Influence of strain relaxation on the structural stabilization of SmNiO3 films epitaxially grown on (0 0 1) SrTiO3 substrates

Materials Science and Engineering B 144 (2007) 32–37

Influence of strain relaxation on the structural stabilization of SmNiO3 films epitaxially grown on (0 0 1) SrTiO3 substrates F. Conchon a,∗ , A. Boulle a , C. Girardot b , S. Pignard b , R. Guinebreti`ere a , E. Dooryh´ee c , J.-L. Hodeau c , F. Weiss b , J. Kreisel b a

Laboratoire Science des Proc´ed´es C´eramiques et de Traitements de Surfaces (SPCTS), CNRS UMR 6638, ENSCI, 47 Avenue Albert Thomas, 87065 Limoges, France b Laboratoire des Mat´ eriaux et du G´enie Physique (LMGP), CNRS UMR 5628, INP Grenoble, MINATEC, 3 parvis Louis N´eel, BP 257, 38016 Grenoble, France c Institut N´ eel (IN), CNRS UPR 2940, 25 avenue des Martyrs, BP 166, 38042 Grenoble, France

Abstract The structural stabilization of SmNiO3 (SNO) films epitaxially grown by an injection MO-CVD process on (0 0 1) SrTiO3 (STO) substrates is investigated. Using high-resolution X-ray diffraction (XRD), we show that SNO can be stabilized on STO with a minor amount of secondary phases and with a layer thickness reaching several hundreds of nanometers. The film quality is discussed by means of the simulation of X-ray reflectivity and XRD profiles that evidence smooth surfaces and interfaces. The actual lattice parameter of bulk (i.e. strain free) SNO is calculated as a function of the deposited thickness. It turns out firstly that the stabilization of SNO is achieved because the lattice mismatch between STO and SNO is not as high as expected (1.6% instead of 2.8%) and secondly the layer chemical composition varies with the film thickness. Finally, the well-known dissociation of the SNO phase into NiO and Sm2 O3 is clearly correlated to the relaxation of epitaxial strain. © 2007 Elsevier B.V. All rights reserved. Keywords: X-ray diffraction; Rare-earth nickelates films; Strain relaxation

1. Introduction The perovskite materials of the family RNiO3 (R = rare-earth) are of interest for their sharp metal to insulator (MI) transition which temperature can be modulated by changing the nature of the rare-earth [1,2]. On one hand, thin films with tunable MI transition can pave the way for a variety of applications such as sensors, modulated switches, or thermochromic coatings [3]; on the other hand, they offer the opportunity to study the relationship between structural and electronic properties. SmNiO3 (SNO) belongs to the orthorhombically distorded perovskites RNiO3 with GdFeO3 structure (space group Pbnm) and can be described in the pseudo-cubic symmetry with a lattice para˚ Most R-nickelates exhibit a complex meter equal to 3.795 A. anti-ferromagnetic ordering in the insulating phase at low temperature allowing the study of the interplay between magnetic and electric instabilities. Unfortunately, the two transitions (magne-

∗

Corresponding author. E-mail address: f [email protected] (F. Conchon).

0921-5107/$ – see front matter © 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.mseb.2007.07.096

tic and metal–insulator) coincide for a number of R-nickelates, and then obscure the straightforward interpretation of experimental results. Among all these R-nickelates, SmNiO3 has the two-fold advantage of having a well-separated MI and magnetic transition [4], thus offering a promising suitability for applications. This separation allows independent studies of these two transitions. Nevertheless, since the nickel must adopt the less stable Ni3+ oxidation state to form RNiO3 , these materials are prepared in the bulk either at low temperatures and atmospheric oxygen pressures or at high temperatures and high oxygen pressures. In this way, Demazeau et al. [5] were the first to describe successful nickelates synthesis. Another way to stabilize RNiO3 phases is to use epitaxial strain [6,7]. Novojilov et al. [6] showed that the single-phase SmNiO3 is only obtained on the perovskitelike substrate with the minimum lattice mismatch (m = −0.13%): LaAlO3 . On perovskite-like substrates with higher mismatch a phase mixture of the perovskite phase with NiO and Sm2 O3 is observed as the film thickness reaches a critical thickness [8] and the larger the mismatch is the lower the critical thickness is. In the latter case, films deposited on SrTiO3 (STO) present a low crystalline quality and the secondary phases Sm2 O3 and

F. Conchon et al. / Materials Science and Engineering B 144 (2007) 32–37

NiO are reported to be in large amount. This was attributed to the high in-plane tensile strain due to the 2.8% lattice mismatch ˚ (aSTO = 3.905 A). Because of their thermodynamic instability and their tunable MI transition, RNiO3 compounds allow us to study the relationship between structural and electronic properties. Nevertheless, one of the long-standing questions associated with RNiO3 is the origin of the structural stabilization of the RNiO3 phase. In the present study, we will focus on the influence of the microstructure on the epitaxial stabilization of SmNiO3 films deposited on SrTiO3 substrates. We will prove that the SNO phase is stabilized because the pseudo-cubic lattice parameter of the bulk SNO phase is higher than its usually assumed value. Consequently, the lattice mismatch between these two perovskites is not as high as expected. Besides, using a careful inspection of XRD data we will investigate the strain state of five layers as a function of the deposited thickness (from 22 to 500 nm) and evidence a correlation between the strain relaxation of the SNO layers and their chemical composition. 2. Experimental 2.1. Film growth SNO films were grown by injection MO-CVD “band flash” [9] on single-crystalline (0 0 1) STO substrates. The liquid started solution is prepared with tris(tetramethylheptanedionato) samarium, Sm(tmhd)3 , and bis(tetramethylheptanedionato) nickel, Ni(tmhd)2 , dissolved in 1,2-dimethoxyethane. Droplets of samarium and nickel precursors of a few microliters are injected in the reactor. A porous band assisted by a gas flow carries the solid reactives to the evaporation zone. Precursors evaporate at 230 ◦ C in a pressure of 10−2 bar. The oxygen and argon gas flow trains precursors to the deposition zone where they are adsorbed, diffused, and decomposed on the surface of the substrate heated at 680 ◦ C. At the end of the deposition, the reactor is filled with oxygen up to 1 bar. Films undergo an in situ annealing at the same temperature for 30 min in a pure oxygen atmosphere. We report here the results obtained for five thicknesses of SNO films: from t ≈ 22 to 500 nm (labeled 1–5). 2.2. Epitaxial relationships The films were characterized by high-resolution XRD. A laboratory diffractometer with a rotating Cu anode, a fourreflection monochromator and a curved position sensitive detector was used to record reciprocal space maps (RSMs) [10,11]. A five-movement sample holder allows precise sample positioning. The X-ray beam impinging on the sample is monochromatic (Cu K␣1 , λ/λ = 1.4 × 10−4 ) and parallel in the detector plane (θ = 12 arcsec). A detailed description of the set-up has been given elsewhere [10,11]. A RSM represents the scattered intensity in the (Qx , Qz ) plane, where Qx and Qz are the components of the scattering vector Q (Q = 4π sin θ/λ) in the film plane and perpendicular to it, respectively. In the following

33

experiments Qx and Qz have been set parallel to the [1 0 0] and [0 0 1] direction of STO, respectively. High-resolution Qx and Qz scans were carried out at the ESRF (Grenoble, France) on the BM2 beam-line [12] (Figs. 2 and 3). The beam is monochromated by a Si(1 1 1) two-crystal monochromator. A single Si(1 1 1) crystal is used as an analyzing crystal. The wavelength was set to ˚ λ = 0.613 A. RSMs recorded in the vicinity of (0 0 l) reciprocal lattice points (RLPs) of STO and ␾-scans performed on the (−1 0 3) RLPs of SNO led us to the following out-of-plane and inplane epitaxial relationships between SNO and STO [15]: (0 0 1)STO //(0 0 1)SNO and [1 0 0]STO //[1 0 0]SNO . 3. Results 3.1. Film thickness The thickness of the SNO layers has been evaluated by means of two techniques depending on the thickness range of the layers. For thin layers, the thickness has been obtained by the simulation of X-ray reflectivity profiles, whereas for thicker layers (t ≥ 100 nm) the thickness has been extracted from the simulation of (0 0 2) scans recorded perpendicular to the sample surface. For samples 1 and 2 RSMs have been recorded in the vicinity of the origin of the reciprocal space (Fig. 1a and c). In Fig. 1a and c the streak inclined towards low Qx values corresponds to the transmittance function of the PSD [11], whereas the streak inclined towards high Qx values correspond to a Yoneda wing [13] (the symmetrical Yoneda wing inclined towards low Qx is hidden by the intense detector streak). The vertical streak corresponds to the specular reflectivity streak, and has been plotted in Fig. 1b and d as a function of the X-ray incidence angle ω. These angular scans exhibit well-defined thickness fringes, the period of which is inversely proportional to the layer thickness. Besides, the fringes contrast and the Q-dependence reflectivity signal are functions of σ s and σ i , respectively, the surface and the interface root-mean-squared roughness. The reflectivity curves have been simulated using the transfer matrix method and the roughness has been accounted for using the Nevot-Croce approximation [13]. The result is shown in Fig. 1b and c. From the simulation of the experimental reflectivity profiles we obtained for sample 1: t = 22 nm, σ s = 0.3 nm, and σ i = 0.5 nm, and for sample 2: t = 38 nm, σ s = 0.4 nm, and σ i = 0.8 nm. For samples 3 and 4 the thickness has been extracted from the simulation of the (0 0 2) Qz scans (recorded perpendicular to the sample surface). Thickness fringes only appear for sample 3 (Fig. 2) and sample 4. In the case of X-ray diffraction, the fringes contrast is not only sensitive to the surface and interface roughness, but also to the crystalline defect density which gives rise to heterogeneous strain. Moreover, Qz scans exhibit an asymmetrical distribution of the scattered intensity which evidences the presence of a vertical strain gradient within the layer. A detailed inspection of Fig. 2 reveals that the fringes pattern is not strictly periodic which prevents one to use the well-known formula, Qz = 2π/t, to extract the film thickness. This aperiodicity can also be attributed to the presence of a vertical strain gradient. The

34

F. Conchon et al. / Materials Science and Engineering B 144 (2007) 32–37

Fig. 1. (a and c) Reflectivity reciprocal space maps recorded for samples 1 and 2 and (b and d) corresponding reflectivity profiles plotted as a function of the incidence angle.

obtaining of the film thickness hence requires the simulation of the data with a model including such a strain gradient. For that purpose we made use of an appropriate model [14] that allows to obtain not only the layer thickness and thickness fluctuations but also the vertical lattice displacement strain profile. The vertical lattice displacement profile u(z) is decomposed into B-spline basis functions: u(z) =

N 

wi Bi,m (z)

i=1

where wi is the weight of the ith B-spline of degree m, Bi,m (x). We choose cubic B-spline functions (m = 3) because of their minimum of curvature which turns out very useful in the case of noisy data because it avoids large amplitude oscillations. Moreover these functions allow the description of a wide range of displacement profiles. A detailed description of the model can be found in [14]. However, in contrast to Ref. [14], in the present case the coherent contribution of the scattering from the substrate has to be taken into account (i.e. we have to add the

amplitudes instead of the intensities as in [14], this point will be detailed elsewhere). The result of the simulation for sample 3 is shown in Fig. 2. The resulting lattice displacement profile is shown in the inset. The thickness is t = 85 nm for sample 3 and t = 150 nm for sample 4. The Qz profile of sample 5 does not exhibit fringes which makes the determination of the thickness unreliable. From the deposition parameters it is estimated to be ∼500 nm. 3.2. Film quality The crystalline quality of SNO films has been previously investigated [15] in terms of relative volume fraction of secondary phases present in the layers as a function of the deposited thickness. Indeed, Qz scans exhibiting several orders of (0 0 l) reflections have been recorded for the five samples. Very weak peaks attributed to NiO have only been observed for thicknesses larger than 150 nm (samples 4 and 5). The simulation of the (0 0 2) NiO peak of sample 5 was performed using Voigt functions. The relative volume fraction of NiO has then been cal-

F. Conchon et al. / Materials Science and Engineering B 144 (2007) 32–37

35

Fig. 3. Qx scans of the (0 0 2) reciprocal lattice point of four SNO layers. The curves are shifted vertically for clarity. Two-component transverse scans are obtained. Fig. 2. Qz scan of the (0 0 2) reciprocal lattice point of sample 3 and corresponding calculated profiles of the scattered intensity along Qz and the vertical lattice displacements.

culated from the volume of each phase as determined with the following equation: V = C(I/PL)(vm /mh k l |F |2 ) [16], where V is the volume of each phase, I the integrated intensity of the XRD peak, P and L the polarization and the Lorentz factors, vm the unit cell volume of each phase, mh k l the multiplicity factor of the considered (h k l) reflection and F is the structure factor of each phase. We obtained a relative volume fraction of NiO of 1.2%, which can be considered as negligible. The high quality of the SNO layers is not only evidenced by a very low tendency toward dissociation of the perovskite but also by the presence of a coherent peak on the (0 0 2) Qx scans (recorded in the surface plane) (Fig. 3). Indeed, the (0 0 2) Qx scans exhibit two intensity components: a narrow Bragg (coherent) peak corresponding to the diffraction from a perfect crystal [17] which attests of a long-range crystalline order, and a broad (diffuse) peak corresponding to local atomic displacements. Such a behavior is now currently observed in epitaxial layers (see [17] and references therein). The high film quality can be further attested by the presence of thickness fringes present on the reflectivity scans for samples 1 and 2 (Fig. 1) and on (0 0 2) longitudinal scans, especially for sample 3 (Fig. 2) and sample 4. As mentioned above, the visibility of thickness fringes strongly depends on the defect density and on the surface and interface smoothness. The weak values obtained for σ s and σ i in the previous section confirm that the SNO layers exhibit smooth surfaces and interfaces. To cut a long story short, Figs. 1–3 hence illustrate that the interface and the surface are smooth and that the layers contain very few structural defects. In the following section, we further discuss the effect of strain and the associated structural defects.

3.3. Strain relaxation The strain state of each layer is investigated by means of reciprocal space mapping of the (−1 0 3) reflection (Fig. 4). A qualitative observation of the maps reveals that increasing thickness (from samples 1 to 5) results in a transverse broadening of the layer RLP concomitantly with a slight displacement of the layer RLP towards low Qx values. From the position of the SNO (−1 0 3) RLP, we calculated a and a⊥ , respectively, the in-plane and the out-of-plane lattice parameters of the strained layers. Contrarily to what is expected, in all cases, the measured value of a⊥ is much higher than the theoretical value of the bulk (i.e. strain free) SNO pseudo-cubic lattice parameter th ˚ ). The lattice spacing of STO (aSTO = 3.905 A) ˚ (aSNO = 3.795 A being larger than the theoretical pseudo-cubic lattice spacing of SNO, the layer is subjected to in-plane tensile strain which should give rise to a contraction of the layer out-of-plane lattice spacing. The measurements of a⊥ led us to the conclusion that the actual lattice parameter of bulk SNO, aSNO , must be higher th . Moreover, as the layer than its commonly assumed value aSNO ˚ and thickness increases a decreases (from 3.905 to 3.876 A) ˚ which seems to evidence a⊥ increases (from 3.805 to 3.810 A) a strain relaxation phenomenon. A careful quantitative interpretation of the (−1 0 3) RSMs allowed us to derive the actual lattice parameter of SNO as well as the strain relaxation rate [18] R using the previously determined values of a and a⊥ , corresponding to each layer. We used the following equations: aSNO = R=

a⊥ + ν2 a ν2 + 1

a − aSTO aSNO − aSTO

36

F. Conchon et al. / Materials Science and Engineering B 144 (2007) 32–37

Fig. 4. (a–e) (−1 0 3) reciprocal space maps of both SNO and STO for samples 1–5, respectively. (f) Plot of aSNO vs. R.

For a fully strained layer, a = aSTO so that R = 0, which gives rise to the alignment of the layer and substrate RLPs along Qz . For a fully strain relaxed layer a = aSNO so that R = 1, which gives rise to the alignment of the layer and substrate RLPs along the scattering vector direction, Q [18]. ν2 is defined as the ratio of the vertical strain ε⊥ to the lateral strain ε and is equal to 2C12 /C11 [19]. Extensive bibliographic research on SNO did not provide any value for ν2 , so, we assumed that the elastic constants C11 and C12 remain the same through the interface, in other words, we neglect the difference of elastic properties between the substrate and the layer. For each sample the following relaxation state has been obtained: R = 0% for sample 1, R = 4% for sample 2, R = 7% for sample 3, R = 12% for sample 4 and R = 42% for sample 5. As the thickness increases the value of ˚ In all cases we indeed aSNO decreases from 3.844 to 3.836 A. th . obtain aSNO > aSNO 4. Discussions Since the nickel must adopt the less stable Ni3+ oxidation state to form SNO, this phase is unstable when prepared under normal temperature and pressure conditions [20] and dissociates into Sm2 O3 and NiO. To circumvent this, epitaxial strain can be used to form metastable SmNiO3 . As R increases the layer tends to recover its bulk lattice parameter. During this relaxa-

tion process, we showed that SNO partially dissociates to form NiO. Complementary measurements on highly dissociated films showed that the NiO phase is textured with respect to the substrate, whereas the Sm2 O3 phase exhibit a random orientation of its crystals. This simple fact clearly explains why Sm2 O3 cannot be detected in the present data. It hence appears that the dissociation of SNO reported many times in the literature is clearly due to the relaxation of epitaxial strain. It is interesting to notice that the evolution of the bulk lattice parameter of SNO is correlated with the relaxation rate, R. The variation of aSNO can be attributed to compositional changes occurring during the dissociation of the SNO phase according to the following equation: 2SmNiO3 → 2NiO + Sm2 O3 + (1/2)O2 + VO 2+ The contraction of the SNO unit cell could be due to the appearance of oxygen vacancies during the dissociation of SNO. However, further studies are needed to confirm this statement. The actual lattice parameter of SNO can be determined by extrapolating the behavior of aSNO for a fully strained hence non-dissociated layer. This is shown in Fig. 4f. The extrapola˚ which coincides with the ted value of aSNO at R = 0 is 3.844 A point corresponding to sample 1. Then, it turns out that m = 1.6% instead of 2.8% (taking into account the commonly published lattice parameter of the bulk SNO phase). This lower lattice

F. Conchon et al. / Materials Science and Engineering B 144 (2007) 32–37

mismatch might very probably be the reason for the successful stabilization of SNO on STO substrates. In order to discover the origin of this higher lattice parameter, wavelength dispersive spectroscopy measurements have been carried out on all samples. It is found that the Sm/Ni ratio varies from 1.3 for sample 1 to 1.2 for sample 5. Extensive TEM studies in simple perovskite compounds (BaTiO3 and SrTiO3 ) [21,22] revealed that A-site excess gives rise to planar defects, the so-called Ruddlesden–Popper (RP) faults and involves an increase of the lattice parameter. From a more general viewpoint, in the case of perovskite-type materials, these defects fall into the broader defects family of defects known as out-of-phase boundaries (OPBs) [23]. It can be expected that similar defects, also occur in SNO which could account for the higher lattice parameter observed. For a RP fault lying in a (1 0 0) plane, the fault is characterized by a 1/2 [0 1 1] translation vector. With such a vector the mean distance between RP faults in the nondissociated sample should be around 15 nm. Preliminary TEM observations seem to indicate the existence of planar defects separated by ∼10–20 nm. However further investigations are required to confirm the existence of these defects. 5. Conclusions In summary, we have succeeded in the stabilization of epitaxial films of SmNiO3 on (0 0 1) SrTiO3 substrates by an MO-CVD process. These films present a high crystalline quality principally attested by a very low dissociation rate of this metastable phase and by smooth surfaces and interfaces. This stabilization process has been achieved because the actual lattice mismatch between SmNiO3 and SrTiO3 is not as high as usually expected (1.6 instead of 2.8%). Using reciprocal space mapping the actual lattice parameter of the bulk and non-dissociated SNO phase and then the actual lattice mismatch between the substrate and the layer have been derived. These results allowed us to demonstrate the correlation between strain relaxation and dissociation of the SmNiO3 phase into NiO and Sm2 O3 . Finally, the high value of lattice parameter obtained for the bulk SNO phase, as compared to the usually assumed value, is interpreted in terms of Ruddlesden–Popper faults.

37

References [1] J.B. Lacorre, A.I. Nazzal, E.J. Ansaldo, Ch. Niedermayer, Phys. Rev. B 45 (1992) 14. [2] F. Capon, P. Ruello, J.F. Bardeau, P. Simon, P. Laffez, B. Dkhil, L. Reversat, K. Galicka, A. Ratuszna, J. Phys.: Condens. Matter 17 (2005) 1137–1150. [3] P. Laffez, R. Zaghrioui, I. Ronot, T. Brousse, P. Lacorre, Thin Solid Films 354 (1999) 50. [4] J. P´erez-Cacho, J. Blasco, J. Garcia, M. Castro, J. Stankiewicz, J. Phys.: Condens. Matter 11 (1999) 405–415. [5] G. Demazeau, A. Marbeuf, M. Pouchard, P. Hagenmuller, J. Solid State Chem. 3 (1971) 582. [6] M.A. Novojilov, O.Yu. Gorbenko, I.E. Graboy, A.R. Kaul, H.W. Zandbergen, N.A. Babushkina, L.M. Belova, Appl. Phys. Lett. 76 (15) (2000) 2041–2043. [7] P. Goudeau, P. Laffez, M. Zaghrioui, E. Elkaim, P. Ruello, Cryst. Eng. 51 (2002) 317–325. [8] O.Yu. Gorbenko, S.V. Samoilenkov, I.E. Graboy, A.R. Kaul, Chem. Mater. 14 (2002) 4026–4043. [9] N. Ihzaz, S. Pignard, J. Kreisel, H. Vincent, J. Marcus, J. Dhahri, M. Oumezzine, Phys. Status Solidi (c) 1 (7) (2004) 1679–1682. [10] O. Masson, A. Boulle, R. Guinebretiere, A. Lecomte, A. Dauger, Rev. Sci. Instrum. 76 (2005) 063912. [11] A. Boulle, O. Masson, R. Guinebreti`ere, A. Lecomte, A. Dauger, J. Appl. Cryst. 35 (2002) 606. [12] J.L. Ferrer, J.P. Simon, J.F. B´erar, B. Caillot, E. Fanchon, O. Ka¨ıkati, S. Arnaud, M. Guidotti, M. Pirocchi, M. Roth, J. Synchrotron Radiat. 5 (1998) 1346–1356. [13] V. Hol´y, U. Pietsch, T. Baumbach, High Resolution X-ray Scattering from Multilayers and Thin Films, Springer Tracts in Modern Physics, vol. 149, Springer, Berlin, 1999. [14] A. Boulle, O. Masson, R. Guinebreti`ere, A. Dauger, J. Appl. Cryst. 36 (2003) 1424–1431. [15] F. Conchon, A. Boulle, C. Girardot, S. Pignard, R. Guinebreti`ere, E. Dooryh´ee, J.-L. Hodeau, F. Weiss, J. Kreisel, J.-F. B´erar, J. Phys. D: Appl. Phys. 40 (2007) 4872–4876. [16] B.E. Warren, X-ray Diffraction, Addison-Wesley, New York, 1969. [17] A. Boulle, R. Guinebreti`ere, A. Dauger, J. Phys. D: Appl. Phys. 38 (2005) 3907–3920. [18] H. Heinke, M.O. M¨oller, D. Hommel, G. Landwehr, J. Cryst. Growth 135 (1994) 41–52. [19] D.J. Dunstan, J. Mater. Sci: Mater. Electron. 8 (1997) 337–375. [20] A. Ambrosini, J.F. Hamet, Appl. Phys. Lett. 82 (5) (2003) 727–729. [21] M. Fujimoto, J. Cryst. Growth 237–239 (2002) 430–437. [22] T. Suzuki, Y. Nishi, M. Fujimoto, Philos. Mag. A 80 (2000) 621–637. [23] M.A. Zurbuchen, W. Tian, X.Q. Pan, D. Fong, S.K. Streiffer, M.E. Hawley, J. Lettieri, Y. Jia, G. Asayama, S.J. Fulk, D.J. Comstock, S. Knapp, A.H. Carim, D.G. Schlom, J. Mater. Res. 22 (6) (2007) 1439–1471.