnO structures by optical and X-ray methods
Applied Surface Science 281 (2013) 123–128
Contents lists available at SciVerse ScienceDirect
Applied Surface Science journal homepage: www.elsevier.com/locate/apsusc
Characterization of ZnO structures by optical and X-ray methods P. Petrik a,b,∗ , B. Pollakowski c , S. Zakel d , T. Gumprecht a,e , B. Beckhoff c , M. Lemberger a , Z. Labadi b , Z. Baji b , M. Jank a , A. Nutsch c a
Fraunhofer Institute for Integrated Systems and Device Technology, Schottkystrasse 10, 91058 Erlangen, Germany Institute for Technical Physics and Materials Science (MFA), Research Center for Natural Sciences, Konkoly Thege Rd. 29-33, 1121 Budapest, Hungary Physikalisch-Technische Bundesanstalt (PTB), Abbestr. 2-12, 10587 Berlin, Germany d Physikalisch-Technische Bundesanstalt (PTB), Bundesallee 100, D-38116 Braunschweig, Germany e Erlangen Graduate School in Advanced Optical Technologies (SAOT), Paul-Gordan-Strasse 9, 91052 Erlangen, Germany b c
a r t i c l e
i n f o
Article history: Received 22 October 2012 Received in revised form 5 December 2012 Accepted 6 December 2012 Available online 18 January 2013 Keywords: Zinc oxide Spectroscopic ellipsometry X-ray fluorescence Raman spectrometry VUV reflectometry Atomic layer deposition Sputtering
a b s t r a c t ZnO thin films doped by Ga and In as well as multilayer structures of ZnO/Al2 O3 have been investigated by X-ray fluorescence, Raman spectrometry, spectroscopic ellipsometry and vacuum ultra violet reflectometry. Systematic changes in the optical properties have been revealed even for Ga concentrations below 1%. The Raman active phonon mode of Ga doping at 580 cm−1 shows a correlation with the Ga concentration. Optical models with surface nanoroughness correction and different parameterizations of the dielectric function have been investigated. There was a good agreement between the dielectric functions determined by the Herzinger–Johs polynomial parameterization and by direct inversion. It has been shown that the correction of the nanoroughness significantly influences the accuracy of the determination of the layer properties. The band gap and peak amplitude of the imaginary part of the dielectric function corresponding to the excitonic transition changes systematically with the Ga-content and with annealing even for low concentrations. © 2013 Elsevier B.V. All rights reserved.
1. Introduction The manufacturing of optoelectronic thin films is of key importance, because a significant number of industries rely on such structures. ZnO is one of the most intensively investigated optoelectronic materials for applications as a transparent conductive oxide or complex semiconductor with different doping, defect budget and varying preparation parameters [1]. Thin ZnO layers can be prepared using a range of methods including sputtering [2], atomic layer deposition (ALD) [3], pulsed laser deposition [4–6], spin coating from nanoparticulates [7], or spray pyrolysis [8]. Optical techniques such as ellipsometry can be used for the sensitive, quick and non-destructive determination of a range of material and structural parameters such as thickness, surface nanoroughness, interface quality, density, homogeneity, band gap and exciton strength (which makes possible the indirect determination of electrical properties) [9–11].
∗ Corresponding author at: Institute for Technical Physics and Materials Science (MFA), Research Center for Natural Sciences, Konkoly Thege Rd. 29-33, 1121 Budapest, Hungary. Tel.: +36 13922502; fax: +36 13922226. E-mail address: [email protected] (P. Petrik). 0169-4332/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.apsusc.2012.12.035
The aim of the studies was the improvement (development of sophisticated models and evaluations) and comparison of optical and X-ray metrologies for the characterization of ZnO structures in single layer as well as in multi-layer stacks prepared by sputtering and ALD. This study is intended to be a step toward the establishment of validated reference methodologies for a reliable characterization of key optoelectronic materials, as a part of the joint research project for optoelectronic thin film characterization (IND07, “Metrology for the manufacturing of thin films”) in the European Metrology Research Program of EURAMET. Furthermore, our investigations aim for the development of reference samples with controlled defect concentration and morphology as well as methods for elemental depth profiling. 2. Experimental details Dual-target sputtering was used to prepare InGaZnO samples on oxidized (10 nm SiO2 ) single-crystalline silicon wafers in thicknesses of nominally 50 nm (for X-ray, ellipsometric and reflectometric measurements) and 500 nm (for Raman spectrometry). The parameters of preparation and numbering of samples used in this study are summarized in Table 1. ZnO/Al2 O3 multi-layer structures of n× {ZnO (10 nm)/Al2 O3 (3 nm)}/ZnO (10 nm) have been prepared for n = 1, . . ., 3 using ALD
124
P. Petrik et al. / Applied Surface Science 281 (2013) 123–128
Table 1 Parameters of preparation and numbering of samples investigated in this study. Notations of ‘X’ and ‘R’ were used for X-ray/optical (with nominal thicknesses of 50 nm) and Raman (with nominal thicknesses of 500 nm) measurements, respectively. Ga (%)
Not annealed
Annealed
0.00 0.17 0.34
X1, R1 X5, R5 X7, R7
X9, R9 X13, R13 X15, R15
[12]. The ALD process took place in a flow type Picosun SUNALE TM R-100 type ALD reactor. Nitrogen was used as purging gas and diethyl-zinc, trimethyl-aluminum and H2 O as precursors. The precursors were of electronic grade purity manufactured by SAFC Hitech. The deposition temperature was 300 ◦ C, the flow rates were 150 sccm. The pulse time for all precursor injections was 0.1 s, while the purging time after each metalorganic precursor pulse was 3 s, and 4 s after the water pulses. The substrates were cleaned in cc.HNO3 and high resistivity water. Ellipsometric measurements were performed using a SOPRA SE5 multichannel rotating polarizer and a Woollam M-2000DI multichannel rotating compensator spectroscopic ellipsometer in the wavelength range of 193–1690 nm. The spot size was 2–3 mm in both cases. The measurement time of one spot was in the several seconds range. The X-ray fluorescence (XRF) measurements on the Ga- and In-doped ZnO samples have been carried out at a four-crystal monochromator (FCM) beamline [13] in the PTB laboratory at the synchrotron radiation facility BESSY II. This beamline provides monochromatic radiation from a bending magnet in the energy range between 1.75 eV and 11 keV. The originating measurements shown here were performed at 11 keV at an angle of incidence of 1◦ . A toroidal mirror in the front and a cylindrical behind of the FCM ensures a beam diameter in the focal plane of about 280 m. As a possible end station at the FCM beamline, an ultra high vacuum (UHV) chamber optimized for X-ray spectrometry measurements in different geometries was placed in the focus. So, each of the samples can be adjusted in the center of the chamber, were the focus is located. For detecting the fluorescence radiation an energy dispersive Silicon Drift Detector (SDD) is placed perpendicular to the incoming beam. The detector response function for the respective energies and the efficiency are well-known [14,15]. The solid angle of detection was determined exactly and the incoming photon flux was measured by means of calibrated photodiodes. The knowledge of the experimental details is necessary for a referencefree quantification of the mass deposition of thin layered structures [16]. The Raman spectrometric measurements were performed using a LabRAM Aramis Raman microscope and LabSpec 5.0 Software (Horiba Jobin-Yvon). The instrument was equipped with a 1200 grooves/mm holographic grating and an excitation laser using a wavelength of 532 nm. The laser power was set to 2 mW. The microscope was used in non-confocal mode with an 50× objective (NA 0.75). Calibration of the spectral line position was carried out prior to each acquisition using the crystalline silicon mode at 520 cm−1 . The integration time was 400 s. As the thin films were sputtered on a c-Si wafer with 30 nm of thermal oxide on top, the spectra of c-Si and c-Si coated with thermal oxide were also acquired and used for baseline subtraction. Vacuum ultra-violet (VUV) reflectometry has been done using a Metrosol VUV 7000 equipment in the wavelength range of 120–800 nm with a spot size of 35 m. The spot size can be important when checking lateral inhomogeneity, but it was not utilized in this study.
Fig. 1. X-ray fluorescence spectrum of the sample X7. The excitation energy was set to 11 keV. The black line is the measured spectrum and the gray line is the fitted spectrum of the convolution of the fluorescence lines energies with the detector response function.
3. Results and discussion The samples X1, X7, X13 and X15 were analyzed regarding their chemical composition by means of reference-free X-ray fluorescence analysis [16]. This approach allows for a determination of the mass deposition of the elements and if the density is known the thickness of the layer can be concluded. The photon energy was set to 11 keV in order to excite the main element of the layer (Zn) and the implanted Ga and In atoms. The analysis of XRF data is based on the deconvolution of the respective spectra by means of detector response functions. The outcome of this is the net count rate for the respective fluorescence lines. Considering the detector efficiency the count rate can be converted into an intensity. For the calculation of the mass deposition m/F (m denoting the mass and F the unit area) only primary excitation are taken into account because no secondary effects are to be expected. The description of the used relation between fluorescence intensity and the number of atoms and mass is described in detail elsewhere [16,17]. This approach requires among the experimental parameters atomic fundamental parameters as photo-electron absorption cross sections, mass attenuation coefficients and fluorescence yield. In this work, the data for the cross sections were taken from data bases of Elam et al. [18] and Ebel et al. [19]. Fig. 1 shows a measured XRF spectrum (black line) and the respective fitted spectrum with detector response function of sample X7 (gray line). This XRF spectrum exhibits the main matrix element fluorescence lines of the single layer Zn K␣, , the substrate material silicon Si K␣, the dopants Ga K␣,  and In L␣,  as well as contamination as nickel Ni K␣,  and iron Fe K␣, . For the further evaluation the fluorescence lines of Zn, Ga and In have been considered. In Fig. 2 a selection of the recorded XRF spectra are plotted. Focusing on the dopant’s fluorescence lines Ga K␣,  and In L␣,  the intensity is considerably increased when increasing the dopant concentration of these elements. All other fluorescence intensities remain almost constant. Considering the intensities of respective fluorescence lines the mass deposition (m/F) of Zn, In and Ga can be determined and the results are shown in Table 2. Considering the presented results in Table 2 the stoichiometry of these samples can also be determined by XRF. The results are listed in Table 3.
P. Petrik et al. / Applied Surface Science 281 (2013) 123–128
125
Fig. 2. X-ray fluorescence spectra of the samples X1, X7 and X13. The excitation energy was set to 11 keV. The counts in the fluorescence lines Ga K␣,  and In L␣,  increase with increasing dopant rate.
The results from Raman spectrometry are shown in Figs. 3 and 4. The crystalline ZnO layer causes two Raman-active E2 phonon modes at 101 cm−1 and 437 cm−1 . An additional mode arises from the doping with Ga at 580 cm−1 . Although in general the E2 -ZnO modes weaken with increasing Ga content, the intensity ratio of the ZnO:Ga band over the ZnO band at 437 cm−1 clearly increases with increasing Ga content (Fig. 4). An investigation of Hasuike et al. [20] showed that for Ga contents of 0.5% and higher this correlation does not hold. Furthermore, the peak slightly broadens from 0.5 to 2% and then finally severely broadens at 5–8%. Annealing of the samples at 400 ◦ C causes smaller bandwidth and lower intensity of the dopant mode (Fig. 3). An additional defect mode with maximum intensity at about 595 cm−1 seems to be present while the sample is not annealed and vanishes when the crystallinity is enhanced. Typical spectra measured by ellipsometry on a nominally 500 nm thick ZnO layer (sample R1) is plotted in Fig. 5 together with the curves of the annealed sample (R9) and the VUV reflectometry measurements on samples X1 and X9. The transparent and opaque photon energy ranges can clearly be distinguished by the interference oscillations characteristic to the transparent range. The layer thickness and surface nanoroughness of the samples
Fig. 3. Raman spectra for different Ga contents determined by XRF.
can be determined by ellipsometry in the transparent spectral range using a simple Cauchy dispersion (n = A + B/2 + C/4 , where n denotes the refractive index, and A, B and C are the Cauchy parameters) using an optical model of c-Si/SiO2 /ZnO, whereas the refractive index of ZnO can be described by the Cauchy model. The layer thickness, the refractive index (Fig. 6) and the surface nanoroughness have been determined in the photon energy range of 1.5–2.5 eV, because no absorption (extinction coefficient k = 0) has to be taken into account in this range, which simplifies the model and reduces parameter correlation.
Table 2 The mass deposition m/F of Zn, Ga and In was determined by reference-free X-ray fluorescence analysis. A relative uncertainty of 15% (k = 1) was estimated. Sample X1 X7 X13 X15
Zn (g cm−2 ) 18 ± 3 16 ± 2 18 ± 3 16 ± 2
Ga (g cm−2 )
In (g cm−2 )
0 0.20 ± 0.03 0.11 ± 0.02 0.20 ± 0.03
0 1.00 ± 0.15 0.35 ± 0.05 1.02 ± 0.15
Table 3 The elemental composition was determined by reference-free XRF and a relative uncertainty of 15% was estimated. The oxygen part was concluded by XRF measurements at lower excitation energies, for example 1.622 keV. The layer thickness was determined assuming a density of 10% lower than for bulk ZnO. Sample X1 X7 X13 X15
Zn (%) 50 ± 8 47 ± 7 50 ± 7 47 ± 7
O (%) 50 ± 8 49 ± 7 49 ± 7 49 ± 7
Ga (%)
In (%)
0 0.34 ± 0.05 0.17 ± 0.03 0.34 ± 0.05
0 2.8 ± 0.4 0.9 ± 0.2 2.8 ± 0.4
Thickness 43 ± 6 39 ± 6 44 ± 7 39 ± 6
Fig. 4. Intensity ratio of the ZnO:Ga band over the ZnO band measured by Raman spectrometry.
126
P. Petrik et al. / Applied Surface Science 281 (2013) 123–128
nanoroughness. The mean squared error of the fit is almost twice as much in case of the fit without roughness as with roughness, although only one additional fit parameter was introduced. Fig. 6 shows that the refractive index of sputtered layers investigated in this study are slightly lower than that of bulk ZnO from Refs. [21,22], however, the refractive index calculated with surface roughness correction is larger by ≈0.05. This shift can partly be explained by the stress in as-deposited films [23]. However, this effect is estimated to be not larger than about 0.013, comparing the refractive index of X1 in Table 4 with its annealed counterpart of sample X9. This reveals good film quality and points out the significance of surface roughness correction. The parameters determined by ellipsometry in the transparent range are summarized in Table 4. Small, but systematic decrease can be observed for both n (equivalent with and increase of the volume fraction of voids, fv ) and the surface nanoroughness dr as a result of annealing. For photon energies near the band gap the dielectric function of ZnO can be described by simple oscillator models. One of the most popular methods is the application of different model dielectric functions using several oscillators with analytical lineshapes [21,24,25]. A further possibility is to use a Tauc–Lorentz oscillator [26] combined with a Drude tail for the near infra red region [10,11]. The third approach is the Herzinger–Johs method to use polynomial functions to fit the line-shape of the dielectric functions in the proximity of different critical point energies [27,28]. Finally, the most general, model-independent approach is the direct inversion [29,28], in which the dielectric function is calculated wavelength-by-wavelength using the layer thickness determined independently or in a wavelength range of no absorption. A cross-check for the proper layer thickness is the criterion that the absorption tail has to decay smoothly to zero toward the longer wavelengths [29]. A great advantage of this method is that no assumption has to be made for the dispersion of the dielectric function. A disadvantage is that taking into account the surface nanoroughness or any probable surface layer is more problematic. In this study the Psemi-M0 parameterization of the Herzinger–Johs method was used, which is an adaptation of the polynomial parameterization to M0 critical points. Using this approach the surface nanoroughness could be taken into account as an effective medium overlayer with 50% layer material and 50% void, fitting only the thickness of this layer. Note that similar to the above case of the transparent spectral region the fit quality improves to a large extent when using a surface roughness correction (e.g. the MSE value for sample X15 with and without a roughness layer is 2.8 and 12.6, respectively). Fig. 7 shows the imaginary part of the dielectric function (2 – which is proportional to the joint density of electronic states) in a photon energy region close to the band gap. The height and broadening of the excitonic peak increases and decreases, respectively, as a result of annealing for each doping level. The slope of the increase of 2 at the band gap decreases with increasing Ga content. The inset
Fig. 5. Measured reflectometric (X1 for as deposited and X9 for annealed) and ellipsometric and (R1 for as deposited and R9 for annealed) spectra. The rectangles show the most relevant gap regions.
Fig. 6. Refractive indices of sample X1 calculated in the transparent spectral range with and without a surface nanoroughness layer. The refractive index values from Refs. [21,22] are also plotted.
Taking into account the surface nanoroughness is a crucial step in terms of both fit quality and determination of accurate bulk layer refractive index (Fig. 6). The surface nanoroughness is modeled as a homogeneous surface layer with 50% layer material and 50% void fitting only the thickness of the layer. The difference of the measured and calculated spectra shown in Fig. 6 is significantly larger for the model that does not take into account the surface
Table 4 Optical properties measured in the transparent spectral range. A, B and C are the Cauchy parameters, n denotes the refractive index at the photon energy of 2 eV (620 nm). fv is the volume fraction of voids using an effective medium composition of material X1 and voids (i.e. the change of density relative to sample X1 – negative volume fraction indicates an optical density larger than the reference). db and dr denote the thicknesses of the bulk layer and the surface nanoroughness, respectively, measured by ellipsometry (the reflectometry results are given in parentheses). The sensitivity of the Cauchy parameters are 0.001 for parameters A and B and 0.0001 for parameter C (‘Ann.’ denotes ‘annealed’). Nr.
Ga (%)
Ann.
A
B (m2 )
C (10−4 m4 )
n
fv (%)
db (nm)
dr (nm)
X1 X5 X7 X9 X13 X15
0.00 0.17 0.34 0.00 0.17 0.34
No No No Yes Yes Yes
1.934 1.933 1.932 1.922 1.916 1.926
0.038 0.035 0.038 0.038 0.040 0.041
−5.9 0.0 1.6 −3.4 −4.8 −0.2
2.030 2.026 2.033 2.017 2.018 2.032
0.0 0.2 −0.2 1.2 1.1 −0.1
50.7 (45.6) 51.6 46.6 (46.6) 50.7 51.3 (52.5) 46.3 (48.5)
9.2 8.4 8.9 8.0 8.0 8.1
P. Petrik et al. / Applied Surface Science 281 (2013) 123–128
Fig. 7. Imaginary part of the dielectric function (2 ) of samples X1 (0% Ga) and X7 (0.37% Ga) with and without annealing measured using the Psemi oscillator model by ellipsometry. The inset shows a comparison of 2 values calculated using the Psemi model with (‘rough.’) and without (‘no rough.’) a surface nanoroughness layer, as well as direct inversion (‘dir. inv.’). As a reference, 2 of an ALD-deposited 10-nm ZnO film from the second series is also plotted in the inset (denoted by ‘ALD’).
shows that taking into account the surface nanoroughness slightly decreases the amplitude of 2 . A justification of the Psemi model is the very good agreement with the result from direct inversion (dotted and dash-dotted lines in the inset of Fig. 7), because the method of direct inversion is model-independent, but in this case only the results without surface roughness correction can be compared. It is also important to point out that a good agreement of 2 was found compared with the spectra obtained for the second series of ZnO films prepared by ALD (as shown by solid and dashed lines in the inset of Fig. 7). The gap energies and the peak values of 2 determined by ellipsometry are compared in Table 5. Although the range of investigated Ga content is small, clear trends can be observed: the gap energy and the peak value of 2 increase with annealing [30] and with decreasing Ga content. According to previous studies of other authors the increase of both the peak amplitude and band gap is consistent with the decrease of specific resistance [9] and the increase of carrier concentration [31]. These results are also consistent with the increase of conductance as a result of annealing [32]. The n× {ZnO (10 nm)/Al2 O3 (3 nm)}/ZnO (10 nm) multilayer series was also evaluated using the Herzinger–Johs model using a layer stack shown in the inset of Fig. 8. The typical fitting errors are plotted at the bottom as differences in the measured and calculated and spectra. The thickness of the native SiO2 , as well as the dielectric functions of the ZnO and Al2 O3 layers were determined using reference samples with only these layers as top films. The value of 2 of the 10-nm ZnO layer of these films is plotted in the inset of Fig. 7 as described above. The dispersion of the refractive index of the Al2 O3 layer and a reference spectrum from Ref. [33] is shown at the left axis of Fig. 8. The refractive index of the Al2 O3 film of this layer stack is significantly lower than the reference value, which reveals that the nominally 3 nm thin Al2 O3 ALD film is not uniform. As a consequence, the Al2 O3 layer could not be revealed
Table 5 Gap energies in eV and peak amplitudes of 2 (in parentheses) calculated using the Tauc–Lorentz parameterization measured by ellipsometry. Ga (%)
As deposited
Annealed
0.00 0.17 0.34
2.95 (0.40) 2.87 (0.38) 2.82 (0.36)
3.02 (0.45) 2.88 (0.39) 2.85 (0.39)
127
Fig. 8. Refractive index of 3 nm Al2 O3 (n – left axis) measured on top of a 10-nm ZnO layer; the fit quality measured as the difference between the measured and calculated ellipsometric angles, and , plotted to the right axis for the sample with a ZnO and an additional Al2 O3 layer at the angle of incidence of 75◦ ; the inset shows the total thickness (of all the ZnO and Al2 O3 layers) as a function of the cycles defined in the schematic representation of the structure shown in the top left corner.
in the ellipsometric measurement on multi-layer stacks as separate layers, most probably due to mixing with the subsequent ZnO layer. However, when fitting the stack as one homogeneous layer, and plotting the total thickness as a function of {ZnO/Al2 O3 } cycles (see the inset of Fig. 8), the slope indicates a period of 12.0 nm with good agreement of the nominally 10 + 3 nm, taking into account a non-compact Al2 O3 layer. The offset of 12.4 nm agrees well with the expected sum of one ZnO and native oxide layer thicknesses. 4. Conclusions In the present study it has been shown that the effect of even small changes in Ga on the layer properties can be traced by optical and X-ray methods. Ga doping concentration in the range of <1% and In concentrations in the range of several percent have been measured by XRF and correlated with the results of several optical characterization techniques. The Raman-active dopant mode from Ga at 580 cm−1 showed a correlation with the Ga concentration. While the dopant mode is narrow for the annealed samples, it is overlaid by an additional defect band when the sample is not annealed. It has been shown that the surface nanoroughness has to be taken into account in the optical model for ellipsometry in order to determine accurate bulk layer optical properties. Both the peak amplitude of 2 and the gap energy (which is consistent with decreasing specific resistance [9]) increase as a result of annealing and decrease with increasing Ga content. The refractive index decreases with annealing and increases with increasing Ga content. The surface nanoroughness was slightly but systematically smaller on the annealed samples. Acknowledgements Support from the European Community’s Seventh Framework Program, European Metrology Research Program (EMRP), ERANET Plus under Grant Agreement No. 217257 (the EMRP is jointly funded by the EMRP participating countries within EURAMET and the European Union) as well as from the National Development Agency Grant TÁMOP-4.2.2/B-10/1-2010-0025 and OTKA Grant No. K81842 is greatly acknowledged. References [1] U. Özgür, Y.I. Alivov, C. Liu, A. Teke, M.A. Reshchikov, S. Dogan, V. Avrutin, S.-J. Cho, H. Morkoc, Journal of Applied Physics 98 (2005) 041301. [2] P.F. Carcia, R.S. McLean, M.H. Reilly, J.G. Nunes, Applied Physics Letters 82 (2003) 1117.
128
P. Petrik et al. / Applied Surface Science 281 (2013) 123–128
[3] P.F. Carcia, R.S. McLean, M.H. Reilly, Applied Physics Letters 88 (2006) 123509. [4] V. Craciun, J. Elders, J.G.E. Gardeniers, I.W. Boyd, Applied Physics Letters 65 (1994) 2963. [5] V. Craciun, S. Amirhaghi, D. Craciun, J. Elders, J.G.E. Gardeniers, I.W. Boyd, Applied Physics Letters 65 (1994) 2963. [6] G. Socol, M. Socol, N. Stefan, E. Axente, G. Popescu-Pelin, D.C.L. Duta, C.N. Mihailescu, I.N. Mihailescu, A. Stanculescu, D. Visan, V. Sava, A.C. Galca, C.R. Luculescu, V. Craciun, Journal of Applied Physics 95 (2004) 4953. [7] M. Baum, S. Polster, M. Jank, I. Alexeev, L.F.M. Schmidt, Applied Physics A 107 (2012) 269. [8] S.A. Studenikin, N. Golego, M. Cocivera, Journal of Applied Physics 84 (1998) 2287. [9] C. Major, A. Nemeth, G. Radnoczi, Z. Czigany, M. Fried, Z. Labadi, I. Barsony, Applied Surface Science 255 (2009) 8907. [10] A.C. Galca, M. Secu, A. Vlad, J.D. Pedarnig, Thin Solid Films 518 (2010) 4603. [11] A.C. Galca, G. Socol, V. Craciun, Thin Solid Films 520 (2012) 4722. [12] S.M. George, Chemical Reviews 110 (2010) 111. [13] M. Krumrey, G. Ulm, Nuclear Instruments & Methods A 467–468 (2001) 1175–1178. [14] F. Scholze, M. Procop, X-ray Spectrometry 30 (2001). [15] M. Krumrey, M. Gerlach, F. Scholze, G. Ulm, Nuclear Instruments & Methods A 568 (2006) 364–368. [16] M. Kolbe, B. Beckhoff, M. Krumrey, G. Ulm, Spectrochimica Acta, Part B 60 (2005) 505–510. [17] B. Beckhoff, R. Fliegauf, M. Kolbe, M. Müller, J. Weser, G. Ulm, Analytical Chemistry 79 (2007) 7873–7882.
[18] W.T. Elam, B. Ravel, J.R. Sieber, Radiation Physics and Chemistry 63 (2002) 121–128. [19] H. Ebel, R. Svagera, M.F. Ebel, A. Shaltout, J.H. Hubbell, X-ray Spectrometry 32 (2003) 442–451. [20] N. Hasuike, H. Fukumura, H. Harima, K. Kisoda, H. Matsui, H. Saeki, H. Tabata, Journal of Physics: Condensed Matter 16 (2004) S5807. [21] H. Yoshikawa, S. Adachi, Japanese Journal of Applied Physics 36 (1997) 6237. [22] W.L. Bond, Journal of Applied Physics 36 (1965) 1674. [23] K. Ide, K. Nomura, H. Hiramatsu, T. Kamiya, H. Honoso, Journal of Applied Physics 111 (2012) 073513. [24] T. Holden, P. Ram, F.H. Pollak, J.L. Freeouf, B.X. Yang, M.C. Tamargo, Physical Review B 56 (1997) 4037. [25] G. Jellison Jr., L.A. Boatner, Physical Review B 58 (1998) 3586. [26] G.E. Jellison Jr., F.A. Modine, Applied Physics Letters 69 (1996) 371. [27] B. Johs, C.M. Herzinger, J.H. Dinan, A. Cornfeld, J.D. Benson, Thin Solid Films 313 (1998) 137. [28] R.W. Collins, A.S. Ferlauto, in: E.G. Irene, H.G. Tomkins (Eds.), Handbook of Ellipsometry, William Andrew, Norwich, NY, 2005, p. 93. [29] D.E. Aspnes, A.A. Studna, E. Kinsbron, Physical Review B 29 (1984) 768. [30] F. Wang, M.Z. Wu, Y.Y. Wang, Y.M. Yu, X.M. Wu, L.J. Zhuge, Vacuum 89 (2013) 127. [31] Y. Hwang, H. Kim, Y. Um, Current Applied Physics 12 (Suppl. 4) (2012) S76–S79. [32] Y.F. Lu, H.Q. Ni, Z.H. Mai, Z.M. Ren, Journal of Applied Physics 88 (2000) 498. [33] T. Lichtenstein, Handbook of Thin Film Materials, College of Engineering and Applied Science, University of Rochester, 1979.
Contents lists available at SciVerse ScienceDirect
Applied Surface Science journal homepage: www.elsevier.com/locate/apsusc
Characterization of ZnO structures by optical and X-ray methods P. Petrik a,b,∗ , B. Pollakowski c , S. Zakel d , T. Gumprecht a,e , B. Beckhoff c , M. Lemberger a , Z. Labadi b , Z. Baji b , M. Jank a , A. Nutsch c a
Fraunhofer Institute for Integrated Systems and Device Technology, Schottkystrasse 10, 91058 Erlangen, Germany Institute for Technical Physics and Materials Science (MFA), Research Center for Natural Sciences, Konkoly Thege Rd. 29-33, 1121 Budapest, Hungary Physikalisch-Technische Bundesanstalt (PTB), Abbestr. 2-12, 10587 Berlin, Germany d Physikalisch-Technische Bundesanstalt (PTB), Bundesallee 100, D-38116 Braunschweig, Germany e Erlangen Graduate School in Advanced Optical Technologies (SAOT), Paul-Gordan-Strasse 9, 91052 Erlangen, Germany b c
a r t i c l e
i n f o
Article history: Received 22 October 2012 Received in revised form 5 December 2012 Accepted 6 December 2012 Available online 18 January 2013 Keywords: Zinc oxide Spectroscopic ellipsometry X-ray fluorescence Raman spectrometry VUV reflectometry Atomic layer deposition Sputtering
a b s t r a c t ZnO thin films doped by Ga and In as well as multilayer structures of ZnO/Al2 O3 have been investigated by X-ray fluorescence, Raman spectrometry, spectroscopic ellipsometry and vacuum ultra violet reflectometry. Systematic changes in the optical properties have been revealed even for Ga concentrations below 1%. The Raman active phonon mode of Ga doping at 580 cm−1 shows a correlation with the Ga concentration. Optical models with surface nanoroughness correction and different parameterizations of the dielectric function have been investigated. There was a good agreement between the dielectric functions determined by the Herzinger–Johs polynomial parameterization and by direct inversion. It has been shown that the correction of the nanoroughness significantly influences the accuracy of the determination of the layer properties. The band gap and peak amplitude of the imaginary part of the dielectric function corresponding to the excitonic transition changes systematically with the Ga-content and with annealing even for low concentrations. © 2013 Elsevier B.V. All rights reserved.
1. Introduction The manufacturing of optoelectronic thin films is of key importance, because a significant number of industries rely on such structures. ZnO is one of the most intensively investigated optoelectronic materials for applications as a transparent conductive oxide or complex semiconductor with different doping, defect budget and varying preparation parameters [1]. Thin ZnO layers can be prepared using a range of methods including sputtering [2], atomic layer deposition (ALD) [3], pulsed laser deposition [4–6], spin coating from nanoparticulates [7], or spray pyrolysis [8]. Optical techniques such as ellipsometry can be used for the sensitive, quick and non-destructive determination of a range of material and structural parameters such as thickness, surface nanoroughness, interface quality, density, homogeneity, band gap and exciton strength (which makes possible the indirect determination of electrical properties) [9–11].
∗ Corresponding author at: Institute for Technical Physics and Materials Science (MFA), Research Center for Natural Sciences, Konkoly Thege Rd. 29-33, 1121 Budapest, Hungary. Tel.: +36 13922502; fax: +36 13922226. E-mail address: [email protected] (P. Petrik). 0169-4332/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.apsusc.2012.12.035
The aim of the studies was the improvement (development of sophisticated models and evaluations) and comparison of optical and X-ray metrologies for the characterization of ZnO structures in single layer as well as in multi-layer stacks prepared by sputtering and ALD. This study is intended to be a step toward the establishment of validated reference methodologies for a reliable characterization of key optoelectronic materials, as a part of the joint research project for optoelectronic thin film characterization (IND07, “Metrology for the manufacturing of thin films”) in the European Metrology Research Program of EURAMET. Furthermore, our investigations aim for the development of reference samples with controlled defect concentration and morphology as well as methods for elemental depth profiling. 2. Experimental details Dual-target sputtering was used to prepare InGaZnO samples on oxidized (10 nm SiO2 ) single-crystalline silicon wafers in thicknesses of nominally 50 nm (for X-ray, ellipsometric and reflectometric measurements) and 500 nm (for Raman spectrometry). The parameters of preparation and numbering of samples used in this study are summarized in Table 1. ZnO/Al2 O3 multi-layer structures of n× {ZnO (10 nm)/Al2 O3 (3 nm)}/ZnO (10 nm) have been prepared for n = 1, . . ., 3 using ALD
124
P. Petrik et al. / Applied Surface Science 281 (2013) 123–128
Table 1 Parameters of preparation and numbering of samples investigated in this study. Notations of ‘X’ and ‘R’ were used for X-ray/optical (with nominal thicknesses of 50 nm) and Raman (with nominal thicknesses of 500 nm) measurements, respectively. Ga (%)
Not annealed
Annealed
0.00 0.17 0.34
X1, R1 X5, R5 X7, R7
X9, R9 X13, R13 X15, R15
[12]. The ALD process took place in a flow type Picosun SUNALE TM R-100 type ALD reactor. Nitrogen was used as purging gas and diethyl-zinc, trimethyl-aluminum and H2 O as precursors. The precursors were of electronic grade purity manufactured by SAFC Hitech. The deposition temperature was 300 ◦ C, the flow rates were 150 sccm. The pulse time for all precursor injections was 0.1 s, while the purging time after each metalorganic precursor pulse was 3 s, and 4 s after the water pulses. The substrates were cleaned in cc.HNO3 and high resistivity water. Ellipsometric measurements were performed using a SOPRA SE5 multichannel rotating polarizer and a Woollam M-2000DI multichannel rotating compensator spectroscopic ellipsometer in the wavelength range of 193–1690 nm. The spot size was 2–3 mm in both cases. The measurement time of one spot was in the several seconds range. The X-ray fluorescence (XRF) measurements on the Ga- and In-doped ZnO samples have been carried out at a four-crystal monochromator (FCM) beamline [13] in the PTB laboratory at the synchrotron radiation facility BESSY II. This beamline provides monochromatic radiation from a bending magnet in the energy range between 1.75 eV and 11 keV. The originating measurements shown here were performed at 11 keV at an angle of incidence of 1◦ . A toroidal mirror in the front and a cylindrical behind of the FCM ensures a beam diameter in the focal plane of about 280 m. As a possible end station at the FCM beamline, an ultra high vacuum (UHV) chamber optimized for X-ray spectrometry measurements in different geometries was placed in the focus. So, each of the samples can be adjusted in the center of the chamber, were the focus is located. For detecting the fluorescence radiation an energy dispersive Silicon Drift Detector (SDD) is placed perpendicular to the incoming beam. The detector response function for the respective energies and the efficiency are well-known [14,15]. The solid angle of detection was determined exactly and the incoming photon flux was measured by means of calibrated photodiodes. The knowledge of the experimental details is necessary for a referencefree quantification of the mass deposition of thin layered structures [16]. The Raman spectrometric measurements were performed using a LabRAM Aramis Raman microscope and LabSpec 5.0 Software (Horiba Jobin-Yvon). The instrument was equipped with a 1200 grooves/mm holographic grating and an excitation laser using a wavelength of 532 nm. The laser power was set to 2 mW. The microscope was used in non-confocal mode with an 50× objective (NA 0.75). Calibration of the spectral line position was carried out prior to each acquisition using the crystalline silicon mode at 520 cm−1 . The integration time was 400 s. As the thin films were sputtered on a c-Si wafer with 30 nm of thermal oxide on top, the spectra of c-Si and c-Si coated with thermal oxide were also acquired and used for baseline subtraction. Vacuum ultra-violet (VUV) reflectometry has been done using a Metrosol VUV 7000 equipment in the wavelength range of 120–800 nm with a spot size of 35 m. The spot size can be important when checking lateral inhomogeneity, but it was not utilized in this study.
Fig. 1. X-ray fluorescence spectrum of the sample X7. The excitation energy was set to 11 keV. The black line is the measured spectrum and the gray line is the fitted spectrum of the convolution of the fluorescence lines energies with the detector response function.
3. Results and discussion The samples X1, X7, X13 and X15 were analyzed regarding their chemical composition by means of reference-free X-ray fluorescence analysis [16]. This approach allows for a determination of the mass deposition of the elements and if the density is known the thickness of the layer can be concluded. The photon energy was set to 11 keV in order to excite the main element of the layer (Zn) and the implanted Ga and In atoms. The analysis of XRF data is based on the deconvolution of the respective spectra by means of detector response functions. The outcome of this is the net count rate for the respective fluorescence lines. Considering the detector efficiency the count rate can be converted into an intensity. For the calculation of the mass deposition m/F (m denoting the mass and F the unit area) only primary excitation are taken into account because no secondary effects are to be expected. The description of the used relation between fluorescence intensity and the number of atoms and mass is described in detail elsewhere [16,17]. This approach requires among the experimental parameters atomic fundamental parameters as photo-electron absorption cross sections, mass attenuation coefficients and fluorescence yield. In this work, the data for the cross sections were taken from data bases of Elam et al. [18] and Ebel et al. [19]. Fig. 1 shows a measured XRF spectrum (black line) and the respective fitted spectrum with detector response function of sample X7 (gray line). This XRF spectrum exhibits the main matrix element fluorescence lines of the single layer Zn K␣, , the substrate material silicon Si K␣, the dopants Ga K␣,  and In L␣,  as well as contamination as nickel Ni K␣,  and iron Fe K␣, . For the further evaluation the fluorescence lines of Zn, Ga and In have been considered. In Fig. 2 a selection of the recorded XRF spectra are plotted. Focusing on the dopant’s fluorescence lines Ga K␣,  and In L␣,  the intensity is considerably increased when increasing the dopant concentration of these elements. All other fluorescence intensities remain almost constant. Considering the intensities of respective fluorescence lines the mass deposition (m/F) of Zn, In and Ga can be determined and the results are shown in Table 2. Considering the presented results in Table 2 the stoichiometry of these samples can also be determined by XRF. The results are listed in Table 3.
P. Petrik et al. / Applied Surface Science 281 (2013) 123–128
125
Fig. 2. X-ray fluorescence spectra of the samples X1, X7 and X13. The excitation energy was set to 11 keV. The counts in the fluorescence lines Ga K␣,  and In L␣,  increase with increasing dopant rate.
The results from Raman spectrometry are shown in Figs. 3 and 4. The crystalline ZnO layer causes two Raman-active E2 phonon modes at 101 cm−1 and 437 cm−1 . An additional mode arises from the doping with Ga at 580 cm−1 . Although in general the E2 -ZnO modes weaken with increasing Ga content, the intensity ratio of the ZnO:Ga band over the ZnO band at 437 cm−1 clearly increases with increasing Ga content (Fig. 4). An investigation of Hasuike et al. [20] showed that for Ga contents of 0.5% and higher this correlation does not hold. Furthermore, the peak slightly broadens from 0.5 to 2% and then finally severely broadens at 5–8%. Annealing of the samples at 400 ◦ C causes smaller bandwidth and lower intensity of the dopant mode (Fig. 3). An additional defect mode with maximum intensity at about 595 cm−1 seems to be present while the sample is not annealed and vanishes when the crystallinity is enhanced. Typical spectra measured by ellipsometry on a nominally 500 nm thick ZnO layer (sample R1) is plotted in Fig. 5 together with the curves of the annealed sample (R9) and the VUV reflectometry measurements on samples X1 and X9. The transparent and opaque photon energy ranges can clearly be distinguished by the interference oscillations characteristic to the transparent range. The layer thickness and surface nanoroughness of the samples
Fig. 3. Raman spectra for different Ga contents determined by XRF.
can be determined by ellipsometry in the transparent spectral range using a simple Cauchy dispersion (n = A + B/2 + C/4 , where n denotes the refractive index, and A, B and C are the Cauchy parameters) using an optical model of c-Si/SiO2 /ZnO, whereas the refractive index of ZnO can be described by the Cauchy model. The layer thickness, the refractive index (Fig. 6) and the surface nanoroughness have been determined in the photon energy range of 1.5–2.5 eV, because no absorption (extinction coefficient k = 0) has to be taken into account in this range, which simplifies the model and reduces parameter correlation.
Table 2 The mass deposition m/F of Zn, Ga and In was determined by reference-free X-ray fluorescence analysis. A relative uncertainty of 15% (k = 1) was estimated. Sample X1 X7 X13 X15
Zn (g cm−2 ) 18 ± 3 16 ± 2 18 ± 3 16 ± 2
Ga (g cm−2 )
In (g cm−2 )
0 0.20 ± 0.03 0.11 ± 0.02 0.20 ± 0.03
0 1.00 ± 0.15 0.35 ± 0.05 1.02 ± 0.15
Table 3 The elemental composition was determined by reference-free XRF and a relative uncertainty of 15% was estimated. The oxygen part was concluded by XRF measurements at lower excitation energies, for example 1.622 keV. The layer thickness was determined assuming a density of 10% lower than for bulk ZnO. Sample X1 X7 X13 X15
Zn (%) 50 ± 8 47 ± 7 50 ± 7 47 ± 7
O (%) 50 ± 8 49 ± 7 49 ± 7 49 ± 7
Ga (%)
In (%)
0 0.34 ± 0.05 0.17 ± 0.03 0.34 ± 0.05
0 2.8 ± 0.4 0.9 ± 0.2 2.8 ± 0.4
Thickness 43 ± 6 39 ± 6 44 ± 7 39 ± 6
Fig. 4. Intensity ratio of the ZnO:Ga band over the ZnO band measured by Raman spectrometry.
126
P. Petrik et al. / Applied Surface Science 281 (2013) 123–128
nanoroughness. The mean squared error of the fit is almost twice as much in case of the fit without roughness as with roughness, although only one additional fit parameter was introduced. Fig. 6 shows that the refractive index of sputtered layers investigated in this study are slightly lower than that of bulk ZnO from Refs. [21,22], however, the refractive index calculated with surface roughness correction is larger by ≈0.05. This shift can partly be explained by the stress in as-deposited films [23]. However, this effect is estimated to be not larger than about 0.013, comparing the refractive index of X1 in Table 4 with its annealed counterpart of sample X9. This reveals good film quality and points out the significance of surface roughness correction. The parameters determined by ellipsometry in the transparent range are summarized in Table 4. Small, but systematic decrease can be observed for both n (equivalent with and increase of the volume fraction of voids, fv ) and the surface nanoroughness dr as a result of annealing. For photon energies near the band gap the dielectric function of ZnO can be described by simple oscillator models. One of the most popular methods is the application of different model dielectric functions using several oscillators with analytical lineshapes [21,24,25]. A further possibility is to use a Tauc–Lorentz oscillator [26] combined with a Drude tail for the near infra red region [10,11]. The third approach is the Herzinger–Johs method to use polynomial functions to fit the line-shape of the dielectric functions in the proximity of different critical point energies [27,28]. Finally, the most general, model-independent approach is the direct inversion [29,28], in which the dielectric function is calculated wavelength-by-wavelength using the layer thickness determined independently or in a wavelength range of no absorption. A cross-check for the proper layer thickness is the criterion that the absorption tail has to decay smoothly to zero toward the longer wavelengths [29]. A great advantage of this method is that no assumption has to be made for the dispersion of the dielectric function. A disadvantage is that taking into account the surface nanoroughness or any probable surface layer is more problematic. In this study the Psemi-M0 parameterization of the Herzinger–Johs method was used, which is an adaptation of the polynomial parameterization to M0 critical points. Using this approach the surface nanoroughness could be taken into account as an effective medium overlayer with 50% layer material and 50% void, fitting only the thickness of this layer. Note that similar to the above case of the transparent spectral region the fit quality improves to a large extent when using a surface roughness correction (e.g. the MSE value for sample X15 with and without a roughness layer is 2.8 and 12.6, respectively). Fig. 7 shows the imaginary part of the dielectric function (2 – which is proportional to the joint density of electronic states) in a photon energy region close to the band gap. The height and broadening of the excitonic peak increases and decreases, respectively, as a result of annealing for each doping level. The slope of the increase of 2 at the band gap decreases with increasing Ga content. The inset
Fig. 5. Measured reflectometric (X1 for as deposited and X9 for annealed) and ellipsometric and (R1 for as deposited and R9 for annealed) spectra. The rectangles show the most relevant gap regions.
Fig. 6. Refractive indices of sample X1 calculated in the transparent spectral range with and without a surface nanoroughness layer. The refractive index values from Refs. [21,22] are also plotted.
Taking into account the surface nanoroughness is a crucial step in terms of both fit quality and determination of accurate bulk layer refractive index (Fig. 6). The surface nanoroughness is modeled as a homogeneous surface layer with 50% layer material and 50% void fitting only the thickness of the layer. The difference of the measured and calculated spectra shown in Fig. 6 is significantly larger for the model that does not take into account the surface
Table 4 Optical properties measured in the transparent spectral range. A, B and C are the Cauchy parameters, n denotes the refractive index at the photon energy of 2 eV (620 nm). fv is the volume fraction of voids using an effective medium composition of material X1 and voids (i.e. the change of density relative to sample X1 – negative volume fraction indicates an optical density larger than the reference). db and dr denote the thicknesses of the bulk layer and the surface nanoroughness, respectively, measured by ellipsometry (the reflectometry results are given in parentheses). The sensitivity of the Cauchy parameters are 0.001 for parameters A and B and 0.0001 for parameter C (‘Ann.’ denotes ‘annealed’). Nr.
Ga (%)
Ann.
A
B (m2 )
C (10−4 m4 )
n
fv (%)
db (nm)
dr (nm)
X1 X5 X7 X9 X13 X15
0.00 0.17 0.34 0.00 0.17 0.34
No No No Yes Yes Yes
1.934 1.933 1.932 1.922 1.916 1.926
0.038 0.035 0.038 0.038 0.040 0.041
−5.9 0.0 1.6 −3.4 −4.8 −0.2
2.030 2.026 2.033 2.017 2.018 2.032
0.0 0.2 −0.2 1.2 1.1 −0.1
50.7 (45.6) 51.6 46.6 (46.6) 50.7 51.3 (52.5) 46.3 (48.5)
9.2 8.4 8.9 8.0 8.0 8.1
P. Petrik et al. / Applied Surface Science 281 (2013) 123–128
Fig. 7. Imaginary part of the dielectric function (2 ) of samples X1 (0% Ga) and X7 (0.37% Ga) with and without annealing measured using the Psemi oscillator model by ellipsometry. The inset shows a comparison of 2 values calculated using the Psemi model with (‘rough.’) and without (‘no rough.’) a surface nanoroughness layer, as well as direct inversion (‘dir. inv.’). As a reference, 2 of an ALD-deposited 10-nm ZnO film from the second series is also plotted in the inset (denoted by ‘ALD’).
shows that taking into account the surface nanoroughness slightly decreases the amplitude of 2 . A justification of the Psemi model is the very good agreement with the result from direct inversion (dotted and dash-dotted lines in the inset of Fig. 7), because the method of direct inversion is model-independent, but in this case only the results without surface roughness correction can be compared. It is also important to point out that a good agreement of 2 was found compared with the spectra obtained for the second series of ZnO films prepared by ALD (as shown by solid and dashed lines in the inset of Fig. 7). The gap energies and the peak values of 2 determined by ellipsometry are compared in Table 5. Although the range of investigated Ga content is small, clear trends can be observed: the gap energy and the peak value of 2 increase with annealing [30] and with decreasing Ga content. According to previous studies of other authors the increase of both the peak amplitude and band gap is consistent with the decrease of specific resistance [9] and the increase of carrier concentration [31]. These results are also consistent with the increase of conductance as a result of annealing [32]. The n× {ZnO (10 nm)/Al2 O3 (3 nm)}/ZnO (10 nm) multilayer series was also evaluated using the Herzinger–Johs model using a layer stack shown in the inset of Fig. 8. The typical fitting errors are plotted at the bottom as differences in the measured and calculated and spectra. The thickness of the native SiO2 , as well as the dielectric functions of the ZnO and Al2 O3 layers were determined using reference samples with only these layers as top films. The value of 2 of the 10-nm ZnO layer of these films is plotted in the inset of Fig. 7 as described above. The dispersion of the refractive index of the Al2 O3 layer and a reference spectrum from Ref. [33] is shown at the left axis of Fig. 8. The refractive index of the Al2 O3 film of this layer stack is significantly lower than the reference value, which reveals that the nominally 3 nm thin Al2 O3 ALD film is not uniform. As a consequence, the Al2 O3 layer could not be revealed
Table 5 Gap energies in eV and peak amplitudes of 2 (in parentheses) calculated using the Tauc–Lorentz parameterization measured by ellipsometry. Ga (%)
As deposited
Annealed
0.00 0.17 0.34
2.95 (0.40) 2.87 (0.38) 2.82 (0.36)
3.02 (0.45) 2.88 (0.39) 2.85 (0.39)
127
Fig. 8. Refractive index of 3 nm Al2 O3 (n – left axis) measured on top of a 10-nm ZnO layer; the fit quality measured as the difference between the measured and calculated ellipsometric angles, and , plotted to the right axis for the sample with a ZnO and an additional Al2 O3 layer at the angle of incidence of 75◦ ; the inset shows the total thickness (of all the ZnO and Al2 O3 layers) as a function of the cycles defined in the schematic representation of the structure shown in the top left corner.
in the ellipsometric measurement on multi-layer stacks as separate layers, most probably due to mixing with the subsequent ZnO layer. However, when fitting the stack as one homogeneous layer, and plotting the total thickness as a function of {ZnO/Al2 O3 } cycles (see the inset of Fig. 8), the slope indicates a period of 12.0 nm with good agreement of the nominally 10 + 3 nm, taking into account a non-compact Al2 O3 layer. The offset of 12.4 nm agrees well with the expected sum of one ZnO and native oxide layer thicknesses. 4. Conclusions In the present study it has been shown that the effect of even small changes in Ga on the layer properties can be traced by optical and X-ray methods. Ga doping concentration in the range of <1% and In concentrations in the range of several percent have been measured by XRF and correlated with the results of several optical characterization techniques. The Raman-active dopant mode from Ga at 580 cm−1 showed a correlation with the Ga concentration. While the dopant mode is narrow for the annealed samples, it is overlaid by an additional defect band when the sample is not annealed. It has been shown that the surface nanoroughness has to be taken into account in the optical model for ellipsometry in order to determine accurate bulk layer optical properties. Both the peak amplitude of 2 and the gap energy (which is consistent with decreasing specific resistance [9]) increase as a result of annealing and decrease with increasing Ga content. The refractive index decreases with annealing and increases with increasing Ga content. The surface nanoroughness was slightly but systematically smaller on the annealed samples. Acknowledgements Support from the European Community’s Seventh Framework Program, European Metrology Research Program (EMRP), ERANET Plus under Grant Agreement No. 217257 (the EMRP is jointly funded by the EMRP participating countries within EURAMET and the European Union) as well as from the National Development Agency Grant TÁMOP-4.2.2/B-10/1-2010-0025 and OTKA Grant No. K81842 is greatly acknowledged. References [1] U. Özgür, Y.I. Alivov, C. Liu, A. Teke, M.A. Reshchikov, S. Dogan, V. Avrutin, S.-J. Cho, H. Morkoc, Journal of Applied Physics 98 (2005) 041301. [2] P.F. Carcia, R.S. McLean, M.H. Reilly, J.G. Nunes, Applied Physics Letters 82 (2003) 1117.
128
P. Petrik et al. / Applied Surface Science 281 (2013) 123–128
[3] P.F. Carcia, R.S. McLean, M.H. Reilly, Applied Physics Letters 88 (2006) 123509. [4] V. Craciun, J. Elders, J.G.E. Gardeniers, I.W. Boyd, Applied Physics Letters 65 (1994) 2963. [5] V. Craciun, S. Amirhaghi, D. Craciun, J. Elders, J.G.E. Gardeniers, I.W. Boyd, Applied Physics Letters 65 (1994) 2963. [6] G. Socol, M. Socol, N. Stefan, E. Axente, G. Popescu-Pelin, D.C.L. Duta, C.N. Mihailescu, I.N. Mihailescu, A. Stanculescu, D. Visan, V. Sava, A.C. Galca, C.R. Luculescu, V. Craciun, Journal of Applied Physics 95 (2004) 4953. [7] M. Baum, S. Polster, M. Jank, I. Alexeev, L.F.M. Schmidt, Applied Physics A 107 (2012) 269. [8] S.A. Studenikin, N. Golego, M. Cocivera, Journal of Applied Physics 84 (1998) 2287. [9] C. Major, A. Nemeth, G. Radnoczi, Z. Czigany, M. Fried, Z. Labadi, I. Barsony, Applied Surface Science 255 (2009) 8907. [10] A.C. Galca, M. Secu, A. Vlad, J.D. Pedarnig, Thin Solid Films 518 (2010) 4603. [11] A.C. Galca, G. Socol, V. Craciun, Thin Solid Films 520 (2012) 4722. [12] S.M. George, Chemical Reviews 110 (2010) 111. [13] M. Krumrey, G. Ulm, Nuclear Instruments & Methods A 467–468 (2001) 1175–1178. [14] F. Scholze, M. Procop, X-ray Spectrometry 30 (2001). [15] M. Krumrey, M. Gerlach, F. Scholze, G. Ulm, Nuclear Instruments & Methods A 568 (2006) 364–368. [16] M. Kolbe, B. Beckhoff, M. Krumrey, G. Ulm, Spectrochimica Acta, Part B 60 (2005) 505–510. [17] B. Beckhoff, R. Fliegauf, M. Kolbe, M. Müller, J. Weser, G. Ulm, Analytical Chemistry 79 (2007) 7873–7882.
[18] W.T. Elam, B. Ravel, J.R. Sieber, Radiation Physics and Chemistry 63 (2002) 121–128. [19] H. Ebel, R. Svagera, M.F. Ebel, A. Shaltout, J.H. Hubbell, X-ray Spectrometry 32 (2003) 442–451. [20] N. Hasuike, H. Fukumura, H. Harima, K. Kisoda, H. Matsui, H. Saeki, H. Tabata, Journal of Physics: Condensed Matter 16 (2004) S5807. [21] H. Yoshikawa, S. Adachi, Japanese Journal of Applied Physics 36 (1997) 6237. [22] W.L. Bond, Journal of Applied Physics 36 (1965) 1674. [23] K. Ide, K. Nomura, H. Hiramatsu, T. Kamiya, H. Honoso, Journal of Applied Physics 111 (2012) 073513. [24] T. Holden, P. Ram, F.H. Pollak, J.L. Freeouf, B.X. Yang, M.C. Tamargo, Physical Review B 56 (1997) 4037. [25] G. Jellison Jr., L.A. Boatner, Physical Review B 58 (1998) 3586. [26] G.E. Jellison Jr., F.A. Modine, Applied Physics Letters 69 (1996) 371. [27] B. Johs, C.M. Herzinger, J.H. Dinan, A. Cornfeld, J.D. Benson, Thin Solid Films 313 (1998) 137. [28] R.W. Collins, A.S. Ferlauto, in: E.G. Irene, H.G. Tomkins (Eds.), Handbook of Ellipsometry, William Andrew, Norwich, NY, 2005, p. 93. [29] D.E. Aspnes, A.A. Studna, E. Kinsbron, Physical Review B 29 (1984) 768. [30] F. Wang, M.Z. Wu, Y.Y. Wang, Y.M. Yu, X.M. Wu, L.J. Zhuge, Vacuum 89 (2013) 127. [31] Y. Hwang, H. Kim, Y. Um, Current Applied Physics 12 (Suppl. 4) (2012) S76–S79. [32] Y.F. Lu, H.Q. Ni, Z.H. Mai, Z.M. Ren, Journal of Applied Physics 88 (2000) 498. [33] T. Lichtenstein, Handbook of Thin Film Materials, College of Engineering and Applied Science, University of Rochester, 1979.