Electrical properties of thin rf sputtered aluminum oxide films

Transcription

1 Materials Science and Engineering Bxxx (2004) xxx xxx Electrical properties of thin rf sputtered aluminum oxide films M. Voigt, M. Sokolowski Institut für Physikalische und Theoretische Chemie der Universität Bonn, Wegelerstraße 12, D Bonn, Germany Abstract Thin films of aluminum oxide (Al 2 O 3 ) were fabricated by rf magnetron sputtering. Different sputter conditions, e.g., composition of the sputter gas (Ar:O 2 ), sputter gas pressure, deposition rate, and preparation of the Al 2 O 3 sputter target before deposition were investigated with the aim to achieve good insulating films with high electrical breakdown fields. The Al 2 O 3 films had a thickness of 160 nm and were deposited on ITO covered glass. By evaporation of Au electrodes on top of the Al 2 O 3 films thin film capacitors were fabricated. Current voltage (I V) measurements were performed under high vacuum and temperatures between 4 and 300 K. Significant scattering of the I V curves and burn-in effects are observed. We find that an admixture of 1% of O 2 in the sputter gas improves the electrical properties, but higher breakdown fields and smaller leakage currents are obtained for sputtering in pure Ar, using a sputter target conditioned in an Ar:O 2 mixture. Impedance spectra, revealed a dielectric constant of 7 for all Al 2 O 3 films. Atomic force microscopy experiments reveal that the surfaces are rather rough with grain sizes in the order of m Elsevier B.V. All rights reserved. Keywords: Aluminum oxide; rf magnetron; Sputtered films; Electrical properties 1. Introduction Aluminum oxide (Al 2 O 3 ) is a widely used electrical insulating material. This is due to its high electrical breakdown field, its large bandgap, and its high dielectric constant. In particular, Al 2 O 3 films with thicknesses in the range of nm are interesting for the preparation of gate insulators in thin film field effect transistors (FETs). Thin films of Al 2 O 3 can be fabricated by dc or rf magnetron sputtering of either an Al target in Ar:O 2 mixtures (reactive sputtering) [1 4], or by sputtering of an Al 2 O 3 target in pure Ar, or Ar:O 2 mixtures [4 12]. One aspect of the investigations in the last years has been the optimisation of the dielectric properties, in particular the electrical resistivity and electrical breakdown fields of Al 2 O 3 films fabricated by magnetron sputtering. However, the electrical properties of thin sputtered Al 2 O 3 films were often only described by the dielectric constants (ε) and breakdown fields [6,8 10,12]. Current voltage (I V) curves were reported only exceptionally [3 5,7,11]. Nevertheless, these are also of high interest, since they contain information about the electrical conduction mechanisms leading to Corresponding author. Tel.: ; fax: address: voigt@thch.uni-bonn.de (M. Voigt). unwanted leakage currents, e.g., across the gate insulator in a FET structure, and may contain information on the mechanism of the electrical breakdown, in addition to the statistical analysis of the breakdown fields [9]. In the present paper, we report I V curves measured for sandwiched Al 2 O 3 films in detail. Our aim is to demonstrate the influence of the sputter parameters and to gain insight in the origin of the leakage currents and their relation to the dielectric break down fields. In addition to the electrical measurements, atomic force microscopy data of the surface morphology will be reported. 2. Experimental The aluminum oxide films were prepared by rf magnetron sputtering (13.56 MHz) from an Al 2 O 3 target (1 in. diameter) with a nominal purity of 99.99% in a high vacuum chamber with a base pressure of typically mbar. The target to sample distance was about 10 cm. Sputter powers were between 20 and 200 W. The sputter gas pressure was varied between and mbar. Either pure Ar (99.998%), or Ar:O 2 mixtures were used as sputter gases. Further details of the sputter conditions are given in Table 1. As substrate we used commercially available indium tin oxide (ITO) covered glass slides (25 mm 25 mm). Prior to sputtering they were cleaned by organic solvents, H 2 O 2, and /$ see front matter 2003 Elsevier B.V. All rights reserved. doi: /j.mseb

2 2 M. Voigt, M. Sokolowski / Materials Science and Engineering Bxxx (2004) xxx xxx Table 1 Summary of sputter conditions of Al 2 O 3 films Sample Sputter gas composition Sputter gas pressure (mbar) Base pressure (mbar) Sputter rate (nm min 1 ) Sputter power (W) Conditioning of target a A Ar Yes 307 B Ar/O 2 (10:1) Yes 307 C Ar/O 2 (100:1) Yes 307 D Ar No c 308 E Ar Yes 265 F Ar Yes 300 G Ar Yes 302 H Ar Yes 255 I Ar Yes 303 T b S (K) All oxide films had a nominal thickness of 160 nm. a Conditioning of target: sputtering in Ar:O 2 (1:1) for 30 min at 200 W. For B and C, the target was conditioned in an Ar:O 2 gas mixture of the same composition, as used during the sputter process, for 30 min. b T S denotes the substrate temperature. c Sputtering of the target in pure Ar for 30 min before starting deposition. Fig. 1. Current density vs. electric field strength of four thin Al 2 O 3 films prepared with different sputter conditions (a d). The films correspond to the samples D (a), B (b), C (c) and A (d) of Table 1. For each film I V curves measured on different contacts are shown in order to demonstrate the scattering of the data.

3 M. Voigt, M. Sokolowski / Materials Science and Engineering Bxxx (2004) xxx xxx 3 deionized H 2 O. ITO was chosen, because of its optical transparency. Typically the substrate was water cooled and held at temperatures between 300 and 308 K during the sputtering. The Al 2 O 3 film thickness (d = 160 nm) was monitored by a quartz microbalance, calibrated by additional measurements with a surface profilometer. Finally, eight Au top contacts (area: mm 2, thickness nm) were deposited on top of the Al 2 O 3 films in the same vacuum chamber using a thermal Au evaporation source and a shadow mask. I V measurements were performed in a He cryostat under high vacuum using a Keithley 485 picoamperemeter. The voltage increase was V s 1. E was calculated from the applied voltage U as E = U/d. All reported I V curves are original data, without any smoothing or averaging over several scans. The I V curves were independent of the polarity. For the determination of the dielectric constant from the capacitance, impedance spectra were recorded by a Schlumberger impedance analyser for f = Hz. Atomic force microscopy (AFM) images were obtained at ambient conditions. 3. Results and discussion 3.1. Variation of the sputter conditions Fig. 1 displays I V curves which were measured for Al 2 O 3 films prepared under four different sputter conditions (a) (d). Irrespectively of the preparation conditions, we observe significant burn-in effects at field above ca. 0.5 MV cm 1, which cause a systematic decrease in the current during each voltage ramp. We speculate that the effect is partially related to the burn out of metallic shorts due to nanopores in the Al 2 O 3 films. We also find significant variations of the I V curves (up to several orders of magnitude in current) measured for different contacts on the same Al 2 O 3 film (see, e.g., Fig. 1(c)). We think that this is due to a local variation of the Al 2 O 3 film quality, including the local thickness variations, in combination with the statistical occurrence of the dielectric breakdown [9]. If the sputtering was performed in pure Ar (Fig. 1(a)), the electrical film quality was very bad, and high, approximately ohmic, currents were observed already at low fields. After sputtering, the Al 2 O 3 target exhibited a grey colour, likely due to the enrichment of Al. This result is in agreement with the finding of Segda et al. [12], who observed that sputtering in pure Ar, yields Al 2 O 3 films which are under-stoichiometric in oxygen and thus exhibit lower electrical quality. Better films could be prepared, if 10% (Fig. 1(b)) or1% of O 2 (Fig. 1(c)) was added to the sputter gas. In this case the Al 2 O 3 sputter target maintains its white colour during the sputter process. For these films the dielectric breakdown can be clearly observed as a sharp increase of the current by about four to five orders in magnitude at a critical breakdown field (E b ) (see Fig. 1(b) and (c)), leading to an irreversible change of the I V curve (not illustrated). As can be seen from Fig. 1(b) and (c), the reduction of the O 2 admixture from 10 to 1%, leads to an increase in E b by about a factor of 3, i.e. from about MV cm 1. After the breakdown, we could observe small spots (diameter about m) on the Au contacts, where the metal was evaporated from the sample, likely due to a local heating at breakdown channels. Films of even higher electrical quality could be fabricated, if the sputter target was sputtered in an Ar:O 2 mixture prior to the sputtering of the film ( conditioning of the target ), and the sputtering of the film was then performed in pure Ar (see Table 1). For such films, a step-like increase of the current, indicating the dielectric breakdown (as seen, e.g., in Fig. 1(b)), could not be observed. Instead, the current more gradually increased with the electric field (see Fig. 1(d)). A typical nominal resistivity (ρ = E/j) ate = 2.5MVcm 1 and room temperature is cm, which is in the range observed by others [3]. The reason, why the conditioning of the target is better than the admixture of O 2, is still under investigation. A very plausible mechanism has been suggested in by Schneider et al. [2], namely that activated O species from the O 2 containing sputter gas react with residual water from the rest gas and form OH ions. These then lead to the formation of aluminium hydroxide in the film and thus to a lower film quality [2]. Presumably, the OH formation is much smaller, Fig. 2. Current density vs. electric field strength curves of one Al 2 O 3 film (sample I of Table 1) on a log log scale at different temperatures. The curves were measured on different contacts starting at low temperatures.

4 4 M. Voigt, M. Sokolowski / Materials Science and Engineering Bxxx (2004) xxx xxx if pure Ar is used as the sputter gas. Thus conditioning of the target in Ar:O 2 mixtures prior to the sputtering appears to be a good compromise which allows to achieve high stoichiometries of oxygen to aluminium in the film, but avoids the detrimental formation of aluminium hydroxide. In addition to the variation of the sputter gas composition, we have also tested the influence of lower substrate temperatures (sample E and H, Table 1), smaller base pressures (samples H I), and smaller sputter powers (samples G and I). However, for all these conditions we find I V curves which are comparable with those of sample A (Fig. 1(d)). We thus conclude that within the accessible parameter space, we have reached an optimum, and that further optimisation to higher values of E b or ρ will require other strategies, e.g. reduction of the film thickness [3,9] or using ultra high vacuum conditions. Very recently Lee et al. [5] have reported I V curves for Al 2 O 3 films on ITO-glass which were prepared by the same technique we used. Within the variation of the sputter conditions tested by Lee et al., their best films are comparable with ours. However, details of the I V curves and of the best sputter parameters differ from those we found. This may possibly be due to small differences in the sputter con- ditions, in particular the preparation of the sputter target. It points to the importance of a very accurate control of these parameters for reproducible results Dielectric constant and temperature dependence of the I V curves According to impedance measurements (not shown) the dielectric constants ε of our Al 2 O 3 films were typically 7.0 ± 0.2 with no systematic variations with the sputter conditions. The range of reported ε values is 7 10 [5,6,8,12,13]. Our low ε values could be related to the low densities of the amorphous films, as suggested in [13]. Fig. 2 shows I V curves on a double log plot at different temperatures for a film of good electrical quality (I of Table 1). We note that these curves also exhibit strong variations for the different contacts and different I V measurements. For clarity, Fig. 2 shows only some typical curves. Principally, the I V curves were first taken at 4 K, and than at 100 and 300 K. As one can see from Fig. 2, there is a considerable scattering in the I V curve for 300 K above 0.25 MV cm 1, which points to the statistical character of the charge transport across the films. Ref. [7] discusses Poole Frenkel emission, field ionisation Fig. 3. AFM image of a thin magnetron sputtered alumina film (sample C of Table 1) for three different magnifications: 1000 nm 1000 nm (a) and (b), 5 m 5 m (c), and 10 m 10 m (d).

5 M. Voigt, M. Sokolowski / Materials Science and Engineering Bxxx (2004) xxx xxx 5 and trap hopping in this context. From about 0.25 MV cm 1 onward, the current increases with high exponents of the order of 10. However, this increase of the current is not an electrical breakdown, because the curves are still reversible, allowing to measure several comparable curves in sequence. As expected for temperature assisted processes [7], the currents at 4 K are considerably smaller compared to 300 K by a factor of up to six orders in magnitude (see Fig. 2) AFM investigation of the surface morphology Fig. 3 displays AFM images of a sputtered Al 2 O 3 film. One can clearly observe a rough surface with small grains of about m diameter. The rms surface roughness of the oxide films is about 4 nm. This type of surface morphology was observed for the samples A to D and F, H, I of Table 1, irrespectively of the preparation conditions. Possibly, it is due to the roughness of the underlying ITO surface, which was measured to be the same. Evidently this surface morphology needs to be improved considerably, if the Al 2 O 3 films should be used as substrates for further growth of films. 4. Conclusions Al 2 O 3 films with the smallest leakage currents (nominal resistance about cm at 2.5 MV cm 1 ) and highest dielectric break down fields were obtained, if the films were grown in pure Ar gas after the sputter target has been conditioned in an Ar:O 2 mixture. This preparation was more successful than sputtering in pure Ar or Ar:O 2 mixtures. For reproducible results, the preparation of the target prior to sputtering turned out to be essential. Acknowledgements This work was supported by the Deutsche Forschungsgemeinschaft. We thank L. Knoth, A. Schmidt, M. Schneider, S. Schmitt and E. Umbach for performing the AFM measurements, and D. Gauer for technical assistance. References [1] M.K. Olsson, K. Macák, U. Helmersson, B. Hjörvarsson, J. Vac. Technol. A 16 (1998) 639. [2] J.M. Schneider, A. Anders, B. Hjörvarsson, I. Petrov, K. Macák, U. Helmersson, J.-E. Sundgren, Appl. Phys. Lett. 74 (1999) 200. [3] Q. Li, Y.-H. Yu, C.S. Bhatia, L.D. Marks, S.C. Lee, Y.W. Chung, J. Vac. Technol. A 18 (2000) [4] W.H. Ha, M.H. Choo, S. Im, J. Noncryst. Solids 303 (2002) 78. [5] J. Lee, S.S. Kim, S. Im, J. Vac. Sci. Technol. B 21 (2003) 953. [6] C.A.T. Salama, J. Electrochem. Soc.: Solid State Sci. 117 (1970) 913. [7] C.A.T. Salama, J. Electrochem. Soc.: Solid State Sci. 118 (1971) [8] R.S. Nowicki, J. Vac. Technol. A 14 (1977) 127. [9] K. Kristiansen, Vacuum 27 (1977) 227. [10] T.A. Mäntyla, P.J.M. Vuoristo, A.K. Telama, P.O. Kettunen, Thin Solid Films 126 (1985) 43. [11] C.S. Bhatia, G. Guthmiller, A.M. Spool, J. Vac. Technol. A 7 (1989) [12] B.G. Segda, M. Jacquet, J.P. Besse, Vacuum 62 (2001) 27. [13] M.D. Groner, J.W. Elam, F.H. Fabreguette, S.M. George, Thin Solid Films 413 (2002) 186.