Site-specific mapping of transition metal oxygen coordination in complexoxides

S. Turner,a) R. Egoavil, M. Batuk, A. A. Abakumov, J. Hadermann, J. Verbeeck,and G. Van TendelooEMAT, Department of Physics, University of Antwerp, B-2020 Antwerp, Belgium

(Received 8 October 2012; accepted 27 November 2012; published online 12 December 2012)

We demonstrate site-specific mapping of the oxygen coordination number for transition metals in

complex oxides using atomically resolved electron energy-loss spectroscopy in an aberration-

corrected scanning transmission electron microscope. Pb2Sr2Bi2Fe6O16 contains iron with a

constant Fe3þ valency in both octahedral and tetragonal pyramidal coordination and is selected to

demonstrate the principle of site-specific coordination mapping. Analysis of the site-specific

Fe-L2,3 data reveals distinct variations in the fine structure that are attributed to Fe in a six-fold

(octahedron) or five-fold (distorted tetragonal pyramid) oxygen coordination. Using these

variations, atomic resolution coordination maps are generated that are in excellent agreement with

simulations. VC 2012 American Institute of Physics. [http://dx.doi.org/10.1063/1.4770512]

In complex transition metal oxides, the bonding and

electronic state of the transition metal cations, i.e., the oxy-

gen coordination, spin state, and oxidation state, is of funda-

mental importance. The spin state, valency, and oxygen

coordination of the transition metal cations are all intricately

related to the structural, electronic, magnetic, catalytic, and

ionic transport properties of the oxide materials.1–3 Informa-

tion on the oxygen coordination of the transition metal cati-

ons (coordination number, bond lengths, bond angles) is

conventionally available from various diffraction or spectro-

scopic methods such as X-ray/neutron/electron diffraction,

extended X-ray absorption fine structure (EXAFS), and

M€ossbauer spectroscopy or their combinations.4 However,

for materials applications, changes in the crystal and elec-

tronic structure often need to be monitored with high spatial

resolution, which is not possible using the above mentioned

methods.5,6 For example, valency changes at surfaces or

under metallic surface particles are of vital importance for

many catalytic processes, while coordination or spin changes

at defects like twin boundaries, grain boundaries, and interfa-

ces can greatly affect the electronic, optical, and transport

properties of bulk materials and thin films.7–9

Atomic resolution elemental mapping by means of spa-

tially resolved electron energy-loss spectroscopy (EELS) and

energy dispersive X-ray spectroscopy in a scanning transmis-

sion electron microscope (STEM-EELS and STEM-EDX) has

become a well-established technique over the past years.10–15

STEM-EDX appears to allow more straightforward data ac-

quisition and interpretation as compared to EELS. However,

recent work has shown that the interpretation of atomic reso-

lution data needs to be combined with simulations for an accu-

rate interpretation in both cases.16 In addition, with the current

commercial spectrometer technology, structural information

like bonding or coordination cannot be obtained from X-ray

spectra.17 Therefore, combining the sensitivity of EELS to

valency, coordination, and spin state through the EELS fine

structure (the energy-loss near-edge structure or ELNES) with

the atomic resolution capabilities of a STEM remains the

most direct method to obtain atomic resolution structure

information.

Oxidation state mapping at atomic resolution was

recently demonstrated, using the correlation between ELNES

and valency.7,18 The main issues hindering atomic resolution

valency and bonding measurements are poor EELS signal-

to-noise ratios due to the need for simultaneous high spatial

and energy resolution of the instrument and problems of sig-

nal intermixing due to elastic scattering in the case of “thick”

crystals.19 These problems are worsened further as, in gen-

eral, changes in the fine structure of the L2,3 edge of transi-

tion metals due to bonding or coordination are supposed to

be far more subtle than changes related to valency.20,21 This

makes studies where bonding and coordination have been

mapped at atomic resolution rare, and in most cases, the

change in bonding coincides with a change in valency which

also affects the EELS fine structure.12,22–24 In recent work,

distinct changes in fine structure of the Fe L2,3 and Co L2,3

edges were found between Fe and Co in octahedral and in

tetrahedral layers in Ca2FeCoO5 brownmillerite using atomic

resolution STEM-EELS. This experiment demonstrated the

problems of atomic resolution oxygen coordination mapping:

even though the octahedral and tetrahedral signatures could

be measured at each distinct plane, the signal to noise ratio

and signal intermixing in the experiments did not allow for a

column by column investigation.25 As stated above, column

by column information is of crucial importance when study-

ing, e.g., point defects or surface sites.

In this work, we demonstrate the sensitivity of transition

metal L2,3 edges to changes in local oxygen coordination,

which can be used to map out the coordination at atomic

resolution. Pb2Sr2Bi2Fe6O16, a perovskite-based material

with a structural incorporation of crystallographic shear

(CS) planes was selected for the experiment. This type of

complex ferrites, combining magnetic transition metal cations

at the B-position of the perovskite structure with lone pair A-

cations, potentially demonstrate a combination of ferroic prop-

erties such as the antiferroelectric and antiferromagnetica)E-mail: Stuart.turner@ua.ac.be.

0003-6951/2012/101(24)/241910/5/$30.00 VC 2012 American Institute of Physics101, 241910-1

APPLIED PHYSICS LETTERS 101, 241910 (2012)

Downloaded 12 Dec 2012 to 139.184.30.132. Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions

orderings found in (Pb,Bi)1�xFe1þxO3�y perovskites.26 Within

the material, the oxygen coordination of Fe changes from an

octahedral (6-fold) coordination in the perovskite blocks to a

distorted tetragonal pyramidal (5-fold) setting at the CS planes,

while keeping the Fe3þ valency constant. For the AnBnO3n-2

(A¼ Pb, Bi, Ba, Sr, B¼Fe, Ti, Sn) family of perovskite-based

oxides with perovskite blocks separated by CS planes, the pre-

sence of iron atoms in two distinct 5 - and 6-fold coordinations

as well as the constant Fe3þ oxidation state were proven by

neutron powder diffraction and M€ossbauer spectroscopy.27–29

Pb2Sr2Bi2Fe6O16 has an orthorhombic crystal structure with

lattice parameters a¼ 5.7199(1) A, b¼ 3.97066(7) A,

c¼ 32.5245(8) A and belongs to the n¼ 6 member of this fam-

ily. The fact that Pb2Sr2Bi2Fe6O16 is a mixed coordination/sin-

gle-valency compound makes it an ideal candidate to

demonstrate coordination mapping at atomic resolution; EELS

fine structure variations between crystallographically distinct

Fe positions can in this way unequivocally be assigned to coor-

dination changes and not to, e.g., valency variation.

The sample with composition Pb2Sr2Bi2Fe6O16 was

synthesized using a high temperature solid state reaction of

PbO, SrCO3, Bi2O3, and Fe2O3. The starting materials were

mixed in the molar ratio 1:1:0.5:1.5, thoroughly ground,

pressed into a pellet and annealed in air at 750 �C for 24 h,

850 �C for 24 h, and 900 �C for 15 h with intermediate

regrinding. The sample was prepared for TEM investigation

by crushing the powder, dispersing it in ethanol, and depos-

iting the dispersion onto a holey carbon grid. The sample

was investigated using a FEI Titan “cubed” microscope

equipped with a probe corrector and a monochromator, oper-

ated at 120 kV. The microscope was operated in STEM

mode using a convergence semi-angle a of 18.5 mrad. The

monochromator was excited to provide an energy resolution

of �250 meV, and the energy slit was chosen to provide a

beam current close to 60 pA for spectroscopy, while keeping

acceptable spatial resolution (probe size of approximately

1.5 A). The high-angle annular dark-field (HAADF) inner

collection semi-angle and spectrometer acceptance semi-

angle b was 160 mrad. The acquisition time per pixel was

80 ms and was chosen to avoid beam damage and provide

the best possible signal-to-noise ratio. The iron elemental

map in Figure 2 was generated by plotting the intensity

under the background subtracted L3 edge using a 9 eV

energy window. The maps in Figure 3 were generated by

back-fitting the 6-fold and 5-fold coordinated Fe EELS ref-

erence components from Figure 2(d) (blue and red spectra)

in a linear combination to the acquired EELS data cube

using the EELS model software package.30,31 A power-law

background (A �E�r) model for the EELS data was used in

the fit. When filtered, a 3� 3 light low-pass filter was used

in the DIGITAL MICROGRAPH software package. The image sim-

ulations were performed with the STEMSIM software package,

a MATLAB based image simulation program capable of han-

dling the complex interplay between elastic and inelastic

scattering in the double channeling approximation.32 Elastic

scattering was simulated with a Bloch wave approach

(max wave vector 2.0 A�1) at a total thickness of 10 nm

using a unit cell sampling of 135*24 pixels per unit cell.

Source size broadening was taken into account using a

Gaussian with 0.7 A FWHM and a Lorentzian with 0.2 A

FWHM.33 Inelastic scattering was simulated using a relativ-

istically corrected dipole approximation with a hard Bethe

ridge cutoff.

Figure 1(a) shows a HAADF-STEM image of the

Pb2Sr2Bi2Fe6O16 sample along the most informative [010]

zone axis orientation. In the image, the a and c directions are

indicated by arrows and the unit cell is marked by a white rec-

tangle. The structural repetition of the perovskite blocks con-

taining Bi, Sr as A cations and 6-fold coordinated Fe as B

cation and the crystallographic shear planes with Bi, Pb

atomic pairs alternating with pairs of 5-fold coordinated Fe is

immediately apparent. The image is taken with the electron

monochromator excited, providing an energy resolution of

approximately 250 meV. Under these conditions, the spatial

resolution is still sufficient to image all the structural details of

the compound. The crystal surface is extremely clean, which

is crucial for high resolution EELS experiments; and when-

ever present, the amorphous surface layer is below 1 nm. Sam-

ple thickness is also likely to be a crucial factor in this type of

experiment. As the EELS signal intermixing due to elastic

scattering in the crystal increases with sample thickness, a

very thin sample region close to the crystal surface was

selected for investigation. A Bloch wave HAADF image sim-

ulation for a 10 nm thick crystal is inserted into Figure 1(a)

and agrees remarkably well with the experimental image. The

projected structural models for the [100] and [010] zone axis

orientations are displayed in (c) and further elucidate the coor-

dination of Fe within the structure.

FIG. 1. Pb2Sr2Bi2Fe6O16 structure. (a) HAADF-STEM image along the

[010] zone axis. The unit cell is indicated by the white rectangle. The inset

image simulation for a 10 nm thick crystal matches well with the experi-

ment. (b) Fourier transform of the image in (a), demonstrating information

transfer beyond 6.0 nm�1 (reflections marked by red circles). (c) Structural

model along the [100] and [010] zone axis orientations. The 6-fold (octahe-

dral) coordinated Fe species are rendered in blue, the 5-fold coordinated (tet-

ragonal pyramidal) Fe species in red, Pb/Bi columns in green, and Bi/Sr

columns in violet.

241910-2 Turner et al. Appl. Phys. Lett. 101, 241910 (2012)

Downloaded 12 Dec 2012 to 139.184.30.132. Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions

The atomic resolution EELS data acquired at high

energy resolution are plotted in Figure 2. To acquire the

spectroscopic data needed for coordination mapping, the

spectrum imaging (SI) technique was adopted.17 In this tech-

nique, the electron probe (in our setup, the probe has an ap-

proximate size of 1.5 A) is scanned over the sample and an

EELS spectrum is acquired in each point together with a

high-angle annular dark-field signal as image reference. The

overview image in Figure 2(a) shows the region selected for

spectrum image acquisition. The simultaneously acquired

HAADF image is displayed in Figure 2(b). The Fe L2,3

edge was acquired to investigate local changes in Fe coordi-

nation. The L3 and L2 “white lines” arise from transitions of

2p3/2! 3d3/23d5/2 (L3) and 2p1/2! 3d3/2 (L2) and are known

to be sensitive to valency and coordination as their intensity

is related to the density of unoccupied states in the 3d

bands.20,34 Using a simple integration window placed over

the background-subtracted L3 edge in the EELS datacube, an

Fe elemental map was generated and is plotted in Figure

2(c). At this point, it is important to note that no post-

processing such as principle component analysis (PCA/

weighted PCA35) of the data was performed. All analyses

were carried out on raw data. Previous studies have shown

that PCA treatment of atomic resolution data can mask small

details in the acquired EELS data.36

The distinction between the 5-fold coordinated Fe3þ

atomic pairs in the CS plane structures and the 6-fold coordi-

nated Fe3þ B cations in the perovskite blocks is immediately

apparent. Simple inspection of the data summed over Fe

positions in the perovskite blocks (blue region) and the Fe

CS plane structure positions (red region) reveals distinct

changes in the fine structure for the two types of Fe coordina-

tion. The data collected from the octahedral Fe3þ sites (blue)

displays a clear splitting of the L3 peak, which results from

the split of the energy levels in the Fe 3d unoccupied states

into t2g and eg levels. The pre-peak to L3 at 708 eV is associ-

ated with transitions from 2p3/2 ! t2g. The main L3 maxi-

mum, associated with a transition from 2p3/2! eg, is present

at 709.5 eV. The fine structure shape and edge onset values

are therefore both in full agreement with literature data for

Fe3þ in an octahedral coordination.20 The data collected

from the tetragonal pyramidal Fe3þ sites show other distinct

features. A first observation is that, even though the ELNES

shape of the peak varies with respect to the octahedral data,

the edge onset remains the same, confirming that the valency

of Fe is not changing from one crystallographic site to

another.19 The L3 maximum is shifted by 0.3 eV to lower

energy. At the same time, the pre-peak to the L3 is subdued

due to the loss in symmetry and/or a decrease in the crystal

field splitting at these crystallographic positions, similar to

the case of 4-fold coordinated Fe3þ.4,25 The sum of both sig-

nals yields the total Fe L2,3 edge (black spectrum). In all, it is

clear that even though the coordination of the Fe cations

changes by only a single oxygen atom, clear differences are

present in the Fe L2,3 ELNES signatures.

In Figure 3, the average octahedral and tetragonalpyramidal internal reference spectra for Fe, the blue and red

spectra from Figure 2(d), are fitted to the entire acquired

EELS datacube in a linear combination, following the proce-

dure from earlier work on atomic resolution valency map-

ping.7,18,30,31 By fitting these two internal reference spectra

for Fe in different oxygen coordinations to each spectrum

in the datacube, 2D component maps of the spectral

weights are generated. The results of the fit are displayed in

Figure 3(a). It can be seen that each component peaks at the

correct atomic columns. The 6-fold coordinated Fe positions

in the perovskite blocks peak in the map of the octahedral

coordination, while the 5-fold coordinated Fe positions peak

in the map of tetragonal pyramidal coordination, even

resolving the Fe3þ atomic pairs that are only separated by

2.3 A in this projection. In order to confirm the direct

FIG. 2. Site-specific ELNES investigation. (a) HAADF-STEM overview

image of the 44� 25 pixel spectrum image region, indicated by the white

rectangle. (b) Simultaneously acquired HAADF signal, (c) Fe map, gen-

erated by integrating a 9 eV wide energy range under the background

subtracted Fe L3 edge in the spectrum image. (d) Summed Fe L2,3 edges

from the 6-fold coordinated Fe sites (blue spectrum, blue region indi-

cated in (b) and (c)), 5-fold coordinated Fe sites (red spectrum, red

region indicated in (b) and (c)), from the sum of both regions (black

spectrum) and a single pixel spectrum from an 6-fold octahedral position

(grey spectrum).

241910-3 Turner et al. Appl. Phys. Lett. 101, 241910 (2012)

Downloaded 12 Dec 2012 to 139.184.30.132. Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions

interpretation of the generated coordination maps, detailed

image simulations were carried out. These simulations are

plotted in Figure 3(c) and agree very well with the experi-

mental maps. When a light smoothing is applied to the

experimental data (Figure 3(b)), the similarity between

experiment and simulation is even more striking. To judge

the match between experiment and simulation more quantita-

tively, a line profile through both is overlaid in Figure 3(d)

for both the octahedral and the tetragonal pyramidal maps.

In all, experiment and simulation are in excellent agreement.

A slight discrepancy in the 5-fold line profile is present

around 2 nm, which is probably caused by a small residual

sample or instrumental instability. Both simulations and

experiment indicate that even though the crystal selected for

experiments was thin and almost free of amorphous surface,

beam spreading due to elastic scattering and inelastic deloc-

alisation is still a significant effect which leads to a large part

of the EELS signal leaking into the background. In all, these

maps clearly demonstrate that oxygen coordination of the

transition metal cations can be mapped in complex oxide

structures down to atomic resolution. Even though the coor-

dination of the two distinct Fe positions only differs by a sin-

gle oxygen atom, and the resulting ELNES variations in the

Fe L2,3 edge are small, the high signal-to-noise ratio in the

individual spectra allow the coordination to be mapped

atomic column by atomic column.

In conclusion, we have demonstrated the principle of ox-

ygen coordination mapping by atomically resolved, high reso-

lution EELS in an aberration-corrected electron microscope,

through the example of single valency/mixed coordination

compound Pb2Sr2Bi2Fe6O16. Detailed analysis of the Fe-L2,3

edge showed subtle and distinct variations in the fine structure

that could be attributed to Fe3þ in tetragonal pyramidal or

octahedral coordination; the results are in excellent agreement

5-fold

(d)

(a)

(b)

(c)

6-fold Overlay

FIG. 3. Column by column coordination mapping. (a) EELS maps obtained by point by point fitting of the internal reference Fe L2,3 data for 5-fold and 6-fold

coordinated iron to the EELS datacube. The colour overlay displays the octahedral iron columns in blue and the tetragonal pyramidal iron columns in red.

(b) EELS maps after low-pass filtering. (c) Simulated inelastic maps. (d) Line profiles over the low-pass filtered data from the positions indicated by the arrows

in (b) at 4 pixel width, with overlaid line profiles from the simulated maps.

241910-4 Turner et al. Appl. Phys. Lett. 101, 241910 (2012)

Downloaded 12 Dec 2012 to 139.184.30.132. Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions

with literature. Using the spectral data as an internal reference,

the coordination of the Fe cations in the compound could be

mapped column by column. These experiments open the gate

for coordination determination at surfaces, defects, and grain

boundaries in a plethora of complex oxide materials that have

been out of reach in the past due to the lack of suitable analy-

sis techniques, and for the study of valence and coordination

as input for structure solving.

D. Batuk is acknowledged for fruitful discussions. S.T.

gratefully acknowledges the Fund for Scientific Research

Flanders (FWO). Part of this work was supported by funding

from the European Research Council under the FP7, ERC

Grant No. 246791 COUNTATOMS and ERC Starting Grant

No. 278510 VORTEX. The EMAT microscope was partially

funded by the Hercules fund of the Flemish Government.

The authors acknowledge financial support from the Euro-

pean Union under the Framework 7 program under a contract

for an Integrated Infrastructure Initiative (Reference No.

312483 ESTEEM2). This work was funded by the European

Union Council under the 7th Framework Program (FP7)

Grant No. NMP3-LA-2010-246102 IFOX. M. B., J. H., and

A. A. acknowledge funding from the FWO under grant num-

ber G.0184.09N.

1S. Stolen, E. Bakken, and C. E. Mohn, Phys. Chem. Chem. Phys. 8(4), 429

(2006).2N. Erdman, O. Warschkow, M. Asta, K. R. Poeppelmeier, D. E. Ellis, and

L. D. Marks, J. Am. Chem. Soc. 125(33), 10050 (2003).3J. A. Rodriguez and D. Stacchiola, Phys. Chem. Chem. Phys. 12(33), 9557

(2010).4K. Tatsumi, S. Muto, I. Nishida, and J. Rusz, Appl. Phys. Lett. 96(20),

201911 (2010).5J. P. Buban, K. Matsunaga, J. Chen, N. Shibata, W. Y. Ching, T. Yamamoto,

and Y. Ikuhara, Science 311(5758), 212 (2006).6W. Zhongchang, M. Saito, K. McKenna, G. Lin, S. Tsukimoto, A. Shluger,

and Y. Ikuhara, Nature 479(7373), 380 (2011).7S. Turner, S. Lazar, B. Freitag, R. Egoavil, J. Verbeeck, S. Put, Y.

Strauven, and G. Van Tendeloo, Nanoscale 3(8), 3385 (2011).8J. Gazquez, W. Luo, M. P. Oxley, M. Prange, M. A. Torija, M. Sharma, C.

Leighton, S. T. Pantelides, S. J. Pennycook, and M. Varela, Nano Lett.

11(3), 973 (2011).9M. Lopez-Haro, J. M. Cies, S. Trasobares, J. A. Perez-Omil, J. J. Delgado,

S. Bernal, P. Bayle-Guillemaud, O. Stephan, K. Yoshida, E. D. Boyes, P.

L. Gai, and J. J. Calvino, ACS Nano 6(8), 6812 (2012).10M. Bosman, V. J. Keast, J. L. Garcia-Munoz, A. J. D’Alfonso, S. D.

Findlay, and L. J. Allen, Phys. Rev. Lett. 99(8), 086102 (2007).

11A. D’Alfonso, B. Freitag, D. Klenov, and L. Allen, Phys. Rev. B 81(10),

100101(R) (2010).12D. A. Muller, Nature Mater. 8(4), 263 (2009).13G. A. Botton, S. Lazar, and C. Dwyer, Ultramicroscopy 110(8), 926

(2010).14K. Kimoto, T. Asaka, T. Nagai, M. Saito, Y. Matsui, and K. Ishizuka,

Nature 450(7170), 702 (2007).15E. Okunishi, H. Sawada, Y. Kondo, and M. Kersker, Microsc. Microanal.

12, 1150 (2006).16B. Forbes, A. D’Alfonso, R. Williams, R. Srinivasan, H. Fraser, D.

McComb, B. Freitag, D. Klenov, and L. Allen, Phys. Rev. B 86(2),

024108 (2012).17R. F. Egerton, Electron Energy-Loss Spectroscopy in the Electron

Microscope (Springer, 2011).18H. Tan, S. Turner, E. Yucelen, J. Verbeeck, and G. Van Tendeloo, Phys.

Rev. Lett. 107(10), 107602 (2011).19H. Tan, J. Verbeeck, A. Abakumov, and G. Van Tendeloo, Ultramicroscopy

116, 24 (2012).20P. A. van Aken and B. Liebscher, Phys. Chem. Miner. 29(3), 188 (2002).21L. A. J. Garvie and A. J. Craven, Phys. Chem. Miner. 21(4), 191 (1994).22K. Suenaga and M. Koshino, Nature 468(7327), 1088 (2010).23M. Varela, M. Oxley, W. Luo, J. Tao, M. Watanabe, A. Lupini, S.

Pantelides, and S. Pennycook, Phys. Rev. B 79(8), 085117 (2009).24S. Lazar, Y. Shao, L. Gunawan, R. Nechache, A. Pignolet, and G. A.

Botton, Microsc. Microanal. 16(4), 416 (2010).25S. Turner, J. Verbeeck, F. Ramezanipour, J. E. Greedan, G. Van Tendeloo,

and G. A. Botton, Chem. Mater. 24(10), 1904 (2012).26A. M. Abakumov, D. Batuk, J. Hadermann, M. G. Rozova, D. V.

Sheptyakov, A. A. Tsirlin, D. Niermann, F. Waschkowski, J. Hemberger,

G. Van Tendeloo, and E. V. Antipov, Chem. Mater. 23(2), 255 (2011).27A. M. Abakumov, J. Hadermann, M. Batuk, H. D’Hondt, O. A. Tyablikov,

M. G. Rozova, K. V. Pokholok, D. S. Filimonov, D. V. Sheptyakov, A. A.

Tsirlin, D. Niermann, J. Hemberger, G. Van Tendeloo, and E. V. Antipov,

Inorg. Chem. 49(20), 9508 (2010).28J. Hadermann, A. M. Abakumov, I. V. Nikolaev, E. V. Antipov, and G.

Van Tendeloo, Solid State Sci. 10(4), 382 (2008).29I. V. Nikolaev, H. D’Hondt, A. M. Abakumov, J. Hadermann, A. M.

Balagurov, I. A. Bobrikov, D. V. Sheptyakov, V. Y. Pomjakushin, K. V.

Pokholok, D. S. Filimonov, G. Van Tendeloo, and E. V. Antipov, Phys.

Rev. B 78(2), 024426 (2008).30J. Verbeeck and S. Van Aert, Ultramicroscopy 101(2-4), 207 (2004).31G. Bertoni and J. Verbeeck, Ultramicroscopy 108(8), 782 (2008).32J. Verbeeck, P. Schattschneider, and A. Rosenauer, Ultramicroscopy

109(4), 350 (2009).33J. Verbeeck, A. Beche, and W. Van den Broek, Ultramicroscopy 120, 35

(2012).34G. Vanderlaan and I. W. Kirkman, J. Phys. Condens. Matter 4(16), 4189

(1992).35K. Dudeck, M. Couillard, S. Lazar, C. Dwyer, and G. Botton, Micron

43(1), 57 (2012).36S. Lichtert and J. Verbeeck, “Statistical consequences of applying a PCA

noise filter on EELS spectrum images,” Ultramicroscopy (published

online).

241910-5 Turner et al. Appl. Phys. Lett. 101, 241910 (2012)

Downloaded 12 Dec 2012 to 139.184.30.132. Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions