Atomic-resolution imaging of oxidation states in manganites

M. Varela, M. P. Oxley, W. Luo, J. Tao, M. Watanabe, A. R. Lupini, S. T. Pantelides, and S. J. Pennycook

Phys. Rev. B 79, 085117 – Published 23 February 2009

Abstract

Aberration corrected electron optics allows routine acquisition of high spatial resolution spectroscopic images in the scanning transmission electron microscope, which is important when trying to understand the physics of transition-metal oxides such as manganites. The physical properties of these perovskites are intimately related to the occupancies of the partially filled $3 d$ bands, which define their oxidation state. In this work, we review procedures to obtain this electronic property in ${La}_{x} {Ca}_{1 - x} {MnO}_{3}$ from atomic-column-resolved electron energy-loss spectra measured in the aberration corrected scanning transmission electron microscope. In bulk samples, several features of both the average $Mn L_{2, 3}$ edge and the $O K$ edge fine structure change linearly with Mn nominal valence. These linear correlations are extracted and used as a calibration to quantify oxidation states from atomic resolution spectroscopic images. In such images, the same fine-structure features exhibit further changes, commensurate with the underlying atomic lattice. Mn valence values calculated from those images show unexpected oscillations. The combination of experiment with density-functional theory and dynamical scattering simulations allows detailed interpretation of these maps, distinguishing dynamical scattering effects from actual changes in electronic properties related to the local atomic structure. Specifically, in ${LaMnO}_{3}$ , the two nonequivalent O sites can be distinguished by these methods.

5 More

Received 23 July 2008

DOI:https://doi.org/10.1103/PhysRevB.79.085117

Authors & Affiliations

M. Varela¹, M. P. Oxley^2,1, W. Luo^2,1, J. Tao^1,*, M. Watanabe^3,†, A. R. Lupini¹, S. T. Pantelides^2,1, and S. J. Pennycook^1,2

¹Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831, USA
²Vanderbilt University, Nashville, Tennessee 37235, USA
³National Center for Electron Microscopy, Lawrence Berkeley National Laboratory, Berkeley, California 94720, USA

^*Present address: Condensed Matter Physics & Materials Science Department, Brookhaven National Laboratory, Upton, New York 11973, USA.
^†Present address: Lehigh University, Bethlehem, PA 18025, USA.

Article Text (Subscription Required)

Click to Expand

References (Subscription Required)

Click to Expand

Issue

Vol. 79, Iss. 8 — 15 February 2009

Reuse & Permissions

Access Options

Author publication services for translation and copyediting assistance advertisement

Images

Figure 1
(Color online) (a) $Mn L_{2, 3}$ edges for a series of ${La}_{x} {Ca}_{1 - x} {MnO}_{3}$ compounds with $x = 1$ (black), $x = 0.7$ (dark yellow), $x = 0.55$ (green), $x = 0.33$ (blue), and $x = 0$ (red). These spectra have been acquired from wide sample areas (by defocusing the electron beam or by averaging a number of spectra from different locations). The energy scale is nominal, and has been shifted and the intensity normalized so the $L_{3}$ lines match. (b) EEL spectra showing the $O K$ edge around 530 eV and the $Mn L$ edge around 640 eV for LMO (black) and CMO (red). The spectra have been displaced vertically for clarity. (c) Sketch showing a generic $Mn L_{2, 3}$ edge (black line) and the approximate position of the windows used for integration of the $L_{3}$ and $L_{2}$ line intensities, and also the window used to scale the Hartree-Slater cross-section step function (red line). After scaling and subtraction of this function, the remaining signals under the $L_{3}$ and $L_{2}$ lines (shaded) are integrated, and their ratio is calculated. (d) Dependence of the $L_{23}$ ratio with the formal oxidation state for a series of LCMO compounds. The red dashed line represents a linear fit to the data.Reuse & Permissions
Figure 2
(Color online) (a) Spatially averaged $O K$ edges for a series of LCMO compounds with $x = 1$ (black), $x = 0.7$ (dark yellow), $x = 0.55$ (green), $x = 0.33$ (blue), and $x = 0$ (red). The energy scale has been shifted so the pre-peaks are aligned, and the intensity normalized. A minor splitting of the CMO pre-peak is observed. Vertical dotted lines show the positions of the pre-peak and the adjacent main peak. The spectra have been displaced vertically for clarity. (b) $O K$ EEL spectrum showing the Gaussian curves used to extract peak intensity and position (pre-peak in red and main peak in blue). In both cases a window 5 eV wide (approximately) was used for the fitting, their approximate positions marked with vertical dashed lines. The pre-peak fitting window was displaced to the left so that just a few channels at the peak maximum were included. This way, the split pre-peak feature observed in CMO was avoided. Again, the error bars have been estimated by shifting laterally a few pixels the Gaussian fitting windows. Since the second peak is very well defined, this procedure has a small effect on its fit. (c) Normalized pre-peak intensity versus nominal oxidation state for the series of LCMO samples. The red dashed line is a linear fit to the data, with a regression coefficient of $r = 0.959$ . (d) Energy separation (calculated as the difference between positions of the second peak and the pre-peak) as a function of the Mn nominal oxidation state for the sample set of samples. Again, the red dashed line represents a linear fit to the data (in this case, $r = 0.990$ ).Reuse & Permissions
Figure 3
(Color online) Effect of thickness on parameter quantification: (a) Annular dark field (ADF) image of a CMO particle. The region marked with a green line was used for acquisition of two line scans, one including the zero loss peak (acquisition time of 0.06 s per pixel) and another one for the core-loss region (using 10 s per pixel). The electron beam was defocused 20 nm to simulate average illumination. (b) Relative thickness along the line scan path in inelastic mean-free path units. (c) $Δ E$ , (d) normalized pre-peak intensity, and (e) the $L_{23}$ ratio along the scan. Horizontal solid red lines show the values of the average parameters for CMO reported in Figs. 1, 2. Horizontal red dotted lines show the error bar amplitude as reported in Figs. 1, 2. (f) Mn/O ratio, after subtracting the background using a power-law fit, and integrating the intensity remaining under the $O K$ and the $Mn L_{2, 3}$ edges. Integration windows of 35 and 40 eV were used, respectively. For the $O K$ edge a Hartree-Slater cross section was employed, while for the $Mn L_{2, 3}$ edge a hydrogenic (white line) cross section was used. When Hartree-Slater cross sections are employed for both edges a similar behavior is found in the data, with values shifted upward around 0.02 Mn/O ratio units (consistent with our error bars). The blue line is the same ratio after treating the EELS data with principal component analysis. Some oxygen deficiency is found near the edge. In fact, extended electron-beam irradiation in some of these ultrathin areas gives rise to the appearance of superstructures in the ADF images, which are typically assigned to ordering of O vacancies (Ref. 53).Reuse & Permissions
Figure 4
(Color online) Scree plot after PCA has been performed on the EELS line scan in Fig. 7. A dotted red line highlights the portion of the plot where a flat behavior is observed (characteristic of uncorrelated noise). The red arrow points to the component index $(n = 7)$ where the deviation from this flat behavior is observed. Inset: raw EEL spectrum showing an $O K$ edge and a $Mn L_{2, 3}$ edge, extracted from Fig. 7b (black dots). The red line is the same spectrum after PCA.Reuse & Permissions
Figure 5
(Color online) (a) $Z$ -contrast image of a CMO sample tilted down the pseudocubic $⟨ 100 ⟩$ zone axis. The cubic unit cell has been marked. A green rectangle highlights the approximate window for spectrum image acquisition. (b) ADF signal acquired simultaneously with the spectrum image in (a). (c) Averaged profile along the region marked with a yellow rectangle in (b). [(d), (f), and (h)] $Ca L_{2, 3}$ , $Mn L_{2, 3}$ , and $O K$ images, respectively. [(e), (g), and (i)] Averaged profiles of (d), (f), and (h), respectively, using a window equivalent to the one depicted in (b). A white dotted circle shows the position of the volcano rim in (h). All EELS images have been generated by integrating a 35 eV wide window after background subtraction using a power-law fit from the EELS data after PCA. Green (blue-yellow) circles mark the approximate positions of Ca (Mn-O) columns along the scan.Reuse & Permissions
Figure 6
(Color online) (a) The top panel ADF signal from Fig. 5b. A yellow rectangle shows the area of the image used for averaging through the whole figure. The scale bar is approximately 0.3 nm. The middle panel shows a smoothed version of Fig. 5h, where the volcano structure is more easily seen. A white dotted circle shows the position of the volcano rim. The lower panel is the $O K$ integrated intensity from Fig. 5i. (b) $Δ E$ image (top), together with the averaged $Δ E$ profile (black data points) and the derived Mn oxidation state (red). (c) Normalized pre-peak intensity image (top), together with the averaged profile (black data points) and the derived Mn oxidation state (red). (d) $L_{2, 3}$ ratio image (top), with the averaged profile (black points) and the derived Mn oxidation state (red). Green (blue-yellow) circles mark the approximate position of Ca (Mn-O) columns along the scan in all cases, and all panels derive from PCA treated EELS data. The error bars through the averaged profiles are those of the original data, estimated from the errors of the Gaussian fits or from shifting the fitting, integration, or scaling windows a few pixels around and comparing the resulting values of the different parameters.Reuse & Permissions
Figure 7
(Color online) (a) $Z$ -contrast image of a LMO sample viewed down the pseudocubic $⟨ 100 ⟩$ zone axis. A pseudocubic unit cell is marked, with a red circle marking the La column position, and blue/yellow circles showing the Mn-O columns. Yellow crosses mark the approximate positions of pure O columns between Mn-O columns. (b) EELS line scan acquired along the direction of the red arrow in (a). [(c)–(e)] Magnified portions of (b), showing the $O K$ , the $Mn L_{2, 3}$ and the the $La M_{4, 5}$ edges, respectively. [(f)–(h)] $O K$ , $Mn L_{2, 3}$ , and $La M_{4, 5}$ edge images in (c)–(e) after PCA, respectively. All of the EELS images show the rippling characteristic of atomic resolution EELS. The $O K$ image shows a double ripple, characteristic of the volcano (as does the $La M_{4, 5}$ image, although the contrast is not so strong). For figures (c)–(h) the approximate position of the La (Mn-O) columns is highlighted on the left with red (blue-yellow) circles.Reuse & Permissions
Figure 8
(Color online) (a) ADF signal acquired simultaneously with the EELS line scan in Fig. 7. Red (blue-yellow) circles mark the approximate position of La (Mn-O) columns along the scan. La (Mn-O) columns show brighter (dimmer) on the ADF image. (b) $O K$ image obtained by integrating the intensity under the $O K$ edge after background subtraction. (c) $Mn L_{2, 3}$ image. (c) $La M_{4, 5}$ image. In all cases the black dots are the result of analyzing the raw EELS data, while the red line is the result after PCA analysis of the EEL spectra. A clear volcanolike structure is observed in the $O K$ image (and just a hint of it in the $La M_{4, 5}$ image)Reuse & Permissions
Figure 9
(Color online) (a) Schematic of the manganite structure down the pseudocubic $⟨ 100 ⟩$ projection. Red circles are the La (Ca) columns. Yellow circles are the O columns, and blue-yellow circles mark the Mn-O columns. The gray dotted squares indicate the nearest neighbors of a La (Ca) column and a Mn-O column, all of them pure O columns. An arrow points in the direction of the scan. (b) $O K$ image frozen phonon simulations for a LMO sample 15 nm thick. The dashed line corresponds to the total $O K$ signal. The blue line represents the contribution from the O atoms in the Mn-O columns, while the green line represents the contributions from the neighboring pure O columns. (c) $O K$ image simulation for the CMO compound. Different line formats are coded as in (b).Reuse & Permissions
Figure 10
(Color online) Averaged raw EEL spectra (black points) showing the $O K$ edge and the $Mn L_{2, 3}$ edge. Each spectrum is the result of averaging eight individual EEL spectra from three different positions along the scan in Fig. 7: La column (bottom), rim of the volcano of the Mn-O column (middle), and dip of the volcano on the Mn-O column (top). The red and blue curves show, for each case, the Gaussian fits for the pre-peak and the main peak, respectively. The green lines are the scaled Hartree-Slater step functions for $L_{23}$ ratio quantification. The curves have been displaced vertically for clarity.Reuse & Permissions
Figure 11
(Color online) Analysis of the EELS fine structures for the LMO line scan in Fig. 7. (a) $O K$ integrated intensity along the $⟨ 110 ⟩$ pseudocubic diagonal [from Fig. 8b]. The positions of the maxima allow the Mn-O columns to be located. Blue/yellow circles show the approximate position of the Mn-O columns, while red circles show the approximate position of La along the scans [from Fig. 8b] through the whole figure. (b) $Δ E$ parameter along the scan together with the derived value of the Mn oxidation state. (c) $O K$ edge normalized pre-peak intensity along the line scan, together with the Mn oxidation state calculated from this measurement. (d) $L_{23}$ ratio along the line scan and the resulting Mn oxidation state (red). A data point at $x = 30$ giving rise to an unphysical Mn oxidation state value around 14 units (resulting from a spurious error during $L_{23}$ quantification) has not been plotted. In all cases, Mn oxidation states have been derived applying the calibration shown in Figs. 1, 2. Black squares are the result of the analysis of raw EELS data, while red lines show the result of analysis after PCA. Horizontal green lines show the values ensuing from the analysis of averaged EEL $O K$ spectra, shown in Table (the $L_{23}$ ratio values have not been shown since they are quite flat along the scan).Reuse & Permissions
Figure 12
(Color online) The top panel shows the LMO structure, and the positions of O1 and O2, from Ref. 54. O1 sits in the apex of the JT distorted O octahedron, while O2 is sitting at the equatorial position. The Mn-O1 bond length is $1.968 Å$ . The O2 is the Jahn-Teller active site and the Mn-O2 bond lengths are $1.907 Å$ (short) and $2.178 Å$ (long). The rest of the figure shows the $O K$ edge simulations using the VASP code and the $Z + 1$ approach (1 eV broadening was used) for the nonequivalent O1 and O2 sites in (a) LMO and (b) CMO. Black is used for O1 and blue for O2, for both materials. The fine structure of these edges is consistent with the experimental data in Fig. 2a. These simulations do not include orientation effects, which could lead to further minor variations [details about the fine-structure interpretation will be discussed elsewhere (Ref. 55)].Reuse & Permissions

Physical Review B

covering condensed matter and materials physics