Materials Chemistry and Physics 58 (1999) 280±284
Materials Science Communication
Crystalline properties of MoS2 thin ®lms grown on metallic substrates a
E. Gourmelona,*, J. Pouzeta, J.C. Bernedea, H. Hadoudab, A. Khelilb, R. Le Nya
EPSE Equipe Couches Minces et MateÂriaux Nouveaux, FaculteÂ des Sciences et des Techniques, 2 rue de la HoussinieÁre, BP 92208, 44322, Nantes Cedex 3, France b Laboratoire de Physique des MateÂriaux et Composants pour l'Electronique, UniversiteÂ d'Oran Es SeÂnia, Oran, Algeria Received 12 March 1998; received in revised form 20 October 1998; accepted 15 December 1998
Abstract MoS2 thin ®lms have been achieved on molybdenum, tantalum and tungsten substrates. The ®lms were obtained by solid state reaction between the constituents in thin ®lm form, the substrates being coated by a thin Ni layer. Multilayer samples Mo/S/Mo±Mo/S were annealed under argon ¯ow at T 1073 K or 1123 K for 30 min. The ®lms were characterized by X-ray diffraction (XRD), scanning electron microscopy (SEM), electron probe microanalysis (EPMA) and X-ray photoelectron spectroscopy (XPS). All ®lms are nearly stoichoimetric and crystallized in the expected 2H-MoS2 structure. (XPS analysis con®rms the stoichoimetry and the small contamination of the ®lms while Ni was detected at the surface of the samples). The ®lms are polycrystalline but the size and the orientation of the crystallites depend on the substrates. While in the case of tungsten substrates the crystallite size reach the total thickness of the ®lm (100 nm) and are highly oriented with their c-axis perpendicular to the plane of the substrate, in the case of the other refractory metals (Ta, Mo), their size is smaller and the texture of the ®lms is not so good. These differences are explained by the in¯uence of the crystalline properties of the substrates on the growth process of MoS2 thin ®lms. The current±voltage characteristic of the W/Ni/MoS2/W structure show that an ohmic contact is obtained. # 1999 Elsevier Science S.A. All rights reserved.
1. Introduction Transition metal dichalcogenides MX2 (M Mo, W; X S, Se) are semiconductors that can act as ef®cient photovoltaic materials. Promising results have been obtained in the realization of photoelectrochemical [1±3] or solid device solar cells [4,5]. For economic reasons, obtaining MX2 in thin ®lms would be interesting. However, though stoichoimetric thin ®lms crystallized in the 2H-MoS2 structure can be easily achieved [6,7], the ®lms were poorly crystallized with small crystallites and the ®lms were not photoactive. Tenne with his group [8,9] has shown that it was possible to obtain textured photoconductive WX2 ®lms by annealing, under an HX2 atmosphere, WO3 ®lms deposited on Ni coated substrates. The same result has been obtained in another laboratory . However, the use of an HX2 atmosphere is quite restricting to obtain ohmic back contact. More recently, it has been shown that textured photosensitive ®lms can be obtained by annealing MX3 ®lms under *Corresponding author.
argon atmosphere whatever the deposition technique used: rf sputtering [11,12] or electron beam evaporation [13±15]. In this paper, we use the technique of solid state reaction between the constituents sequentially deposited in thin ®lms form, developed in the laboratory [14,15], to growth MoS2 textured ®lms. The need of an ohmic back contact forces the use of a metal giving a small resistance with MX2 compound and which does not diffuse strongly in the semiconductor at the temperature used to grow the MX2 compound. Moreover, the crystalline properties of the ®lms should be preserved. It is well known that refractory metals are often used in thin ®lms as barrier diffusion contact. Therefore, we have grown MoS2 ®lms on molybdenum, tantalum and tungsten substrates which are also very stable at high temperature. The crystallization quality and the properties of the ®lms have been checked by X-ray diffraction (XRD), scanning electron microscopy (SEM), electron probe microanalysis (EPMA), X-ray photoelectron spectroscopy (XPS). It is shown that W allows the growth of high quality ®lms while in the case of Ta and Mo the results are not so promising. In the case of W, the J±V characteristic of the W/Ni/MoS2/W structure exhibits an ohmic contact.
0254-0584/99/$ ± see front matter # 1999 Elsevier Science S.A. All rights reserved. PII: S0254-0584(99)00010-3
E. Gourmelon et al. / Materials Chemistry and Physics 58 (1999) 280±284
Sheets of Mo, Ta, W have been used as substrates. For comparison silica substrates have also been used. First, a thin (10±20 nm) Ni ®lm was deposited onto the metallic substrates with an electron beam, then Mo and S layers were sequentially deposited. The sulphur was evaporated by a thermal evaporation using a speci®c crucible. Mo was deposited by electron beam evaporation. The evaporation rates and the ®lm thicknesses were measured in situ by two speci®c hf vibrating quartz oscillator. The number of layers varied from ®ve to nine in order to deposit Mo/S/ Mo Mo/S sequences. The last sulphur layer was about 100±200 nm thick, it was used to protect the last molybdenum layer from oxidation during transfer from the deposition apparatus to the oven for the post annealing treatment. The thicknesses of others layers were calculated to achieve the desired atomic ratio Mo/S 1/3, varying from 8 to 25 nm and from 60 to 250 nm for Mo and S respectively, while the deposition rates were 1 and 10 nm sÿ1. The MoS2 ®lms were synthesized by the solid state reaction from thin Mo and S layers during the post annealing treatment. It has been shown earlier [14,15] that the optimum annealing conditions were 1073 to 1123 K for 30 min in an argon ¯ow (40 l hÿ1). These conditions were used in the present work. The structure quality of the ®lms was examined using an analytical X-ray system type Diffract AT V3.1 Siemens instrument which used an EVA graphics program. The wavelength was 0.15406 nm. The preferential orientation and the grain size of the crystallites were deduced from the X-ray diffraction patterns and estimated from the full-width at half-maximum (FWHM) of the diffraction peaks respectively. XPS measurements were performed with a magnesium X-ray source (1253 eV) operating at 10 kV and 10 mA. Data acquisition treatment was realized using a computer and a standard program. The quantitative studies were based on the determination of the Mo 3d and S 2p peak areas with 2.5 and 0.125, respectively, as sensivity factors. The sensivity factors of the spectrometer were given by Leybold, the manufacturer. At the surface of the ®lm there is a carbon±carbon bond corresponding to surface contamination. In the apparatus used, this C±C bond has a well de®ned position at 284.6 eV and the carbon peak was used as a reference to estimate the electrical charge effect. The surface topography was observed with a ®eld effect scanning electron microscope JEOL F 6400. Electronic microanalyses were performed using a JEOL F 5800 LV scanning electron microscope equipped with a PGT X-ray microanalysis system in which X-rays were detected by a germanium crystal. After this characterization an upper electrode of tungsten was deposited by sputtering onto the MoS2 ®lms. The J±V characteristic was measured with an electrometer Keithley 617.
The experimental results presented below have been measured on samples obtained in the same run. The thickness of the ®lms, determined by a stylus type pro®lometer, was 600 nm. The duration of all the annealings was 30 min. For an annealing at 1073 K, the XRD diagrams are presented in Fig. 1 and they are summarized in Fig. 2. It can be clearly seen that only W substrates allow to achieve MoS2 ®lms with a crystalline quality comparable to that obtained with silica substrates. In the case of Ta substrates, the maximum intensity of the (0 0 2) peak is far smaller than those obtained on the silica substrate while the full width at the half maximum of this peak is twice. This means that the grain size is smaller since the FWHM increases when the grain size decreases . The crystallite size, deduced from the FWHM, can be estimated to about 17 nm. Using Mo substrates, there is an improvement of the crystalline quality of the MoS2 ®lms. Though the intensity of the (0 0 2) peak is only a little higher, the FWHM is smaller and the grain size could be estimated to 30 nm. In both cases, the crystalline quality of the ®lms appears too far from the ®lms grown onto silica substrate which allow some photovoltaic applications. In the case of W substrates, the intensity of the (0 0 2) peak is similar to those obtained with silica substrates while the FWHM, even if a little more broad, is not very different from that obtained with ®lms onto silica substrates. Moreover, the texturation coef®cient F00` which is nearly one in the case of silica substrate is about 0.8 in the case of W substrate, which shows that even if high crystalline quality ®lms can be achieved on tungsten substrate their texturation is not so perfect. The surface visualization at a small magnitude show a smooth and homogeneous surface (Fig. 3(a)). For a larger magnitude it can be seen that if large grains with their van der Waals planes parallel to the plane of the substrate are clearly visible, some smaller crystallites, with their c-axis parallel to the plane of the substrate are also present which justi®es the small decrease of F00` (Fig. 3(b)). Moreover, it should be noted that the results described above have been obtained in the case of textured W substrates (W(2 0 0)). When randomly oriented substrates (W(2 0 0),(1 1 0)) are used, the crystalline quality is far poorer (Fig. 1(e) and Fig. 2). All the results described have been measured after annealing at 1073 K. Some experiments have been done at 1123 K. The results obtained on the (0 0 2) peak intensity are summarized in Table 1. It can be seen that even here the crystalline quality of the ®lms achieved on Mo and Ta substrate is very poor. Moreover, in the case of W, the higher temperature (1123 K) does not improve the peak intensity and therefore, a temperature of 1073 K for the annealing appears to be the optimum value. It should be recalled that with silica substrate, if the intensity of the
E. Gourmelon et al. / Materials Chemistry and Physics 58 (1999) 280±284
Fig. 1. XRD diagrams of the structure of (a) silica/Ni/MoS2; (b) Mo/Ni/MoS2; (c) Ta/Ni/MoS2; (d) W(2 0 0)/Ni/MoS2; (e) W(2 0 0)(1 1 0)/Ni/MoS2.
(0 0 2) peak is higher at T 1123 K, the homogeneity of the ®lms is smaller . The surface analysis of the samples by XPS shows that the bonding energy of the elements corresponds to the binding energies expected in MoS2. The Table 1 Evolution of the (0 0 2) peak intensity with temperature and nature of the substrate Temperature
Intensity of the (0 0 2) peak (counts) Silica
1073 K 1123 K
10 338 17 294
10 466 10 376
contamination of the samples is small after 1 min Ar etching and the carbon and oxygen present on the ®lms have disappeared. Therefore, the SO2 peak measured by XRD should be attributed to the surface contamination during the annealing process since some impurity are present in the argon gas used (H2O 3 ppm, O2 3 ppm). The Fig. 4 show that nickel is present at the surface (1% at.) and all over the thickness of the ®lms. The ®lms are systematically stoichiometric as shown by XPS surface analysis but also by microprobe analysis (33 at.% < Mo < 34 at.%). Moreover, the Fermi level has been estimated from the Fig. 5 to be 0.43 eV which indicates that the ®lms are p-type as expected for the photovoltaic applications.
E. Gourmelon et al. / Materials Chemistry and Physics 58 (1999) 280±284
Fig. 4. XPS spectrum of the Ni in an MoS2 film grown on W(2 0 0). Fig. 2. Variation of intensity and FWHM of the (0 0 2) diffraction peak with different substrates used.
After this characterization an upper electrode of W was deposited by sputtering onto the MoS2 ®lms obtained on W(2 0 0) substrates. Fig. 6 show that the J±V characteristic at room temperature is linear, so the contact is ohmic. 4. Discussion As discussed before it has been shown that when a Ni coated substrate is used, MX2 textured ®lms with high crystalline properties are obtained. The same result has been obtained with silica substrate , mica substrate  and also graphite substrates . Therefore, it appears that whatever the substrate used the presence of nickel allows the achievement of high crystalline quality ®lms. However, in the present work, textured ®lms are obtained only on W substrates while Ni coated substrates are systematically used. Even if the solubility of Ni in Ta is 50 times to that of Ni in W at the same temperature , which can allow some nickel diffusion into the Ta, the fact that Ni is detected at the surface of the ®lms after the annealing excludes an exhaustive diffusion of Ni in the substrate.
Fig. 5. XPS spectrum of the valence band of an MoS2 film grown on W(2 0 0).
It can be concluded that metallic substrates have some in¯uence on the crystalline properties of the MoS2 ®lms. Up to now, the substrates used were either amorphous (silica) or lamellar with van der Waals planes as surface (mica, graphite) which are chemically inert, therefore the interaction between the substrate and the ®lms was very small, which allows the thin Ni layer to control the crystallization of the MoS2 ®lms.
Fig. 3. Micrographs of an MoS2 film grown on W(2 0 0) substrate: (a) small magnitude; (b) large magnitude.
E. Gourmelon et al. / Materials Chemistry and Physics 58 (1999) 280±284 Table 2 Crystal parameters of the metal substrates Substrate Structure Crystal parameter a (nm)
Fig. 6. J±V characteristic of the W(2 0 0)/Ni/MoS2/W structure at room temperature.
In the present case, metallic substrates are used with very active surface and therefore, there is some competition between the in¯uence of nickel and that of the substrate. It has already be shown that Ni has a positive in¯uence on the crystalline properties of MoS2 ®lms . In the case of metallic substrate it appears that the in¯uence depends on the substrate properties. In the case of W it should be noted that the results are far better when the crystallites of the substrates are oriented in such a way that only the peak corresponding to the (2 0 0) direction is detected, when the (1 1 0) peak is also detected the crystallization quality is far poorer. Therefore, the use of a textured substrate is very important to obtain textured ®lms, however, when textured (2 0 0) Ta substrates are used no such good results are obtained. It can be seen in Table 2 that, though all the metallic substrates used in the present work have a cubic structure, the a parameter of Ta is greater than that of W. Therefore, textured substrates with adapted parameter (a 0.316 nm) are necessary to obtain textured MoS2 ®lms. The molybdenum substrates being not textured, the quality of the crystallization is poor even if the a parameter of Mo is smaller than that of Ta. From the discussion above it can be concluded that not only do the metallic substrates have some in¯uence on the crystalline properties of MoS2, but it is higher than that of Ni, even if the ®lms grown on Ta and Mo with a thin layer of Ni have poor crystalline properties. 5. Conclusions MoS2 thin ®lms have been achieved on Ni coated metallic substrates by post-annealing of M/Ni/Mo/S/Mo /Mo/S (M Mo, Ta, W) samples. It is shown that, though stoichoimetric MoS2 ®lms are systematically obtained, they are
textured only on textured W substrates because of the crystalline properties of this substrate. The J±V characteristic exhibits an ohmic contact with W. Therefore, the W/MoS2 contact appears very promising for use as back ohmic contact in photovoltaic thin ®lm cells. Acknowledgements The authors wish to thank Mr Barreau for performing MEB studies. This work was supported by a contract between France and Algeria (CMEP 95 DPU 337). References  H. Tributsch, Solar Energy Mater. 1 (1979) 257.  G. Kline, K.K. Kam, R. Ziegler, B.A. Parkinson, Solar Energy Mater. 6 (1982) 337.  R. Tenne, A. Wold, Appl. Phys. Lett. 47 (1985) 707.  G. Prasad, O.W. Srivastava, J. Phys. D 21 (1988) 1028.  A. Seguro, M.C. Martinez Tomas, B. Mori, A. Casanovas, A. Chevy, Appl. Phys. A 44 (1987) 1249.  J. Pouzet, H. Hadouda, J.C. Bernede, R. Le Ny, J. Phys. Chem. Sol. 57 (1996) 1363.  A. Jager-Waldau, M.C. Lux-Steiner, G. Jager-Waldau, E. Bucher, Appl. Surface Sci. 70/71 (1993) 731.  G. Salitra, G. Hodes, E. Klein, R. Tenne, Thin Solid Films 245 (1994) 180.  E. Galun, T. Tsirlinah, H. Cohen, L. Margulis, G. Hodes, R. Tenne, A. Matthaus, S. Tiefenbacher, C. Koelzow, M. Kunst, K. Ellmer, W. Jaergermann, Thin Solid Films 217 (1992) 91.  A. Ennaoui, S. Fiechter, K. Ellmer, R. Scheer, K. Diesner, Thin Solid Films 261 (1995) 124.  C. Ballif, M. Regula, P.E. Schmid, M. Remskar, R. Sanjines, F. Levy, Appl. Phys. A 62 (1996) 543.  O. Lignier, G. Couturier, J. Teed, D. Gonbeau, J. Salardenne, Thin Solid Films 299 (1997) 45.  E. Gourmelon, O. Lignier, H. Hadouda, G. Couturier, J.C. Bernede, J. Tedd, J. Pouzet, J. Salardenne, Solar Energy Mater. and Solar Cells 46 (1997) 115.  H. Hadouda, J.C. Bernede, E. Gourmelon, J. Pouzet, J. Mater. Sci. 32 (1997) 4019.  E. Gourmelon, H. Hadouda, J.C. Bernede, J. Pouzet, Vacuum 48 (1997) 509.  E.F. Kaeble, Handbook of X-rays, McGraw-Hill, New York, 1967.  T.B. Massalski (Ed.), Binary Alloy Phase Diagrams, 2nd ed., ASM International, 1992.