Research Article www.acsami.org
Layer-Resolved Cation Diffusion and Stoichiometry at the LaAlO3/ SrTiO3 Heterointerface Probed by X‑ray Photoemission Experiments and Site Occupancy Modeling Gabriele Salvinelli,† Giovanni Drera,† Alessio Giampietri,† and Luigi Sangaletti*,† †
Interdisciplinary Laboratories for Advanced Materials Physics (I-LAMP) and Dipartimento di Matematica e Fisica, Università Cattolica del Sacro Cuore, via dei Musei 41, 25121 Brescia, Italy ABSTRACT: The layer-resolved cation occupancy for different conducting and insulating interfaces of LaAlO3 (LAO) thin films on SrTiO3 (STO) has been determined by angle-resoled X-ray photoelectron spectroscopy (AR-XPS). Three STO interfaces with LAO have been considered, namely, a conducting interface with a 5 unit cell (u.c.) LAO layer, an insulating interface with a 5 u.c. LAO layer, and an insulating interface with a 3 u.c. LAO layer. Considering inelastic and elastic scattering processes in the transport approximation, the core-level signal attenuation has been modeled on the basis of Monte Carlo calculations of the electron trajectories across the heterostructures. Different effects involving cation stoichiometry and diffusion through the interface have been considered to interpret data. Beyond a mere abrupt interface modeling, the LaAlO3/SrTiO3 heterojunction is shown to host cation diffusion processes within 3−4 unit cells in the bulk layer, along with a clear Sr substoichiometry, an issue so far virtually neglected in the analysis of these systems. The present results show the capability of the AR-XPS modeling to explore element-sensitive properties at the oxide interfaces, matching and completing the information that can be provided by probes based on electron microscopy or X-ray scattering. KEYWORDS: oxides, photoemission, heterointerfaces, LaAlO3, SrTiO3, cation interdiffusion, intermixing
■
INTRODUCTION Interfaces between perovskite oxides display properties at the nanometer scale which are unexpected considering their parent compounds alone, allowing one to engineer novel functionalities through the growth of epitaxial heterostructures.1−3 An extraordinary attention has been focused on the heterojunction based on a thin film of LaAlO3 (LAO) grown on a SrTiO3 (STO) TiO2-terminated substrate4−32 due to the wide variety of unexpected properties arising at the interface, from the twodimensional electron gas (2DEG)4,8,12 to the coexistence of superconductivity and ferromagnetism,22−25 as well as the electric-field-induced metal−insulator phase transition.8 Despite a long-standing fundamental research focused to control the interface properties and their origin, many scientific questions yet remain to be answered before the goal can be shifted toward the applications. One of the most controversial issues at the center of the debate is the presence of the cation diffusion across the n-type LAO/STO and its role in determining the physical properties of the system. Interdiffusion is expected between similar cations (La−Sr as A-type cations and Al−Ti as B-type cations in ABO3 perovskites). Furthermore, lanthanum ions on strontium sites 5,14,33−35 and oxygen vacancies in the STO substrate9−11,16,36 have been proposed as predominant defects at the LAO/STO-interface, since both types of defects can cause © 2015 American Chemical Society
n-type conductivity in STO. Exposure to oxygen at high temperatures, aimed to reach a control over the oxygen content, is also expected to induce the formation of cation vacancies, as discussed in ref 37, suggesting an interplay between the cation and anion stoichiometries in LAO-STO. Cation stoichiometry within the LAO layer is also recognized to play a relevant role in the onset of the 2DEG at the interface, as demonstrated by Warusawithana et al.38 So far, several studies have evidenced a cation disorder at the junction using different techniques, such as angular dark field (ADF) imaging in scanning transmission electron microscopy (STEM),5,6,34,39 surface X-ray diffraction (SXRD),14,40 angleresolved X-ray photoelectron spectrocopy (AR-XPS),33,41 crystal truncation rod (CTR),40 Rutherford backscattering (RBS),33 cross-sectional STEM with electron energy loss spectroscopy (STEM-EELS),33,38 X-ray reflectivity (XRR),40 medium-energy ion spectroscopy (MEIS),33,34,42 time-of-flight secondary ion mass spectrometry (ToF-SIMS),33 and Kelvin probe force microscopy (KPFM).42 The cation interdiffusion is found to be present for different deposition techniques generally pulsed laser deposition (PLD) and molecular beam Received: July 7, 2015 Accepted: November 2, 2015 Published: November 2, 2015 25648
DOI: 10.1021/acsami.5b06094 ACS Appl. Mater. Interfaces 2015, 7, 25648−25657
Research Article
ACS Applied Materials & Interfaces
Table 1. Growth Parameters and Cation Disorder Effects Related to the LAO/STO n-Type Samples Characterized in Refs 14, 33, 38, 40, and in the Present Study LAO/STO n-type
growth technique
temperature during the growth
5 u.c. 4 u.c. 4 u.c. 1−5 u.c. 8 u.c.
PLD PLD PLD PLD MBE
770 700 770 770 680
3 u.c.
PLD
800 °C
1·10−3
5 u.c.
PLD
800 °C
1·10−3
5 u.c.
PLD
800 °C
1·10−1
°C °C °C °C °C
O2 partial pressure (mbar) 5·10−6 1·10−5 8·10−6 3·10−5 1.6·10−6
cation disorder across the LAO film and at the interface
ref
La ↔ Sr and Al ↔ Ti interdiffusion La ↔ Sr and Al ↔ Ti interdiffusion La ↔ Sr and Al ↔ Ti interdiffusion surface La-vacancy, La ↔ Sr interdiffusion La ↔ Al substitution, La ↔ Sr and Al ↔ Ti interdiffusion, Alcomplex vacancy La↔Al substitution, La↔Sr intermixing, and Al↔Ti interdiffusion, Sr-vacancy La ↔ Al substitution, La ↔ Sr intermixing, and Al ↔ Ti interdiffusion, Sr-vacancy La ↔ Al substitution, La ↔ Sr intermixing, and Al ↔ Ti interdiffusion, Sr-vacancy
14 33 33 40 38 this work this work this work
Ti−Al intermixed layers or Ti-vacancies at the interface are suggested to structurally compensate the diverging electrostatic energy build-up. Recently, a study of Warusawithana et al.38 related the arise of the interfacial conductivity to the La/Al stoichiometry ratio of the film. In facts, on the basis of STEM-EELS measurements on LAO/STO 8 u.c. samples (grown by MBE at 680 °C in pO2 = 1.3·10−6 mbar, monitored by RHEED), both insulating and conductive samples have evidenced cation interdiffusion across the junction, but no correlation between conductivity and a specific exchange of cation pairs has been observed. On the other hand, the insulator-to-metal transition has been obtained only for La/Al ≤ 0.97 ± 0.03, having ruled out the extrinsic origin of defects during the growth. This interpretation has been found to be consistent with the polar catastrophe model; indeed, combining the DFT calculations and the results of STEM-EELS, the formation of Al2O3-vacancy-complexes at the interface has been suggested for insulating samples, thereby removing the divergent potential without the transfer of electronic charge. For conductive samples, an Al−La substitution process should prevent the cation migration, suggesting an electronic charge transfer to justify the 2DEG origin. From the overview so far outlined, and summarized in Table 1, it is clear that LAO/STO heterointerfaces host a relevant number of problems typical of the physical and chemical properties of oxide junctions. Some of these issues were, to some extent, unexpected or overlooked in the early investigations, which were mainly focused on the control of the 2DEG through band engineering. In this study, an AR-XPS investigation is presented for a 5 u.c. conductive (C-5uc), 3 u.c. insulating (I-3uc), and 5 u.c. insulating (I-5uc) LAO/STO interfaces. The present approach is not meant to be a quick characterization of large batches of samples; rather, it is aimed to address a specific issue (cation stoichiometry and diffusion) in few selected and well-characterized samples. By improving the modeling for the AR-XPS analysis on the basis of a refined Monte Carlo (MC) depth distribution function (DDF) approach, this study shows the possibility of the AR-XPS method to address a set of problems in the interface characterization, with the aim to track the cation profiles across the heterojunction. The abrupt interface and the ideal 1:1 cation stoichiometric ratio (La/Al = Ti/Sr = 1) are excluded from the AR-XPS results, evidencing a complex cation distribution across the junction within 3−4 u.c. from the
epitaxy (MBE)even though the surface quality of the LAO film during the growth is monitored by reflection high-energy electron diffraction (RHEED) to verify the two-dimensional layer-by-layer growth of the film. Usually, the low surface roughness is further confirmed by atomic force microscopy (AFM), showing smooth terraces with clear unit cell steps which follow the surface morphology of the TiO2-terminated STO substrates.43 However, it is noteworthy that the growth conditionsespecially the annealing procedure and the oxygen pressure during/after the film depositioncan significantly affect the two-dimensional layer-by-layer growth of the film,43 in terms of stoichiometry and/or surface roughness. Nevertheless, in the 10−5−10−3 mbar range of oxygen partial pressures and for growth temperatures between 750 and 850 °C, a LAO film can be grown with an interface of high quality, as demonstrated by different groups.6,13 In this complex scenario, we can retrieve several hints about the distribution and the role of cation disorder at the LAO/ STO interface, equally supported by ab initio DFT calculations. For instance, Willmott et al.14 have revealed by SXRD the layerby-layer cation concentration on a 5 unit cells (u.c.) thick interface (grown by PLD at 770 °C in pO2 = 5·10−6 mbar), showing a nonabrupt junction where the cation interdiffusion occurs at different depths for two cation pairs, i.e. Sr−La and Ti−Al. From the resulting intermix, a layer of about 1−2 u.c. of La1−xSrxTiO3 is evidenced at the interface, leading to a lattice expansion due to presence of Ti3+ ions, which may induce the interfacial conductivity. A similar cation distribution has been evidenced by Chambers et al.33,41 on 4 u.c. thick interfaces (grown by PLD at 770 °C in pO2 = 1·10−5 and 8·10−6 mbar), the most accurate AR-XPS study presented so far. The deposition was monitored by RHEED, and the terrace-step surface morphology was verified by AFM. This work suggests both the Sr and Ti diffusion in LAO and La/Al diffusion in the bulk, although the model is not layer-resolved, and it is based on a simple exponential approximation of core level intensities. Subsequently, Vonk et al.40 have evaluated a La-interdiffusion depth of 2 u.c. by XRR on 1−5 u.c. thick interfaces (grown by PLD at 770 °C in pO2 = 3·10−5 mbar; RHEED and AFM interface quality check). This result is corroborated by a CTR investigation,40 which detects a significant amount of Lavacancies in the filmespecially in the overlayerup from the deposition of the first LAO layer; the latter outcome is interpreted as a La↔Sr exchange during the deposition in both the STO bulk and the topmost layer of the LAO film, thereby forming charge-neutral LaxSr1−1.5xO layers. As a consequence, 25649
DOI: 10.1021/acsami.5b06094 ACS Appl. Mater. Interfaces 2015, 7, 25648−25657
Research Article
ACS Applied Materials & Interfaces nominal interface and a Sr-substoichiometry so far virtually neglected in the physics of LAO-STO interfaces.
■
EXPERIMENTAL AND COMPUTATIONAL DETAILS
Sample Preparation. The LAO/STO heterostructures were grown by PLD at the MESA+ Institute for Nanotechnology, University of Twente. A couple of n-type 3 u.c. insulating (I-3uc) and 5 u.c. conductive (C-5uc) LAO/STO samples were grown at the same pO2 ∼ 10−3 mbar oxygen partial pressure. In addition, an insulating n-type 5 u.c. (I-5uc) LAO/STO sample was grown at 10−1 mbar O2 partial pressure. The I-3uc and I-5uc samples showed a sheet resistance above 1 GΩ/□, while the C-5uc sample exhibited a sheet resistance of 5.5 kΩ/□ at 300 K. The electronic properties of the present samples have been carefully characterized in ref 44 (all three samples) and ref 45 (band alignment of I-3uc and C-5uc across the interface; 2DEG states of the C-5uc sample), whereas the transport properties of the C-5uc sample are presented in ref 44. Experimental Setup. AR-XPS measurements were collected at the Surface Science and Spectroscopy Laboratory of the Università Cattolica del Sacro Cuore (Brescia, Italy) using a nonmonochromatized dual-anode PsP X-ray source with an Al Kα line (hν = 1486.6 eV) to maximize the probing depth. The electron analyzer was a SCIENTA R3000 (calibrated according to the procedure described in ref 46), operating either in transmission-mode, which maximizes the transmittance, or in angular-mode, which provides the highest angularresolution in the polar direction. All experiments were carried out with a base pressure of 2·10−10 mbar. Method and Modeling. The method described below is aimed to evaluate the cation occupancy in a layer-resolved model of the LAOSTO heterointerface represented in Figure 1, which shows the layered
Figure 1. Layered-model for a 5 u.c. LAO/STO interface used for the DDF calculations. The height of the bars represents the cation occupancy of the layer. One = 100% occupancy; 1 u.c. = 2 atomic layers; in the ABO3 perovskite structure of LAO and STO, the A sites are La and Sr, while B sites are Al and Ti.
structure of SrO-TiO2 planes in the STO substrate and LaO-AlO2 planes in the LAO film. AR-XPS experiments consist in collecting the core-level photoemission peak intensity as a function of the polar angle θ, which is defined by the surface normal and the direction of the photoelectron detection, as displayed in Figure 2a. This method is widely used to probe nondestructively the chemical composition as a function of depth and evaluate the thickness of thin films in multilayers. The physical origin of the angular dependence attenuation of the core-level photoemission intensity stems from the inelastic and elastic scattering processes, which the photoelectrons undergo during the propagation from the emitter to the surface.
Figure 2. (a) Schematic of the experimental geometry: polar angle θ, azimuthal angle φ. The α angle between the X-ray source and the electron analyzer is fixed at 54.5°, eliminating the asymmetry correction. (b) XPD pattern for the Sr 3d core level in the I-3uc LAO/STO sample. The green dotted line points out the polar scan at φ = 0°, while the green arrow indicates the angular φ-range used for the average. (c) Angular-mode φ-averaged AR-XPS profiles for the La 4d (red solid line), Sr 3d (green solid line), and Al 2s (gray solid line) peaks. Angular-mode AR-XPS profiles at φ = 0° are displayed for comparison (filled circles). XPS spectra of La 4d, Al 2s, and Sr 3d core 25650
DOI: 10.1021/acsami.5b06094 ACS Appl. Mater. Interfaces 2015, 7, 25648−25657
Research Article
ACS Applied Materials & Interfaces
where n is the label for the atomic layer, N is the number of the unit cells in the LAO film, and Cjn is the occupation factor of the cation j in the atomic layer n. In this way, the photoemission intensity for each core-level is the sum over all the layers involving the selected cation. In the following analysis, the coefficient of determination R2 has been used for the fitting-accuracy evaluation, as
Figure 2. continued levels at φ = 0°, θ = 70° (d), φ = 0°, 0° < θ < 70° (e) and φ = 0°, θ = 0° (f).
R2 = 1 −
Considering in first approximation only the inelastic scattering processes, this phenomenon can be treated as an isotropic damping in relation with the distance crossed by the photoelectrons inside the matter, i.e. the inelastic mean free path (IMFP), λi.47 In this frame, the probability of an electron to reach the surface is modeled as a simple exponential P(z) ∝ exp[−z/λi·cos θ ], so that the effective total depth is approximated as 3(λi·cos θ). Such an approximation leads to a simple analytical expression for the peak areas, which is defined as straight line motion by Tilinin et al.48 However, this formulation may result in an overestimation of the top layer thickness, especially for thin overlayers. Indeed, different studies indicate that some electrons can change their direction of motion due to elastic scattering processes (accounted for by introducing the transport mean free path, TMFP, with a characteristic λelastic), thereby modifying both the angular predictionbased on only inelastic collisionsand the depth distribution function, even for samples consisting of low atomic number elements.48,49 As a result, the Monte Carlo technique has been widely used to achieve accurate evaluations in the effective study of transport processes.49−51 If both the elastic and inelastic electronic scattering processes are taken into account, the XPS core-level peak intensity I(Ek; θ) of a selected layer at a depth d with a thickness t can be evaluated through the following formula:
I(E k , θ) = K
∫d
Table 2. Calculated Values of IMFP(λi), TMFP(λelastic), λeff for the Cation Core Levels Considered in the Model Simulations LaAlO3 (6.41 g/cm3)
N n=0
ϕ(E k ; θ ; z)dz
N
ILa(E k ; θ) = K · ∑ C 2Lan + 1· n=0 100
ITi(E k ; θ ) = K · ∑ C2Tin · n=N
ISr(E k ; θ ) = K · ∑ n=N
C2Srn + 1·
ϕ(E k ; θ ; z)dz
(3)
2n + 1
∫2n
100
(2)
2n + 2
∫2n+1
ϕ(E k ; θ ; z)dz
ϕ(E k ; θ ; z)dz
KE (eV)
IMFP (Å)
TMFP (Å)
λeff (Å)
IMFP (Å)
TMFP (Å)
λeff (Å)
Ti 2p Sr 3d Al 2s La 4d Al 2p
1027.8 1347.4 1364 1382.6 1409
19.45 23.93 24.15 24.41 24.75
62.07 86.96 88.39 90.07 92.38
14.81 18.76 18.96 19.20 19.53
22.12 27.26 27.51 27.79 28.20
75.28 109.18 110.98 113.05 116.12
17.10 21.81 22.04 22.31 22.71
■
(4)
RESULTS AND DISCUSSION Film Stoichiometry Evaluation. The dependence of the La 4d/Al 2s XPS peak area ratio on the growth conditions for
2n + 2
∫2n+1
core level
0° in Figure 2a) about 95% of the signal comes from photoelectrons emitted within a depth 3λeff below the surface, since the signal attenuation with depth z is assumed to vary as I(z) = I0exp(−z/λeff). Data Treatment. In periodic lattices, a relevant modulation of the photoemission intensity occurs in relation with polar θ and azimuthal φ angles (see Figure 2a) due to the X-ray photoelectron diffraction (XPD). In principle, a fully quantitative analysis of the elastic scattering effects, based on a multiple scattering formalism,54 can provide local structural information around each photoemitter. The manifold of phenomena occurring at the LAO/STO heterojunction (e.g., cation substitution, diffusion, vacancies) make the match of the measured XPD profiles with a model of the local crystal structure quite demanding. The present study is focused on the overall dependence on θ of the XPS peak intensity, where the XPD effects appear as modulations on top of the background signal, which is modeled though eqs 3−6. In order to reduce the θ-dependent photodiffractive modulation, an averaging of the θ-profiles over the azimuthal φ-angles for each core-level has been carried out. For this purpose, we have collected the integrated area for Sr 3d, Al 2s and La 4d core-levels in the 0°−65° range of the polar angle θ and 0°−45° for the azimuthal angle φ, thereby obtaining the XPD patterns, as shown in Figure 2b for the Sr 3d peak. Subsequently, we have averaged each polar scan on the azimuthal angle φ, obtaining the averaged AR-XPS profiles. Thus, the sharp XPD modulations are reduced, as shown in Figure 2c by the comparison between the φ = 0° (dots) and the φ-averaged profiles (thick lines). Because the photoemission intensity of the Ti 2p core level is weaker than those of Sr 3d and La 4d (because of the photoemission cross section and the signal attenuation due to the capping LAO), the photodiffraction modulations are significantly reduced, allowing to collect a simple representative polar scan at φ = 0° and 0° < θ < 65°.
2n + 1
∫2n
SrTiO3 (5.12 g/cm3)
(1)
where Φ(Ek ; θ; z) is the generic escape probabilityalso known as depth distribution function (DDF)of an electron generated at a depth z with a kinetic energy Ek at an angle θ with respect to the surface normal. K is a normalization constant which includes the total photoemission cross section,52 the photon flux density and the atomic density for the selected species. In this study, the DDF function has been calculated with a Monte Carlo (MC) approach based on the algorithm described by Werner,49 in order to include both inelastic and elastic electronic scattering processes in the so-called transport approximation (TA).53 The photoemission asymmetry parameter is also taken into account, as well as the analyzer acceptance angle. The MC calculations of the electron trajectories, which are needed for an effective treatment of the θangular dependence of XPS peak intensities, have been carried out with the BriXias package (see ref 46 for further details), designed to calculate XPS spectra on complex multilayer structure, carefully considering both the experimental geometry and the features of the Xray source (energy, polarization). On this basis, from the multilayer model shown in Figure 1which reproduces the layered structure of SrO-TiO2 in the STO substrate and LaO-AlO2 in the LAO filmthe DDF on the LAO/STO heterostructure has been calculated by using the following formulas:
IAl(E k ; θ ) = K · ∑ C 2Aln ·
(6)
with SSres = ∑i(yi − f i)2 and SStot = ∑i(yi − y)̅ 2, where y ̅ is the mean value of data. As mentioned above, the calculation method for the AR-XPS method implies the direct numerical evaluation of the DDF from the inelastic and elastic mean free paths (IMFP and TMFP), i.e. from the probabilities of the photoelectron to perform an elastic or inelastic scattering before reaching the analyzer. It is possible to give a rough estimate of the probing depth by using the combined scattering length λeff = λi·λelastic/(λi + λelastic), shown in Table 2. At normal takeoff (θ =
d+t
Φ(E k , θ , z)dz
SSres SStot
(5) 25651
DOI: 10.1021/acsami.5b06094 ACS Appl. Mater. Interfaces 2015, 7, 25648−25657
Research Article
ACS Applied Materials & Interfaces
Figure 3. (a) 5 u.c.conductive, (b) 5 u.c. insulating, (c) 3 u.c. insulating LAO/STO sample: φ-averaged AR-XPS data of the LAO/STO samples for La 4d (red line) and Al 2s (gray line) core levels for a stoichiometry evaluation of the film compounds. Calculated curves (black dotted lines) for a model based on (i) the abrupt interface, (ii) the exact stoichiometry of the bulk compounds, and (iii) the nominal thickness of the LAO film.
Figure 4. Top panels: (a) 5 u.c. conductive, (b) 5 u.c. insulating, (c) 3 u.c. insulating LAO/STO sample. φ-averaged AR-XPS data of the LAO/STO samples for La 4d (red line), Al 2s (gray line), Sr 3d (green line), and Ti 2p (blue dots) core levels. The calculated profiles (dashed lines) are obtained on the basis of (i) the abrupt interface, (ii) the nominal thickness of the LAO film, (iii) the exact stoichiometry for the bulk compounds, and (iv) the correction of the La/Al stoichiometry of the film. Bottom panels (d−f) display the stack of layers (blue, TiO2; gray, AlO2; green, SrO; red, LaO) related to (a−c) panels, respectively.
stoichiometry ratio of 0.71 for C-5uc (a), 1.26 for I-5uc (b) and 0.68 for I-3uc (c) samples. Then, by holding the La and Al cation occupancies just found, we have calculated the AR-XPS profiles also for Sr 3d and Ti 2p core levels. The Sr and Ti occupancy was set to 1. The results are shown in Figure 4a−c for the AR-XPS profiles and in panels (d−f) for the schematics of the resulting layered-models. In all samples (a−c), we can observe a quite good match for the Ti 2p profile (dotted line) and a rather pronounced overestimation for the Sr 3d case (dot-dashed line). The latter evidence suggests that the experimental AR-XPS data cannot be modeled in the abrupt interface approximation, since any additional layer on the surface (homogeneous or islands) would decrease both the Sr and Ti profile intensity. In other words, within the restriction of the abrupt interface is not possible to correct the Sr 3d profile leaving the intensity of the Ti 2p unchanged. Therefore, other structural modifications should be considered either at the interface (e.g., cation substitution, vacancies,
the present LAO/STO samples was qualitatively evidenced in a previous study (Figure 1a of ref 44). For this reason, we have started the analysis from the AR-XPS profiles of La 4d and Al 2s alone, in order to evaluate first the average La/Al stoichiometry ratio of the films. On the basis of the layeredmodel of Figure 1, we have assumed that (i) the La/Al ratio is not dependent on the LAO thickness (ii) the film thickness is as thick as the nominal growth value (i.e., 5 u.c. and 3 u.c.) and (iii) the junction between LAO and STO is abrupt. Using the overall cation concentration of La and Al as fitting variables, we have taken into account the conclusions of Warusawithana et al.38 in order to change the La/Al stoichiometry in the layeredmodel through AlO2-complex vacancies for the I-5uc sample, which shows an insulating character with a La/Al ratio >1, and La↔Al substitutions for the other samples, because they exhibit interfacial conductivity above the LAO critical thickness (i.e., 4 u.c.) and show a La/Al ratio
