Mater. Res. Soc. Symp. Proc. Vol Materials Research Society

Transcription

1 Mater. Res. Soc. Symp. Proc. Vol Materials Research Society 1165-M05-24 Hydrogen diffusion in zinc oxide thin films W. Beyer 1,3, U. Breuer 2, F. Hamelmann 3, J. Hüpkes 1, A. Stärk 2, H. Stiebig 3, U. Zastrow 1 1 IEF5-Photovoltaik, Forschungszentrum Jülich GmbH, D Jülich, Germany. 2 Zentralabteilung für Chemische Analysen, Forschungszentrum Jülich GmbH, D Jülich, Germany. 3 Malibu GmbH & Co.KG, Böttcherstrasse 7, D-33609, Bielefeld, Germany. ABSTRACT Hydrogen diffusion in zinc oxide thin films was studied by secondary ion mass spectrometry (SIMS) measurements, investigating the spreading of implanted deuterium profiles by annealing. By effusion measurements of implanted rare gases He and Ne the microstructure of the material was characterized. While for material prepared by low pressure chemical vapour deposition an interconnected void structure and a predominant diffusion of molecular hydrogen was found, sputter-deposited ZnO films showed a more compact structure and long range diffusion of atomic hydrogen. Hydrogen diffusion energies of ev, i.e. higher than reported in literature were found. The results are discussed in terms of a H diffusion model analogous to the model applied for hydrogen diffusion in hydrogenated amorphous and microcrystalline silicon. INTRODUCTION Thin films of zinc oxide (ZnO) are of interest as transparent conductive oxide layers for application in thin film solar cells and other devices. One ubiquitous impurity in such layers is hydrogen which has been identified as a shallow donor [1]. A strong influence of hydrogen on conductivity of single crystalline ZnO was first reported more than 50 years ago [2,3]. Various efforts were undertaken to measure H diffusion and to determine the H diffusion coefficient, both in ZnO crystals as well as in polycrystalline ZnO thin films. However, most experiments confined to the in- and out-diffusion of H [4] and not to the diffusion of H within a given sample. Here we report on a study of H diffusion using SIMS depth profiling of deuterium implantation profiles prior to and after annealing at various annealing temperatures and annealing time. We focus on polycrystalline ZnO films prepared by sputtering (SP) and by low pressure chemical vapour deposition (LPCVD). Doped as well as (nominally) undoped films were investigated. The microstructure of these films was characterized by effusion of implanted helium and neon [5]. EXPERIMENTAL DETAILS The ZnO films investigated were deposited by sputtering using ceramic target (undoped or 1 wt.% Al 2 O 3 ) and by LPCVD using flows of diethylzinc and water vapor as process gases and diborane ( 0.2 % in gas phase) for doping. As a substrate, crystalline silicon wafers were used. For hydrogen diffusion measurements, deuterium (D + ) was implanted at a dose of /cm 2 and at an energy of 25 and 40 kev. For microstructure analysis, He + and Ne + were implanted at a

2 Figure 1. SIMS depth profiles of ZnO, H and D for LPCVD ZnO implanted with deuterium for various annealing temperatures T a (annealing time 10 min) (n.a.: not annealed). Figure 2. Effusion rate dn/dt of hydrogen, and implanted He and Ne versus temperature T for (a) LPCVD (T S = 200 C), (b) sputtered (SP,T S = 60 C) and (c) single crystalline ZnO. dose of 3x10 15 /cm 2 and at energies of 40 or 100 kev, respectively. For all implantations, a mass separator was employed. SIMS measurements were performed using a time-of-flight setup with a cesium sputtering beam at an energy of 2 kev. In the effusion experiment, the samples were heated in a turbomolecular-pumped vacuum (base pressure 10-7 mbar) at a rate of 20 K/min up to 1050 C and the effusing gases were detected by a quadrupole mass analyzer. RESULTS AND DISCUSSION In Fig. 1, SIMS depth profiles of a typical (B-doped) LPCVD ZnO film is shown which was implanted with 25 kev deuterium (D + ) and was annealed (annealing time 10 min) at various temperatures. This film was deposited at a substrate temperature of 200 C. Up to an annealing temperature of T a =200 C, no changes in the D and H profiles were observed. At higher annealing temperatures, deuterium shows (at low concentration levels) plateau-like profiles extending into the not implanted material, indicating fast H(D) migration processes. With increasing T a, the D concentration in the implantation peak decreases without a spreading of the implantation profile and the H concentration decreases, too. Similar D/H diffusion profiles have been observed for hydrogenated amorphous silicon (a-si:h) with a columnar microstructure and were attributed to the diffusion and effusion of molecular deuterium and hydrogen through an interconnected void network [6]. This latter interpretation is supported by the results of effusion measurements of implanted He and Ne, as shown in Fig. 2 for the material of Fig. 1 as well as for single crystalline ZnO and sputtered ZnO (T S = 60 C) [7]. As demonstrated for amorphous and microcrystalline silicon materials, such measurements are known to give information of a material s microstructure [5], since rare gas atoms like He and Ne atoms do not form bonds to the host material. It is presumed that these atoms diffuse in a so-called doorway diffusion process

3 [8]. Lower effusion temperatures then indicate a more open microstructure, which may be due to the presence of a less dense material or of grain boundaries or interconnected voids [5]. A strong dependence of the diffusion energy on the size of the diffusing atoms is predicted [8]. Note that according to literature data the atomic diameter of He is about 0.2 nm and for Ne and H 2 about 0.25 nm, while other molecules and atoms like O 2, H 2 O, CO, CO 2 or Ar have sizes of 0.3 nm and more [9]. The result for crystalline ZnO of no Ne effusion up to 1050 C then indicates that up to Figure 3. SIMS depth profile of ZnO Figure 4. Deuterium diffusion and deuterium for sputter-deposited coefficient D in SP ZnO:Al (T S = ZnO:Al (T S =60 C), as implanted (n.a.) 60 C) as a function of diffusion and after annealing (5 min.) at 450 C. length L= (Dt) 1/2 for various annealing temperatures T a. high temperatures H 2 cannot diffuse through the ZnO lattice, while it can diffuse at rather low concentration level down to near 300 C in the LPCVD ZnO (in agreement with the results of Fig.1) and to near 800 C in the sputtered ZnO. In all three samples, however, the major fraction of the implanted neon is stable up to 900 C and above, suggesting that in all three materials rather dense regions exist. Helium effusion shows for the three samples a maximum effusion rate near 700 C. Assuming a diffusion-limited process, the diffusion energy E D defined by the Arrhenius dependence D = D 0 exp (-E D /kt)) can be estimated from the temperature T M of maximum effusion rate, using the formula [5,6] D/E D = (d/πt M ) 2 (ß/k). (1) Here, d is the film thickness, k the Boltzmann constant and ß the heating rate. For the He effusion maximum of about 700 C (see Fig.2b), a helium diffusion energy of E D 1.6 ev results if for the diffusion prefactor D 0 the theoretical prefactor of D 0 = 10-3 cm 2 s -1 is assumed [5,6]. While for all three materials a large fraction of the implanted helium seems to follow these diffusion parameters, for LPCVD ZnO the enormous width of the He effusion distribution (extending from 200 C to about 1000 C) demonstrates a strong structural inhomogeneity of this material. Interconnected voids are likely causing the low temperature effusion while isolated voids lead to high helium stability and to the high temperature He effusion [5]. In contrast to the

4 LPCVD material, the results for the sputtered ZnO indicate a more homogeneous dense material. The more dense structure of this latter material also shows up in the hydrogen effusion maximum near 700 C, which is much higher than observed for the LPCVD material (T M 500 C). Fig. 3 shows the results of SIMS depth profiling of an implanted (40 kev) deuterium profile in sputtered polycrystalline ZnO material prior to and after annealing at 450 C for 5 min. After annealing, a spreading of the D implantation profile is observed, unlike what is found for ZnO crystals [10]. Fitting the diffusion profile by a complementary error function (or superpositions of error functions), the diffusion length L = (Dt) 1/2 and thus the (atomic) D (H) diffusion coefficient in ZnO is obtained. Similar as observed for a-si:h [6], a time dependence Figure 5. Deuterium diffusion coefficient for L 6x10-6 cm in sputter-deposited ZnO films (doped with 1 wt. % Al) for various substrate temperatures T S and undoped ((UD); T S = 330 C)). Figure 6. Hydrogen diffusion prefactor D 0 as a function of diffusion energy E D. ( ) this work, ( ) ref. (2), ( ) ref. (3), (, ) ref. (4), ( ) ref. (12). Open symbols: polycrystalline ZnO; closed symbols:zno single crystals. D = D 0 t -α (with α 0.25 for the sample of Fig. 3) shows up. In a-si:h, this time dependence has been associated with the presence of interconnected voids [6]. In Fig. 4, for the ZnO material of Fig. 3 (SP ZnO:Al (T S = 60 C)), the dependence of the diffusion coefficient D on diffusion length L = (Dt ) 1/2 is shown, demonstrating that the time dependence is largely independent of the annealing temperature. The results for the temperature dependence of the H (D) diffusion coefficient are shown in Fig. 5 for three Al-doped sputtered ZnO films of different substrate temperature and one (nominally) undoped sputtered film (T S = 330 C). In order to account for the time dependence, a fixed diffusion length of (L 6x10-6 cm) was evaluated. As is seen in Fig. 5, the diffusion coefficient follows an Arrhenius dependence with a diffusion energy of E D = ev. This diffusion energy is considerably higher than other hydrogen diffusion energies reported for ZnO [4]. The experimental diffusion prefactors D 0 (D cm 2 s -1 ) are also higher than literature data [4]. In Fig. 6, the experimental hydrogen diffusion prefactors D 0 of the present samples are plotted as a function of the respective diffusion energies E D. In addition, various literature values of diffusion parameters of hydrogen for polycrystalline (open symbols) ZnO and for ZnO single crystals (closed symbols) are shown. While diffusion energies of about 1 ev were associated in

5 literature with a trap-limited diffusion of H [11], the low H diffusion energies of about 0.2 ev observed for in-diffusion from a H (D) plasma were attributed to interstitial diffusion of hydrogen without H trapping [12]. However, the small experimental diffusion prefactors of about cm 2 /s observed together with the low diffusion energies of about 0.2 ev point against an interstitial H diffusion, as the theoretical diffusion prefactor [4,6] D H0 = (1/6) ν ph a 2 (2) with a phonon frequency of about ν ph =10 13 s -1 would give unrealistic short jump lengths of a 0.02 Å or less. However in a band model of H diffusion, as applied for the H diffusion in a-si:h [6], such strong deviations of the experimental diffusion prefactors from the theoretical prefactor D H cm 2 /s (3) (see eq. (2), assuming for a realistic mean jump length the Zn-O bond length) can be easily understood. In such a band model, the hydrogen diffusion energy E D is related to the energetic distance between the energy E tr of the H transport path and the H chemical potential µ H, by setting D = D 0 exp (-E D /kt) = D H0 exp (-(E tr - µ H )/kt). (4) Temperature shifts of (E tr - µ H ) can then explain the wide variation of the experimental diffusion prefactor D 0, since for a linear temperature shift of E tr -µ H, i.e. E tr - µ H (T) = (E tr -µ H ) 0 + γ T, (5) the experimental diffusion prefactor becomes D 0 = D H0 exp (-γ /k). Note that the (experimental) diffusion energy E D then equals (E tr -µ H ) 0. The results of Fig. 6 suggest a correlation between experimental diffusion prefactors and diffusion energies, i.e. that a Meyer-Neldel rule is valid. Similar effects were observed for H diffusion in amorphous and microcrystalline Si:H [6,13], and for all these materials a similar explanation of the Meyer-Neldel rule may apply, namely a transition from a high temperature behaviour where intrinsic properties of a material prevail, to a low-temperature behaviour, where extrinsic effects like the in-diffusion of H from a plasma or the freezing of solid state chemical reactions become active. In this model, (E tr -µ H ) MN = E D at D 0 = D H0 characterizes a given diffusion system in the intrinsic state [6]. From the data of Fig.6 and eq. (3), (E tr -µ H ) MN 1.0 ev is obtained for hydrogen in zinc oxide. In analogy to H diffusion a-si:h, it is tempting to associate this latter energy with activation energies of the reaction constants of the solid state reactions: ZnH + HO 2 H + Zn _ + _ O (6) Zn _ + _ O ZnO (7) Thus, it is assumed that H trapping proceeds predominantly by an interruption of ZnO bonds by hydrogen, in agreement with experimental data showing that exposure of ZnO to hydrogen leads

6 to the formation of both, Zn-H and O-H species [14]. With literature data [15] for the binding energies Zn-O (2.85 ev), Zn-H (0.9 ev), O-H (4.4 ev) one obtains µ H approximately 1.25 ev below the vacuum level in reasonable agreement with (E tr - µ H ) MN 1.0 ev. However, in view of the large scatter of data points in Fig. 6 and the possibility that, e.g. due to solubility effects [6] the H diffusion mechanisms in polycrystalline and single crystalline ZnO may be different, more work on H diffusion in ZnO appears necessary. CONCLUSIONS While microstructure effects in polycrystalline ZnO can lead to a high instability of hydrogen due to effusion of H 2 through interconnected voids or grain boundaries, compact ZnO thin films can be grown (e.g.) by sputtering with rather stable incorporation of hydrogen. In this latter material, long range diffusion of atomic hydrogen is detected. The diffusion effects of atomic hydrogen can be explained by a band model of H diffusion, similar as applied for H diffusion in hydrogenated amorphous and microcrystalline silicon. ACKNOWLEDGEMENTS The authors thank H. Siekmann, A. Doumit, H. Kurz for the preparation of the ZnO samples, A. Dahmen for the ion implantation and W. Hilgers, L. Niessen and D. Lennartz for technical assistance. The support by funds from the state of Nordrhein-Westfalen, Germany (TRISO Project) and BMU (LIMA Project) is gratefully acknowledged. REFERENCES 1. C.G. Van de Walle, Phys. Rev. Lett. 85, 86 (2000). 2. E. Mollwo, Z. Phys. 138, 478 (1954). 3. D.G. Thomas, J.J. Lander, J. Chem. Phys. 25, 1136 (1956). 4. N.H. Nickel, Phys. Rev. B 73, (2006). 5. W. Beyer, Phys. Status Solidi C 1, 1144 (2004). 6. W. Beyer, in: Semiconductors and Semimetals, Vol.61, edited by N.H. Nickel (Academic Press, San Diego, 1999) p W. Beyer, J. Hüpkes, H. Stiebig, Thin Solid Films 516, 147 (2007). 8. O.L. Anderson, D.A. Stuart, J. Am. Ceram. Soc. 37, 573 (1954). 9. A. Eucken, H. Hellwege, (Eds.) Landoldt-Börnstein, Atom- und Molekularphysik I (Springer Verlag, Berlin, Germany, 1950) p K. Ip, M.E. Overberg, Y.W. Heo, D.P. Norton, S.J. Pearton, S.O. Kucheyev, C. Jagadish, J.S. Williams, R.G. Wilson, J.M. Zavada, Appl. Phys. Lett. 81, 3996 (2002) 11. M.G. Wardle, J.P. Gross, P.R. Briddon, Phys. Rev. Lett. 96, (2006). 12. K. Ip, M.E. Overberg, Y.W. Heo, D.P. Norton, S.J. Pearton, C.E. Stutz, B. Luo, F. Ren, D.C. Look, J.M. Zavada, Appl. Phys. Lett. 82, 385 (2003). 13. W. Beyer, Solar Energy Mat. & Solar Cells 78, 235 (2003). 14. M. Kunat, S.G. Girol, T. Becker, U. Burghaus, C. Wöll, Phys. Rev. B 66, (R) (2002). 15. J.A. Kerr, A.F. Trotman-Dickenson, in: CRC Handbook of Chemistry and Physics, 58 th Edition (CRC Press, West Palm Beach, Fl. 1978) F-219.