Transmission electron microscopy of NdNiO 3 thin films on silicon substrates

Transcription

1 Eur. Phys. J. AP 12, (2000) THE EUROPEAN PHYSICAL JOURNAL APPLIED PHYSICS c EDP Sciences 2000 Transmission electron microscopy of NdNiO 3 thin films on silicon substrates P. Laffez 1,a,R.Retoux 2, P. Boullay 3, M. Zaghrioui 1, P. Lacorre 2, and G. van Tendeloo 3 1 Laboratoire de Physique de l État Condenséb,Université du Maine, Le Mans Cedex 9, France 2 Laboratoire des Fluorures c,université du Maine, Le Mans Cedex 9, France 3 EMAT, University of Antwerp, groeonenborgerlaan 171, Antwerp 2020, Belgium Received: 12 May 2000 / Revised: 18 September 2000 / Accepted: 19 September 2000 Abstract. The microstructure of NdNiO 3 thin films deposited on Si (100) has been investigated by high resolution electron microscopy. Deposition at 250 C and 600 C and several annealing at high temperature under oxygen pressure were performed. Depending on the deposition temperature and annealing conditions, different texture and microstructure were observed. Relationships between microstructure and transport properties are discussed. The differences of grain boundaries are suggested to be responsible for the difference in transport properties of the films. PACS h Electronic transport phenomena in thin films and low-dimensional structures h Metal insulator transitions and other electronic transitions 1 Introduction Electronic transport properties of the perovskite series ReNiO 3 (Re 3+ = La, Pr, Nd, Sm, Eu) are well known. These compounds can exhibit metal to insulator transition versus temperature. The mechanism of the metalinsulator transition is attributed to a charge transfer and changing the Re 3+ size can modulate the transition temperature T MI [1,2]. From a technological point of view, such materials are of great interest if one is able to control this T MI temperature. This would open numerous applications such as temperature-modulated switches, sensors, or thermochromic coatings if deposited as thin films over appropriate substrates. DeNatale et al. first demonstrated the feasibility of growing NdNiO 3 films on LaAlO 3 and have studied the optical properties of these films in the infrared range [3,4]. We have previously optimized the deposition parameters of NdNiO 3 polycrystalline thin films by RF magnetron sputtering on Si (100) [5,6] and various single crystal perovskites substrates [14]. We have shown that for a given annealing process, the deposition temperature has an important role on the composition and the orientation of the films as observed by X-ray diffraction. Films deposited at 250 C are single phase, polycrystalline and randomly oriented. They exhibit a broad transition between 150 K and 170 K. Films deposited at 600 C are also polycrystalline but with a strong preferential orientation and with smaller grains than the previous a patrick.laffez@univ-lemans.fr b UMR CNRS 6087 c UMR 6010 samples. They also show a sharp metal-insulator transition around 200 K. In order to understand the behavior of these films we performed quantitative texture analysis on various samples [7]. The strong preferential orientation along the unit cell axis of the pseudo cubic perovskite was confirmed although the direct relationship between transport properties and texture was not clearly evidenced. A study of the local structure of such films has not been performed yet, and the relationship between twin boundaries, dislocations, stacking faults and other defects with the transport properties is not established. The present study compares the microstructure of NdNiO 3 films deposited at 250 C and 600 C respectively and then annealed under oxygen pressure. Several annealing processes were tested. Three samples were selected for the present microstructural characterization and were synthesized using the following processes: (i) Sample deposited at 250 C, annealed 2 days at 800 C and 170 bar O 2 ; (ii) Sample deposited at 250 C, annealed 30 minutes at 950 C and 200 bar O 2 ; (iii) Sample deposited at 600 C, annealed 2 days at 800 C under 170 bar O 2. The samples grown at 600 C were single phase only after annealing at 800 C for two days. For others heat treatment at 800 C or 950 C impurities or uncompleted reaction were observed. In our previous work we performed X-ray analysis of the films using the Bragg-Brentano geometry [5]. Two main differences between the samples were evidenced, depending on the preparation conditions.

2 56 The European Physical Journal Applied Physics (i) Samples grown at 250 C always exhibit X-ray patterns corresponding to a randomly oriented material equivalent to diffraction patterns usually observed on powder samples. Samples deposited at 600 Cshowed a systematic preferential orientation along (101) and (010) axis in the Pnma space group (fiber texture). (ii) For a given annealing condition, the width of the diffraction peak is much larger for samples grown at 600 C, than for samples deposited at 250 C. This phenomenon is related to the size of the coherent domains. On the basis of these X-ray experiments we already pointed out a strong difference in the microstructure of the samples depending on the process used for the film synthesis. The present investigation of the microstructure by transmission electron microscopy (TEM) allows the identification of the microstructural parameters that could be responsible of the differences in transport properties observed in such thin films. 2 Experimental RF sputtering using an argon-oxygen atmosphere and subsequent annealing in oxygen pressure were used for the deposited films. Details of the sample preparation can be found in references [5 7]. Resistivity measurements as a function of temperature between 80 and 320 K were carried out using a four probe technique with silver paste contacts. Specimens for TEM observations were prepared by two different techniques. The first one consists in scraping off the thin films in ethanol using a diamond knife. A drop of the suspension is deposited and dried onto a carbon coated copper grid. TEM study of the scraped samples was performed in a 200 kv side entry JEOL 2010 Transmission Electron Microscope (tilt ±30 ). In the second way, samples oriented with the foil normal along [100] of the Si substrate (cross section) were prepared by mechanical polishing and ion milling. The TEM study of ion milled samples was carried out on a JEOL 4000EX top-entry microscope (double tilt ±20 ) operating at 400 kev. The chemical composition of the deposited material was determined by energy dispersive X-ray analysis (EDX) using a KEVEX and a LINK EDX spectrometer coupled with the 200 kv microscope. Conventional bright field image and selected area electron diffraction were used for the microstructure study. SEM micrographs were taken on a Philips FEG XL30 scanning microscope at 20 kv. 3 Results 3.1 Resistivity versus temperature The resistance of the three samples was measured from 70 K to 320 K and is shown in Figure 1. Fig. 1. Resistance versus temperature for NdNiO 3 films deposited at (a) 250 C and annealed 950 C 30 minutes at 170 bar oxygen pressure (b) 250 C and annealed at 800 C for two day, (c) 600 C and annealed at 800 C for two days. The transition temperatures corresponding to the point where dr/dt becomes negative were 150 K, 165 K and 210 K respectively. Since the measuring geometry of the different samples is the same, the relative resistance approximates their resistivity. The onset of the transition is taken at the temperature where dr/dt becomes negative. The detail of the measurement versus temperature shows two different behaviors depending on the preparation process. Films deposited at 250 C gave a transition ranging from 150 K to 165 K (depending on the annealing) and did not always exhibit metallic state over the whole temperature range above T MI. On the other hand, samples deposited at 600 C clearly show that above T MI,theR(T ) slope is positive indicating a metallic state. When the temperature decreases, a huge increase in the resistance, up to three orders of magnitude is obtained. The onset of the transition is extremely abrupt and was measured at 215 K. Details concerning the transport properties of these samples can be found in references [5,6]. 3.2 Microstructural characterization Differences between the samples will be highlighted here in order to elucidate the different contributions of the microstructure to the transport properties Grain size distribution and grain boundaries Plane view Observations of the surface of the annealed films by SEM shows the typical microstructure of the NdNiO 3 films for films deposited at 250 C and 600 C and annealed at 800 C. In Figure 2 the difference of the grain size of the two samples is clearly evidenced. The sample deposited at

3 P. Laffez et al.: Transmission electron microscopy of NdNiO 3 thin films on silicon substrates 57 Fig. 2. SEM observations of the surface of NdNiO 3 films showing the microstructure depending on the deposition temperature. (a): film grown at 250 C, (b) film grown at 600 C. The annealing process was two days at 800 C and 170 bar oxygen pressure. Fig. 3. Cross-section TEM micrograph of NdNiO 3 film on Si/SiO 2. (a) Film grown at 250 C, annealed at 800 C2days under O 2 (170 bar), (b) film grown at 600 C, annealed at 800 C 2 days under O 2 (170 bar). 250 C (Fig. 2a) shows grains ranging from 50 nm and up to 500 nm. Moreover many cracks are present, randomly dispersed all over the surface of the sample and domains within each crack are several micrometers wide. The sample deposited at 600 C and annealed two days at 800 C (Fig. 2b) exhibits a very different microstructure with a very different grain size distribution. This distribution is more homogeneous than the previous one with a smaller grain size (maximum 10 nm). In this sample, cracks are evidenced only from time to time. Cross section The TEM of cross section samples displayed in Figure 3 confirm the microstructure and the grain distribution deduced from the plane view. The corresponding films show a granular structure. The film grown at 250 Candannealed at 800 C for 2 days is composed of large grains (Fig. 3a), (size up to 300 nm along the perpendicular direction of the substrate, corresponding to the total thickness of the film). The presence of cracks that was observed by SEM is also observed by TEM. Additionally, the TEM micrographs show that these cracks can sometime run from the top of the surface to the interface of the film with the substrate. Moreover it is pointed out that the grain boundaries are quite different within the two different samples. The grain boundaries are relatively large and well defined in the samples deposited at 250 C (Fig. 3a) while they are diffuse but sharp in the samples grown at 600 C (Fig. 3b). This observation will be detailed in a next section Interfaces Figure 3 also shows the thickness of the SiO 2 layer at the interface for a given annealing process. In a previous work we have shown that a more or less important growth of amorphous SiO 2 interface can be evidenced between the thin film and the Si substrate [5]. An important feature is that the thickness of this amorphous layer is strongly dependent on the annealing conditions. The sample annealed for 30 minutes showed a SiO 2 layer of approximately 30 nm thick while longer annealing lead to a drastic increase of the thickness of the SiO 2 layer (450 nm for 2daysannealingat800 C). The increase of this interfacial layer is related to the oxidation time. However, for the sample grown at 600 C and annealed 2 days at 800 C the thickness of the SiO 2 layerisonly50nm(fig.3). The initial temperature deposition is thus a key parameter for the growth of this layer. The growth of the SiO 2 layer is enhanced for samples deposited at 250 C. This phenomenon is probably related to the reactivity of the as synthesized samples. Our previous studies [6] had shown that films deposited at 250 C are amorphous before annealing while films deposited at 600 C are already crystallized after the sputtering process (mixture of Nd 2 NiO 4 and NiO). It seems that films with two different textures at the initial step of the annealing do not react similarly under oxygen pressure. The actual growth of the SiO 2 layer in the amorphous samples deposited at 250 C suggests that their microstructure enhances the oxygen diffusion through the film leading to thicker SiO 2 layers compared to those observed in the 600 C deposited films.

4 58 The European Physical Journal Applied Physics Fig. 5. Image of a grain boundary (white arrow) in a sample grown at 250 C and annealed at 800 C for 2 days at 170 bar showing that no crystalline orientation appears between the grains. Fig. 4. High resolution images of NdNiO 3 film grown at 250 C and annealed at 800 C 2 days at 170 bar oxygen pressure along different zone axis. Insets are the corresponding electron diffraction patterns confirming the Pnmasymmetry. (a) [010] zone axis, (b) [001] zone axis, (c) [100] axis. (Note that extra reflection are sometime observed due to double diffraction phenomena and to existence of oriented domains.) Local structure of the NdNiO 3 phase Sample deposited at 250 C In our previous study, we have observed a film deposited at 250 C and annealed at 950 C 30 minutes. The corresponding sample showed a lot of defects involving very small homogenous domains ( 10 nm). These small domains are probably the cause of the relatively poor quality of the M-I transition observed in this sample. In the present study, a sample deposited at at 250 Candannealed 2 days at 800 C is shown in Figure 4. The [010], [001], [100] electron diffraction patterns of such sample are shown in inset in Figures 4a 4c. A reconstruction of the reciprocal space allows the cell parameters to be determined as a 2a p, b 2a p, c a p (with a p 0.38 nm, the cubic parameter of the perovskite subcell). The limiting reflection conditions are hk0: h =2n, 0kl: k + l = 2n, which confirm the space group to be Pnma. The corresponding images show large areas without defects in the majority of the crystal. This typical microstructure was observed in domains up to 300 nm wide in most of the grains. The number of defects (stacking faults, dislocations, twining) usually observed in perovskite thin films is very low in this sample. The images in Figure 4 are very similar to those observed for bulk NdNiO 3.Figure5 shows a grain boundary confirming that no simple structural correlation in the orientation of the domains appears in this sample. Finally HREM also pointed out that the interface between SiO 2 and NdNiO 3 is more complex and shows (Fig. 6a) an intermediate amorphous layer between 5 nm and 10 nm (darker contrast). As shown in Figure 6b we performed conventional EDX analyses of the intermediate amorphous layer. For analyses on such small areas (about 5 to 10 nm wide), the spot size should ideally be as small as possible but then the number of counts received by the detector becomes drastically low, resulting in the necessity to increase the counting time. Under such conditions the drifting of the sample is a severe limitation in the possibility to analyse small areas. As a result a compromise between an acceptable counting time and a small spot size has to be found. We choose to limit ourselves to a real counting time of about 85 s (25% deadtime). We try (moving slightly the spot) to go from the SiO 2 layer to the NdNiO 3 film, to see if any gradient in the chemical composition can be observed. In Figure 6b the radiation damages corresponding to the analysed areas are clearly visible. In particular we can

5 P. Laffez et al.: Transmission electron microscopy of NdNiO 3 thin films on silicon substrates 59 Fig. 6. (a) Interface of NdNiO 3/SiO 2 in a sample grown at 250 C and annealed at 800 C 2 days under oxygen showing the intermediate layer between SiO 2 and NdNiO 3 (white arrow); (b) on this figure are pointed out the areas analysed by conventional EDX analysis and the Nd/Ni ratio observed is given in insert. Silicon is present in all analysis except in the NdNiO 3 layer. Fig. 7. Typical microstructure of a film deposited at 600 C and annealed at 800 C for two days under oxygen. (a) Cross-section illustrating the columnar structure of such films. The growth of the grains is oriented along the white arrows in a direction perpendicular to the surface of the silicon substrate, (b), (c) crystallites with different orientations are observed. notice than the areas noted 1, 2 and 3 at the interface correspond to areas closer and closer to the NdNiO 3 film. While it seems to us over-optimistic to have a full confidence in the ratios given by quantitative analyses (see table in insert Fig. 6b), it seems that the darker amorphous layer between the film and SiO 2 is richer in Nd than in Ni. This confirms that the darker contrast of this small layer originates from a Z-contrast due to the diffusion of Nd (preferably) and Ni into the SiO 2 layer. Samples deposited at 600 C For this deposition temperature a sample annealed at 800 C for two days was analyzed. The average size of the coherent domains is estimated around ten nanometer; the microstructure is highly inhomogeneous and Figure 7a shows that the film has a columnar structure. The electron diffraction patterns from such areas exhibit spotty rings, that can be indexed with the Pnma perovskite based material. Figures 7b 7c show details of two different microstructures characteristic of this sample. In Figure 7b grains boundaries are very thin, and the crystalline order nearly absent. Figure 7c presents an image where the grains are correlated with simple crystallographic relationship between each grain. Twinning is evidenced and the domains observed are well crystallized. The microstructure of Figure 7b however is the most often observed in this type of sample. 4 Discussion We have characterized the microstructures NdNiO 3 thin films obtained by RF sputtering with interesting electrical properties. These different microstructures may help to understand the electrical behavior of these films. Different features are evidenced and concern the structure of the interface and the size of the coherent domains respectively. The grain distribution is summarized below: (i) Low temperature synthesis and short time annealing leads to the formation of large grains with a lot of defects within each grain; (ii) Low temperature synthesis and long annealing produces large single crystal grains, with no correlation between neighboring grains. Many cracks are evidenced; (iii) High temperature deposition with long annealing involves small crystallites with a dense structure. For films deposited at 250 C, the highest conductivity of the films is found after an annealing at 800 Cfortwo days. It may be related to a better ordering in these films compared to films annealed at 950 C for 30 min. The presence of cracks in low temperature deposited films and the larger grain boundaries can explain the higher resistivity of the films deposited at 250 Cand annealed 800 C for two days in comparison with the ones grown at 600 C and annealed under the same conditions. However the reason why the transition temperature T MI

6 60 The European Physical Journal Applied Physics is in general lower in films made at 250 C than at 600 C still has to be explained. Two possibilities can be pointed out: (a) The films are stressed with different intensity of the strain: it is well known in this system that the temperature of the metal insulator transition can be drastically decreased under pressure with a rate of dt MI /dp = 4.2 Kkbar 1 [8,9]. If one considers that the stress generated by the substrate could be different in the two kinds of samples, it could explain the change of T MI. This behavior suggests that the nature of the substrate and the thermal process should strongly influence the temperature of the metal-insulator transition. In order to address this issue, further investigations to characterize the strain of the films by X-ray diffraction are now in progress [10]; (b) The oxygen stoichiometry of the films can change: it is well known that oxygen deficient could also influence the transport properties of NdNiO 3. Previous studies have evidenced that oxygen losses lead to an increase of resistance and a decrease of T MI [11,12]. Although the films are deposited in the same conditions of oxygen partial pressure and the annealing process is the same for the two samples annealed 2 days at 800 C under 170 bar O 2, we cannot rule out the possibility of an oxygen deficit in the films deposited at 250 C. We have in fact observed that the growth of the amorphous SiO 2 layer during annealing is enhanced in the film deposited at low temperature. For the same temperature deposition, the thinnest SiO 2 layer is obtained for the shorter time annealing. As the oxygen diffuse through the silicon substrate, to form the amorphous SiO 2 intermediate layer, it could then take off the oxygen of the film and lead to a NdNiO 3 δ composition. Complementary studies are in progress using the RBS (Rutherford Back Scattering) technique in order to attempt to estimate the oxygen concentration in the films [13]. These analysis will help to estimate if the SiO 2 layer (thicker in one case) as a role in the oxygen stoichiometry of the thin films. 5 Conclusion In this study we tried to quantify the influence of sample preparation conditions on the resistance behavior of NdNiO 3 thin films. One of the most important information is that the quality of the transport properties of our films is, as expected, strongly influenced by the deposition temperature and annealing conditions. It appears that in relation with the transport properties the microstructure of our films is also an important parameter to optimize if one want to obtain NdNiO 3 thin films with rather good properties. Films of textured NdNiO 3 are deposited on Si (100) substrates. A huge resistance change of three orders of magnitude is obtained near 200 K for textured films grown at 600 C, and around 150 K for polycrystalline films deposited at 250 C. For the same annealing conditions 800 C under oxygen, films grown at 250 C are polycrystalline, randomly oriented, granular with wide grains and present a high density of cracks while films grown at 600 C are polycrystalline, textured, columnar with nanosized grains and dense. Films grown at 600 C presents the best and quite good transport properties compared to the bulk ones and the differences in grain boundaries and the presence of cracks are supposed to be responsible of the difference of transport properties observed between these films and the 250 C deposited ones. A good correlation between the microstructure and the transport properties has been established and an optimized approach of the synthesis process is proposed to get good NdNiO 3 thin films on silicon. The authors are grateful to L. Rossou (EMAT) for TEM samples preparation, to Dr. G. Trolliard (SPCTS-Limoges) for valuable help in EDX analysis and to Dr. I. Monot-Laffez (CRISMAT-Caen) for SEM observations. References 1. J.L. Garcia-Munoz, J. Rodriguez-Carvajal, P. Lacorre, Phys. Rev. B 50, 978 (1994). 2. J.L. Garcia-Munoz, J. Rodriguez-Carvajal, P. Lacorre, J.B. Torrance, Phys. Rev. B 46, 4414 (1992). 3. J.F. DeNatale, P.H. Kobrin, J. Mater. Res. 10, 2992 (1995). 4. J.F. DeNatale, P.H. Kobrin, Mater. Res. Soc. Symp. Proc. 479, 145 (1997). 5. P. Laffez, M. Zaghrioui, R. Retoux, P. Lacorre, J. Magn. Magn. Mater. 211, 111 (2000). 6. P. Laffez, M. Zaghrioui, P. Lacorre, T. Brousse, I. Monot, Vide, Sci. Tech. Appl. 289, 607 (1998). 7. N. Boisard, M. Zaghrioui, J. Ricote, D. Chateignier, P. Laffez, P. Lacorre, EMRS 1999 Spring Meeting 1st European Materials Conference, 1 4 June 1999, Strasbourg, France. 8. X. Obradors, L.M. Paulius, M.B. Maple, J.B. Torrance, A.I. Nazzal, J. Fontcuberta, X. Granados, Phys. Rev. B 47, (1993). 9. P.C. Canfield, J.D. Thompson, S.W. Cheong, L.W. Rupp, Phys. Rev. B 47, (1993). 10. P. Goudeau, M. Zaghrioui, P. Laffez, D. Thiaudière, M. Gailhanou, in preparation. 11. R. Mahesh, K.R. Kannan, C.N.R. Rao, J. Solid State Chem. 114, 294 (1995). 12. A. Tiwari, K.P. Rajeev, Solid State Commun. 109, 119 (1999). 13. I. Vickridge, P. Laffez, M. Zaghrioui, in preparation. 14. P. Laffez, M. Zaghrioui, I. Monot, T. Brousse, P. Lacorre, Thin Solid Films 354, 50 (1999).