Epitaxial Growth of CaMnO3–y Films on LaAlO3 (112¯0) by Pulsed Direct Current Reactive Magnetron Sputtering

CaMnO3 is a perovskite with attractive magnetic and thermoelectric properties. CaMnO3 films are usually grown by pulsed laser deposition or radio frequency magnetron sputtering from ceramic targets. Herein, epitaxial growth of CaMnO3–y (002) films on a (112¯0) ‐oriented LaAlO3 substrate using pulsed direct current reactive magnetron sputtering is demonstrated, which is more suitable for industrial scale depositions. The CaMnO3–y shows growth with a small in‐plane tilt of <≈0.2° toward the (200) plane of CaMnO3–y and the ( 1¯104 ) with respect to the LaAlO3 (112¯0) substrate. X‐ray photoelectron spectroscopy of the electronic core levels shows an oxygen deficiency described by CaMnO2.58 that yields a lower Seebeck coefficient and a higher electrical resistivity when compared to stoichiometric CaMnO3. The LaAlO3 (112¯0) substrate promotes tensile‐strained growth of single crystals. Scanning transmission electron microscopy and electron energy loss spectroscopy reveal antiphase boundaries composed of Ca on Mn sites along <101> and <002>, forming stacking faults.

DOI: 10.1002/pssr.202100504 CaMnO 3 is a perovskite with attractive magnetic and thermoelectric properties. CaMnO 3 films are usually grown by pulsed laser deposition or radio frequency magnetron sputtering from ceramic targets. Herein, epitaxial growth of CaMnO 3-y (002) films on a ð1120Þ-oriented LaAlO 3 substrate using pulsed direct current reactive magnetron sputtering is demonstrated, which is more suitable for industrial scale depositions. The CaMnO 3-y shows growth with a small in-plane tilt of <%0.2 toward the (200) plane of CaMnO 3-y and the (1104) with respect to the LaAlO 3 ð1120Þ substrate. X-ray photoelectron spectroscopy of the electronic core levels shows an oxygen deficiency described by CaMnO 2.58 that yields a lower Seebeck coefficient and a higher electrical resistivity when compared to stoichiometric CaMnO 3 . The LaAlO 3 ð1120Þ substrate promotes tensile-strained growth of single crystals. Scanning transmission electron microscopy and electron energy loss spectroscopy reveal antiphase boundaries composed of Ca on Mn sites along <101> and <002>, forming stacking faults.

Experimental Section
CaMnO 3-y films were grown by pulsed reactive magnetron sputtering of elemental 99.9% purity Ca and Mn targets (Plasmaterials, Livermore, CA, USA) on 10 Â 10 mm 2 LaAlO 3 (1120) substrates at 600 C in an ultrahigh vacuum system with a base pressure of 10 À6 Pa (10 À8 mbar) described in detail elsewhere. [11] The LaAlO 3 substrates were cleaned by successively immersing in Hellmanex for 3 min and in deionized water for 2 min. This process was repeated once, followed by ultrasonic cleaning in acetone and ethanol for 10 min each, and blow-drying with N 2 gas. The Ca and Mn magnetrons were operated at 75 and 45 W, respectively, using a 50 kHz pulsing frequency and an on time of 2 μs. The reverse voltage and crowbar delay were set to 10% and 30 μs, respectively. CaMnO 3-y deposition was carried out at 0.46 Pa (3.5 mTorr) using 1:20 oxygen/argon mixture with 3 sccm oxygen and 60 sccm Ar for 240 min to obtain a 73 nm-thick film.
Core-level spectra were obtained by X-ray photoelectron spectroscopy (XPS) using a Kratos Axis Ultra DLD instrument with a monochromatic Al Kα radiation source (hν ¼ 1486.6 eV) with the anode power set to 150 W and a 1.1 Â 10 À9 Torr (1.5 Â 10 À7 Pa) base pressure. A 3 Â 3 mm 2 raster area was obtained by sputter cleaning the sample surface with a 0.5 keV Ar þ beam incident onto the sample surface at a 20 angle, and the spectra were collected from a 0.3 Â 0.7 mm 2 area at normal emission angle. A low-energy electron gun operated at filament current 1.83 A, charge balance 2.06 V, and filament bias 1 V was used to compensate sample charging effects. The analyzer pass energy was set to 20 eV, yielding a 0.55 eV full width at half maximum for the Ag 3d 5/2 peak. The KolXPD software [12] was used to fit the peaks using the pseudo-Voigt function (70% Gaussian, 30% Lorentzian) formalism after a Shirley background subtraction. The spin-orbit splitting intensity ratios were fixed at ½ for the Ca 2p 3/2 and Ca 2p 1/2 core levels separated by 3.5 eV.
For cross-sectional transmission electron microscopy (TEM) analysis, the sample was cut along two in-plane directions [1104] and [1102 into 0.6 Â 1.8 mm 2 pieces and glued to a Ti grid using araldite. The Ti grid was mechanically thinned down to %50 μm using diamond polishing papers. The samples were ion milled in a Gatan's Precision Ion Polishing system to create electron-transparent regions using 5 keV Ar þ ions incident on to the surface at 5 and a follow-up polishing by a 2 keV Ar þ beam.
The film microstructure was probed using high-angle annular dark field (HAADF) STEM imaging, selective area electron diffraction (SAED), and EELS. Characterization was performed using the Linköping double C s -corrected FEI Titan 3 60-300, operated at 300 kV. We obtained sub-Ångström resolution in HAADF-HRSTEM images using a 21.5 mrad convergence semiangle with %60 pA probe beam current and an angular detection range of 46-200 mrad. STEM-EELS spectrum images of 32 Â 32 pixels were acquired for 1 min using a 0.25 eV/channel energy dispersion, 0.2 s pixel dwell time, and a collection semiangle of 55 mrad of the employed Gatan GIF Quantum ERS postcolumn imaging filter. Elemental Ca and Mn distribution maps were extracted from EELS spectrum images by background subtraction, using a power law, and choosing characteristic edges Ti-L 23 (340-280 eV) and Mn-L 23 (640-490 eV) energy loss integration windows.
The X-ray diffractometry (XRD) and X-ray reflectivity (XRR) were carried out in a PANalytical Empyrean diffractometer equipped with a copper Cu Kα (λ ¼ 1.54 Å) source operated at 45 kV and 40 mA, a hybrid mirror on the incidence beam path, and a triple-axis Ge 220 analyzer on the diffracted beam path, and a PIXcel3D detector operated in open detection mode. The XRR data were fitted using the X'Pert Reflectivity program. Scanning electron microscopy (SEM) was carried out using a LEO Gemini 1550 Zeiss instrument operated at 3 kV using an in-lens secondary electron detector.
The Seebeck coefficient α was measured in a home-built setup, previously described, [13] equipped with two Peltier heat sources and two 0.1 mm-diameter K-type thermocouples, for creating a temperature gradient and for measuring the temperature, respectively. Two, 1.5 Â 4 cm 2 in size, Cu electrodes were attached to, but electrically isolated from, the thermocouples, in contact with the hot and cold spots, and connected to a Keithley 2001 multimeter. The electrical resistivity was determined from four-point-probe sheet resistance measurements using a Jandel RM3000 probe and the film thickness, as determined by XRR measurements, shown in the Supporting Information.

Results and Discussion
Fitting XRR data from as-deposited CaMnO 3 (see Figure S1, Supporting Information) showed a film thickness of 73 nm with a 4.28 g cm À3 density and a 2.7 nm surface roughness. The estimated film density is %7% lower than the theoretical value [14] of 4.59 g cm À3 . SEM micrographs (see Figure S2, Supporting Information) show a film surface with small linear voids, consistent with the estimated surface roughness. X-ray diffractograms ( Figure S3, Supporting Information) show two sets of peaks that are also seen from a bare substrate (inset in Figure S3, Supporting Information), i.e., all observed peaks are from the substrate.
Cross-sectional HAADF-STEM images ( Figure 1) show a film thickness of %70 nm thick, corroborating the XRR analysis. The www.advancedsciencenews.com www.pss-rapid.com homogeneous contrast indicates no apparent grain boundaries in the low magnification images ( Figure 1a); however, dark lines appear throughout the entire film. The film is grown with a small angular offset (<%0.2 ) with respect to the substrate, as indicated by the superimposed lines drawn across the film-substrate interface marked 1 (following the film lattice) and 2 (following the substrate lattice) in Figure 1b. This offset between film and substrate crystal structure is further verified by SAED patterns as small radial smear of the film and the substrate patterns (insets in Figure 1d), additional STEM analysis (see Figure S4, Supporting Information). The film-substrate interface is sharp (Figure 1d) with no apparent defects, suggesting that the film is of high epitaxial quality. Based on the SAED patterns, we can describe the crystallographic relationship between the film and the substrate as CaMnO 3-y (002) // LaAlO 3 (1120) and CaMnO 3-y (200) // LaAlO 3 (1104). Furthermore, correlations between plan-view images of the film and the SAED patterns (see Figure S5, Supporting Information) indicate that the dark lines follow the <101> directions in the CaMnO 3-y film. We observed only one domain of the CaMnO 3 films on LaAlO 3 (1120), in contrast to three different orientations reported [7] for CaMnO 3 films grown on (0112)-oriented LaAlO 3 substrate.
Examples of the dark lines are shown in (Figure 1b-d). The applied mass contrast imaging conditions yield that these lines are associated with a locally lower atomic number. In addition, we observe an extended bond length across the lines (%14%). EELS elemental maps of the dark lines (Figure 1f-g) reveal a significantly higher Ca signal and a lower Mn signal that indicate agglomeration of defects on specific planes, where Ca substitutes for Mn. Consequently, the comparatively longer Ca─O bonds also expand the lattice. Figure 2 shows the corresponding analysis of a sample prepared with a 90 in-plane rotation that shows similar results, but with some notable differences. The dark lines appear in the <002> growth direction, and some of the lines span the entire film thickness (see Figure 2a). Also, in this projection the lines correlate to defects and a locally increased bond length (Figure 2b-c). The width of the dark lines is approximately 33% larger in relation to the ambient lattice and is much larger than for the dark lines in Figure 1. The SAED pattern (Figure 2d inset) confirms the epitaxial relationship described above (Figure 1). The defects are correlated with the formation of antiphase boundaries, which has been reported as a relaxation mechanism in perovskites [15] and other structures. [16] Following the off angle tilt growth found in STEM, an additional XRD measurement was performed as shown in Figure 3 where two different measurements on the same film were performed. The difference between the two measurements is the tilt in ω (incidence angle) and χ (tilt angle). ω and χ are aligned on the substrate peak for the red curve and aligned on the film on the blue curve. As clearly seen, the 002 film peak of CaMnO 3 is observed for the tilted alignment, matching the tilt observed in STEM. It is apparent that the film peak is hidden beneath the substrate peak if not properly aligned on. Additionally, the intensity scale difference between the red (substrate) and blue (film) curve is on the order of 100 where the red curve has higher intensity.
Rocking curves on the substrate (b) and film (c) are obtained at the alignment on respective material. The full width at half maximum of the two rocking curves is similar, implying that the crystal quality of the film and substrate is similar.
XPS was used to obtain compositional and bonding information. The Ca 2p, Mn 3s, and O 1s core level peaks are shown in Figure 4. The XPS analyses showed an overall composition of  , respectively, and their satellite peaks at 5.6 and 4.7 eV, respectively, indicating manganese oxides [17][18][19] also indicate oxygen vacancies [18,19] that induce decrease in the Mn oxidation number from þ4 to þ3. The O 1s peak (Figure 4c) consists of two core level subbands at 528.7 eV (FWHM ¼ 1.26 eV) from the oxygen in the bulk material and 530.9 eV (FWHM ¼ 1.31 eV), corresponding to oxygen in the bulk and the surface, respectively. Oxygen vacancies are known to form in CaMnO 3 by the reduction of Mn 3þ ions from Mn 4þ ions. [20] Thus, it is unlikely that the observed Mn 3þ would be an artifact of sputter cleaning prior to XPS. This inference is supported by the fact that the compositions obtained from both EDX and XPS indicate Mn 3þ content than Mn 4þ . These results indicate that the LaAlO 3 (1120) substrates support the epitaxial growth of substoichiometric CaMnO 3-y films. The film growth is likely strained to enable lattice matching with the substrate. A possible explanation for the low oxygen content is that the formation energy of oxygen vacancies is decreased when CaMnO 3 is under tensile stress and the film relaxes by the socalled chemical strain relaxation. [7,19,21,21,22] The oxygen vacancies can also form due to insufficient oxygen supply during growth, [20] which could be improved by increasing the oxygen flow during the synthesis process or through a postannealing step.

Conclusions
In summary, this study has shown that epitaxial CaMnO 3-y films can be grown on LaAlO 3 (1120) substrates by pulsed DC magnetron sputtering. A well-defined single-crystal structure with occasional stacking faults where Ca ions occupy Mn ion sites was observed. STEM and EELS reveal antiphase boundaries composed of Ca on Mn sites along <101> and <002>. XPS and EDS revealed a substoichiometric composition of around CaMnO 2.60 , likely due to oxygen vacancies. The Seebeck coefficient of À208 μV K À1 and the electrical resistivity of 4 Ω cm are consistent with other substoichiometric CaMnO 3 films. The LaAlO 3 ð1120Þ substrate promotes tensile-strained growth of single crystals.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.