Vacuum 68 (2003) 31–38
TiO2 films prepared by DC magnetron sputtering from ceramic targets Henryk Tomaszewskia,*, Hilde Poelmanb, Diederik Deplab, Dirk Poelmanb, Roger De Gryseb, Lucien Fiermansb, Marie-Franc-oise Reyniersc, Geraldine Heynderickxc, Guy B. Marinc a
b
Institute of Electronic Materials Technology (ITME), Wolczynska 133, 01-919 Warsaw, Poland Department of Solid State Sciences, Surface Physics Division, Ghent University, Krijgslaan 281 S1, B-9000 Gent, Belgium c Department of Chemical Engineering, Laboratory for Petrochemical Techniques, Ghent University, Krijgslaan 281 S5, B-9000 Gent, Belgium Received 8 February 2002; received in revised form 1 April 2002; accepted 1 April 2002
Abstract The deposition of stoichiometric TiO2 films for V2O5/TiO2 anatase catalysts was investigated. DC magnetron sputtering from ceramic oxide targets in an argon/oxygen atmosphere was chosen as deposition technique. The target behaviour upon sputtering was followed by means of the target voltage. The layers were prepared with different oxygen concentrations in the plasma and the influence upon electronic and optical properties of the deposited layers was investigated. Surface XPS measurements showed that stoichiometry is obtained for low oxygen mole fractions, while for higher oxygen contributions, Ti3+ species are increasingly present. This departure from stoichiometry is ascribed to active resputtering of the deposited layer by O ions. Due to the higher sputter yield of oxygen as compared to titanium, this resputtering resulted in the observed reduction. From optical transmission measurements, the refractive index was found to decrease with increasing oxygen mole fraction, while the band gap increased considerably. The variation of the optical absorption with the oxygen concentration in the plasma was found to confirm the XPS analysis regarding the stoichiometry of the layers. Polycrystalline anatase layers were obtained at elevated substrate temperatures. r 2002 Elsevier Science Ltd. All rights reserved. Keywords: Sputter deposition; Titanium oxide; X-ray photoelectron spectroscopy; Visible/ultraviolet absorption spectroscopy; Anatase layers
1. Introduction TiO2 films are since long highly desired for their optical properties. The high refractive index of TiO2 and its excellent transparency in the visible *Corresponding author. Tel.: +48-22-83530441/472; fax: +48-22-834-9003. E-mail address: tomasz
[email protected],
[email protected]. edu.pl (H. Tomaszewski).
range make it very appropriate as compound in dielectric interference filters. Specific multilayer combinations with e.g. SiO2 or Si3N4 turn out to be very efficient antireflective coatings [1]. Chemical applications involve use in photocatalysis, photochemical solar cells [2] and as a gas sensor, e.g. for the detection of NO2 [3]. Quite recently, it was found that titanium compounds are biocompatible, making them attractive as implant material [4].
0042-207X/03/$ - see front matter r 2002 Elsevier Science Ltd. All rights reserved. PII: S 0 0 4 2 - 2 0 7 X ( 0 2 ) 0 0 2 7 9 - 8
32
H. Tomaszewski et al. / Vacuum 68 (2003) 31–38
TiO2 also plays an important role in catalytic processes. The system V2O5/TiO2 anatase is widely used as redox catalyst in partial oxidation reactions for hydrocarbons and de-NOx reactions. In this system the anatase state of the substrate is of importance, since it enhances the activity and selectivity of the catalyst system [5]. In the framework of the development of V2O5/ TiO2 anatase catalysts, the present paper reports on the sputter deposition of nearly stoichiometric TiO2 layers. DC magnetron sputtering from ceramic targets was chosen as deposition technique, because the sputter process can be readily controlled and the deposited layers show good adhesion as well as good coating uniformity. Oxide targets can indeed be used in DC operated magnetrons, provided they are reasonably electrically conducting. Operation with a metal target implies several drawbacks in the reactive mode, such as process instability and low deposition rates [6]. Transparent, i.e. stoichiometric layers of TiO2 are only obtained when sputtering in the oxide mode, i.e. with a high O2 partial pressure. In this mode the target surface is highly oxidized, giving rise to arcing while within the race track a modified surface layer is created, which leads to a lower sputter yield and a correspondingly lower deposition rate. Higher sputter rates can be obtained when operating in the transition mode, closer to the metallic mode (i.e. low O2 partial pressure), but this is a highly unstable operation zone which needs continuous feed back control, e.g. by means of a plasma emission monitor (PEM) and a controllable and fast reacting valve for O2 introduction. Moreover, in this regime and in order to obtain stoichiometric layers, the substrate-to-target distance has often to be increased, so that Ti and O or O2 particles, either neutral or ionised, reach the substrate in the ratio necessary for stoichiometric deposition [7]. In other words, a substrate-totarget distance such that the sputtered and deposited Ti has the time to react with the oxygen available at the substrate. All of these measures are technically complicated and make it worthwhile exploring stoichiometric deposition from oxide targets.
2. Experimental Titanium oxide targets were prepared from TiO2 rutile powder (type 3025, Kronos, Germany) following a standard ceramic preparation procedure, e.g. cold pressing and sintering at 17001C for 1 h in vacuum (1.3 104 Pa). This preparation was done in the Institute of Electronic Materials Technology (IEMT, Warsaw). Sintering in vacuum of typically white titania powder resulted in black targets, which proved their nonstoichiometry. Indeed, Rutherford Backscattering Spectrometry (RBS) measurements showed that the composition of the titanium oxide target corresponded with TiO1.75 instead of TiO2. RBS measurements were taken with a 2 MeV He+4 beam produced by the Van de Graaff accelerator of the IEMT (Warsaw). XRD revealed that the only phase present in the targets is rutile. Their density was 3.87 g/cm3, which is 91% of the single crystal density of rutile (4.25 g/cm3 [8]). Traces of Mg, Al, Si and Ca were present according to an EDX (Energy Dispersive X-ray Analysis) measurement. Van Der Pauw type measurements yielded a resistivity of 0.7 O cm, which proved to be low enough for the material to be used as oxide target in a DC magnetron sputtering system. Depositions were performed from 6 mm thick ceramic targets with a diameter of 50 mm onto unheated glass substrates. The substrate-target distance was 100 mm. The magnetron-sputtering source was designed in the lab and was operated by means of a Huttinger . power supply (PFG 1500 DC). Before deposition the chamber was pumped down by a rotation pump (Balzers DUO100) and diffusion pump (Balzers DIF320, BFA 320 W baffle) to a base pressure of 1 103 Pa. The pressure was measured by Pirani and Penning pressure gauges. Sputtering gases Ar and O2 were introduced through stainless steel tubes at the top of the system and standard mass flow controllers (AERA ROD4) controlled both flows. The input power was set at 100 W. The power density at the ceramic target determined from the size of the race track was 7 W/cm2. Each reactive sputter deposition was preceded by target sputter cleaning in pure argon.
H. Tomaszewski et al. / Vacuum 68 (2003) 31–38
3. Results and discussion 3.1. Sputter process
Deposition rate, A/min
increasing argon flow (Fig. 1). As the chamber pressure varied linearly with the argon flow, the observed decrease in deposition rate can be interpreted as follows. Increasing the argon sputter pressure results in a smaller mean free path of the sputtered species and consequently they are more scattered. However, other effects can also influence the deposition rate. Although, no direct experimental evidence can be given for these effects, we believe that their influence cannot be excluded. Increasing the argon pressure results in more volume ionisation. This will reduce, at constant power, the target voltage, i.e. the ion energy and will increase the total current, or stating differently more ions will bombard the target surface. As shown in Fig. 2, over the argon flow range covered, the target potential is seen to decrease by about 100 V. However, in this ion energy range, the influence of the ion energy on the sputter yield is small. Therefore, we expect that the higher flux of bombarding ions will tend to increase the deposition rate. Another effect which can influence the deposition rate is the reduction of the target due to argon ion bombardment. Indeed, our XPS measurements and the results of Zhang et al. [11] show that the surface of a TiO2x target is reduced by argon ion bombardment due to preferential oxygen sputtering and the bombarded surface
80 60 40 20 0 10
20 30 40 Argon flow rate, sccm
10
20
50
8 Sputtering press, x10 -2 mbar
The stoichiometry of the layers was checked through ex situ XPS measurements (Perkin-Elmer PHI ESCA 5500 system) by evaluation of the Ti3+/Ti4+ ratio. No additional cleaning or sputtering was done prior to XPS characterisation in order not to disturb the layer. The photoemission intensities of C1s, Ti2p, Si2p and O1s coreelectrons were recorded for the as deposited layers and binding energies are reported with reference to the C1s line at 284.6 eV. Si was essentially present as SiO (Si2+ binding energy 101.9 eV) [9]. The Ti2p3/2 lineshape was examined by deconvolution into its components. In literature, a binding energy of 458.5 eV is reported for Ti4+ in sputtered TiO2 layers, while the reduced species Ti3+ has a binding energy smaller by a value between 1.3 and 1.7 eV as compared with Ti4+ [10]. In order to establish the fitting parameters for Ti3+ and Ti4+ photoemission components, the Ti2p3/2 line of a clearly reduced layer was decomposed. This curve fitting resulted in a binding energy difference of 1.6 eV, value that was kept constant for all decompositions. Based on the Ti3+/Ti4+ ratio, the oxygen deficiency d for the deposited TiO2x layers was calculated. The layers were characterised optically by means of specular transmission measurements (Varian Cary 5), from which absorption coefficient a, refractive index and band gap were determined. After characterisation of the layers as deposited, annealing was performed in an O2 flow at atmospheric pressure for 30 min at 5001C and its influence upon the layers’ properties was determined. Finally, deposition experiments were performed at increasing substrate temperatures (200–5001C). For this purpose the substrate holder was equipped with a resistive heater. The crystallinity of the layers was checked by X-ray diffraction (Siemens diffractometer Kristalloflex D5000).
33
6 4 2 0 30
40
50
Argon flow rate, sccm
During sputtering of TiO2 in a pure argon atmosphere, the deposition rate decreases with
Fig. 1. Deposition rate of TiO2 layers and sputtering pressure as a function of argon flow rate.
Sputtering voltage, V
650
550
450
10
20 30 40 Argon flow rate, sccm
50
Sputtering voltage, V
H. Tomaszewski et al. / Vacuum 68 (2003) 31–38
34
600 500 ∆V=26V 400 300 0
Fig. 2. Sputtering voltage as a function of argon flow rate.
0.1 0.2 Oxygen mole fraction
0.3
Deposition rate, A/min
Fig. 4. Sputtering voltage as a function of oxygen concentration in plasma at a constant argon flow rate (13.6 sccm).
80 60 40 20 0 0
0.1 0.2 Oxygen mole fraction
0.3
Fig. 3. Deposition rate of TiO2 layers as a function of oxygen concentration in plasma at a constant argon flow rate (13.6 sccm).
changes to metal rich comparing to the original surface. The Ti 2p3/2 peak becomes asymmetric and there are contributions at lower binding energies, which show the presence of Ti cations with lower oxidation state. In the case of a very high degree of reduction even Ti1 can be observed. The effect of this target reduction on the deposition rate is not easily determined, because it affects several parameters. At this stage of the investigation, it is concluded that the different effects, initiated by argon flow increase, add up to a general decrease of the deposition rate. When oxygen is added to the sputtering atmosphere instead of increasing the argon pressure, the deposition rate is also seen to decrease continu( ( ously from 70 A/min down to 13 A/min when the oxygen flow rate is increased (Fig. 3). However, the effect is much stronger than the argon pressure effect discussed above. This indicates that another mechanism is also important. We can also confirm this statement from the influence of oxygen addition on the target voltage (Fig. 4). In contrast
to the pressure increase, which leads to a target voltage decrease, the target voltage on oxygen addition first increases from 544 V (0 sccm O2) to 572 V for 1.0 sccm oxygen flow. Further increasing the oxygen flow rate leads to a significant decrease of the sputtering voltage. This latter decreasing part is probably related to an increasing sputtering pressure, combined with an oxidation of the target surface, which will enhance the ion induced secondary electron emission (ISEE) from the target. The oxidation of the target surface can also explain the strong effect of oxygen addition on the deposition rate. Indeed, in contrast to pure argon sputtering, for which the target surface will be reduced to metal rich, the target surface in argon/oxygen plasma will be more oxidized. It is well known that the sputter yield of compounds (with a fully oxidized surface) is generally lower than the sputter yield of metals (or metal rich target surfaces). The origin of the increasing part of the voltage pattern however has to date remained unexplained. Surely, what plays here is not a pressure effect, because this has an opposite behaviour. Also, oxidation of the target is not at play. One certainly starts with a somewhat reduced TiO2x target surface due to pre-sputtering in pure argon. However, oxidation would reduce the target potential because of the higher ISEE of oxides. On the other hand, chemisorption is known to influence the ISEE in a different way than true oxide formation [12]. Independent of the oxygen content of the Ar-O2 sputter gas, the deposited films are found to be amorphous. Heating the substrate during deposition changes the X-ray characteristics of the
H. Tomaszewski et al. / Vacuum 68 (2003) 31–38
titanium oxide layers. X-ray diagrams (Fig. 5) show that at substrate temperatures higher than 2001C, TiO2 is increasingly present in anatase form.
[001] [200]
[211] [204]
XRD intensity, arb.units
[105]
500°C
400°C
300°C
200°C
35
40
45
50
55
60
65
70
75
80
2theta, degrees
Fig. 5. X-ray diagrams of TiO2 layers deposited at elevated substrate temperatures.
35
3.2. Electronic properties In Fig. 6 oxygen deficiency versus oxygen mole fraction is presented. At low oxygen mole fraction (0.03–0.07) the deposited layers were almost stoichiometric, while the layers deposited without or with more oxygen added to the plasma, displayed a substantial Ti3+ component in the photoemission intensity, interpreted in terms of departure from stoichiometry. The observed minimum in the oxygen deficiency can be readily interpreted when looking at the deposition rate for different O2 mole fractions. Indeed, the oxygen content of the deposited layer depends on the arrival rate of the sputtered material (Ti, Ti-O fragments) and their reaction on the substrate with oxygen species present in the chamber. As shown in Fig. 3, the deposition rate dropped when the oxygen mole fraction was increased. At low deposition rate, the oxidation of the deposited layer is favoured, because less fragments are arriving while more oxygen is present for the oxidation reaction [13]. Comparison of Figs. 6 and 3 confirms the correlation between deposition rate and oxygen deficiency at low oxygen mole fraction.
0.10
Oxygen deficiency δ
0.08
0.06
0.04
0.02
0.00 0.00
0.10
0.20
0.30
Oxygen mole fraction
Fig. 6. Oxygen deficiency d versus the O2 mole fraction; full line—as deposited layers, dotted line—after O2 anneal.
H. Tomaszewski et al. / Vacuum 68 (2003) 31–38
However, further increasing the oxygen mole fraction yielded higher oxygen deficiency while the deposition rate remained constant. The origin for these observations must be sought in negative oxygen ion bombardment of the substrate. As shown by Zeuner et al. [14], electron attachment of oxygen atoms in the bulk plasma is not relevant in the context of negative ion bombardment. However, negative ions formed at the target surface during reactive sputtering are accelerated away from the cathode [13–15]. As shown by Tachibana et al. [16], O originated from the ceramic target is less important for resputtering and only the plasma oxygen can get ionized and hence give rise to substrate resputtering. Moreover, as it was found by Tachibana et al. [16], Ti and O-related species sputtered from TiO2x target reach the substrate with rather low energies. However oxygen must be supplied from the sputter gas for filling up the oxygen lack in order to obtain stoichiometric TiO2 films. Part of the O2 sputter gas is decomposed and changed to O ions in the plasma and these are accelerated by the electric field around the target. Upon reaching the deposited film, the negative ions cause damage to the layer because of the higher penetration depth. Substrate bombardment by these negative ions will reduce the oxygen content due to preferential sputtering of oxygen atoms from the deposited layer [17]. As the oxygen content in the plasma increases, the target surface becomes more oxidised and more negative ions will be produced. Hence, the observed changes in film stoichiometry with the oxygen mole fraction in the plasma, is the result of two competing effects. On the one hand, addition of oxygen to the chamber results in a decrease of the deposition rate and an enhanced oxidation of the deposited layer. On the other hand, at higher oxygen concentration in the plasma, negative oxygen ion bombardment will preferentially sputter the oxygen atoms from the deposited layer. The annealed samples were also characterised by means of XPS. In general, the oxygen deficiency d was seen to decrease, especially for the higher O2 mole fractions (Fig. 6). Hence, additional oxygen was built into the layer by the annealing treatment so that stoichiometry of the layers was improved.
3.3. Optical properties As already mentioned, transmission measurements were used to calculate the absorption coefficient a; refractive index, n; extinction coefficient k and thickness d; using a classical oscillator fitting procedure [18] over an extended wavelength range (500–2000 nm). The absorption coefficient as a function of oxygen mole fraction is displayed in Fig. 7. Its pattern is quite similar to the one found for the oxygen deficiency of the layers (Fig. 6), showing a minimum absorption for a small amount of oxygen and a higher absorption for increasing oxygen mole fractions. Since the absorption coefficient of a layer relates directly to its stoichiometry, the observed pattern for a confirms the previous observations from the XPS line shape analysis (Section 3.2). The refractive index at l ¼ 550 nm was found to be quite low: 2.27 and lower (Fig. 8). This can be ascribed to the deposition technique used—DC magnetron sputtering—which produces particles with a relatively low energy, resulting in layers with a low packing density. Indeed, the refractive index is known to depend on the energy of the deposited particles and is lower for less energetic deposition methods [1]. For increasing O2 content in the plasma, the n value decreased considerably. This behaviour is again correlated with the deposition rate: from comparison of Figs. 3 and 8, it is obvious that the refractive index increases for higher deposition rates, i.e. for lower O2 mole fractions. Similar behaviour of n versus oxygen mole fraction for TiO2 layers has been found for
Absorption coeff. α, cm -1
36
700 650 600 550 500 450 400 0
0.1 0.2 Oxygen mole fraction
0.3
Fig. 7. Absorption coefficient a of TiO2 layers as a function of oxygen concentration in plasma at a constant argon flow rate (13.6 sccm).
H. Tomaszewski et al. / Vacuum 68 (2003) 31–38
37
2.30
Refractive index n
2.20
2.10
2.00
1.90 0.00
0.05
0.10
0.15
0.20
Oxygen mole fraction
Fig. 8. Refractive indexes versus the O2 mole fraction; full line—as deposited layers, dotted line—after O2 anneal.
Optical band gap, eV
3.60
3.40
3.20
3.00 0.00
0.10
0.20
0.30
Oxygen mole fraction
Fig. 9. Optical band gap versus the O2 mole fraction; full line—as deposited layers, dotted line—after O2 anneal.
mid-frequency sputtering using a double magnetron arrangement [19] and for reactive DC plasmatron sputtering from a Ti target [20]. The increase probably reflects a higher packing density
in the layer, which tends towards bulk values. The annealing treatment in oxygen appeared to lower further the refractive index. This points towards still lower packing densities, possibly resulting
38
H. Tomaszewski et al. / Vacuum 68 (2003) 31–38
from microstructural changes in the layers upon thermal treatment [21]. The absorption coefficient near the band edge was calculated and was found to follow closely the typical shape for an indirect allowed band gap: aBðE Eg Þ2 [22]. The width of the band gap was seen to increase with increasing oxygen mole fraction (Fig. 9): from about 3.2 eV at zero oxygen mole fraction to 3.5 eV at a mole fraction of about 0.3. These values can be compared with the ones for crystalline anatase and rutile: 3.2 and 3.05 eV [22]. It appears that the layers showing the best stoichiometry yielded a band gap value closest to the single crystal one for anatase. Annealing in oxygen, which caused the recrystallization of the layers into anatase, gave rise to a significant drop of the band gap.
4. Conclusions TiO2 thin films have been deposited by DC magnetron sputtering, using ceramic targets. It was found that nearly stoichiometric layers could be obtained with a small amount of oxygen in the sputtering gas, with reasonable deposition rate. Higher oxygen mole fractions caused reduction of the deposited layers through negative ion resputtering. Crystalline anatase layers were obtained when the depositions were made at elevated temperatures. Variations of refractive index were interpreted in terms of the packing density of the layers.
Acknowledgements The authors are grateful to Geert De Doncker and Ulric Demeter for performing the XPS measurements. One of us (H.T.) acknowledges support from an exchange project of the Flemish government (AWI).
This work was performed in the framework of a Concerted Research Action (GOA) financed by the Ghent University, and of the Belgian Programme on Interuniversity Poles of Attraction initiated by the Belgian State, Prime Minister’s Office, Science Policy Programming. The scientific responsibility is assumed by its authors.
References . M, Malkomes N, Staedler T, Matth!ee T, Richter [1] Vergohl U. Thin Solid Films 1999;351:42–7. [2] Graetzel M. Comments Inorg Chem 1991;12:93. [3] Ferroni M, Guidi V, Martinelli G, Nelli P, Sberveglieri G. Sensors Actuators B—Chemical 1997;44:499–502. [4] Hanawa T, Ota M. Appl Surf Sci 1992;55:263. [5] EUROCAT, Catal Today 1994;20:1–184. [6] Szczyrbowski J, Br.auer G, Ruske M, Teschner G, Zmelty A. J Non-Cryst Solids 1997;218:262–6. [7] Schiller S, Heisig U, Steinfelder K, Strumpfel . J. Thin Solid Films 1979;63:369–75. [8] Rao KVK, Nagender Naidu SV, Iyengar L. J Am Ceram Soc 1970;53:124. [9] Grunthaner FJ, Grunthaner PJ, Vasquez RP, Lewis BF, Maserjian J. Phys Rev Lett 1979;43:1683. [10] Sanjin"es R, Tang H, Berger H, Gozzo F, Margaritondo G, L!evy F. J Appl Phys 1994;75:2945. [11] Zhang Z, Henrich WE. Surf Sci 1992;277:263–72. [12] Kaminsky M. Atomic and ion impact phenomena on metal surfaces. Springer, Berlin, 1965. p. 14–6. [13] Aita K. Crit Rev Solid State Mater Sci 1998;23:205. [14] Zeuner M, Neumann H, Zalman J, Biederman H. J Appl Phys 1998;83:5083. [15] Tucek JC, Walton SG, Champion RL. Phys Rev B 1996;53:14127. [16] Tachibana Y, Ohsaki H, Hayashi A, Mitsui A, Hayashi Y. Vacuum 2000;59:836–43. [17] Kester DJ, Messier R. J Mater Res 1993;8:1928. [18] Nenkov M, Pencheva T. J Opt Soc Am 1998;A15:1852. [19] Br.auer G, Szczyrbowski J, Teschner G. Surf Coat Technol 1997;94–95:658–62. [20] Schiller S, Beister G, Sieber W. Thin Solid Films 1981;83:239–45. [21] Guenther KH. Appl Opt 1984;23:3612. [22] Tang H, Prasad K, Sanjin"es R, Schmid PE, L!evy F. J Appl Phys 1994;75:2042.