Synthesis, Structure, and Spectroscopy of Epitaxial EuFeO3 Thin Films

Jan 21, 2015 - A. K. Choquette , C. R. Smith , R. J. Sichel-Tissot , E. J. Moon , M. D. Scafetta , E. Di Gennaro , F. Miletto Granozio , E. Karapetrov...
0 downloads 0 Views 6MB Size
Download PDF
Article pubs.acs.org/crystal

Synthesis, Structure, and Spectroscopy of Epitaxial EuFeO3 Thin Films Amber K. Choquette,† Robert Colby,‡,∥ Eun Ju Moon,† Christian M. Schlepütz,§ Mark D. Scafetta,† David J. Keavney,§ and Steven J. May*,† †

Department of Materials Science and Engineering, Drexel University, Philadelphia, Pennsylvania 19104, United States Environmental Molecular Sciences Laboratory, Pacific Northwest National Laboratory, Richland, Washington 99352, United States § X-ray Science Division, Argonne National Laboratory, Argonne, Illinois 60439, United States ‡

S Supporting Information *

ABSTRACT: Rare earth iron perovskites RFeO3, where R is a rare earth cation, exhibit an array of magnetic, catalytic, optical, and electrochemical properties. Here we study EuFeO3 films synthesized by molecular beam epitaxy to improve our understanding of the optical properties of ferrites. A combination of X-ray diffraction, X-ray reflectivity, Rutherford backscattering spectroscopy, and scanning transmission electron microscopy was used to characterize the film structure and cation composition. X-ray absorption spectroscopy confirms the nominal 3+ valence states of Eu and Fe. The optical properties of EuFeO3 were investigated using variable-angle spectroscopic ellipsometry between the photon energies of 1.25 and 5 eV. We find that EuFeO3 is a semiconductor with an onset of optical absorption near 2.5 eV. The absorption spectrum of EuFeO3 is blue-shifted with respect to LaFeO3 films, a result that is attributed to the structural differences between the two materials. attention.23−25 In perovskites, rotations and distortions of the corner-connected BO6 octahedra act to reduce electronic bandwidth, which is often defined as the energy width of valence and/or conduction bands. In semiconducting materials, a reduction in electronic bandwidth would be expected to increase the band gap. Indeed, previous ab initio and experimental studies have shown that the optical band gap of d0 perovskites can be increased by decreasing the B−O−B bond angles.23 However, similar studies of non-d0 perovskites, in which electronic correlations play a critical role in electronic structure, are lacking. We focus on epitaxial EuFeO3 (EFO) thin films and compare the optical properties of this material to those of LaFeO3 (LFO). In bulk, EFO exhibits larger octahedral rotations than LFO because of the reduced size of Eu3+ compared to that of La3+; the average Fe−O−Fe bond angles are 147.9° in EFO and 157.6° in LFO.11,26 The difference in cationic size is also reflected in the pseudocubic lattice parameters (ap) of the two materials, where ap = 3.869 Å for EFO and ap = 3.931 Å for LFO. We report on the synthesis of epitaxial EFO thin films and the characterization of their structural, chemical, and optical properties. The quality of the films is established using a combination of X-ray diffraction, X-ray reflectivity, Rutherford backscattering spectroscopy, and scanning transmission electron microscopy. We then move to spectroscopic probes to

1. INTRODUCTION Interest in rare earth iron perovskites has grown considerably in recent years because of the wide array of ferroic, catalytic, electrochemical, and optical properties present in these compounds.1−8 The most studied material within this family is LaFeO3, in which Fe exhibits a high-spin 3d5 electronic configuration. The Hund’s splitting of the hybridized Fe 3d−O 2p states leads to a band gap on the order of 2−2.6 eV, depending on how the optical absorption spectra are analyzed.9,10 In addition to promoting an insulating state, the d5 configuration results in G-type antiferromagnetism arising from superexchange interactions between the full eg−full eg states.1 The substitution of isovalent rare earth cations in ferrites does not alter the basic magnetic interactions or insulating behavior; instead, the main consequences of such substitutions are structural. Substituting smaller cations, for example, Eu for La, increases the magnitude of the octahedral rotations and distortions present within the orthorhombic Pbnm crystal structure.11 While there have been advances in the optical studies of the orthoferrites,12−15 the implications of changes in atomic structure on the electronic structure and resultant optical properties of rare earth ferrites have not been fully explored. In this study, we analyze the sensitivity of the optical band gap to the structural distortions in rare earth ferrites. While previous studies of other perovskite systems have demonstrated the relationship among magnetic order, metal−insulator transitions, and subtle structural changes,16−22 the impact of octahedral rotations on optical properties has received less © XXXX American Chemical Society

Received: September 18, 2014 Revised: January 13, 2015

A

DOI: 10.1021/cg501403m Cryst. Growth Des. XXXX, XXX, XXX−XXX

Article

Crystal Growth & Design

Figure 1. (a) (001) data of EFO on STO and (b) (001) data on LSAT, with a corresponding fit. (c) 2θ−θ scan from 15° to 60° showing the absence of any secondary phases. (d) Substrate (top) and film (bottom) omega scans about (001). The fwhm of the substrate is 0.022°, while it is 0.046° for the film. The RSMs about the (113) peak were measured from an EFO film on STO (e) and on LSAT (f); the in-plane lattice parameters of the films match those of the substrates. Photon Source, using a PILATUS 100K area detector and a photon energy of 15 keV. Cross sections for STEM analysis were prepared using an FEI Helios dual-beam focused ion beam/scanning electron microscope (FIB/SEM) equipped with an Omniprobe, using the standard lift-out technique.29 High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images were recorded using a probe-corrected FEI Titan 80-300 instrument operated at a 300 kV. The optical absorption spectra were recorded with a J. A. Woolam M-2000U variable-angle spectroscopic ellipsometer at room temperature over the photon energy range of 1.25−5 eV. Ellipsometry measurements were taken at five angles from 65° to 75° to improve the data fitting. The measured Ψ and Δ were fit using WVASE software to extract the index of refraction (n) and the extinction coefficient (k) of the film. Optical absorption was ascertained using the equation α = 4πk/λ, where α is the absorption coefficient and λ is the incident photon wavelength.

gain insight into the electronic structure. X-ray absorption spectroscopy confirms nominal 3+ valence states of both Eu and Fe. The optical absorption of the EFO films is presented and compared with that of LFO films to probe the spectral differences arising from the structural differences of the materials. We find that the absorption edge of EFO is blueshifted relative to that of LFO, consistent with a structurally induced increase in the band gap.

2. EXPERIMENTAL METHODS Thin films of EFO were deposited by molecular beam epitaxy (MBE) (Omicron modified LAB-10 system). All films were deposited on single-crystal SrTiO3 (001) and (LaAlO3)0.3(Sr2AlTaO6)0.7 (100) substrates (MTI Corp.). The SrTiO3 substrates were prepared using the procedure described by Connell et al. involving two sets of annealing steps at 1000 °C in air for 1 h, each followed by a 30 s deionized water etch.27 A substrate temperature of 600 °C was maintained during deposition, and reflection high-energy electron diffraction (RHEED) was used to monitor the growth in situ. The films were grown in an O2/O3 environment (approximately 5% O3 in O2) with a dosing valve used to fix the chamber pressure at ∼1.0 × 10−6 Torr, approximately 2−3 orders of magnitude higher than the base pressure of the system when the effusion cells are at temperature. The metallic cations were sublimed from elemental sources, with the flux monitored pre- and postgrowth by a quartz crystal monitor that was calibrated using Rutherford backscattering spectroscopy and X-ray reflectivity. Al2O3 and pyrolytic boron nitride crucibles were used for the evaporation of Fe and Eu, respectively. The metal cations were coevaporated with a 15 s pause following each unit cell, leading to an overall deposition rate of ∼35−45 s/unit cell. The films used in this study are 14−24 nm thick. X-ray diffraction and reflectivity were measured using a Rigaku SmartLab diffractometer to probe the material crystallinity, roughness, and thickness. Rutherford backscattering spectroscopy (RBS) was performed at the Laboratory for Surface Modification at Rutgers University (Piscataway, NJ). RBS simulations were conducted with SIMNRA.28 Resonant soft X-ray spectroscopy was performed at beamline 4-ID-C of the Advanced Photon Source at Argonne National Laboratory in total electron yield (TEY) mode to probe the Eu M5edge, Fe L2−3-edge, and O K-edge at 300 K. Reciprocal space maps and (00L) scans were measured at Sector 33-BM-C of the Advanced

3. RESULTS AND DISCUSSION 3.1. Structural Characterization. X-ray diffraction was used to investigate the crystallinity of the EFO films. In panels a and b of Figure 1, we show a scan of the (001) peak of STO and LSAT with a corresponding GenX30 fit of the 38 u.c. films (∼14 nm). From these fits, we obtain a c-axis lattice parameter of 3.841 Å on STO and 3.864 Å on LSAT. Bulk EFO has an orthorhombic crystal structure with a = 5.371 Å, b = 5.589 Å, and c = 7.681 Å,31 corresponding to an average pseudocubic lattice parameter of 3.869 Å. The observed c-axis parameter on STO is smaller than the bulk pseudocubic lattice parameter, consistent with a tensile strain state induced by the SrTiO3 (STO; a = 3.905 Å) substrate. In total, 10 films on STO were measured with an average c-axis lattice parameter of 3.837 ± 0.010 Å extracted from the GenX fits,30 where the error bar is the standard deviation of the values obtained from all the samples. A long 2θ−θ scan was performed to ascertain the phase purity of the film. As shown in Figure 1c, the STO (00L) peaks and the lower-intensity EFO film peaks are clearly seen, with no evidence of secondary phases. In Figure 1d, representative rocking curves of the film and substrate taken at the (001) peak are shown. For the films measured, the B

DOI: 10.1021/cg501403m Cryst. Growth Des. XXXX, XXX, XXX−XXX

Article

Crystal Growth & Design average full width at half-maximum (fwhm) for the substrates is 0.021 ± 0.005°, and for the films, it is 0.043 ± 0.022°. Reciprocal space maps about the (113) peak were measured for films on STO (Figure 1e) and on LSAT (Figure 1f) to probe the in-plane lattice parameters. The films on both STO and LSAT are strained and coherent with the substrate, and no indication of lattice relaxations or diffuse scattering due to noncoherency with the substrate is seen. To investigate the thickness and surface quality of the films, we performed X-ray reflectivity. The interference fringes that arise from scattering length density contrast between the film and substrate are very sensitive to thickness, allowing for the determination of the film thickness to within a few angstroms. Figure 2 shows a typical reflectivity curve with a fit obtained

Figure 3. Measured (○) and simulated (red line) RBS spectra.

The microstructures of the EFO thin films were investigated using HAADF-STEM imaging. Figure 4a shows a lowmagnification cross-sectional image revealing a relatively uniform thickness and the flat interface between the STO substrate and EFO film. Some regions of the film exhibit a 1− 2° tilt in the growth direction from the STO [001] direction, which can be seen in Figure 4b. There is no obvious preferential direction to the tilt. The presence of edge dislocations was observed at the STO− EFO interface, as shown in Figure 5. We find these edge dislocations at the boundaries between domains exhibiting different rotational patterns of the FeO6 octahedra. In bulk, EFO is orthorhombic with the a−a−c+ rotational pattern, where the superscript + indicates in-phase rotations of the cornerconnected FeO6 octahedra and the superscript − indicates outof-phase rotations.34 For the thin films in this study, we define the c-axis as the [001] direction, which is also the growth axis of the films. The film exhibits a mixed a+a−c−/a−a−c+ structure, consisting of domains in which the in-phase axis is perpendicular to the film growth direction (a+a−c−) and domains with the in-phase axis parallel to the growth direction (a−a−c+). The in-phase rotation axis can be more easily distinguished through fast Fourier transformations (FFTs) of the HAADF-STEM images (Figure 5), which reveal half-order peaks arising from the alternating Eu displacements and FeO6 rotations in the EFO film. As seen in Figure 5, the edge dislocation is present at the boundary between a+a−c− and a−a−c+ regions of the film. 3.2. Spectroscopy. Resonant soft X-ray absorption spectroscopy, which probes the unoccupied electronic density of states, was used to determine the nominal valence states of the cations and the crystal-field splitting of the Fe 3d states. The Eu M4,5-edge, Fe L2,3-edge, and O K-edge are shown in Figure 6 along with relevant simulations. Simulations were conducted using CTM4XAS35 to gain insight into the valence states of Eu and Fe. In Figure 6a, the measured absorption spectrum (black) of the Eu M5-edge is shown. The red and blue curves show the simulated Eu3+ and Eu2+ spectra, respectively. The simulated spectra were shifted energetically on the basis of experimentally measured Eu3+ and Eu2+ spectra.36−38 The measured data are in good agreement with the simulated Eu3+ curve, which consists of three peaks: a small peak at 1125.2 eV, the main peak at 1130.2 eV, and a third small peak at 1133.8 eV. This result confirms the Eu cations are 3+ in the EFO films. Figure 6b shows the measured Fe L2,3-edge spectrum, which is consistent with previous reports of Fe3+ in perovskite oxides.39−41 The two strong peaks in the L3-edge arise from t2g states (at 707.7 eV) and eg states (at 709.4 eV). The energetic

Figure 2. Measured X-ray reflectivity data (○) with a simulated fit (red line). The top right inset shows the scattering length density depth profile obtained from the XRR fit. The bottom left inset shows a typical RHEED pattern from the end of film growth.

using the GenX software package.30 The scattering length density (SLD) depth profile obtained from the fit is shown in the top inset of Figure 2. The transition between the STO and the EFO is relatively abrupt. This fit produces an SLD of 4.64 × 10−5 Å−2 for EFO, slightly lower than the theoretical value of 5.02 × 10−5 Å−2. The surface roughness is 6.1 Å, which is defined as the root-mean-square roughness. A qualitative measure of surface quality can be seen in the RHEED pattern obtained following film growth. Shown in the inset of Figure 2, the RHEED pattern is typical of smooth perovskite (001) surfaces. Given the tolerance of perovskites to off stoichiometry and the difficulty of exactly matching atomic fluxes in MBE growth of oxides, the cation stoichiometries of the films were measured with RBS. In Figure 3, a representative RBS data set obtained from a film on STO with a corresponding SIMNRA simulation is shown. The composition of each cation was quantified by minimizing the summation of the difference squared (Δ2) between the simulated and measured counts at each channel for the Eu and Fe peaks.32 For a representative film shown in Figure 3, the processed fitting yielded a Eu:Fe ratio of 1:0.97. The compositional errors associated with the fits are 0.9 and 3.0% for Eu and Fe, respectively. Here the error is defined as the compositional range that yields (Δ2 − Δmin2)/Δmin2 values from −0.05 to 0.05.33 Thus, the cation composition of the film is stoichiometric within the uncertainty of the RBS measurement. The thickness values obtained from the same film using XRD, XRR, and RBS are within 9% of each other. C

DOI: 10.1021/cg501403m Cryst. Growth Des. XXXX, XXX, XXX−XXX

Article

Crystal Growth & Design

Figure 4. HAADF-STEM images (300 kV) of the cross section of the film. (a) Low-magnification image showing the interface between the STO and the EFO. (b) Tilt observed between the film and the substrate.

Figure 5. HAADF-STEM image (300 kV) of a boundary in the EFO film, with a correlating edge dislocation at the EFO−STO interface. FFTs (inserts) show the corresponding patterns for the STO and the EFO film on either side of the defect. The small arrows denote the half-order peaks from the alternating Eu displacements in the EFO film, revealing the presence of an a−a−c+ pattern to the left of the dislocation and an a+a−c− pattern to the right of the dislocation, where the c-axis is the growth direction of the film. The Eu displacements observed in the STEM image, highlighted with green circles, are consistent with these rotation patterns. A schematic of the crystal structure is presented for reference with the Eu atoms colored green, oxygens red, and the FeO6 octahedra blue with the in-phase rotation axis labeled.

D

DOI: 10.1021/cg501403m Cryst. Growth Des. XXXX, XXX, XXX−XXX

Article

Crystal Growth & Design

difference between these two peaks, approximately equal to the Fe crystal-field splitting (10Dq), is 1.7 eV. To further quantify the crystal-field splitting, the Fe3+ spectrum with various values of 10Dq was simulated. The simulated spectra were adjusted such that the simulated eg L3 peak is at the same energy as the measured eg L3 peak. As shown in Figure 6b, the simulation with a 10Dq value of 1.5 eV yielded the best qualitative agreement with the experimentally measured spectrum, based on the L3 peak positions. This value is consistent with previous analysis of crystal-field splitting in perovskites and oxides with octahedrally coordinated Fe3+, such as LFO with a reported 10Dq value of 1.8 eV41 and α-Fe2O3 with values from 1.45 eV42 to 1.8 eV.39 The O K-edge was also investigated via XAS. Figure 6c shows the O K prepeak, with the inset showing the full O Kedge spectra. The O K-edge spectrum is a result of promoting electrons from the 1s edge into the empty 2p band. This can be used as a first-order approximation to the oxygen 2p unoccupied density of states.43,44 The oxygen 2p prepeak features display the character of the empty Fe (3d) states because of hybridization with the unoccupied bands; therefore, the O K-edge prepeak is attributed to O 2p hybridization with Fe 3d states.43 As such, there are two main features to the prepeak associated with the Fe t2g and eg states. The O K-edge prepeak was fit to two Gaussians, the centers of which are separated by 2 eV, which is similar to the experimentally measured splitting of the t2g and eg peaks (1.7 eV) from the Fe L3-edge. Room-temperature optical absorption was probed to determine the absorption edge and investigate the optical transitions within the visible spectrum. The optical absorption spectrum was obtained using ellipsometry, which measures the phase difference after the reflection of light reflected from the film, Δ, and the ratio of the reduction of the amplitudes, Ψ. The absorption spectra are obtained through modeling the collected data. Prior to growth, the substrates were measured and fit to a Cauchy function to ensure the accurate representation of the substrate contributions. The films were modeled as homogeneous smooth layers. The data were fit using a Levenberg− Marquart algorithm of Lorentz oscillators to yield the best fits on average for each film. The Lorentz oscillator model used by WVASE is

Figure 6. (a) Soft X-ray absorption spectrum of the Eu M5-edge measured from an EFO film on STO at 300 K. For comparison are shown the simulations of Eu2+ (top, blue) and Eu3+ (center, red) with dotted lines as a guide to the eye. (b) Fe L2,3-edge spectra probing the unoccupied Fe 3d states via the 2p−3d dipole transition for the EFO thin film. The dotted lines are guides to the eye. Three different simulations, obtained using different values of 10Dq (in electronvolts), are shown for comparison. (c) O K-edge prepeak fit to two Gaussians, with a splitting of 2 eV. The inset shows the full O K-edge spectra, with the prepeak highlighted.

m

ε=

∑ j=1

Aj BrjEj 2

Ej − E2 − i BrjE

Figure 7. (a) Optical absorption of an EFO and a LFO thin film on SrTiO3. The EFO is 60 u.c. thick, and the LFO is 70 u.c. thick. Region A corresponds to the energies below the band gaps of the two materials; region B is the onset of absorption, and region C is the spectral regime of higher-energy excitations. (b) Absorption of two additional EFO films on STO and LSAT substrates. E

DOI: 10.1021/cg501403m Cryst. Growth Des. XXXX, XXX, XXX−XXX

Article

Crystal Growth & Design where ε is the dielectric constant, Ej is the center energy of the given oscillator, Aj is the oscillator amplitude fitting parameter, E is the incident photon energy, and Brj is the broadening fitting parameter. Eight oscillators were needed for a good correlation to the collected data. In Figure 7a, we show the absorption data from a 60 u.c. EFO (red) and a 70 u.c. LFO (black) thin film. The physical properties of LFO films were previously reported by Scafetta et al.10 The absorption curve has three main regions, marked A−C in Figure 7a. In region A (1−2 eV), the absorption is less than 0.1 × 105 cm−1, which is expected when the incident photon energy is smaller than the energy needed to excite a carrier across the EFO band gap energy. Region B (2.5−3.5 eV) is the onset of absorption, where light induces an excitation of a charge carrier into the conduction band. In this region, the EFO peak is blue-shifted from that of the LFO. For the two highest-amplitude oscillators in region B, there are 0.03 and 0.04 eV shifts to higher center energies for the EFO compared to the LFO, values consistent with the increase in the band gap discussed below. We note that the energy of the first oscillator is slightly smaller for LFO; however, the difference is only 7 meV, the magnitude of which is comparable to the error of the fit. A full list of oscillator fitting parameters for both EFO and LFO can be found in Table S1 of the Supporting Information. Region C (3.5−5 eV) includes the higher-energy excitations in the material. In this range, oscillator 4 defines the leading edge of the higher-energy absorption curve, and oscillators 5−7 define the fine structure. Oscillator 4 of EFO is shifted to a lower center energy by 0.01 eV (comparable to the error of the fit), whereas the energies of oscillators 5 and 6 are blue-shifted by 0.03 and 0.04 eV, respectively, compared to those of the oscillators from LFO. Tauc analysis of the EFO and LFO spectra shown in Figure 7a is presented in Figure S1 of the Supporting Information, but we note that there is significant uncertainty in the use of Tauc analysis with perovskite oxides as the exact nature of the transitions for most perovskites is unknown. Through a combination of DFT calculations and optical measurements, LFO was demonstrated to have a direct forbidden band gap of ∼2.34 eV.45 By conducting Tauc analysis and assuming that EFO also has a direct forbidden transition, we find that the band gap of the EFO film is blue-shifted by 0.15 eV compared to that of the LFO film. We note that the strain state of LFO on STO is 0.7% compressive while that of EFO on STO is 0.9% tensile. Therefore, the optical spectra of EFO on STO and LSAT substrates were compared to gain insight into the role that epitaxial strain may play in the observed optical differences between EFO and LFO. Strain can alter rotations and distortions of the BO6 octahedra and thus may also change optical absorption; however, the modifications induced by strain to B−O−B bond angles are anisotropic compared to bulk cation substitution effects. For example, substitution of Eu for La in bulk ferrites decreases the bond angles along both the inand out-of-plane directions, whereas strain tends to reduce (or increase) out-of-plane bond angles and bond lengths while having the opposite effect on in-plane bond angles and lengths.46,47 To verify the validity of a comparison of LFO and EFO, which are in different strain states, a comparison of the absorption curves of EFO on different substrates is shown in Figure 7b. For EFO on STO, the film is in a tensile stain state of 0.9%, whereas EFO is lattice-matched to LSAT (3.868 Å), resulting in