Chemical etching of Tungsten thin films for high ...

216 downloads 0 Views 1MB Size Report
May 12, 2016 - Chemical etching of Tungsten thin films for high-temperature surface acoustic wave-based sensor devices. M. Spindler a,⁎, S. Herold b, ...
Thin Solid Films 612 (2016) 322–326

Contents lists available at ScienceDirect

Thin Solid Films journal homepage: www.elsevier.com/locate/tsf

Chemical etching of Tungsten thin films for high-temperature surface acoustic wave-based sensor devices M. Spindler a,⁎, S. Herold b, J. Acker b, E. Brachmann a, S. Oswald a, S. Menzel a, G. Rane a a b

IFW Dresden, SAWLab Saxony, P.O. Box 270116, D-01171 Dresden, Germany BTU Cottbus – Senftenberg, Faculty of Sciences, P.O. Box 101548, 01968 Senftenberg, Germany

a r t i c l e

i n f o

Article history: Received 23 September 2015 Received in revised form 11 April 2016 Accepted 22 April 2016 Available online 12 May 2016 Keywords: SAW devices Tungsten electrodes Magnetron sputtering Wet-chemical etching Lift-off structuring

a b s t r a c t Surface acoustic wave devices are widely used as wireless sensors in different application fields. Recent developments aimed to utilize those devices as temperature sensors even in the high temperature range (T N 300 °C) and in harsh environmental conditions. Therefore, conventional materials, which are used for the substrate and for the interdigital transducer finger electrodes such as multilayers or alloys based on Al or Cu have to be exchanged by materials, which fulfill some important criteria regarding temperature related effects. Electron beam evaporation as a standard fabrication method is not well applicable for depositing high temperature stable electrode materials because of their very high melting points. Magnetron sputtering is an alternative deposition process but is also not applicable for lift-off structuring without any further improvement of the structuring process. Due to a relatively high Ar gas pressure of about 10− 1 Pa, the sidewalls of the photoresist line structures are also covered by the metallization, which subsequently prevents a successful lift-off process. In this study, we investigate the chemical etching of thin tungsten films as an intermediate step between magnetron sputtering deposition of thin tungsten finger electrodes and the lift-off process to remove sidewall covering for a successful patterning process of interdigital transducers. © 2016 Elsevier B.V. All rights reserved.

1. Introduction During the last two decades, the development of surface acoustic wave devices had seen a boost due to a major progress in radio frequency (rf) filters, wireless communication techniques and microfluidic research. However, the use of surface acoustic wave (SAW) devices at temperatures above 300 °C or under harsh environments require particular materials with a high thermal stability, reliability and lifetime for the piezoelectric substrate as well as for interdigital transducer (IDT). Current SAW-based high temperature sensors favourize langasite (LGS) as substrate material in combination with Platinum based electrode materials [1–4]. Investigations on the temperature stability of these Platinum based electrodes showed diffusion of Gallium from the substrate into the electrode [5] which can lead to phase formation. Tungsten seems to be an alternative compared to Platinum and in combination with diffusion or covering layers an appropriate metal for high temperature applications not only because of a higher melting point but also a reduced activation energy for self-diffusion and lower electrical resistivity ⁎ Corresponding author. E-mail address: [email protected] (M. Spindler).

http://dx.doi.org/10.1016/j.tsf.2016.04.035 0040-6090/© 2016 Elsevier B.V. All rights reserved.

[6,7]. Due to its high melting point, the deposition of tungsten by electron beam evaporation is very hard to control which seems to be a big disadvantage. Hence, magnetron sputtering is more suitable than electron beam evaporation although the process also involves difficulties especially regarding patterning of the tungsten films. The higher gas pressure during sputtering (≈ 10− 1 Pa compared to 10− 3 Pa for E-beam) implies an increased collision probability of particles that are moving towards the substrate. The resulting wide angular distribution of the particle drift velocity leads to a covering of the sidewalls of the photoresist lines (see Fig. 1). In this case, the removal of the subsequent lift-off process is not successful i.e. the metal lines are not fully structured. As it can be seen from Fig. 1b, the thickness “d” of this sidewall covering is usually smaller than that of the electrode height “h”. An intermediate etch process for instance by wet chemical etching can help to remove this sidewall covering, which should enable the following lift-off structuring step. Therefore, a proper etchant and well-known characteristics of the etching process have been evaluated to etch thin tungsten layers deposited by magnetron sputtering. The big advantage of this method is to combine sputtering technique with the low-cost lift-offprocess, which is still a standard method to achieve structured metal lines in SAW technology.

M. Spindler et al. / Thin Solid Films 612 (2016) 322–326

323

Fig. 1. Schematic demonstration of the effect of the sidewall deposition intensity of the photoresist mask by deposition of metal using a) electron beam evaporation and b) magnetron sputtering. The sidewall covering of the resist line avoids their removal by the solvent during the lift- off process.

Fig. 2. Schematic cross section view of the etching test samples to study the tungsten etching process. (a) before and (b) after chemical etching.

Table 1 Chemical composition of etching solutions (in deionized water).

2. Experimental methods

Notation

% (w/w) H2O2

% (w/w) NH3

W30 W21A0 W21A1 W21A7 W10

30.0 21.3 21.3 21.3 10.0

0.0 0.0 1.0 7.3 0.0

The following test structure was designed to determine tungstenetching rates under different conditions: Tungsten films with a thickness of 350 nm have been deposited onto Si (100) wafers with a 1 μm thick SiO2 film by magnetron sputtering under a process pressure of 1.7 ⋅ 10−4 Pa and an Ar flow rate of 30 sccm. The wafer was notched into square shaped samples each with dimensions of 10 × 10 mm2.

Fig. 3. Etching rates for 350 nm thick tungsten layers on Si substrates using different compositions of the etching solution. (a) H2O2–based etchant without ammonia (b) 21.3% (w/w) H2O2–based etchant with varying amount of NH3.

324

M. Spindler et al. / Thin Solid Films 612 (2016) 322–326

Fig. 4. XPS spectra of a sputtered tungsten layer on Si substrate, (a) as deposited where the spectrum is dominated by metallic W, (b) after 90 s etching in 21% (w/w) hydrogen peroxide solution. Here, the W+4and the W+6 signal indicates the oxidation of metallic W to WO2 and WO−2 4 , respectively [3–5].

Then, a photoresist line with a thickness of 380 μm was printed onto the sample surface by using a Nordson EFD Dispenser. This photoresist line allows the determination of the etching depth by using a mechanical step-profilometer (see Fig. 2). Therefore, the samples were treated at 80 °C for 1 h under normal pressure in air to bake the photo resist. Then they were etched in five different acid solutions based either on a variable proportion of hydrogen peroxide [8] or ammonia at a fixed hydrogen peroxide (H2O2) concentration of 21.3% (w/w) (see Table 1). In order to find the most appropriate etchant, the amount of hydrogen peroxide and ammonia has been varied between 0% and 7.5% (w/w). Three samples were etched by hand in each etching solution under stirring for a certain time (up to 300 s). Subsequently, within b 1 s they were rinsed immediately using deionized water, followed by a lift-off process in an ethanol–acetone mixture to remove the photoresist. The etch depths were obtained by using a mechanical surface profiler Alpha Step 500. The Scan velocity was set to 20 μm/s with a scan length of 500 μm and a sampling rate of 200 Hz. In addition, the samples were analysed by X-ray photo Electron Spectroscopy (XPS) to analyse the chemical mechanism of the etching process. Focused Ion Beam (FIB) method was used to obtain cross section SEM pictures from the photoresist and the electrodes. A Platinum protection bar has been deposited on a photoresist edge-structure to prevent the region of interest (ROI) on the sample surface from Ga ion implantation. Finally, finger electrodes were investigated by atomic force microscopy (AFM) to compare roughness before and after wet chemical etching. After the lift-off process, the electrical response (S11 scattering parameter) of the tungsten IDT on Lithium niobate has been measured by network analyzer.

3. Results and discussion 3.1. Etching rates as a function of hydrogen peroxide content Fig. 3a reveals that etching of tungsten layers on Si substrate in H2O2 etch solutions of given concentration proceeds at a constant etch rate whereas that the etching rate increases with a higher H2O2 concentration of the etchant. At the beginning of the etching process the reaction shows a typical induction period with initial etching rates considerably lower compared to the etch rate observed at bigger depths. This is the reason why the approximated linear function does not meet the origin of ordinates. This observation is explained by the XPS spectra in Fig. 4 which compare the tungsten surface before and after the etch attack. The spectrum of the as deposited W layer is dominated by metallic W (see Fig. 4a). The deconvolution of the spectrum reveals the presence of additional peaks which are assigned to W4+ representing a passivating layer with a thickness of only some angstrom on top of the metallic tungsten consisting of tungsten(IV)-oxide, WO2. The Peak assignment is based on the work by Shankoff et al. [9], Warren et al. [10], and Perry et al. [11]. Etching with H2O2 leads to the formation of a relatively thick oxidation layer on top of the tungsten metal. As shown in Fig. 4b the deconvoluted XPS spectra is now dominated by signals that can be assigned as W6+, representing tungsten(VI)-oxide, WO3, as well as by the signals of W4+ (WO2). The XPS spectra reveal that the dissolution of tungsten in H2O2 solution proceeds via the formation of a WO3 layer on top of the metallic tungsten and WO2 seems to play the role of an intermediate step of W oxidation process. WO3 is insoluble in water, however, slowly dissolved in the slightly acidic H2O2 etch solutions under formation of tungstate ions, WO2− [12–14] (see Fig. 5). In 4

Fig. 5. Model of the dissolution of thin W films in H2O2 and H2O2–NH3 solutions.

M. Spindler et al. / Thin Solid Films 612 (2016) 322–326 Table 2 Etching rates for 350 nm thick tungsten layers on Si substrate. Concentration in % (w/w) H2O2

NH3

Etching rate in nm/min

Delay time in min

10 21,3 21,3 21,3 30

0 0 1 7 0

17,3 ± 0,6 24,9 ± 0,6 60,0 ± 1,9 48,1 ± 1,8 25,3 ± 0,3

0,80 0,77 0,63 0,77 0,47

this pH range the dissolution proceeds via a solid hydrated WO3 phase which rate is limited by the OH– concentration [12–14]. The formation of the initial WO3 layer might explain the observed induction period of a low etch rate at the beginning of the reaction. It can also be seen from Fig. 3a that the etching rate does not change significantly in case of concentrations between 20% and 30% H2O2 (w/w). The reason is not clarified yet. It is reasonable to assume that the etching is limited either by the growth rate of the WO3 layer or by its dissolution rate. This assumption is supported by an experimentally determined activation energy of 57 kJ mol− 1, that excludes a limitation of the reaction by mass transport in the aqueous solution. In case of long etching times (N180 s), we observe a bigger deviation of the etching depth, which can be explained by a bigger roughness of the sample surface. 3.2. Effect of ammonia content in hydrogen peroxide based etching solution The influence of ammonia in a hydrogen peroxide based etching solution on the etching rate is depicted in Fig. 3b. Here, a different etching behavior compared to the pure hydrogen peroxide based etching solutions is observed. Most significantly, the etching rate increases by increasing the ammonia concentration that varies from 0% (w/w) to 1% (w/w) (see Table 2). The addition of ammonia leads to a significantly higher reaction rate. The presence of NH3 increases the pH to values between 9 and 10 that turns the dissolution of WO3 via the hydrated WO3 phase into a pH

325

independent regime as stated out by Anik and co-workers [6–8]. Consequently, the passivating WO3 layer is removed faster from the surface (Scheme 1). Furthermore, ammonium tungstate has a quite high solubility in water [15]. On the contrary, a further increase of the ammonia concentration leads to an accelerated decay of hydrogen peroxide as previously described by Zambova et al. [16], which was indicated by an enhanced formation of gas bubbles inside the stock solution due to the decomposition of H2O2 into oxygen and water This leads to the consequence, that the etching rate decreases with concentrations of ammonia higher than 1% (w/w) in a hydrogen peroxide based solution.

3.3. Etching of tungsten finger electrodes After verification of the etching rates, an appropriate etching solution consisting of 10% (w/w) H2O2 was chosen for etching of sputtered tungsten layers on Lithium niobate (LiNbO3) substrates to form IDT finger electrode structures (for λ/4 fingers) with an electrode width of 15 μm and a pitch of 30 μm i.e. for a wavelength of 60 μm. This etchant had the lowest etching rate from all solutions investigated in our experiments. The photoresist mask had a film thickness of approximately 700 nm. The tungsten films with a thickness of 350 nm were deposited using magnetron sputtering, which lead to a covering tungsten layer thickness at the sidewalls of the photoresist lines of approximately 200 nm and at the thinnest point of approximately 70 nm (see Fig. 6a). According to the etching rate determination the samples were etched for 200 s. The electrode edges have an increased roughness, resulting by lift-off artefacts (see Fig. 6c). For the IDTs which were used in this study, this roughness is neglectable compared to the electrode's width (see Fig. 6b, d). Further AFM investigations show the difference in surface roughness between the as deposited and the etched electrodes after the lift-off process (see Fig. 7a, b). The roughness was measured at the IDT contact pads. The wet chemical etching leads to a slightly increased RMS (root mean square) value up to 3.7 nm, which is small compared to electrode dimensions. The measurement of the electrical reflexion coefficient shows a distinctive minimum at

Fig. 6. FIB/SEM images of tungsten finger electrodes on lithium niobate, (a) cross-section view of a photoresist line after sputtering of tungsten, (b) tungsten electrode after etching for 200 s in 10% (w/w) hydrogen peroxide solution and lift-off structuring process, (c) lift-off artefacts at the electrode edges, (d) details of tungsten electrodes as a part of an IDT (λ = 60 μm).

326

M. Spindler et al. / Thin Solid Films 612 (2016) 322–326

Fig. 7. AFM measurements on of tungsten films on 128°YX Lithium niobate. The measured area has a size of 5 × 5 μm2. (a) as deposited state. RMS = 2 nm, (b) after 200 s of wet chemical etching, RMS = 3.7 nm.

the resonance frequency at about 64 MHz, which indicates an acceptable quality of the tungsten IDT (see Fig. 8).

4. Conclusion The etching of 350 nm magnetron sputtered thin tungsten layers to fabricate IDT electrodes structures for high temperature surface acoustic wave devices using the lift-off patterning method in combination with wet-chemical etching has been investigated. Different chemical etching solutions based on hydrogen peroxide but also with an additional content of ammonia were used. The obtained etching rates significantly depend on the concentration of hydrogen peroxide and ammonia. Finally, an etching solution with 10% (w/w) hydrogen peroxide has been selected for removing the Tungsten layer at the sidewall to apply lift-off process. This solution provides a low etching rate, which is also needed to find the etch stop. The Tungsten, which was also deposited at the sidewalls of the photoresist lines has been removed successfully by wet chemical etching. The surface roughness of the electrodes was determined to values of approximately RMS ≈ 4 nm. In contrast to that, the lift-off process incurs artefacts at the finger edges which also increase the roughness here. Nevertheless, we could show that the electrical behaviour of the IDT is quite acceptable when we use individual electrode widths of some μm. If the height of the electrodes is bigger than the nominal thickness of the sidewall covering, this process can generally be used as a structuring method of refractory IDT metallizations.

Fig. 8. Electrical measurement of the reflexion coefficient after lift-off process.

Acknowledgements The authors wish to thank A. Winkler, S. Kaschube and T. Wiek for their assistance in sample preparing and analytics. We gratefully acknowledge the funding by the German Federal Ministry of Education and Research (BMBF), project InnoProfile-Transfer grant 03IPT610Y. References [1] B. François, D. Richter, H. Fritze, Z.J. Davis, C. Droit, B. Guichardaz, V. Pétrini, G. Martin, J.-M. Friedt, S. Ballandras, Wireless and passive sensors for high temperature measurements, Procedia Eng. 47 (2012) 1227–1230. [2] S. Sakharov, A. Zabelin, S. Kondratiev, D. Richter, H. Fritze, D. Roshchupkin, A. Shvetsov, S. Zhgoon, Optimization of wafer orientation and electrode materials for LGS high-temperature SAW sensors, Ultrasonics Symposium (IUS), 2012 IEEE International 2012, pp. 1525–1528. [3] M. Schulz, E. Mayer, I. Shrena, D. Eisele, M. Schmitt, L.M. Reindl, H. Fritze, Correlation of BAW and SAW properties of langasite at elevated temperatures, J. Sens. Sens. Syst. 4 (2015) 331–340. [4] J.A. Thiele, M.P. Cunha, Platinum and palladium high-temperature transducers on langasite, IEEE Trans. Ultrason. Ferroelectr. Freq. Control 52 (2005) 545–549. [5] S.C. Moulzolf, D.J. Frankel, M. Pereira da Cunha, R.J. Lad, High temperature stability of electrically conductive Pt–Rh/ZrO2 and Pt–Rh/HfO2 nanocomposite thin film electrodes, Microsyst. Technol. 20 (2013) 523–531. [6] G.K. Rane, S. Menzel, T. Gemming, J. Eckert, Microstructure, electrical resistivity and stresses in sputter deposited W and Mo films and the influence of the interface on bilayer properties, Thin Solid Films 571 (2014) 1–8. [7] G. Rane, M. Seifert, S. Menzel, T. Gemming, J. Eckert, Tungsten as a chemically-stable electrode material on Ga-containing piezoelectric substrates Langasite and Catangasite for high-temperature SAW devices, Materials 9 (2016) 101. [8] S. Zimmermann, R. Ecke, M. Rennau, S.E. Schulz, M. Hecker, A. Voß, J. Acker, H.-J. Engelmann, N. Mattern, E. Zschech, T. Gessner, in: G.W. Ray, T. Smy, T. Ohta, M. Tsujimura (Eds.),Advanced Metallization Conference 2003, pp. 397–402. [9] T.A. Shankoff, E.A. Chandross, High-resolution tungsten patterning using buffered, mildly basic etching solutions, J. Electrochem. Soc. 122 (1975) 294–298. [10] A. Warren, A. Nylund, I. Olefjord, Oxidation of tungsten and tungsten carbide in dry and humid atmospheres, Int. J. Refract. Met. Hard 14 (1996) 345–353. [11] S.S. Perry, H.C. Galloway, P. Cao, E.J.R. Mitchell, D.C. Koeck, C.L. Smith, M.S. Lim, The influence of chemical treatments on tungsten films found in integrated circuits, Appl. Surf. Sci. 180 (2001) 6–13. [12] M. Anik, K. Osseo-Asare, Effect of pH on the anodic behavior of tungsten, J. Electrochem. Soc. 149 (2002) B224–B233. [13] M. Anik, T. Cansizglu, S. Cevik, Diffusion effect on the anodic reactions of tungsten, Turk. J. Chem. 28 (2004) 425–439. [14] M. Anik, T. Cansizoglu, Dissolution kinetics of WO3 in acidic solutions, J. Appl. Electrochem. 36 (2006) 603–608. [15] H. Kempel, M. Saradshow, Löslichkeit und stabile Kristallhydrate im System Ammoniumparawolframate-Wasser, Krist. Tech. 2 (1967) 437–445. [16] A. Zambova, L. Zambov, K. Stantchev, Mechanism and kinetics of molybdenum films etching in peroxide-ammonia solution, J. Electrochem. Soc. 139 (1992) 2470–2477.