Three Dimensional Chemical Analysis of

Nanoparticles Using Energy Dispersive

X-ray Spectroscopy

A thesis submitted to the University of Manchester for the degree of Doctor of

Philosophy in the Faculty of Engineering and Physical Sciences

2015

Thomas Slater

School of Materials
List of Contents

List of Figures ........................................................................................................................................... 4

List of Abbreviations ................................................................................................................................. 7

ABSTRACT .............................................................................................................................................. 8

Declaration ............................................................................................................................................... 9

Copyright Statement .............................................................................................................................. 10

Acknowledgements ................................................................................................................................ 11

Publications ............................................................................................................................................ 12

1. Introduction ..................................................................................................................................... 14

1.1. Nanoparticle Introduction ............................................................................................ 14

1.2. Nanoparticle Characterisation ..................................................................................... 16

1.3. Thesis Structure .......................................................................................................... 17

2. The Scanning Transmission Electron Microscope ......................................................................... 18

2.1. Electron Interactions with Matter ...................................................................................... 20

2.1.1. Elastic Scattering ....................................................................................................... 20

2.1.2. Inelastic Scattering .................................................................................................... 23

2.2. Electron Imaging and Spectroscopy of Nanoparticles ..................................................... 24

2.2.1. Bright Field Imaging ................................................................................................... 24

2.2.2. High-Angle Annular Dark Field Imaging .................................................................... 26

2.2.3. Electron Energy Loss Spectroscopy .......................................................................... 27

2.3. Energy Dispersive X-ray Spectroscopy ............................................................................ 28

2.3.1. X-ray Generation........................................................................................................ 28

2.3.2. X-ray Detection .......................................................................................................... 31

2.3.3. Quantification ............................................................................................................. 35

2.3.4. Nanoparticle Studies .................................................................................................. 38

1
 2.4. Spatial Resolution in the STEM ........................................................................................ 38

2.5. Electron Beam Damage ................................................................................................... 42

2.6 Summary ........................................................................................................................... 44

3. X-ray Absorption Correction in Spherical Nanoparticles ................................................................ 45

3.1. Introduction ....................................................................................................................... 45

Paper: X-Ray Absorption Correction for Quantitative STEM-EDX of Spherical Nanoparticles

................................................................................................................................................. 47

4. Electron Tomography...................................................................................................................... 64

4.1. Introduction to Tomography ............................................................................................. 64

4.2. The Tomography Procedure ............................................................................................ 67

4.3. Tilt Series Acquisition ....................................................................................................... 68

4.3.1. Imaging Modes .......................................................................................................... 68

4.3.1.1. HAADF Tomography .............................................................................................. 69

4.3.1.2. EFTEM/EELS Tomography .................................................................................... 69

4.3.1.3. EDX Tomography ................................................................................................... 71

4.4. Alignment of Projections ................................................................................................... 73

4.5. Reconstruction .................................................................................................................. 76

4.5.1. Backprojection ........................................................................................................... 76

4.6. Resolution in Electron Tomography ................................................................................. 80

4.8. 3D Imaging Beyond STEM Tomography.......................................................................... 81

4.9. Summary .......................................................................................................................... 84

5. Investigating STEM-EDX Tomography Methodology ..................................................................... 85

5.1. Introduction ....................................................................................................................... 85

Paper: STEM-EDX Tomography of Bimetallic Nanoparticles: A Methodological Investigation

................................................................................................................................................. 87

6. STEM-EDX Tomography of AgAu Nanoparticles ......................................................................... 124

2
 6.1. Introduction ..................................................................................................................... 124

Paper: Correlating Catalytic Activity of Ag-Au Nanoparticles with 3D Compositional Variations

............................................................................................................................................... 126

7. Summary and Future Work ........................................................................................................... 151

7.1. Correlative Tomography ................................................................................................. 154

8. References ....................................................................................................................................... 164

Appendix A - Conditions for successful STEM-EDX tomography ....................................................... 186

Optimising individual projections ........................................................................................... 186

Reconstruction algorithms and number of iterations ............................................................. 187

Word Count 54267

3
List of Figures*

Figure 2.1. Illustration of the components of a scanning transmission electron microscope and

the associated signals, energy dispersive X-ray (EDX), bright field (BF), high angle annular dark

field (HAADF) and electron energy loss (EEL).

Figure 2.2. Diagram illustrating the principal of reciprocity between a) CTEM and b) STEM for

elastic scattering. The source and detection planes are interchanged and the position of the

objective aperture with respect to the sample is reversed.

Figure 2.3. Example EELS spectrum showing the characteristic features of the zero-loss peak,

the plasmon-loss peaks and core-loss edges that have had their intensity multiplied by 50.

Elemental identification is possible through the characteristic energies of core-loss edges

associated to each element.

Figure 2.4. a). Diagram of the process in which an incident electron causes a characteristic X-

ray to be emitted from an atom. b) Diagram of the possible electron transitions in the atom with

associated X-rays labelled (reproduced from (Williams & Carter, 2009b)).

Figure 2.5. Spectra from a sample containing elements Al, O and Co from vacuum (blue),

carbon film (green) and the sample (red). There are few counts from vacuum, demonstrating

that stray electrons are not a major contributor to spurious X-rays.

Figure 2.6.. Schematics of cross-sections of EDX detectors. a) Si(Li) detector, in which the

anode covers the whole back face. b) SDD detector, in which concentric p-type rings promote

“drift” of generated electrons to a small central anode.

Figure 2.7. Plot of normalised Au Mα peak intensity from an AgAu nanocube as a function of tilt

angle away from the <100> zone-axis.

Figure 4.1. Diagram of two coordinate systems related by a set of Euler angles. The blue

coordinate system is equivalent to the coordinate system of the object and the red coordinate

system rotated by the Euler angles (θ, φ, ψ) is equivalent to that of a projection.

4
 j
Figure 4.2.a) Illustration of the projection of a small dot f(x,z) on to a coordinate system (x ,θ) to
j
form a projection pj(x ,θj). b)The sinogram associated with a tilt series obtained for a point; the
j
slice highlighted in the sinogram corresponds to the projection in (a), p j(x ,θj).

Figure 4.3. Representation of methods for tilt axis determination, using a sample of Au and Pd

nanoparticles on graphene. a) Projection at 0° tilt displaying the lines (in red) at which

reconstructions are performed to display the arcing in (c-e). b) Summation of the entire tilt series,

displaying streaking in the direction perpendicular to the tilt axis at an angle α to the horizontal.

c-e) Reconstructions of individual slices (i-iii) represented in (a) for c) a tilt axis shifted from the

correct orientation by 40 pixels, d) a tilt axis tilted from the correct orientation by 15° and e) a tilt

axis approximately at the correct orientation. The sample represented here is Au and Pd

nanoparticle on graphene, supplied by Dr Peter Toth (University of Manchester).

Figure 4.4. Backprojection of an object along a varying number of directions. a) Image of a two-

phase object to be back-projected, b) backprojection of object (a) along a single direction, c)

backprojection of object (a) along 3 equally-spaced directions and d) backprojection of object

(a) along 10 equally-spaced directions.

j
Figure 4.5. a) The projected intensity p (x) of an object f(x,z) at an angle of θj. b) An illustration

of the angular sampling of an object in Fourier space under a single tilt acquisition scheme with

a missing wedge. The red line in (b) corresponds to the Fourier transform of the projected
j
intensity p (x) at the same angle θj.

Figure 4.6. Reconstructions of the object in (a) from projections at 1° intervals using b) an

unweighted backprojection operator from projections over the full tilt range, c) a weighted

backprojection operator from projections over the full tilt range and d) a weighted backprojection

operator with projections from between ± 70° displaying the missing wedge.

Figure 7.1. a) SEM image within the plasma-FIB dual beam instrument showing the pillar

sample after annular milling. b) The 30 μm diameter pillar attached to a pin for further analysis.

c) An SEM image of the cross-sectional slab cut from the nano-CT pillar before liftout. d) An ion-

beam image of one polished face of the slab attached to a TEM grid. e) Ga ion-beam image of

the “block” cut from the cross-sectional slab and attached to the lift-out needle. f) SEM image of

the final STEM needle.

5
Figure 7.2. X-ray nanotomography of a pillar of stainless steel 316. a) Volume visualisation

displaying an isosurface rendering of the outer surface of the pillar, traces along the three grain

boundaries and the surface of the segmented cavities. b) Slice through the X-ray

nanotomography reconstruction displaying the three grain boundaries and the intensity changes

associated with cavities and large intergranular carbides.

Figure 7.3. STEM-EDX elemental maps of the STEM needle displaying the intensity of X-ray

counts associated with the a) Fe Kα, b) Cr Kα, c) Ni Kα, d) Mo Lα, e) Nb Kα, f) O Kα, g) C Kα

and h) Si Kα peaks. The austenite phase at top and bottom is associated with Fe, Ni and to a

lesser extent Mo and Mn, the ferrite phase is predominantly Fe, the M23C6 carbides contain Cr,

Mn, Mo, C and O, the G-phase precipitate contains Ni, Mo and Si. The surface of the needle

has high carbon and silicon intensities that are associated with organic contamination.

Figure 7.4. Visualisation of the three dimensional reconstruction of elemental distributions and

the associated segmented phases. Volume visualisations of a) the Fe Kα signal (red), b) the Cr

Kα signal (yellow) and c) the Ni Kα signal (blue). Surface visualisations of the segmented

phases showing the d) ferrite phases (green), e) carbide precipitates (purple) and f) austenite

(orange) and G-phase (cyan).

Figure 7.5. Images displaying the faces of the extracted block and the correlation between a)

SEM images and b) slices through the CT data of faces i-iv. c) Faces (i) and (ii) displayed in

three dimensions from the CT data with the X-ray CT coordinate system indicated.

Figure 7.6. a) Visualisation of the STEM-EDX pillar with respect to the SEM image of face (iii)

(flipped) and b-c) with respect to the top-down SEM image taken after annular milling with (c)

and without (b) segmented cavities. The coordinate systems indicated all correspond to the X-

ray CT coordinates.

Figure 7.7. Correlation of the orientation of the STEM-EDX needle embedded in to the micro-CT

data (a) with that from the SEM image of the final pillar (b) at the same orientation in the SEM

as in Figure 5ai.

*Please note that the figures for each manuscript are self-contained and

so are not listed here.

6
List of Abbreviations

2D – Two dimensional

3D – Three dimensional

ACF – Absorption correction factor

AFM- Atomic force microscopy

APT – Atom probe tomography

BF – Bright field

CT – Computed tomography

CTEM – Conventional transmission electron microscope

EDX – Energy dispersive X-ray

EELS – Electron energy loss spectroscopy

EFTEM – Energy filtered transmission electron microscopy

FEG – Field emission gun

FIB – Focussed ion beam

HAADF – High angle annular dark field

IR - Infrared

SDD – Silicon drift detector

SEM – Scanning electron microscope

SIRT – Simultaneous iterative reconstruction technique

STEM – Scanning transmission electron microscope

STM – Scanning tunnelling microscopy

TDS – Thermal diffuse scattering

TEM – Transmission electron microscope

UV-Vis – Ultra Violet - Visible

WBP – Weighted backprojection

XPS – X-ray photoelectron spectroscopy

7
 Three Dimensional Chemical Analysis of Nanoparticles Using Energy
Dispersive X-ray Spectroscopy
A thesis submitted to The University of Manchester for the degree of Doctor of
Philosophy in the Faculty of Engineering and Physical Sciences

2015

Thomas Slater
School of Materials, The University of Manchester

ABSTRACT

The aim of this thesis is to investigate the methodology of three dimensional

chemical imaging of nanoparticles through the use of scanning transmission
electron microscope (STEM) – energy dispersive X-ray (EDX) spectroscopy.
In this thesis, an absorption correction factor is derived for spherical
nanoparticles that can correct X-ray absorption effects. Quantification of EDX
spectra of nanoparticles usually neglects X-ray absorption within the
nanoparticle but may lead to erroneous results, thus an absorption correction is
important for accurate compositional quantification. The absorption correction
presented is verified through comparison with experimental data of Au X-ray
peaks in spherical Au nanoparticles and is found to agree excellently. This
absorption correction allows accurate compositional quantification of large (>
100 nm) particles with STEM-EDX.
Three dimensional chemical mapping is achievable through the use of EDX
spectroscopy with electron tomography. Here, the methodology of STEM-EDX
tomography is fully explored, with a focus on how to avoid artefacts introduced
through detector shadowing and low counts per pixel. A varied-time acquisition
scheme is proposed to correct for detector shadowing that is shown to provide a
more constant intensity over a series of projections, allowing a higher fidelity
reconstruction. The STEM-EDX tomography methodology presented is applied
to the study of AgAu nanoparticles synthesized by the galvanic replacement
reaction. The elemental distribution as a function of the composition of the as-
synthesized nanoparticles is characterised and a reversal in the element
segregated to the surface of the nanoparticles is found. The composition at
which the reversal takes place is shown to correlate with a peak in the catalytic
yield of a three component coupling reaction. It is hypothesized that a
continuous Au surface results in the optimum catalytic conditions for the
reaction studied, which guides the use of galvanically prepared AgAu
nanoparticles as catalysts.

8
Declaration

No portion of the work referred to in the thesis has been submitted in support of an application

for another degree or qualification of this or any other university or other institute of learning.

9
Copyright Statement

i. The author of this thesis (including any appendices and/or schedules to this thesis)

owns certain copyright or related rights in it (the “Copyright”) and s/he has given The

University of Manchester certain rights to use such Copyright, including for

administrative purposes.

ii. Copies of this thesis, either in full or in extracts and whether in hard or electronic copy,

may be made only in accordance with the Copyright, Designs and Patents Act 1988 (as

amended) and regulations issued under it or, where appropriate, in accordance with

licensing agreements which the University has from time to time. This page must form

part of any such copies made.

iii. The ownership of certain Copyright, patents, designs, trade marks and other intellectual

property (the “Intellectual Property”) and any reproductions of copyright works in the

thesis, for example graphs and tables (“Reproductions”), which may be described in this

thesis, may not be owned by the author and may be owned by third parties. Such

Intellectual Property and Reproductions cannot and must not be made available for use

without the prior written permission of the owner(s) of the relevant Intellectual Property

and/or Reproductions.

iv. Further information on the conditions under which disclosure, publication and

commercialisation of this thesis, the Copyright and any Intellectual Property and/or

Reproductions described in it may take place is available in the University IP Policy (see

http://documents.manchester.ac.uk/DocuInfo.aspx?DocID=487), in any relevant Thesis

restriction declarations deposited in the University Library, The University Library’s

regulations (see http://www.manchester.ac.uk/library/aboutus/regulations) and in The

University’s policy on Presentation of Theses.

10
Acknowledgements

Firstly, I would like to express the sincerest gratitude to my supervisor, Dr Sarah Haigh, for all of

her help, support and guidance throughout my studies. I have been lucky to have a supervisor

who takes such genuine interest and care in the progress and well-being of her students.

I would also like to thank all members of the Haigh group who have provided me with a great

deal of support and friendship over my time in the group. In particular, I would like to thank

Edward Lewis and Yiqiang (Kelvin) Chen, with whom I have shared the entirety of my studies

and who have helped me on uncountable occasions. In addition, I would like to thank Dr Eric

Prestat for many helpful discussions on the STEM and his ever present optimism, particularly to

do with the Manchester weather. I am thankful to both Dr Alan Harvey and Matthew Smith for

their tuition and guidance on the topic of the TEM.

Thank you to all those who I have had the pleasure of collaborating with throughout the course

of my studies, including but not limited to Dr Pedro Camargo, Dr Nestor Zaluzec, Prof Grace

Burke and Dr Tim Burnett.

I would like to extend my thanks to all of my family and friends for all of their support throughout

my studies. Thank you to my parents and sister for their continued support and unconditional

love. My sincerest thanks go to my girlfriend Sarah for her everlasting patience and for her

generosity, care and support. Thank you also to Fred and Henry for their help, and hindrance, in

dealing with all of life’s problems.

Thank you to the EPSRC and the School of Materials for their support in funding my studies.

11
Publications

SLATER, T.J.A., CAMARGO, P.H.C., BURKE, M.G., ZALUZEC, N.J. & HAIGH, S.J. (2014).

Understanding The Limitations Of The Super-X Energy Dispersive X-Ray Spectrometer As A

Function Of Specimen Tilt Angle For Tomographic Data Acquisition In The S/TEM. Journal of

Physics: Conference Series 522(1), 012025.

SLATER, T.J.A., MACEDO, A., SCHROEDER, S.L.M., BURKE, M.G., O'BRIEN, P., CAMARGO, P.H.C. &

HAIGH, S.J. (2014). Correlating Catalytic Activity Of Ag-Au Nanoparticles With 3D Compositional

Variations. Nano Letters 14(4), 1921-1926. (Included as Chapter 6)

SLATER, T.J.A., CHEN, Y.Q., AUTON, G, ZALUZEC, N.J., HAIGH, S.J. X-Ray Absorption Correction

For Quantitative STEM-EDX Of Spherical Nanoparticles. Submitted To Microscopy And

Microanalysis. (Included as Chapter 3)

SLATER, T.J.A., JANSSEN, A., CAMARGO, P.H.C., BURKE, M.G., ZALUZEC, N.J., HAIGH, S.J. STEM-

EDX Tomography of Bimetallic Nanoparticles: A Methodological Investigation. Ultramicroscopy,

Accepted Manuscript. (Included as Chapter 5)

NEWTON, L., SLATER, T., CLARK, N. & VIJAYARAGHAVAN, A. (2013). Self Assembled Monolayers

(SAMs) On Metallic Surfaces (Gold And Graphene) For Electronic Applications. Journal Of

Materials Chemistry C 1(3), 376-393.

BARTON, C.W., SLATER, T.J.A., ROWAN-ROBINSON, R.M., HAIGH, S.J., ATKINSON, D. & THOMSON, T.

(2014). Precise Control Of Interface Anisotropy During Deposition Of Co/Pd Multilayers. Journal

Of Applied Physics 116(20), 203903.

BURNETT, T.L., MCDONALD, S.A., GHOLINIA, A., GEURTS, R., JANUS, M., SLATER, T., HAIGH, S.J.,

ORNEK, C., ALMUAILI, F., ENGELBERG, D.L., THOMPSON, G.E. & W ITHERS, P.J. (2014). Correlative

Tomography. Scientific Reports 4, 4711.

CHEN, Y.Q., SLATER, T.J.A., LEWIS, E.A., FRANCIS, E.M., BURKE, M.G., PREUSS, M. & HAIGH, S.J.

(2014). Measurement Of Size-Dependent Composition Variations For Gamma Prime (Gamma ')

Precipitates In An Advanced Nickel-Based Superalloy. Ultramicroscopy 144, 1-8.

12
HARRIS, P.J.F., SLATER, T.J.A., HAIGH, S.J., HAGE, F.S., KEPAPTSOGLOU, D.M., RAMASSE, Q.M. &

BRYDSON, R. (2014). Bilayer Graphene Formed By Passage Of Current Through Graphite:

Evidence For A Three-Dimensional Structure. Nanotechnology 25(46), 465601.

LEWIS, E.A., HAIGH, S.J., SLATER, T.J.A., HE, Z., KULZICK, M.A., BURKE, M.G. & ZALUZEC, N.J.

(2014). Real-Time Imaging And Local Elemental Analysis Of Nanostructures In Liquids.

Chemical Communications 50, 10019-10022.

LEWIS, E.A., SLATER, T.J.A., PRESTAT, E., MACEDO, A., O'BRIEN, P., CAMARGO, P.H.C. & HAIGH,

S.J. (2014). Real-Time Imaging And Elemental Mapping Of Agau Nanoparticle Transformations.

Nanoscale 6(22), 13598-13605.

VAN DE LOCHT, R., SLATER, T.J.A., VERCH, A., YOUNG, J.R., HAIGH, S.J. & KROEGER, R. (2014).

Ultrastructure And Crystallography Of Nanoscale Calcite Building Blocks In Rhabdosphaera

Clavigera Coccolith Spines. Crystal Growth & Design 14(4), 1710-1718.

13
 1. Introduction

1.1. Nanoparticle Introduction

Nanoparticles display vastly different properties to their bulk counterparts due to their incredibly

small size (1 – 100 nm). The size of nanoparticles results in a large surface to volume ratio,

which is of particular interest in catalysis, where active catalytic sites are solely related to the

surface of the material (Serp & Philippot, 2013; Xia, et al., 2013). Size effects also promote

properties such as localised surface plasmon resonances in metallic nanoparticles (Hutter &

Fendler, 2004; Raether, 1988; Willets & Van Duyne, 2007) and quantum confinement in

semiconductor quantum dots (Wroggen, 1997), both of which have seen great interest for

optical purposes, such as new generation solar materials (Hillhouse & Beard, 2009; Kim, et al.,

2012; Yu & Chen, 2009).

The precise size and shape of nanoparticles has a large impact on many of their properties. As

well as an increase in catalytic performance in general for decreasing nanoparticle size (Laoufi,

et al., 2011; Thiele, 1939), the precise morphology of nanoparticles is important for their

catalytic properties. Nanoparticle catalysis is highly facet dependent, i.e. the catalytic activity is

dependent on the crystal facets available at the surface of nanoparticle (Cuenya, 2013;

Narayanan & El-Sayed, 2004), which provides a key dependence on nanoparticle shape. The

size and shape of plasmonic nanoparticles also has a large effect on their plasmon resonance

wavelength, which offers the potential to tune the wavelength to a desired value (Liz-Marzan,

2006; Willets & Van Duyne, 2007). Thus, accurate characterisation of both size and shape of

nanoparticles is important to both predicting and understanding their properties.

In nanoparticles of more than one element, the precise composition and the distribution of

elements both play key roles in the particles’ properties. Nanoparticles of more than one

constituent element have become particularly important for catalytic applications, where the

combination of properties from different elements can improve catalytic properties (Ferrando, et

al., 2008; Wang & Li, 2011). For example, there has been great interest in combining additional

metals with Pt nanoparticles for their use in hydrogen fuel cells where the addition of Ni or Co to

form bimetallic particles acts to reduce the cost of expensive Pt catalysts, enhances the oxygen

reduction reaction and improves the stability through many thousands of reaction cycles (Liu, et

14
al., 2013; Mukerjee & Srinivasan, 1993). In synthesizing multi-element nanoparticles it is vital to

ensure that the surface is catalytically active, either through segregation of the more catalytically

active element or precise control of a core-shell structure so that the electronic distribution is

optimised (Jiang, et al., 2011; Toshima & Yonezawa, 1998). As the crystallography of surface

facets is important for catalytic properties, the composition of each facet is therefore directly

relevant to the catalytic activity of multi-element nanoparticles (Cui, et al., 2012; Shao, et al.,

2006). The distribution of elements within nanoparticles is therefore another key property for

characterisation, although typically more challenging to determine than the size and shape

analysis.

The transition from single-element to multi-element nanoparticles also has large effects on the

plasmonic properties of noble metal nanoparticles. The wavelength and linewidth of localised

surface plasmon resonances are tuneable through the addition of a second metal (Cortie &

McDonagh, 2011; Link, et al., 1999), such that the resonance wavelength can be tuned to

wavelengths of lower absorption in key materials, such as tissue (Jain, 2008). The distribution of

the elements within nanoparticles plays a key role in their plasmonic properties; fully alloyed

nanoparticles display sharper resonance peaks than core-shell systems for example (Prodan, et

al., 2003).

Extending the definition of nanoparticles to include those encapsulated in a solid matrix, often

termed nanoprecipitates, vastly extends the scope of nanoparticle studies. Any secondary

phase of material, distinct from the matrix surrounding it, with dimensions of 1 -100 nm can be

considered a nanoparticle. In fact, characterisation of this type of particle often involves removal

from the matrix via etching, in order to obtain unambiguous information on the particles’

structure and composition without contribution from the surrounding matrix, at which point the

nanoprecipitate becomes a freestanding nanoparticle.

Nanoprecipitates play an important role in many of the properties of bulk materials. For example,

in the case of carbide formation in stainless steels, the formation of intragranular carbides

greatly increases the strength of the steel due to improved creep resistance (Sourmail, 2001;

Yamamoto, et al., 2007). Carbides formed at grain boundaries also play an important role in the

corrosion behaviour of stainless steels, causing Cr depletion in the matrix which lowers the

corrosion resistance (Armijo, 1968; Joshi & Stein, 1972; Tedmon, et al., 1971). Thus,

15
characterising the number, distribution and the composition of precipitates has an important

function in the understanding of materials’ properties.

1.2. Nanoparticle Characterisation

The size, shape and elemental distribution of nanoparticles can be characterised by a wide

variety of analytical methods. For example, Ultra Violet – Visible (UV-Vis) and Infrared (IR)

spectroscopies (Amendola & Meneghetti, 2009; Haiss, et al., 2007; Higgins, et al., 2003) and X-

ray diffraction (Dorofeev, et al., 2012; Patterson, 1939) can be used to determine nanoparticle

size and shape. X-ray photoelectron spectroscopy (Baer, et al., 2010) and X-ray absorption

spectroscopy (Hwang, et al., 2005) are used to probe the surface and bulk composition of

nanoparticle populations respectively. However, these techniques are only able to obtain

average values from a large ensemble of nanoparticles and do not offer direct visualisation of

individual nanoparticles or their properties.

Scanning probe techniques, which include atomic force microscopy (AFM) and scanning

tunnelling microscopy (STM), have shown great utilisation in examining individual nanoparticle

size and morphology (Grobelny, et al., 2011; Henry, 2005; Kano, et al., 2015). Characterisation

of individual particles is particularly important due to the large dispersity in properties that is

often found amongst samples of synthesized nanoparticles. However, the nature of scanning

probe techniques means that it is not possible to characterise nanoparticles embedded inside a

material. Examination of chemical variations is possible also with STM, but it is not

straightforward to quantify chemical composition (Kano, et al., 2015; Suzuki, 2015).

Electron microscopy has proved an invaluable technique in nanoparticle characterisation, with

nanoparticle size and shape determination straightforward in the scanning electron microscope

(SEM) (Bootz, et al., 2004; Liu, 2005). However, the use of energy dispersive X-ray (EDX)

spectroscopy in the SEM is limited in its resolution, due to the large interaction volume of X-rays

from nanoparticles mounted on a thick substrate (Goldstein, et al., 2003). Techniques utilising

the transmission electron microscope (TEM) and scanning transmission electron microscope

(STEM) are incredibly versatile in characterising the size, morphology and composition of

nanoparticles. In comparison to the use of scanning probe techniques, mapping the extent of

nanoparticle composition is becoming a routine process in the STEM due to the complementary

16
spectroscopic techniques of electron energy loss spectroscopy (EELS) and EDX spectroscopy

(Brydson & Hondow, 2011; Watanabe, 2011; Williams, et al., 2002).

1.3. Thesis Structure

The aim of this thesis is to thoroughly investigate the methodology of STEM-EDX tomography in

order to provide three dimensional elemental imaging of nanoparticles. The results included in

this thesis have been published in, or have been submitted to, a number of peer-reviewed

journals and the submitted manuscripts are included in full. These chapters therefore have self-

contained figure numbering and referencing. The information presented in this thesis is the

result of work with a number of collaborators and their contributions are made clear at the start

of each chapter where appropriate. Contributions not assigned to collaborators were performed

by the author.

An overview of scanning transmission electron microscopy and its application to elemental

imaging of nanoparticles is included as chapter 2, with a focus on EDX spectroscopy. Chapter 3

presents a novel method to determine X-ray absorption in STEM-EDX elemental maps of

spherical nanoparticles. An absorption correction factor is derived and is verified through

comparison to elemental maps and spectra acquired from a number of Au nanoparticles.

An introduction to electron tomography is presented in chapter 4, with a review of the imaging

modes used to characterise three dimensional elemental distributions within nanoparticles.

Chapter 5 contains a study on the methodology of EDX tomography, with consideration of a

number of factors influencing the robustness of the technique. In particular, the effect of the

sample holder shadowing the X-ray detectors is investigated and compensation methods are

discussed and implemented. EDX tomography is applied to understand the catalytic behaviour

of AgAu nanoparticles in chapter 6. The three dimensional elemental analysis reveals a change

in the distribution of Ag and Au with overall composition that correlates to their catalytic

properties.

17
 2. The Scanning Transmission Electron Microscope

Electron microscopy has become one of the most important materials characterisation

techniques since its inception in the 1930s (Knoll & Ruska, 1932). Its strength comes from the

very small wavelength of the electron, particularly in comparison to that of light. The minimum

resolvable distance between two objects is limited by diffraction to the order of the wavelength

of the illuminating radiation, as was first proposed by Lord Rayleigh in 1879 (Rayleigh, 1879).

Conventional light microscopy is therefore typically limited to a resolution of near optical

wavelengths (on the order of 100 nm). Electron wavelengths are typically on the order of 1 x
-3
10 nm for typical transmission electron microscope energies (100 - 300 keV) and therefore

have the potential to resolve not just nanometre size features but features down to the atomic

level.

Types of TEM are generally split in to two categories, the conventional TEM (CTEM) and the

scanning TEM (STEM). The more well-known technique, the CTEM, is akin to a conventional

light microscope in that a broad parallel illumination is used that can be treated as a plane wave.

In the scanning transmission electron microscope a very finely focussed probe is scanned

across a sample in a pixel-by-pixel manner (Fig 2.1). The two techniques are interrelated by the

principle of reciprocity, which will be explained in the next section, but leads to direct

comparison between imaging modes in the two techniques. In this chapter, all concepts will be

introduced in relation to the STEM but relation of imaging modes back to those of the CTEM is

discussed where appropriate.

18
Figure 2.1. Illustration of the components of a scanning transmission electron microscope and

the associated signals: energy dispersive X-ray (EDX) spectroscopy, bright field (BF), high

angle annular dark field (HAADF) and electron energy loss (EEL) spectroscopy.

In short, in the STEM, electrons are extracted from a source in the electron gun and an image of

this source is effectively demagnified to a small spot on the sample by the condenser lens

system and the objective lens. This spot can be raster scanned across the sample by deflection

coils located immediately before the specimen. After undergoing interactions with the sample,

transmitted electrons are collected by a number of detectors, dependent on the electron

scattering angle. Electron detectors include bright field and annular dark field detectors, as well

as spectrometers for electrons and X-rays (Fig 2.1). One of the great advantages of the STEM

is that a number of different signals can be simultaneously collected by different detectors, with

the detected signals having ideally originated at the scanned position of the beam. This type of

raster scanned approach is particularly useful in spectroscopy, where a full spectrum can be

collected at each pixel in a method known as spectrum imaging.

This chapter introduces the concept of scanning transmission electron microscopy with relation

to analysing elemental distributions in nanoparticles. Firstly, electron interactions are discussed

and the concepts of elastic and inelastic scattering, as well as coherent and incoherent imaging

are introduced. Secondly, imaging modes are discussed, with a particular focus on nanoparticle

19
analysis. As the focus of this thesis is on energy dispersive X-ray spectroscopy, the key

developments of the technique are reviewed that have allowed spectrum imaging of

nanoparticle systems. At the end of the chapter, factors limiting STEM imaging are discussed, in

sections detailing the resolution of the STEM and electron beam damage.

2.1. Electron Interactions with Matter

In the STEM, practically all of the detected signals are due to the interaction of the electron

probe with the sample. Electrons interact strongly with matter, with a very short mean free path

compared to photons (on the order of nanometres or tens of nanometres depending on the

electron energy and material (Powell & Jablonski, 2010; Sondheimer, 2001)), which limits the

thickness of the sample that electrons may be transmitted through. Electron interactions are

termed either elastic interactions, in which the transmitted electron does not lose an appreciable

amount of energy, or inelastic interactions, in which the transmitted electron loses an

appreciable amount of energy. In reality, elastic interactions also result in a transfer of energy

but, particularly for the small-angle scattering considered in the CTEM or STEM, this energy is

so small when compared to the energy of the beam electrons (< 1eV compared to 100 – 300

keV) that it is not considered (Egerton, 2011a). Both elastic and inelastic interaction

mechanisms are important to different imaging modes in the STEM and are discussed here

independently.

2.1.1. Elastic Scattering

Elastic scattering can be described by considering the Coulombic interaction of the probe

electrons with the potential distribution of the sample. The electron is both a particle and a wave

and therefore attempts to describe this interaction may make use of either the particle or wave

nature of the probe electrons.

If the electron is considered as a particle, the elastic interaction can be described in the simplest

terms by considering the scattering of the probe electrons by the point potentials of the atomic

nuclei. This scattering is known as Rutherford scattering and has a differential cross section, the

probability of scattering in to a small solid angle dΩ at an angle θ, that is defined by

𝑑𝜎(𝜃) 𝑒 4𝑍2 (2.1)

= 𝜃 4
,
𝑑Ω
16(𝐸0 )2 (sin )
2

20
where e is the unit electric charge, Z is the atomic number of the nucleus and E0 is the initial

electron beam energy (Williams & Carter, 2009b). Thus, in this scheme, the scattered intensity
2
at a scattering angle θ is proportional to Z of the nucleus. In practice, shielding of the nuclear

potential/charge by atomic electrons modifies this scattering to some degree. This can be

𝜃 2 𝜃 2 𝜃 2
represented by replacing (sin ) with (sin ) + ( 0) in the Rutherford cross section to give
2 2 2

𝑑𝜎(𝜃) 𝑒 4 𝑍2 (2.2)
= 2 ,
𝑑Ω 𝜃 2 𝜃 2
16(𝐸0 )2 ((sin ) +( 0 ) )
2 2

where θ0 is the screening parameter and can be thought of as the scattering angle below which

the electron screening affects the scattering cross section. The screening parameter is also

dependent on Z and in practice acts to reduce the power of the Z dependence, particularly for

low-angle scattering.

The particle description of electron scattering is useful to a certain degree, where the differential

cross section can be thought of as the probability of scattering a single electron by a single

atom to certain scattering angles. In this description we would expect this intensity distribution to

sum for many electrons scattered from an array of atoms so that if each electron could only

scatter once, known as kinematic scattering, we would retrieve an intensity as defined by

equation 2.2 at a particular scattering angle θ. Simply by placing an aperture or detector at

some scattering angle it should therefore be possible to include only electrons scattered to a

particular range of angles, the intensity of which will show a clear dependence on the atomic

number of the material.

In actuality, in order to fully describe the intensity distribution at detectors the electron probe

must be treated as a wave, in which the scattering from atomic potentials is described by a

scattering factor. When considering the electron as a wave, we have to take in to account that

the scattering process may be coherent or incoherent in nature. Coherent scattering involves

the summing of the amplitudes of scattered waves, which, in a crystalline lattice, will result in

constructive and destructive interference from points of scatter. This is the basis of both phase

contrast and diffraction contrast in bright field and dark field STEM and CTEM.

21
The intensity of coherent imaging in the STEM, Icoh, can be described by a convolution of the

wavefunction of the probe ψp with the transmission function of the sample φ, which describes

how the sample alters the electron wavefunction, for a probe position Rp (Nellist, 2011):

2
𝐼𝑐𝑜ℎ (𝑹𝑝 ) = |𝜓𝑝 (𝑹𝑝 )⨂𝜙(𝑹𝑝 )| . (2.3)

If the imaging mode can be shown to be incoherent, i.e. the scattering intensities from individual

atoms can be summed, we can approximate the scattering using the Rutherford description.

This is the basis for contrast in bright field imaging of non-crystalline materials, so called mass-

thickness contrast and, as we will expedite later, in high-angle annular dark field imaging. In

contrast to coherent imaging, the intensity of incoherent imaging is given by a convolution of the

intensity of the STEM probe and the object function of the specimen, that describes how the

sample scatters intensity to a specific detector (Nellist, 2011):

2 (2.4)
𝐼𝑖𝑛𝑐𝑜ℎ (𝑹𝑝 ) = |𝜓𝑝 (𝑹𝑝 )| ⨂𝑂(𝑹𝑝 ).

The intensity of incoherent imaging is simply the convolution of the square of the object function

with the point spread function of the electron probe, which is the intensity distribution in the

probe.

Another key point to make about purely elastic scattering is that it can be viewed as reversible

with respect to time, i.e. the scattering process is independent of the order of events. This leads

to the principle of reciprocity in the STEM, that the same scattering process occurs, and

therefore the same images are obtained, if the source and detector are exchanged (illustrated in

Fig 2.2). This allows the transfer of the detailed mechanistic understanding of contrast formation

processes in the CTEM to the STEM.

22
Figure 2.2. Diagram illustrating the principal of reciprocity between a) CTEM and b) STEM for

elastic scattering. The source and detection planes are interchanged and the position of the

objective aperture with respect to the sample is reversed. Adapted from (Brown, 1981).

2.1.2. Inelastic Scattering

In addition to the elastic scattering described above, probe electrons may also undergo inelastic

collisions, in which an appreciable amount of energy is transferred from/to an excited state in

the sample. Inelastic interactions include those with phonons, plasmons and the ionisation of

core-level electrons. All of these interactions have the potential to be used spectroscopically,

however, the energies associated with phonon or plasmon modes have historically been too

small to give accurate spectroscopic measurements. The excitation of core-level electrons is

therefore the most intensely scrutinised inelastic scattering process in the STEM.

Ionisation of atomic electrons is usually described by the Bethe cross-section σI (Bethe, 1930),

as given in equation 2.4, or one of a number of variants of the Bethe cross-section (Powell,

1976):

𝜋𝑒 4 𝑏𝑠 𝑛𝑠 𝑐𝑠 𝐸0 (2.5)
𝜎𝐼 = ( ) log ( ),
𝐸0 𝐸𝑐 𝐸𝑐

where Ec is the energy required to ionise the atom, ns is the number of electrons in the ionised

subshell and bs and cs are known as Bethe parameters and are constant for ionisations of the

23
same sub-shell. Here, a total cross section is presented as opposed to the differential cross

section presented for elastic scattering. The total cross section is generally a reasonable

description of spectroscopic processes in the STEM, for EDX spectroscopy (EDXS) due to the

fact that there is no angular dependence on characteristic X-ray production (Bethe, 1930) and

for EELS when the collection aperture is large enough to collect almost all inelastically scattered

electrons (Egerton, 2009).

Whilst the interaction of the electron beam with phonons has only recently become a

spectroscopically measurable quality (Krivanek, et al., 2014), it does play an important role in

electron imaging modes in the STEM. Thermal diffuse scattering (TDS), which is usually defined

as interaction of the beam electrons with pre-existing phonons in the crystal lattice, is

particularly important in HAADF imaging where it has been shown that the majority of the

electrons impinging on the detector are from thermal diffuse scattering if a large detector inner

angle is used (Jesson & Pennycook, 1995). TDS in CTEM has conventionally been treated as

“absorption” of electrons, after which the absorbed electrons do not contribute to diffraction or

image formation. Similarly, in the treatment of HAADF images the effect of TDS can be

considered using an absorptive potential approach (Allen, et al., 2003; Findlay, et al., 2003), in

which electrons undergoing TDS are no longer elastically scattered.

2.2. Electron Imaging and Spectroscopy of Nanoparticles

2.2.1. Bright Field Imaging

Bright field (BF) images in the STEM are acquired by collecting electrons with a detector

mounted on the optic axis (Fig 2.1). Due to the principle of reciprocity, the bright field technique

in the STEM can be treated in the exact same manner as the more commonly used bright field

imaging in the CTEM. In the strictest sense, true BF imaging in the STEM, in which only

undiffracted electrons are collected, is only achieved when the convergence angle of the probe,

determined by the objective aperture, is smaller than the scattering angle of the first Bragg

scattered disk. When the convergence angle matches this criterion there is no contribution from

phase contrast, just as in CTEM BF imaging. Typically, however, STEM BF imaging is operated

with a large convergence angle that includes information from at least the first Bragg diffracted

disks and phase contrast is generated in a similar manner to high resolution TEM (HRTEM)

24
imaging in the CTEM. Contrast from the bright field detector is therefore due to diffraction

contrast, phase contrast and mass thickness contrast, which will each be discussed here briefly

in turn.

The coherent summing of electron wave amplitudes in a crystalline material results in so-called

diffraction contrast. Changes in contrast due to coherent scattering come from changing sample

thickness or orientation (including effects due to defects or bending) and all come under the

umbrella term of diffraction contrast. Whilst diffraction contrast is a very important contrast

mechanism for traditional CTEM and STEM of thin foils and has extensive associated literature

(Hirsch, et al., 1977; Williams & Carter, 2009a), in the study of nanoparticles it has made little

impact. In fact, if attempting to analyse elemental distributions it is useful to avoid conditions

conducive to diffraction contrast as it may obscure information from mass thickness contrast.

The coherent nature of the electrons impinging on the BF detector also lead to Fresnel fringes

at the edge of structures in the presence of defocus in a similar manner to BF CTEM (Colliex, et

al., 1977). Fresnel fringes can obscure the analysis of the surfaces of nanoparticles, especially

if confused with mass thickness effects.

The probe wavefunction represented in equation 2.3 is described by a complex aperture

function that has a phase, at a given wavevector, given by the aberration phase shift –χ(Kp),

where Kp is the transverse component of the wavevector of the partial plane wave included by

the aperture (Nellist, 2011). In the weak phase object description of phase contrast it is

assumed that the transmission function of the object results in only a small phase shift with no

change in amplitude and thus the intensity of coherent scattering is dependent on the factor

sin(χ(Kp)). This factor is known as the phase contrast transfer function and determines the

intensity of contributions from features of certain spatial frequencies. The aberration phase shift

is dependent on all aberrations, including defocus, and therefore small changes in focus can

result in changes in the intensity of periodic features, including reversals in contrast. The

complex relationship between the properties of the objective lens and the contrast in a phase

contrast image mean that interpretation of these images relies on comparison with accurate

image simulation (Kirkland, 2010).

Bright field imaging in the STEM or CTEM can be used to distinguish between different

elements through mass thickness contrast, particularly when the mass thickness difference is

25
large. For instance, BF imaging has been used to verify the encapsulation of La in carbon

nanoparticles (Ruoff, et al., 1993) and Pt in carbon nanotubes (Kyotani, et al., 1997). With

respect to bimetallic nanoparticles STEM BF imaging can be used to distinguish a

lighter/heavier shell, as shown in studies of AgAu (Rodriguez-Gonzalez, et al., 2005).

2.2.2. High-Angle Annular Dark Field Imaging

The high-angle annular dark field (HAADF) imaging mode is that in which an annular detector is

used to collect electrons scattered above a threshold scattering angle (see Fig 2.1). HAADF is

generally considered to be an incoherent imaging mode, as briefly discussed in section 2.1.1.

Incoherence of HAADF imaging is assumed either due to exclusion of Bragg scattered beams

and domination of thermal diffuse scattering (Howie, 1979; Jesson & Pennycook, 1995) or

through averaging over a large number of Bragg scattered discs (Nellist, 2011). Thus, the

intensity of the HAADF signal is often described by Rutherford scattering and is largely

dependent on the Z of the material and its thickness. The exact exponent of the Z dependence

depends upon the element in question and the angles that the detector subtends (Treacy, 2011).

Through the principle of reciprocity the HAADF mode is related to the use of a large annular or

tilted electron source in the CTEM, which would provide a largely incoherent source of electrons

(Spence, et al., 1989).

The strong dependence of HAADF imaging on the atomic number of the sample has led to its

wide-spread use in assessing the elemental distribution within nanoparticles. While a slowly

varying change in intensity in HAADF images cannot be unambiguously separated from

thickness changes, a step change in intensity is often related to a change in elemental

composition. Bimetallic or multimetallic nanoparticles are therefore often confirmed as being of

a core-shell nature via HAADF imaging. A wide range of bimetallic nanoparticles have been

studied in this way, including AgAu (Belic, et al., 2011), FeAu (Cho, et al., 2005), AgSi (Cassidy,

et al., 2013) and AuRh (Chantry, et al., 2012). The interface between regions of different

elements may even be probed down to atomic resolution, as has demonstrated in AuPd (Ferrer,

et al., 2008; Mayoral, et al., 2012) and GaPd (Leary, et al., 2013). Although HAADF imaging

may distinguish between different regions of known composition, it cannot determine the

elements present within a nanoparticle, for that a truly spectroscopic technique, such as EELS

or EDXS, is needed.

26
2.2.3. Electron Energy Loss Spectroscopy

Electron energy loss spectroscopy involves the collection and analysis of electrons that have

lost energy due to inelastic scattering processes within the sample (Egerton, 2009). The EEL

spectrometer works in much the same way as a prism of light, separating electrons of different

energy spatially on to a CCD camera by a magnetic field normal to the electron beam. The

electron energy loss spectrum is generally split in to two regions, the low-loss region that

contains energy losses associated with phonons and plasmons and the core-loss region that

contains energy losses associated with ionisation events (Fig 2.3). At a core-loss edge, after

subtracting a background at each pixel, integrating over the intensity of a number of channels

will give an intensity related to the elemental composition (Egerton, 2011b). The elemental

composition can be quantified in a similar manner to the EDXS quantification that is described

later. One of the great benefits of EELS is the high-energy resolution of the technique (down to

10s of meV in the most recent cases (Krivanek, et al., 2013)) in comparison to EDX

spectroscopy (≈ 130 eV for silicon drift detectors (Eggert, et al., 2004)). The ability to resolve

differences in energy loss on the order of nanometres allow visualisation of the fine structure of

core-loss edges that reveals information on the chemical bonding of elements (Batson, 1993;

Silcox, 1998). Similarly, the energy resolution allows measurement of subtle shifts in edge

energy that can be associated to bonding environment (Potapov & Schryvers, 2004).

Figure 2.3. Example EEL spectrum showing the characteristic features of the zero-loss peak,

the plasmon-loss peaks and core-loss edges, the latter have had their intensity multiplied by 50

times to ensure they are visible at this scale. Elemental identification is possible through the

characteristic energies of core-loss edges associated to each element.

27
A similar imaging method to EELS is energy filtered TEM (EFTEM), which is the broad-beam

CTEM equivalent of EELS (Reimer, 1995). In EFTEM a slit is placed in the dispersion plane of

the electron spectrometer so that only electrons of a certain energy are allowed to form an

image on the camera at the end of the spectrometer.

Elemental mapping via EELS can provide direct confirmation of the core-shell structure of

nanoparticles, for example in CoMn (Lee, et al., 2011). In terms of elemental segregation, EELS

has directly shown Pt surface segregation in electrochemically dealloyed CuPt (Wang, et al.,

2012) and acid leached PtNi nanoparticles (Gan, et al., 2013), as well as Pt segregation to

certain edge sites in PtNi nanoparticles (Cui, et al., 2013), helping to explain each system’s

increased performance in catalysis of the oxygen reduction reaction. Ordering of elements in

nanoparticles can also be observed, with an ordered structure seen in PtCo nanoparticles

(Wang, et al., 2013). Through observation of shifts in the edge energy, it is also possible to

monitor phase changes within a single nanoparticle for a single element, such as that in Pd

nanoparticles (Baldi, et al., 2014).

Atomic resolution EELS mapping, routinely performed in bulk oxides (D'Alfonso, et al., 2010;

Kimoto, et al., 2007), has also been demonstrated in an oxide nanoparticle (Rossell, et al.,

2012). In this study atomic resolution EELS mapping of Ba-doped SrTiO3 nanoparticles was

used to reveal the extent of dopant clustering within these nanoparticles. Dopants down to the

single atom level can be detected via EELS in bulk materials (Varela, et al., 2004), which should

be extendable to nanoparticle studies.

2.3. Energy Dispersive X-ray Spectroscopy

2.3.1. X-ray Generation

2.3.1.1. Characteristic X-rays

As discussed previously, inelastic scattering of probe electrons through interaction with atomic

electrons may result in the ionisation of atoms within the sample, in which a core-level electron

is ejected. The decay of atomic electrons as the electronic configuration of the atom proceeds

back to the ground state may result in the production of X-rays of a specific energy (Fig 2.4).

This energy is dependent on the energy difference between electronic states and is therefore

specific to each element. Detection of these characteristic X-rays allows identification of

28
elements within a sample, which is the underlying principle of EDX spectroscopy. Characteristic

X-rays are labelled based on the final and initial energy levels of the transitioning electron, these

are summarised in Fig 2.4b.

Figure 2.4. a). Diagram of the process in which an incident electron causes a characteristic X-

ray to be emitted from an atom. b) Diagram of the possible electron transitions in the atom with

associated X-rays labelled (reproduced from (Williams & Carter, 2009c)).

X-ray generation can be characterised as a secondary process to the primary process of

electron scattering. It is important to note that X-ray generation is always in competition with

another secondary process, namely the emission of Auger electrons. The fraction of ionised

atoms which decay through X-ray emission is known as the fluorescence yield, ω, and can be
-3
as low as 10 for low energy X-rays (Krause, 1979). Characteristic X-ray detection is therefore

known to be a highly inefficient detection method for light elements, particularly when compared

to electron energy loss spectroscopy (EELS), in which the primary process is detected.

However, there are advantages to characteristic X-ray generation over the collection of

inelastically scattered electrons. X-rays are known to be emitted isotropically over all angles,

allowing a somewhat arbitrary position of X-ray detectors and no need for consideration of

differential cross sections (Bethe, 1930). Characteristic X-ray generation is also an inherently

incoherent process as there is no phase relationship between atomic decays.

2.3.1.2. Bremsstrahlung X-rays

One of the drawbacks of X-ray detection is that the X-rays that reach the detector are not just

produced through the radiative decay of ions. X-rays are also produced through the

29
deceleration of the electron beam as it undergoes inelastic scattering, emitting X-rays that are

known as Bremsstrahlung, or braking, radiation. Bremsstrahlung X-rays are emitted

continuously over an energy spectrum, as electrons in the beam may undergo any amount of

deceleration. The number of Bremsstrahlung photons produced, N, at an energy, E B, is given by

Kramers’ “cross section” (Kramers, 1923):

𝐾𝑍(𝐸0 −𝐸𝐵 ) (2.6)

𝑁(𝐸𝐵 ) = ,
𝐸𝐵

where E0 is the energy of the incident electrons and K is Kramers’ constant. The continuous

nature of Bremsstrahlung X-rays mean that characteristic X-rays in an EDX spectrum are

always superimposed on to a background of Bremsstrahlung X-rays. Fortunately, the angular

distribution of Bremsstrahlung radiation is strongly forward peaked and X-ray detectors are

generally placed at large scattering angles, therefore the Bremsstrahlung background does not

swamp the characteristic X-ray peaks.

2.3.1.3. Spurious X-rays

In the detection of electrons in the CTEM/STEM, the projection system, including the pre- and

post-specimen apertures and the position of the detectors, ensures that only electrons

associated with small angle scattering from the electron beam are detected. For EDX detectors,

a collimator is used to try to ensure the detection of X-rays only from the illuminated specimen

area, but invariably X-rays generated from outside of this area are also detected. X-rays

generated from outside of the illuminated area are known as spurious or system X-rays, these

terms have slightly differing definitions, and can compromise accurate spectroscopic analysis.

Spurious X-rays may be generated either by stray electrons exciting characteristic X-rays away

from the intentionally illuminated area, or through emitted X-rays (either characteristic or

Bremsstrahlung) fluorescing X-rays away from the illuminated area. X-rays generated by stray

electrons can often be determined through collection of a ‘hole’ spectrum from a region in which

there is no sample (Fig 2.5).

Typically, system X-rays may occur from the support grid and holder (typically Cu), the pole-

piece (Fe and Co), the detector (Si and Pb) and, counterproductively, the X-ray collimator (Zr in

the case of the Super-X detector system presented here). It is possible to avoid, or at least to

limit, the production of spurious X-rays through the use of holders and grids manufactured from

30
Be, which causes production of low energy X-rays which are not resolved by the EDX detector

and cannot fluoresce higher energy X-rays elsewhere. Limiting the tilt of the sample if mounted

on to a grid also helps to limit spurious X-ray production from grid bars, as large sample tilts

lead to grid bars interacting with a larger number of stray electrons and Bremsstrahlung X-rays,

although tilting the sample is a necessity for some detector-holder setups in order to avoid

shadowing of the X-ray detector.

Figure 2.5. Spectra from a sample containing elements Al, O and Co from vacuum (blue),

carbon film (green) and the sample (red). There are few counts from vacuum, demonstrating

that stray electrons are not a major contributor to spurious X-rays in this particular case. Sample

courtesy of Cat Yeoh (Cambridge University).

2.3.2. X-ray Detection

The proportion of generated X-rays that are ultimately collected is dependent on a number of

factors, including the solid angle that the detector subtends, the absorption of X-rays in the

sample, and at the detector, and the proportion of X-rays that pass all the way through the

detector. A large amount of detector development has aimed to increase this value, often known

as the detection efficiency, as EDX spectroscopy in the STEM has typically been limited by low

count rates, particularly for nanoparticle samples.

2.3.2.1. EDX Detector Development

All commercially available X-ray detectors for the CTEM/STEM work on the principle of electron-

hole generation in a semiconducting material. As an X-ray traverses the detector it has a

probability of exciting a number of electron-hole pairs with energy equal to the bandgap of the

detection material (1.1 eV in the most commonly used case of Si). The generated electrons are

collected at an anode where they are then detected as a sharp step in voltage from a field effect

31
transistor. Noise, whether electronic or thermal noise, is detected as small steps in voltage in

much the same way as characteristic X-rays. This noise may be detected in isolation, where it is

assigned to channels around 0 eV and leads to a large peak at this value, or it may be detected

on top of a step due to a characteristic X-ray, which leads to broadening of the detected

energies of the characteristic X-rays, thus limiting the energy resolution of the technique. The

detection mechanism of the detector may also result in artefacts in the spectra generated. For

example, if two X-rays are incident on the detector within the read time of the detector a sum

peak is generated at an energy that is the sum of the two X-rays. Detectors may also suffer

from incomplete charge collection which results in a shoulder on the characteristic peak at the

low energy side of the peak. For most commercial detectors the acquisition software

automatically corrects for these artefacts.

Most modern EDX detectors in the TEM are now silicon drift detectors (SDDs), as opposed to

the Li drifted silicon detectors (Si(Li)), which were standard until recently. The SDD features a

large n-doped silicon crystal with concentric p-doped rings embedded on the back side of the

detector (Fig 2.6b). The design of the SDD, with a small area anode (typically less than 100 μm

in diameter) to collect electrons, enables shorter discharge times than the Si(Li) detector, in

which the anode covers the whole back face (Fig 2.6a). This leads to much lower processing

times in the SDD which has allowed much higher X-ray count rates. It also means that the

throughput achievable is somewhat independent of detector size, allowing the development of

larger and larger detectors (Newbury, 2006). The small anode also removes an important

source of noise in the detector, which allows good energy resolution at relatively modest

temperatures (above -100°C). This removes the necessity to cool with liquid nitrogen and

instead electronic Peltier cooling can be used.

32
Figure 2.6. Schematics of cross-sections of EDX detectors. a) Si(Li) detector, in which the

anode covers the whole back face. b) SDD detector, in which concentric p-type rings promote

“drift” of generated electrons to a small central anode.

When high count rates are present (> 100,000 counts), it is noted that increasing the detector

size has a large negative impact on the resolution of the detector (Eggert, et al., 2004). For the

low count rates that are characteristic of the STEM however, large area SDDs can play an

important role in increasing the solid angle of the detector system. The SDD also lends itself to

the fabrication of arrays of single detectors (Fiorini, et al., 2002; Zaluzec, 2004). It seems

entirely possible that arrays of SDDs that are shaped to fit specifically within the STEM stage

will be available in the near future.

One drawback of SDDs is that they are typically much thinner than the Si(Li) detector, which

means that they do not detect all incident X-rays above a certain X-ray energy. Thus, the

detection efficiency of an SDD decreases above a certain threshold energy. The detection

efficiency of detectors is also affected by the positioning of a window in front of the EDX

detector. Traditionally, a thin window (10 – 25 μm) of Be was placed in front of the EDX detector

in order to prevent ice or contamination build up on the cold detector. Over the years these

windows have had their thickness gradually reduced to less than 100 nm through the

incorporation of other materials such as polymer films (Williams, et al., 1995). It is now common

for SDDs to be “windowless” in that they do not have a traditional window but are shuttered

when not in use (Anderhalt & Sandborg, 2011). Even when detectors are windowless they will

typically have a very thin (on the order of 100 nm) light shield to reflect/absorb incident light and

thus have a degraded detection efficiency for low energy X-rays. The detection efficiency of a

single detector is therefore largely dependent on the X-ray energy, with detection efficiency

33
degraded at both low and high X-ray energies. The precise detection efficiency can be vastly

different for each type of detector and highly dependent on X-ray energy (Fladischer, 2013).

2.3.2.2. Detector geometry

The traditional geometry of an EDX detector in the STEM is a single detector, inserted from

outside the column to as close a distance as possible to the sample, typically subtending a solid

angle of approximately 0.1 - 0.3 sr. A solid angle of between 0.1 – 0.3 sr results in detection of

only 0.8 – 2.4 % of characteristic X-rays. When we consider this poor collection efficiency in

combination with potentially low fluorescence yields described in section 2.3.1, it is obvious that

EDX spectroscopy is a very inefficient method to detect inelastic electron scattering. For

example, a typical EDX detector of 0.2 sr is able to detect a maximum of approximately 0.005 %

of inelastic electron scattering events from carbon atoms, without consideration of X-ray

absorption, compared to a value often larger than 50 % for EELS. It should be noted, as the

positive case for EDX, that the nature of EDX spectroscopy means that peak to background

ratios are often far greater than those of EELS and heavy element detection is often much more

favourable in EDXS than EELS (von Harrach, et al., 2010). A large part of EDX spectroscopy

development in the STEM in recent years has therefore focussed on increasing the solid angle

of detectors to increase their collection efficiency.

There have been many suggestions of detector geometries that increase the solid angle of the

EDX detector. In most cases, the introduction of the SDD has allowed increases in detector size

or efforts have been made to position the detector closer to the sample (Williams & Carter,

2009c). However, novel detector geometries have also been suggested to increase detector

solid angle. These include a dual detector setup (Myers, et al., 2010), an annular detector

positioned above the sample (Kotula, et al., 2008) and a π-sr detector in a post-specimen

position (Zaluzec, 2009). The design that has found arguably the greatest commercial success,

and the one used almost exclusively here, is the Super-X detector, as included in FEI’s

ChemiSTEM technology (von Harrach, et al., 2009). This detector is in fact four separate SDDs

that are positioned symmetrically within the polepiece of the STEM. Each detector is similar to a

standard SDD, although positioned very close to the sample, but the fact that there are four

detectors results in an increased solid angle of up to 0.9 sr (von Harrach, et al., 2009). The

precise solid angle achievable depends on the gap within the objective lens, as this determines

34
how close to the sample the detectors can be positioned. The positioning of the detectors at

regular 90° azimuthal intervals around the optic axis also allow for X-ray detection when the

sample is tilted to any angle.

2.3.3. Quantification

The quantification of elemental composition from an EDX spectrum of a thin sample is a

straightforward process that can yield important quantitative information at near atomic

resolution (Kothleitner, et al., 2014). Quantification in CTEM/STEM has principally followed the

method set out by Cliff and Lorimer, using their Cliff-Lorimer k-factors that relate the intensity of

a pair of X-ray peaks (IA and IB) to their composition, in this case the ratio in weight fraction (CA

and CB) (equation 2.6) (Cliff & Lorimer, 1975), where A and B are two elements of interest:

𝐶𝐴 𝐼
= 𝑘𝐴𝐵 𝐴 . (2.7)
𝐶𝐵 𝐼𝐵

Historically, this type of intensity ratio was preferred to absolute intensities due to instabilities in

the CTEM/STEM, primarily instability in the beam current. Cliff-Lorimer factors are dependent

on the precise detector-microscope combination, specifically on a number of factors, including

the accelerating voltage and the collection/detection efficiency of the detector. Cliff-Lorimer

factors can be determined in a number of ways. The most robust method, in terms of accurate

determination, is to experimentally measure the k-factors from standard samples of known

composition (Sheridan, 1989). Standard samples are available that contain a large number of

elements but invariably it can be difficult to obtain accurate standards for all elements. A less

accurate but widely used method is to calculate approximate k-factors from microscope

parameters. This calculation is often performed via equation 2.7, which is obtained through a

derivation from first principles (Williams & Goldstein, 1991):

(σ ωaε)B AA (2.8)
k AB = (σI ,
I ωaε)A AB

where ω is the fluorescence yield, aX is the relative transition probability of the peak, εX is the

detection efficiency of the detector and AX is the atomic weight of the element X.

Another quantification method gaining increasing popularity is the zeta-factor method proposed

by Watanabe (Watanabe & Williams, 2006). This method uses single element factors to

determine composition and has become possible due to the stable beam currents achieved in

35
the modern day STEM. The intensity of X-rays of an element A are related to the composition of

the sample by a zeta factor, ζA, through the following:

𝜁𝐴 𝐼𝐴 (2.9)
𝐶𝐴 = ,
𝜌𝑡

where ρ is the density of the sample and t is the sample thickness. There are many practical

benefits to the zeta-factor method that derive from its use of only one X-ray peak. Firstly,

experimental determination of zeta-factors can be done from single element standards of known

thickness or volume, which are easy to obtain for a large number of elements. Secondly, the

zeta-factor offers a method to determine sample thickness or inelastic cross-sections when the

composition of the sample is already known.

2.3.3.1. Background Subtraction

In both the Cliff-Lorimer and zeta-factor methods an important initial step is to separate

characteristic X-rays from the Bremsstrahlung background. This is typically known as

background subtraction and there are a number of well-developed methods to do this (Williams

& Carter, 2009c). The most basic of background subtraction methods is the simple window

method. Here, the intensity of the background in two windows, one above and one below the

characteristic peak, is integrated and subtracted from the intensity integrated over a window of

the same width on the characteristic peak. This very simple technique is very effective for well

separated peaks on a slowly varying background, typical of the high-energy regime, but it is

often not possible to perform this type of background subtraction in the low-energy regime.

A slightly more advanced method is to model the Bremsstrahlung background using Kramers’

law (equation 2.5), so that it is possible to subtract the modelled counts in the peak regions.

Kramers’ law can be fit to background regions of spectra by varying the constant K to match the

intensity and the modelled intensity can simply be subtracted from the regions underneath the

peaks.

An entirely different approach to background subtraction comes in the form of filtering out the

smooth background. A top-hat filter can be convoluted with the EDX spectrum and produces a

function that is similar to a second differential of the spectrum. Thus, the slowly varying portions

of the spectrum are close to zero and the peaks remain. However, this form of filtering does not

produce a physically meaningful background subtracted spectrum, with regions of negative

36
intensity present at the base of peaks, and careful comparison with element standards is

needed to extract peak intensities. Nevertheless, this method is still used in some commercial

software packages, such as those of Oxford Instruments.

2.3.3.2. Absorption Correction

In the quantification of EDX spectra from nanoparticles the absorption of X-ray intensity through

the sample is often neglected. This is often a reliable assumption, although not always,

particularly when considering low energy X-rays. The absorption of X-rays in matter follows the

exponential attenuation law put forward by Beer and Lambert for light, where X-ray intensity is

given by

𝜇
−𝜌 ]𝑙 (2.10)
𝐼 = 𝐼0 𝑒 𝜌 ,

𝜇
where I0 is the intensity of X-rays before absorption, ρ is the density of the sample, ] is the
𝜌

mass attenuation coefficient of the sample for X-rays of the energy in question and l is the path

length through the material. The values of both density and mass attenuation coefficient are

strongly dependent on the atomic number of the material so that higher atomic number

materials typically absorb X-rays more strongly, although the exact dependence of the mass

attenuation coefficient on Z is quite complicated due to absorption edges. The mass attenuation

coefficient is also highly dependent on X-ray energy; lower energy X-rays are absorbed more

strongly than higher energy X-rays.

For an analysed point the absorption of X-rays must be considered for all X-ray paths passing

from all points on the electron beam to the detector. Assuming the detector is a point detector,

consideration of all X-ray paths is performed by integrating all X-ray paths along the electron

beam. This is straightforward in thin film samples where the X-ray path length is given by

𝑙 = 𝑧 csc(𝛼) (Philibert, 1963), where z is the depth of X-ray production in the sample and α is the

take-off angle. This leads to an expression for X-ray absorption of a single energy of X-ray from

a thin film of

𝜇
𝜇 −( ]𝜌𝑡 csc(𝛼)) (2.11)
𝐴𝐶𝐹 = ] (1 − 𝑒 𝜌 ).
𝜌

The absorption correction factor (ACF) can be combined with the zeta-factor in the form above

(equation 2.11) or two factors can be combined in ratio with k-factors. Calculating the X-ray

37
absorption from free standing nanoparticles is certainly not so straightforward and so far has not

been examined in the STEM. However, X-ray beam paths have been determined for a variety of

particle morphologies for electron probe microanalysis (EPMA) (Armstrong & Buseck, 1975;

Zreiba & Kelly, 1988) and it seems reasonable that these may be applied in a similar manner to

STEM studies. The application of absorption correction to spherical particles is the subject of

chapter 3.

2.3.4. Nanoparticle Studies

The improved collection efficiency of the latest EDX detector systems has generated an

increasing number of studies in which EDX spectroscopy is applied to elemental mapping of

single nanoparticles. For example, EDX spectrum imaging has revealed the structure of AuPd

nanoparticles synthesized by microbes (Tran, et al., 2012) and after calcination (Wilson, et al.,

2013). EDXS elemental mapping of catalysts after electrochemical aging has also been applied

to reveal Pt surface segregation in Pt/PdAu nanoparticles (Sasaki, et al., 2012) and the

elemental distribution in AgAu nanoparticles (Rodriguez-Gonzalez, et al., 2005).

Elemental mapping can also provide key information to aid understanding of the long term

stabilities or reactivities of nanoparticle systems with respect to oxidation. Alloyed NiCr

nanoparticles have been observed by EDX spectroscopy to form a Cr 2O3 layer, a result which is

driven by the greater ability of Cr to oxidise compared to Ni (Wang, et al., 2014). Similarly, in

core-shell CuSe-CuS, the CuSe component has been found to preferentially oxidise, with a

similar result found in the inverse core-shell CuS-CuSe structure (Miszta, et al., 2014).

Elemental redistributions during in situ heating and beam induced oxidation of AgAu

nanocrystals have been mapped with in situ EDX spectrum imaging, demonstrating alloying

during heating and the formation of hollow Au-core Ag2O-shell nanostructures under electron

beam irradiation (Lewis, et al., 2014a).

2.4. Spatial Resolution in the STEM

The spatial resolution of techniques in the STEM is not an entirely straightforward topic. The

resolution of coherent imaging modes is particularly complex, with the angles of both the probe

forming aperture and the detector playing an important role in defining the resolution limits

(Nellist, 2011). The situation is somewhat less complex for incoherent modes, and in particular

38
inelastic scattering, where the resolution of the technique is simply related to the size of the

electron beam within the sample. For this type of inelastic scattering it is pertinent to split the

topic into two parts, factors influencing the size of the electron beam before the sample and

those affecting the size of the beam within the sample.

The size of the electron probe is generally known to be influenced by three factors (Craven,

2011; Watanabe, 2011). These are:

2 𝐼
 The diameter of the probe due to the finite source size, 𝑑𝑠 = √ , where α is the
𝜋𝛼 𝐵

probe-forming semi-angle, I is the probe current and B is the source brightness.

 The diameter of the first Airy disk due to diffraction from the probe-forming aperture,
1.22𝜆
𝑑𝑑 = , where λ is the wavelength of the probe electrons.
𝛼

 The diameter of the disk of least confusion caused by lens aberrations, where each

spherical aberration of the nth order (Cn,0) contributes as 𝑑𝑔 = 𝐴𝑛 𝐶𝑛,0 𝛼 𝑛 .

1
The first two terms are dependent on , with the prefactor determining which term will be the
𝛼

more significant. For thermionic sources, the low brightness of the source leads to a large d s

term which dominates all others and has meant that recent advances in the STEM have been

coupled to the availability of the field emission gun (FEG) source. For cold FEGs in particular,

this term tends to be outweighed by the diffraction term, dd, although for a Schottky FEG when

using beam currents associated with analytical work (100s of pA) it is often the ds term that

dominates at low values of α. The finite source size and diffraction terms diminish with

increasing α, so that for standard imaging conditions spherical aberrations of the third order

typically dominate at a value of approximately 10 mrad. A third order spherical aberration of

approximately 1 mm, typical of uncorrected instruments leads to a disk of least confusion of

diameter ds = 5 Å at a probe forming semi-angle of 10 mrad. Any increase of the probe-forming

angle will result in poorer resolution. This description is largely simplified from discussions of the

electron probe diameter, which require a full wave optical treatment for accurate probe size

determination (Colliex & Mory, 1984; Watanabe, 2011).

Aberration correction of electron optical systems has produced rapid improvements in STEM

imaging by reducing the size of the electron probe via the addition of non-spherical electron

39
lenses (Bleloch & Ramasse, 2011; Haider, et al., 1998; Krivanek, et al., 1999). With the advent

of electron optical aberration correction in the STEM the third order spherical aberration is

essentially reduced to zero. Thus, the probe forming angle can be increased until the limits of
th
aberrations of higher order, which are typically 5 order spherical aberrations or chromatic
th
aberrations, depending on the specifics of the microscope. At this point, either 5 order

spherical aberration correction or monochromation of the probe electrons is needed to improve

the probe size further (Krivanek, et al., 2013). Aberration correction is particularly useful for

analytical electron microscopy studies, as it allows an increased probe current inside a smaller

probe size.

Through aberration correction the electron probe size can be reduced to sub-Ångstrom size at

the surface of the sample. However, as the electron beam traverses the specimen it does not

stay at the same size. Firstly, the convergent nature of the beam ensures that the beam

intensity spreads after the focal point even if there is no sample present, although the

convergence angles associated with the STEM tend to be so small (10s of mrads) that this

effect is relatively small. Assuming a 100 nm sample with the focal point at the midpoint of the

sample thickness, a 20 mrad semi-angle will result in a beam spread of 1 nm.

In a sample at most orientations, interactions of the electron beam with the sample atoms will

result in the scattering of the beam that leads to beam broadening as the beam traverses the

sample. This beam broadening is dependent on the density and thickness of the sample and

can be estimated using the single-scattering equation of Goldstein (Goldstein, et al., 1977)

modified by Reed (Reed, 1982),

1
𝑍̅ 𝜌 2 3 (2.12)
𝑏 = 7.21 × 105 ( ̅) 𝑡 2 ,
𝐸0 𝐴

where 𝑍̅ is the average atomic number of the sample, E0 is the energy of the probe electrons, ρ

is the sample density, 𝐴̅ is the average atomic weight of the sample and t is the sample

thickness. This beam broadening term takes in to account 90% of the beam intensity and for

100 nm thick Ag and Au samples, and a beam energy of 200 kV, would result in broadening to

17 nm and 29 nm respectively. The beam broadening is largely independent of initial beam

diameter, such that in thick samples the initial beam diameter has little impact on the overall

resolution. Whilst use of this type of equation is commonplace in considerations of resolution in

40
EDX spectroscopy, electron scattering is a largely stochastic process and therefore cannot be

accurately described using fixed terms. Many treatments of beam broadening involve use of

Monte Carlo codes that generate random scattering processes to estimate beam broadening

within a sample (Drouin, et al., 2007).

However, when the electron beam direction is oriented parallel to a major crystallographic zone

axis of a material the major consideration is no longer beam broadening but is instead a

phenomenon known as electron channelling (Fertig & Rose, 1981; Pennycook & Jesson, 1990).

Here, the line of positive potentials along the beam direction confines the electron beam to the

atomic columns, with the intensity distribution “beating” down the column in the “Pendelossung”

manner. Electron channelling is most often associated with atomic resolution imaging in the

STEM, as it a key component to inhibiting electron beam broadening that would remove any

atomic resolution features. However, electron channelling is also an important consideration for

STEM imaging in which pixel sizes are above that at which atomic columns could be resolved,

as it alters the effective interaction parameter of electron scattering processes. That is, electron

channelling will act to increase the number of electron interactions within the sample over a

given distance. This can be seen as an increase in the intensity of electron scattering (Tafto &

Lehmpfuhl, 1982) and characteristic X-ray production (Duncumb, 1962; Liao & Marks, 2013)

along certain atomic columns in nanoparticles oriented along a major zone axis. This can be

seen from EDX spectra collected from an AgAu nanocube tilted about its <100> zone axis,

where the intensity of the Au Mα peak increases by 50% on the zone axis compared to a 5°

misorientation (Fig 2.7).

41
Figure 2.7. Plot of normalised Au Mα peak intensity from a 50 nm size AgAu nanocube as a

function of tilt angle away from the <100> zone-axis. Sample courtesy of Pedro Camargo

(University of Sao Paulo).

2.5. Electron Beam Damage

Whilst the strong interaction of the electron beam with the sample is of great use for detecting

signals from very small sample volumes, it can also lead to relatively high levels of damage

(Egerton, et al., 2004). If sufficient damage occurs to provide structural changes to the

nanoparticle, then the nanoparticles may no longer be representative of the initial state. It is

therefore important to find methods to mitigate against beam damage, firstly by considering the

processes under which beam damage can occur. Electron beam damage in inorganic

nanoparticle samples becomes a particular problem when using techniques that require a high

dose, or dose rate, such as EDX spectroscopy and electron tomography.

Beam damage in inorganic samples can occur via both the elastic and inelastic scattering

processes outlined in section 2.1. Rutherford scattering has up until this point been considered

as elastic scattering, as at the small scattering angles considered in electron imaging there is

immeasurable energy transfer. However, this is not the case for large angle scattering which, for

a 200 keV electron beam scattering from an Al atom, can transfer energy of up to 20 eV

(Egerton, et al., 2010). Rutherford scattering gives rise to “knock-on” damage, where

backscattering of an electron can result in a large enough transfer of energy to remove an atom

42
from its position in the atomic lattice. This occurs for incident electron energies up to a threshold

energy, which is the energy required to break atomic bonds at specific lattice sites.

Knock-on damage is element dependent, in that a greater amount of energy can be transferred

to lighter atoms. For the vast majority of elements the energy transferred through high-angle

elastic scattering at CTEM/STEM energies is not sufficient to displace samples in the bulk

material. Knock-on damage in inorganic thin films is known to be primarily through sputtering

from the exit surface of the electrons, where the energy required to break surface bonds is

lower than that in the bulk (Egerton, et al., 2010). The lowest knock-on energies are found at

defect or kink sites, where it is expected that sputtering rates are greatest (Egerton, et al., 2010).

Considering that nanoparticles possess a large number of these sites, it is anticipated that

sputtering from nanoparticles occur at lower energies than that of defect free surfaces, as has

been seen for the sputtering of Ag from AgAu nanoparticles (Braidy, et al., 2008). The addition

or removal of atoms from the surface of nanoparticles is known to cause “quasi-melting”

(Krakow, et al., 1994; Marks, 1994), in which the crystallinity of nanoparticles can be temporarily

lost. In order to reduce the effect of knock-on damage it is possible to reduce the energy of the

incident electrons below the threshold energy, potentially eliminating damage altogether, as is

seen in the study of carbon nanomaterials at low electron energies (Bell & Erdman, 2013;

Suenaga & Koshino, 2010). It is also possible to inhibit knock-on damage via sputtering by

encapsulating the sample in a thin layer of a conducting material such as carbon, so that it is

the carbon that is preferentially sputtered (Egerton, 2013; Muller & Silcox, 1995).

Decreasing the incident energy of the electron beam to a few 10s of keV may limit damage from

Rutherford scattering, however, for typical S/TEM beam energies this increases the cross

section of inelastic scattering, which peaks at approximately five times the critical ionisation

energy (Williams & Carter, 2009b). Damage due to the ionisation of atoms through inelastic

scattering is often termed radiolysis, and the precise damage mechanism after initial bond

breaking depends on the material in question (Egerton, et al., 2004). It has been seen, however,

that radiolysis does not proceed in conducting samples (Egerton, et al., 2006), which it is

suggested is due to the presence of conduction band electrons that are able to rapidly fill any

empty core-levels. In insulating materials where radiolysis is often the primary damage

mechanism, coating with a conductive layer has been shown to provide some protection to

43
damage (Egerton, 2013), although it is not known whether this is due to the supply of free

electrons or to forming a barrier to reactive species.

In non-conducting specimens, the prominent damage mechanisms are identified as electron

beam heating or charging. Charging due to the electron interactions, primarily inelastic

scattering, has been shown to cause significant damage in inorganic samples at high beam

currents (Cazaux, 1995). Similarly, inelastic collisions may cause deposition of thermal energy

in a sample. However, it is known that for thin samples at typical STEM accelerating voltages

that the small fraction of the energy deposited results in only minor temperature rises (Egerton,

et al., 2004). Reducing the dose rate, i.e. the beam current, has been shown to be an effective

method to limit electron beam induced charging (Cazaux, 1995) and should also reduce

possible damage due to heating.

For all damage mechanisms a reliable method of limiting degradation is to lower the total dose,

although this may preclude a sufficient signal to noise ratio at a low dose. Particularly in the

techniques used here, EDXS and its extension to electron tomography (reviewed in Chapter 4),

a high dose is currently a necessity, although every effort should be made to reduce the dose

as far as possible.

2.6 Summary

The scanning transmission electron microscope is a key tool to probing the elemental

distribution of elements in single nanoparticles, or small populations of nanoparticles. The

majority of imaging modes are able to distinguish regions of different elements, although BF and

HAADF modes do not provide the truly spectroscopic signal of EELS and EDXS. Advances in

EDX detector systems over the past decade have led to a vastly improved capability for

spectroscopy of nanoparticles, particularly in spectrum imaging. However, there are still

challenges to performing EDX spectrum imaging, particularly the low count rates typically

associated with the technique. To improve count rates, thicker samples or higher electron doses

are favourable, albeit with inherent associated drawbacks such as decreased resolution or

beam damage.

44
 3. X-ray Absorption Correction in Spherical Nanoparticles

3.1. Introduction

This chapter details a method for the determination and application of X-ray absorption

correction to spherical nanoparticles. There are many studies which describe quantitative EDX

analysis of nanoparticles in the STEM, all of which assume that the particles are small enough

that X-ray absorption may be ignored. This is a good assumption for many nanoparticles but

should be carefully considered when analysing larger nanoparticles, particularly when using low

energy X-ray peaks.

The presented manuscript deals specifically with X-ray absorption in spherical nanoparticles.

There are many examples of spherical or near-spherical nanoparticles in the literature and the

specific motivation for this work came from the analysis of spherical nanoprecipitates extracted

from Ni-based superalloys (Chen, et al., 2014). The correction factors here for spherical

nanoparticles could be used to approximate X-ray absorption for a number of concave

nanoparticle morphologies, although care should be taken in the quoted precision of the

composition in these cases.

Derivation of X-ray path lengths could be extended to a number of nanoparticle shapes but X-

ray absorption correction in the traditional sense is only possible for nanoparticles with well-

defined morphologies. If the nanoparticles have a complex morphology, or if the morphology of

the nanoparticle is not discernible from a two dimensional STEM image, then a full

determination of X-ray path length could be obtained from HAADF electron tomography.

Characterisation of morphology would only be sufficient to provide fully quantitative absorption

correction if the nanoparticle is known to have a homogeneous composition. For nanoparticles

with compositional variations, a potential method is to use EDX tomography in an iterative

fashion to determine the composition at each voxel. Chapter 5 attempts to provide a platform for

this type of quantitative work through thorough investigation of the methodology of STEM-EDX

tomography, although full quantification on a voxel by voxel basis is beyond the scope of this

thesis.

45
A portion of the experimental data was collected by Yiqiang Chen (approximately half of that

collected on the Tecnai), who also assisted in the derivation of the X-ray path length. The Au

films were deposited by Greg Auton.

The results of this project have been submitted to Microscopy and Microanalysis. The original

manuscript for “X-Ray Absorption Correction for Quantitative STEM-EDX of Spherical

Nanoparticles” is presented below.

46
Paper: X-Ray Absorption Correction for Quantitative STEM-EDX of

Spherical Nanoparticles
1 1 2 1,3 1
Thomas Slater , Yiqiang Chen , Gregory Auton , Nestor Zaluzec , Sarah Haigh

1. School of Materials, University of Manchester, Manchester, M13 9PL, UK

2. School of Computer Science, University of Manchester, Manchester, M13 9PL, UK

3. Nanoscience and Technology Division, Argonne National Laboratory, Argonne, Illinois

60439, US

Abstract

A new method to perform X-ray absorption correction for spherical particles in quantitative

energy dispersive X-ray spectroscopy in the scanning transmission electron microscope is

presented. An absorption correction factor is derived and simulated data is presented

encompassing a range of X-ray absorption conditions. Theoretical calculations are compared to

experimental data of X-ray counts from Au nanoparticles to verify the derived methodology. The

effect of detector elevation angle is considered and a comparison with thin film absorption

correction is included.

Key words: EDX, XEDS, STEM, nanoparticles, quantification, absorption correction

Introduction

The physical and chemical properties of nanoparticles are known to be sensitive to their precise

elemental composition. Studies of individual bimetallic nanoparticles have observed that

composition has a large effect on their localised surface plasmon resonance wavelength

(Hostetler, et al., 1998; Kariuki, et al., 2004) and the stoichiometry of iron oxide nanoparticles

plays a vital role in determining their magnetic properties (Carvalho, et al., 2013; Salazar, et al.,

2011). Properties can also depend on the elemental segregation within a particular nanoparticle.

For example, the segregation of certain elements to the surface or to particular facets of

nanoparticles has been shown to greatly affect their catalytic performance (Cui, et al., 2012;

Slater, et al., 2014b). The average composition of large ensembles of nanoparticles can be

accurately obtained through time of flight mass spectrometry or flame atomic absorbance

spectroscopy and average surface composition through X-ray photoelectron spectroscopy (Baer

47
& Engelhard, 2010; Baer, et al., 2010). However, nanoparticle samples are often polydisperse in

composition, such that elemental analysis of individual nanoparticles provides important

information, complementary to ensemble techniques.

Energy dispersive X-ray (EDX) spectroscopy within the scanning transmission electron

microscope (STEM) provides a robust method to quantitatively analyse the elemental

composition of a wide range of materials at nanometre length-scales. Quantitative EDX

spectroscopy in the STEM has traditionally been based on the use of Cliff-Lorimer factors (k-

factors) (Cliff & Lorimer, 1975) that relate the relative intensity of X-ray peaks for two elements

(Ij) to their relative composition (cj) via the following equation,

cA I (1)
=kAB A ,
cB IB

Cliff-Lorimer factors, kAB, can be measured experimentally by preparing a set of suitable

standard samples (Sheridan, 1989; Wood, et al., 1984), or can be theoretically calculated

(Goldstein, et al., 1977; Zaluzec, 1979), and can provide accurate quantitative information of

thin samples. The relationship in equation 1 was derived assuming that the specimen region of

interest is sufficiently thin so that the combined effects of electron energy loss, x-ray absorption

and x-ray fluoresence can be ignored. While the effects of energy loss on the ionization cross-

section are negligible for typical TEM specimens, the same is not generally true for x-ray

absorption effects, particularly for low energy x-rays in high atomic number materials. In order

to correct for the effect of X-ray absorption on relative peak intensities, an absorption correction

factor (ACF) was introduced which attempts to correct for the differences in the X-ray absorption

of the two X-ray peaks analysed. This absorption correction factor is given by (Philibert, 1963;

Zaluzec, 1979; Zaluzec, 1981)

μB (2)
- ] ρl(z)
∫e ρ dz
ACF= μA
,
- ] ρl(z)
∫e ρ dz

where l(z) is the X-ray path length through the sample to the detector as a function of the
𝜇
distance, z, along the optic axis, ρ is the density of the sample, is the mass attenuation
𝜌

coefficient of a particular X-ray peak within the sample and the integral is performed over all

possible X-ray paths to the detector. The ACF term can be simply multiplied by the initial Cliff-

48
Lorimer factor to provide a modified Cliff-Lorimer factor, that takes absorption into account (Qiu,

et al., 2013, Wather & Wang, 2015),

k*AB =kAB ×ACF (3)

,
The modified Cliff-Lorimer factor, k*AB, is therefore a function of X-ray path lengths to the

detector through the sample and thus varies with the experimental configuration. Historically,

thin film analysis has dominated EDX spectroscopy due to the prevalence of polished metal and

semiconductor foil samples and the simple geometry of this system. For a thin film, the x-ray

path length to the detector for an x-ray generated at a depth z in the specimen is determined by

sin α
the specimen-detector geometry. This path length (Fig 1a) is given by l(z)=z , where α is
cos(α-θE )

the beam incidence angle, i.e. the angle between the electron beam and the surface of the

sample, in the vertical plane subtending the centre of the detector, θE is the elevation angle of

the detector, defined as the angle between the horizontal plane subtending the specimen and

the point at the centre of the detector face and t is the sample thickness (Zaluzec, 1981).

Inserting this expression into equation 2 and assuming a point detector allows closed form

evaluation of the integral in the ACF (equation 4), giving the result expressed in equation 5,

μB sin α (4)
(-ρ] ρz )
t cos(α-θE ) dz
∫0 e
ACF = μA sin α
,
(-ρ] ρz )
t cos(α-θE ) dz
∫0 e

μB sin α (5)
μ A (-ρ] ρt )
] 1-e cos(α-θE )
ρ
ACF = μ B
( μA sin α
).
] (-ρ] ρt )
ρ 1-e cos(α-θE )

Figure 1. Schematics illustrating the geometry of X-ray detection from a) a thin film sample of
𝑡
thickness t and b) a spherical particle of radius R= .
2

49
This absorption correction equation has been widely used for many years in thin film analysis

and is incorporated into commercial S/TEM-EDX analysis software packages.

X-ray Absorption in Spherical Nanoparticles

Spherical nanoparticles can often have dimensions and compositions such that X-ray

absorption cannot be ignored during quantitative EDX analysis. Whilst the X-ray path length in a

uniform thin film is constant for any position of the beam, for a spherical geometry this is not the

case, as illustrated in Fig 1. For this reason, the path length in a sphere must be defined using a

three-coordinate system. X-ray path lengths for particles of a number of different geometries

have been previously calculated considering detector geometries for both EDX spectroscopy

and wavelength dispersive X-ray spectroscopy performed in an electron probe microanalyser

(EPMA) (Armstrong & Buseck, 1975; Zreiba & Kelly, 1988). Here, we present a derivation of X-

ray path length in a sphere similar to that of Zreiba and Kelly, using a cartesian coordinate

system in which z is the direction of the electron beam, x is the direction towards the X-ray

detector in the plane perpendicular to the path of the electron beam and y is the direction

perpendicular to the plane in which the X-ray path sits (Fig 2). Unlike for the thin film geometry

the specimen thickness t in a spherical particle depends strongly on the position (x,y).

50
Figure 2. Schematics illustrating the geometry of an X-ray path through a spherical particle of

radius R at spatial coordinates x and y, where x is the direction towards the X-ray detector in the

horizontal plane and y is the direction perpendicular to the plane in which the X-ray path sits. a)

schematic subtending the xy plane, i.e. viewed along the optic axis, b) schematic subtending

the xz plane, i.e. within the plane in which the X-ray travels from the sample to the detector.

Based on a figure from (Zreiba and Kelly 1988).

An X-ray is generated at point P with coordinates (x,y,z). Variations in the value of y, the

distance away from the centre of the particle perpendicular to the plane in which the X-ray path

sits, act only to change the effective radius of a circular slice of the sphere in which the X-ray

travels, so that the modified radius, Rm, is simply calculated as a chord a distance y from the

centre of a circle of radius R,

(6)
Rm =√R2 -y2 .

The X-ray path through the spherical particle, l(z), is then calculated similarly as a chord in a

circle of radius Rm at a distance of |OQ| from the circle centre, minus the distance |PQ|,

(7)
l(z) =√Rm 2 -|OQ|2 − |PQ|.

The distances |OQ| and |PQ| are found through simple trigonometry to be,

t (8)
|OQ| = (z- +x tan θE ) cos θE
2
t (9)
|PQ|= (x-(z- ) tan θE ) cos θE .
2

The equation for X-ray path length is then,

51
 2
(10)
𝑡 t
l(z) =√Rm 2 - (𝑥 sin 𝜃𝐸 − ( − 𝑧) cos θE ) -x cos θE - ( -z) sin θE ,
2 2

where the thickness through the sphere at coordinates (x, y) is given by

t=2√Rm 2 -x2 =2√R2 -y2 -x2 . Substituting our result from equation 10 into equation 2 we arrive at an

equation for the absorption correction factor in a spherical particle of

2 (11)
μB t t
-ρ] ρ(√Rm 2 -((z- +x tan θE ) cos θE ) -x cos θE -( -z) sin θE )
2 2
t
∫0 e dz
ACF = 2
.
μA √ t t
-ρ] ρ( Rm 2 -((z- +x tan θE ) cos θE ) -x cos θE -( -z) sin θE )
2 2
t
∫0 e dz

Equation 11 is not analytically solvable and therefore must be solved numerically for each

individual case. We note that the final form of the ACF has no dependence on the tilt of the

sample, due to the rotational symmetry of a sphere, although changing the tilt angles may act to

partially shadow the detector and therefore change the effective elevation angle of the detector.

In the special case of x = y = 0, i.e. when the beam is positioned on the centre of the particle,

t
Rm = R = and the absorption correction factor becomes
2

μB 2 (12)
-ρ] ρ(√R2 -((z-R) cos θE ) -(R-z) sin θE )
2R
∫0 e dz
ACF = .
μA 2
-ρ] ρ(√R2 -((z-R) cos θE ) -(R-z) sin θE )
2R
∫0 e dz

For the case in which the detector is positioned at zero elevation angle, 𝜃𝐸 = 0, the ACF at x = y

= 0 is simplified to

μB (13)
2R -ρ] ρ√z(2R-z)
∫0 e dz
ACF= μA
.
2R -ρ] ρ√z(2R-z)
∫0 e dz

If more than one detector is used, then the numerator and denominator in the ACF simply

become a sum of the X-ray absorption terms for each detector using a coordinate system

(xi,yi,z) associated with each individual detector, di,

2 (14
μB t t
-ρ] ρ(√R2 -yi 2 -((z- +xi tan θE ) cos θE ) -xi cos θE -( -z) sin θE )
2 2
t
∑d ∫0 e dz
i
ACF= 2
. )
μA t t
-ρ] ρ(√R2 -yi 2 -((z- +xi tan θE ) cos θE ) -xi cos θE -( -z) sin θE )
2 2
t
∑d ∫0 e dz
i

52
The X-ray absorption due to the derived path lengths is examined for three specific cases here,

to show the effect on nanoparticles with largely different ratios in mass attenuation coefficient.

For the case of the absorption of Lα and Mα X-rays in Au nanoparticles, experimental data is

acquired in order to verify the distribution and extent of X-ray absorption in this particular case.

Au nanoparticles were chosen as they have a known fixed composition, in comparison to multi-

element nanoparticles, and they have X-ray peaks at energies that provide appreciable

absorption correction (2.1 keV for Au Mα and 9.7 keV for Au Lα).

Experimental Methods

Theoretical calculations were performed using Python, with numerical integration performed

using the quad function from the SciPy package (Jones, et al., 2001). Experimental net count

maps were rotated by 45° using the rotate function in the SciPy package. Mass-attenuation

coefficients were taken from (Henke, et al., 1993) where available or through linear interpolation

when the value was not listed and are presented in the results section. Densities for each

system were taken as a weighted average of the densities of constituent elements and the
-3 -3 -3
values used were ρAu = 19.32 gcm , ρAuPt = 20.39 gcm and ρNiO = 6.67 gcm .

A range of sizes of surfactant-stabilized Au nanoparticles were purchased from Sigma Aldrich

(10 nm, 30 nm, 50 nm, 80 nm, 100 nm, 200 nm and 400 nm radius). Nanoparticles were

deposited dropwise on to holey carbon coated copper TEM grids and subsequently washed

dropwise with methanol. Au films were deposited on to 50 nm thick SiN membranes (Agar

S171-2T) using a Moorfield MiniLab evaporator. Au films of thickness 23 nm, 27 nm, 41 nm, 64

nm and 82 nm were deposited after first depositing a 3 nm thick Cr adhesion layer.

EDX spectrum, single spectra, HAADF images and low-loss EELS spectrum images of Au

nanoparticles and Au thin films were acquired on a probe corrected Titan G2 80-200 (S)TEM

microscope operated at 200 kV with a beam current of 600 pA. EDX spectrum images of size

256x256 pixels were acquired for a single 200 nm diameter particle for approximately 30 mins

using multiple passes with a 100μs per pixel dwell time. Single spectra were acquired for 10 -

60 s for 10 nanoparticles of each nanoparticle size, with the electron beam positioned in the

centre of each nanoparticle and drift correction applied. EDX spectra of Au nanoparticles were

also acquired on a Tecnai TF30 microscope operating at 300kV with a beam current of 500pA.

53
HAADF STEM images for each particle were acquired before spectrum acquisition and used to

measure the nanoparticle radius. Au thin film thickness was measured from low-loss EELS

spectrum images using a t/λ method in Digital Micrograph and the measurement was averaged

over a 2 μm x 2μm area for each film. The contribution from the SiN support membrane was

measured independently and its contribution subtracted to obtain the net t/λ value for the Au film.

The t/λ ratio was calibrated using measurements from low-loss spectrum images of Au

nanoparticles where the spherical shape allowed the thickness to be accurately known.

Analysis of EDX spectra and spectrum images was conducted using Bruker Esprit software

(version 1.9.4.3448). Net counts were extracted using the physical TEM model background with

constant fit ranges. Net count maps were acquired using the same settings but with a 2x binning.

All errors in counts and sample sizes are the standard deviation of the measurements taken in

each case.

Detector elevation angles were found by acquiring EDX maps of spherical nanoparticles over a

range of sample tilt values and selecting the angle at which there was a minimum in

characteristic X-ray counts (Slater, et al., 2014a).

Results and Discussion

The proposed equations for X-ray absorption correction in spherical particles were used to

model the variation in the ratio of counts of different nanoparticle systems. For simulation of the

absorption behaviour, three different test cases are presented: AuPt, pure Au and NiO. The test

cases have been chosen as examples for which the difference in mass attenuation coefficient
𝜇 2 -1 𝜇 2 -1
can be characterised as small (Au Lα ( = 125 cm g ) vs Pt Lα ( = 137 cm g )), medium (Au
𝜌 𝜌

𝜇 2 -1 𝜇 2 -1 𝜇 2 -1 𝜇
Lα ( = 128 cm g ) vs Au Mα ( = 1020 cm g )) and large (Ni Kα ( = 48 cm g ) vs O Kα ( =
𝜌 𝜌 𝜌 𝜌

2 -1
4650 cm g )). A value of normalised count ratio is used to represent the extent of X-ray

absorption throughout, where a value of 1 for the count ratio corresponds to no difference in

absorption between the two X-ray peaks.

Simulations of the ratio in counts across a 200 nm spherical particle for each of the three

systems show the same characteristic distribution of counts but at different intensity scales (Fig

3). The values displayed in Fig 3 are calculated with a detector elevation angle of 18° in all

54
cases. Simulation of the ratio of counts with one detector present reveals the largest variation in

counts across a particle (Fig 3(a-c)i), with the absorption of the lower energy X-ray at its

greatest furthest from the detector. A single detector setup such as this is employed in the

majority of S/TEM systems. While the difference in the ratio of counts between Au Lα and Pt Lα

reaches a low of only 0.995 (Fig 3ai), the difference in the ratio of counts between Ni Kα and O

Kα has a minimum value of 0.543 (Fig 3bi). This ratio equates to the fact that in a 200 nm

diameter particle of NiO, quantification without the use of absorption correction in this case

would lead to identification of the composition of one side of the nanoparticle as close to Ni 2O.

Figure 3. Simulated maps of normalised X-ray counts of a) Au Lα and Pt Lα X-rays in an AuPt

particle, b) Au Lα and Au Mα X-rays in an Au particle and c) Ni Kα and O Kα X-rays in a NiO

particle, where all particles are 200 nm diameter spheres. Maps are shown for i) a single

detector, ii) two detectors at 180° to each other azimuthally and iii) four detectors at an

azimuthal interval of 90° around the optic axis. The detector configuration is shown

schematically in (d) for each detector setup (i-iii). A detector elevation angle of 18° was used for

each simulated map.

When two detectors on opposite sides of the particle are employed, a detector geometry

previously employed by Wu et al (Wu, et al., 2010), the maximum variation is from points at the

surface closest to the detector to points at the surface at 90° to the X-ray path (Fig 3(a-c)ii). The

55
difference in the absorption of Au Lα and Pt Lα is very low (a minimum of the ratio in counts of

0.998) in this case (Fig 3aii). In the case of NiO the ratio in counts reaches a minimum of 0.771

at points on the nanoparticle closest to each detector (Fig 3bii).

Finally, a simulation was performed for four detectors that are at 90° to one another azimuthally

around the optic axis, as is the case in the Super-X detector (von Harrach, et al., 2009) (Fig 3(a-

c)iii). Here, the difference in the ratio of counts is smaller than that for a one or two detector

setup. In the most severe absorption case considered here, that of NiO, the normalised count

ratio varies from 0.788 in the centre of the particle to 0.869 at the points closest to each detector.

This demonstrates that care should be taken when analysing EDX maps of large nanoparticles

when using a four detector setup that variations in X-ray absorption are not misattributed to a

rise in the percentage composition of the element with the lower energy X-ray at the edge of the

particle. Although, we note that the errors are much smaller, even for the most extreme NiO

case presented here, lack of absorption correction would only lead to a rise in the apparent

atomic percentage of O in NiO from 44 % in the centre of the particle to 46 % at the edge.

These values are still somewhat different from the 50 % true value if no absorption occurred.

In order to experimentally verify the proposed models, the simulated maps of the Au Mα vs Au

Lα count ratios are compared to experimentally acquired equivalent data from a ~ 200 nm

diameter approximately spherical Au nanoparticle. The experimental values presented here

have been normalized by fitting a linear equation to the data points and dividing all of the count

ratios by the value found for x = 0 for each microscope. A Titan microscope with a Super-X EDX

detector was used, in which four detectors are positioned at 90° azimuthal intervals around the

optic axis, rotated 45 degrees with respect to the primary tilt axis of the specimen holder. This

also allows us to investigate the effect of different detector configurations on X-ray absorption

within a nanoparticle. The agreement between experimental and simulated data is excellent (Fig

4b). Each detector configuration shows the same qualitative variation across a particle as the

simulated data (Fig 4a) and, in addition, the values of the normalised count ratio in the

experimental data appear qualitatively similar to the simulated data.

56
Figure 4. Maps of the normalized count ratio between Au Mα and Au Lα X-ray counts acquired

from an approximately 200 nm diameter Au particle showing a) simulated absorption maps and

b) experimental maps using i) one detector, ii) two detectors facing each other and iii) four

detectors at 90° to each other. c) Schematics showing the position of detectors either on (black)

or off (white).

In addition to investigating how the X-ray absorption in spherical particles varies across a single

particle, the dependence on particle size has also been investigated. The theoretical models for

AuPt, Au and NiO, considering the same three pairs of X-ray peaks are plotted in Fig 5 for both

spherical particles and parallel sided thin films. While the difference between the particle and

thin film cases is small for the Pt Lα vs Au Lα case, 1 % difference even at a thickness of 400

nm, the difference between the particle and thin film absorption is large when considering Ni Kα

vs O Kα X-rays in NiO, above a 10 % difference at thicknesses of only 10 nm. Experimental

verification was again provided through the use of pure Au samples. The count ratios obtained

from point spectra at a central beam position on a number of spherical Au nanoparticles were

acquired for a range of sizes of nanoparticle (Fig 5), with all four detectors of the Super-X

detector. The normalized count ratios measured at the centre of the nanoparticles (black points)

57
appear to qualitatively match the theoretical relationship defined in equation 12 (black line). The

error bars represent the standard deviation of values for each point. Point spectra were

acquired at an arbitrary position on untilted thin films of a range of thicknesses and the

normalized ratio between Au Mα and Lα X-ray counts is represented by red points in Fig 5. The

experimental data points appear to qualitatively provide a good fit to the theoretical calculation

(red line) and confirm the predicted difference in X-ray absorption between the particle and thin

film specimen geometries.

Figure 5. Comparison of X-ray absorption at the centre of spherical nanoparticles and that of

thin films. Experimental data points of the ratio between Au Mα and Lα X-ray counts are

compared to the theoretical values calculated.

The effect of detector elevation angle on the absorption correction behaviour has also been

investigated for spectra acquired at the centre of spherical nanoparticles (Fig 6). The change in

X-ray absorption at the centre of a nanoparticle due to a change in elevation angle is a relatively

small change over the range of detector elevation angles commonly used, represented here by

simulations at elevation angles of 4° and 18° to match experimentally acquired data. For the

extreme case of NiO particles, the difference in count ratio for the two elevation angles reaches

10 % at a particle diameter of 250 nm. Again experimental verification is achieved by comparing

the Au Mα/Lα count ratio obtained from point spectra collected from the centre of Au

nanoparticles, using two different detector geometries (four detectors mounted at approximately

18° elevation angle, Super-X detector on Titan, and one detector mounted at approximately 4°

58
elevation angle, Oxford XMAX 80T on Tecnai F30). The change in elevation angle has little

effect on the absorption of our test case of Au Mα/Lα X-rays, the theoretical absorption

correction in this case only reaching a 1% variation at a radius of 250 nm, which results in no

discernible difference in the experimentally acquired data presented here.

Figure 6. Comparison of X-ray absorption in spherical nanoparticles for X-ray detectors at

elevation angles of 4° (blue) and 18° (black). Experimental data points of the ratio between Au

Mα and Lα X-ray counts are compared to the theoretical values calculated.

These results, while illustrating the correction factors for a number of test cases, should not be

used to generalize the magnitude of absorption effects in spherical particles, as the degree of a

specific correction is a function of the composition and importantly, the relative mass absorption

coefficients for the x-ray lines analyzed in the particle. The formulae presented in the

introduction should be used to calculate the magnitude of absorption correction in each

individual case.

Summary

A method for performing absorption correction in the EDX analysis of spherical nanoparticles

has been presented and experimentally verified. In many cases nanoparticles are small enough

and X-rays of sufficiently high energy that X-ray absorption can be neglected when trying to

quantify the composition of a nanoparticle with EDX spectroscopy. However, when

characterising large nanoparticles (the difference in absorption is larger than 10% for radii

59
greater than 80 nm for the Au particles in this study) or when considering low energy X-ray

peaks (for example carbon or oxygen peaks) neglecting X-ray absorption effects has the

potential to introduce large errors. We also note that errors in absorption correction will be

particularly serious when analyzing nanoparticles containing both light and heavy elements

where low energy X-rays are generated in a high density material.

The absorption correction required for EDX spectroscopy of spherical nanoparticles is

significantly different to that required for thin film samples. The standard thin film analysis is

therefore not appropriate when performing quantitative analysis of nanoparticle composition and

we provide an alternative methodology.

This absorption correction should allow accurate quantification of the elemental composition of

homogeneous spherical or approximately spherical nanoparticles. For single or double-detector

systems this theory could also be applied to cylindrical nanorods with the correct orientation

(long axis perpendicular to the vertical plane subtending the sample and the X-ray detector). For

nanoparticles of complex morphology an accurate absorption correction could be applied

through the application of electron tomography to provide a full three dimensional shape,

although this process is relatively long and not entirely straightforward. The absorption

correction factor given here can be automatically applied with knowledge of the detector

geometry and the sample density. The method outlined here can also be incorporated in to the

ζ-factor method (Watanabe & Williams 2006), which would also provide a measure of the

sample density.

Acknowledgments

T.J.A. S and S.J.H and gratefully acknowledge support from the North-West Nanoscience

Doctoral Training Centre, EPSRC grant EP/G03737X/1, EPSRC grant EP/M010619 and the

Defence Threat Reduction Agency grant number HDTRA1-12-1-0013. N.J.Z. also

acknowledges support from the Electron Microscopy Center at the Center for Nanoscale

Materials of Argonne National Laboratory, a U.S. Department of Energy, Office of Science,

Office of Basic Energy Sciences User Facility under Contract No. DE-AC02-06CH11357, as well

as, a visiting appointment in the School of Materials at the University of Manchester. The

authors wish to acknowledge the support from HM Government (UK) for the provision of the

60
funds for the FEI Titan G2 80-200 S/TEM associated with research capability of the Nuclear

Advanced Manufacturing Research Centre. The data associated with the paper is openly

available from The University of Manchester eScholar Data Repository:

http://dx.doi.org/10.15127/1.269245. The computer code associated with this paper is available

from Github: http://dx.doi.org/10.5281/zenodo.21030.

References

ARMSTRONG, J.T. & BUSECK, P.R. (1975). Quantitative chemical analysis of individual

microparticles using the electron microprobe. Anal Chem 47, 2178-2192.

BAER, D.R. & EENGELHARD, M.H. (2010). XPS analysis of nanostructured materials and

biological surfaces. J Electron Spectrosc Relat Phenom 178, 415-432.

BAER, D.R., GASPAR, D.J., NACHIMUTHU, P., TECHANE, S.D. & CASTNER, D.G. (2010).

Application of surface chemical analysis tools for characterization of nanoparticles. Anal Bioanal

Chem 396, 983-1002.

CARVALHO, M.D., HENRIQUES, F., FERREIRA, L.P., GODINHO, M. & CRUZ, M.M. (2013).

Iron oxide nanoparticles: the Influence of synthesis method and size on composition and

magnetic properties. J Solid State Chem 201, 144-152.

CLIFF, G. & LORIMER, G.W. (1975). Quantitative-analysis of thin specimens. J Microsc (Oxford,

UK) 103, 203-207.

CUI, C., GAN, L., LI, H.-H., YU, S.-H., HEGGEN, M. & STRASSER, P. (2012). Octahedral PtNi

nanoparticle catalysts: exceptional oxygen reduction activity by tuning the alloy particle surface

composition. Nano Lett 12, 5885-5889.

GOLDSTEIN, J.I., COSTLEY, J.L., LORIMER, G.W. & REED, R.J.B. (1977). Quantitative X-ray

analysis in the electron microscope. In Scanning Electron Microscopy, JOHARI, O. (Ed.), pp.

315. Chicago, IL: IITRI.

HENKE, B.L., GULLIKSON, E.M. & DAVIS, J.C. (1993). X-ray interactions - photoabsorption,

scattering, transmission and reflection at E=50-30,000 eV, Z=1-92. At Data Nucl Data Tables 54,

181-342.

61
HOSTETLER, M.J., ZHONG, C.J., YEN, B.K.H., ANDEREGG, J., GROSS, S.M., EVANS, N.D.,

PORTER, M. & MURRAY, R.W. (1998). Stable, monolayer-protected metal alloy clusters. J Am

Chem Soc 120, 9396-9397.

JONES, E., OLIPHANT, T.O. & PETERSON, P.P. (2001). SciPy: Open Source Scientific Tools

for Python.

KARIUKI, N.N., LUO, J., MAYE, M.M., HASSAN, S.A., MENARD, T., NASLUND, H.R., LIN,

Y.H., WANG, C.M., ENGELHARD, M.H. & ZHONG, C.J. (2004). Composition-controlled

synthesis of bimetallic gold-silver nanoparticles. Langmuir 20, 11240-11246.

PHILIBERT, J. A method for calculating the absorption correction in electron-probe

microanalysis. In 3rd International Congress on X-ray Optics and Microanalysis, PATTEE, H. H.,

COSSLETT, V. E. and ENGSTROM, A. (Eds.), pp. 379. Academic.

QIU, Y., NGUYEN, V.H., DOBBIE, A., MYRONOV, M. & WALTHER, T. (2013). Calibration of

thickness-dependent k-factors for germanium X-ray lines to improve energy-dispersive X-ray

spectroscopy of SiGe layers in analytical transmission electron microscopy. J Phys Conf Ser

471, 012031.

SALAZAR, J.S., PEREZ, L., DE ABRIL, O., LAI TRUONG, P., IHIAWAKRIM, D., VAZQUEZ, M.,

GRENECHE, J.-M., BEGIN-COLIN, S. & POURROY, G. (2011). Magnetic iron oxide

nanoparticles in 10-40 nm range: composition in terms of magnetite/maghemite ratio and effect

on the magnetic properties. Chem Mater 23, 1379-1386.

SHERIDAN, P.J. (1989). Determination of experimental and theoretical KαSi factors for a 200-kV

analytical electron-microscope. J Electron Microsc Tech 11, 41-61.

SLATER, T.J.A., CAMARGO, P.H.C., BURKE, M.G., ZALUZEC, N.J. & HAIGH, S.J. (2014a).

Understanding the limitations of the Super-X energy dispersive X-ray spectrometer as a function

of specimen tilt angle for tomographic data acquisition in the S/TEM. J Phys: Conf Ser 522,

012025.

SLATER, T.J.A., MACEDO, A., SCHROEDER, S.L.M., BURKE, M.G., O’BRIEN, P., CAMARGO,

P.H.C. & HAIGH, S.J. (2014b). Correlating catalytic activity of Ag–Au nanoparticles with 3D

compositional variations. Nano Lett 14, 1921-1926.

62
VON HARRACH, H.S., DONA, P., FREITAG, B., SOLTAU, H., NICULAE, A. & ROHDE, M.

(2009). An integrated silicon drift detector system for FEI Schottky field emission transmission

electron microscopes. Microsc Microanal 15, 208-209.

WALTER, T. & WANG, X. (2015). Self-consistent method for quantifying indium content from X-

ray spectra of thick compound semiconductor specimens in a transmission electron microscope.

J Microsc (Oxford, UK), Early View.

WATANABE, M. & WILLIAMS, D.B. (2006). The quantitative analysis of thin specimens: a

review of progress from the Cliff-Lorimer to the new zeta-factor methods. J Microsc (Oxford, UK)

221, 89-109.

WOOD, J.E., WILLIAMS, D.B. & GOLDSTEIN, J.I. (1984). Experimental and theoretical

determination of kαFe factors for quantitative X-ray microanalysis in the analytical electron

microscope. J Microsc (Oxford, UK) 133, 255-274.

WU, J., KIM, A., MARVIN, R., MYERS, B., WOODRUFF, T., O'HALLORAN, T., DRAVID, V.,

MCILWRATH, K. & LI, S. (2010). Imaging and elemental mapping of biological specimens with

the hitachi HD-2300A dual-EDS scanning transmission electron microscope. Microsc Microanal

16(S2), 884-885.

ZALUZEC, N.J. (1979). Quantitative x-ray microanalysis: instrumental considerations and

applications to materials science. In Introduction to Analytical Electron Microscopy, HREN, J. J.,

GOLDSTEIN, J. I. and JOY, D. C. (Eds.), pp. 121-155. New York: Plenum Press.

ZALUZEC, N.J. (1981). On the geometry of the absorption correction in analytical electron

microscopy. Microbeam Anal (San Francisco), 167.

ZREIBA, N.A. & KELLY, T.F. (1988). Absorption and fluorescence corrections of characteristic

X-rays from thin spheres. X-Ray Spectrom 17, 229-238.

63
 4. Electron Tomography

4.1. Introduction to Tomography

Images collected in the STEM are two dimensional (2D) projections of a three dimensional (3D)

volume, with information only collected in directions perpendicular to the electron beam. When

determining the structure of objects in the STEM, often assumptions are made about the

distribution of information in the direction parallel to the electron beam, which may or may not

be accurate. In order to fully determine the structure of a 3D object in the STEM information is

needed in the third dimension, this is where electron tomography plays a role.

Tomography is the process of reconstructing an n-dimensional object from a number of n-1

dimensional images of the object. Typically, this involves reconstructing a three dimensional

volume from two dimensional images. This can take the form of reconstruction from a series of

projections through an object at a number of different angles (as is the case in X-ray computed

tomography (CT) (Hsieh, 2009) and electron tomography (Midgley, 2007)), or from a series of

slices through the object in question (as is the case in focussed ion beam (FIB) scanning

electron microscope (SEM) slice and view tomography (Dunn & Hull, 1999; Sakamoto, et al.,

1998)).

The form of tomography in which an object is reconstructed from its projections has its roots in a

1918 paper by Johann Radon (reprinted in (Deans, 2007)), in which he provides a proof that an

n dimensional object can be fully reconstructed from an infinite set of n-1 dimensional

projections. The Radon transform, R, of a two-dimensional object f(x) is a series of line integrals

through all possible lines running through the object,

𝑅𝑓(𝒙) = ∫𝑙 𝑓(𝒙)𝑑𝒙. (4.1)

To reconstruct the object an inverse Radon transform must be found and implemented on

collected data. In practice, a finite number of line integrals are collected at a finite number of

angles. Implementations of this form of tomography therefore offer an approximation to the full

solution of the object as defined by Radon. Projections in STEM are usually similarly assumed

to be formed from a series of line integrals through the object to each pixel in an image. In this

64
case, the intensity (I) in an image pixel is equivalent to the projection pj which, in the notation of

(Frank & Radermacher, 2006), is given by

𝐼(𝑥 𝑗 , 𝑦 𝑗 ) = 𝑝𝑗 (𝑥 𝑗 , 𝑦 𝑗 ) = ∫ 𝑓(𝑥 𝑗 , 𝑦 𝑗 , 𝑧 𝑗 )𝑑𝑧 𝑗 , (4.2)

j j j j
where (x , y , z ) = r is the rotated coordinate system of the projection. The coordinate system of

the projection, rj, is related to the coordinate system of the object, r, through the three Euler

angles (θj, φj, ψj) of the projection, shown in Fig 4.1, via

𝑟 𝑗 = 𝐷𝜓𝑗 𝐷𝜃𝑗 𝐷𝜙𝑗 𝑟, (4.3)

where the rotation matrices of each angle are given by

cos 𝜃𝑗 0 − sin 𝜃𝑗 (4.4)

𝐷𝜃𝑗 = ( 0 1 0 ),
sin 𝜃𝑗 0 cos 𝜃𝑗

cos 𝜙𝑗 sin 𝜙𝑗 0 (4.5)

𝐷𝜙𝑗 = (− sin 𝜙𝑗 cos 𝜙𝑗 0),
0 0 1

1 0 0 (4.6)
and 𝐷𝜓𝑗 = (0 cos 𝜓𝑗 − sin 𝜓𝑗 ).
0 sin 𝜓𝑗 cos 𝜃𝑗

65
Figure 4.1. Diagram of two coordinate systems related by a set of Euler angles. The blue

coordinate system is equivalent to the coordinate system of the object and the red coordinate

system rotated by the Euler angles (θ, φ, ψ) is equivalent to that of a projection.

The line integral assumption is a simplification of the true case, in which a projection is formed

via an integral over a double-cone through the object with its apex at the focal point of the probe.

The line integral approach is valid for small probe convergence angles and will be assumed

throughout this work. However, the double-cone approach may be necessary as the probe

convergence angle increases, which is becoming practically achievable due to the influence of

aberration correction, and has been successfully employed when combining tilt and focal series’

(Dahmen, et al., 2014). For thicker specimens the effect of beam spreading has an even greater

effect on the probe distribution than the probe forming angle, as discussed in section 2.4. This

effect could become important in obtaining accurate reconstructions from thick specimens,

particularly if voxel intensities are to be quantified.

Assuming that projections are acquired with the sample only rotated along a single axis so that

only θ is not zero, then the Radon transform can be seen to transform an object in a Cartesian
j j
coordinate system (x,y,z) to that of a coordinate system (x ,y ,θ) that is a representation of the

66
object in Radon space. A representation of an object in this way is often termed a sinogram, as

a point in Cartesian coordinates is translated to a sinusoidal curve in Radon space (Fig 4.2).

j
Figure 4.2.a) Illustration of the projection of a small dot f(x,z) on to a coordinate system (x ,θ) to
j
form a projection pj(x ,θj). b)The sinogram associated with a tilt series obtained for a point; the
j
slice highlighted in the sinogram corresponds to the projection in (a), pj(x ,θj).

4.2. The Tomography Procedure

Electron tomography in the STEM begins with the acquisition of a tilt series using one of the

imaging modes detailed in sections 2.2 and 2.3. Subsequent to acquisition, the tilt series’

images must be accurately aligned, as is detailed in section 4.4. After alignment, the tomogram

is reconstructed via one of a number of different algorithms (section 4.5). Finally, the tomogram

can be visualised in one of a number of ways.

Visualisation is performed either (i) through extraction of two dimensional slices through the

tomogram, termed orthoslices if they are orthogonal to one another, (ii) through creating a

surface across a number of connected voxels, termed an isosurface if all voxels of the surface

have the same intensity, or (iii) through direct visualisation of the entire volume with different

intensities represented through different screen intensities or colours.

67
4.3. Tilt Series Acquisition

Electron tomography is most often performed through acquisition of a series of images with the

sample tilted around a single axis, with angular increments typically between 1° and 10°,

depending on the imaging mode and sample limitations. The tilt increments are almost always

determined by the dose that the specimen can tolerate, due to the damage mechanisms

described in section 2.5, with imaging modes that require a larger dose per projection (EELS

and EDXS) often requiring a larger tilt increment to limit damage (Mobus, et al., 2003; Saghi, et

al., 2007).

For samples mounted on a standard TEM grid, the size of the holder within the polepiece of the

microscope and the position of grid bars often limits the extremes of the tilt to ± 70° or even

lower angles. This limit in the extreme tilt angles restricts the possible data collected and results

in an artefact in reconstruction known as the ‘missing wedge’ (Radermacher, 1988). Efforts to

minimise the missing wedge have led to specimen holders which allow novel acquisition

schemes, primarily the dual-axis acquisition scheme that reduces the missing wedge to a

‘missing pyramid’ (Arslan, et al., 2006; Penczek, et al., 1995). The achievable angular range

can be extended to the full 180° when using a needle shaped sample attached to an on-axis

tomography holder (Kato, et al., 2008; Koguchi, et al., 2001).

4.3.1. Imaging Modes

A number of imaging modes allow visualisation of the elemental distributions within

nanoparticles, as discussed in section 2.2 and 2.3. To provide an interpretable reconstruction

each imaging mode must conform to the projection requirement in order to obtain an accurate

reconstruction. The projection requirement states that the intensity detected must be a

monotonic function of the physical quantity to be reconstructed (Frank & Hawkes, 2006). Ideally,

the intensity should be a linear summation, as defined in equation 4.2, but nonlinearity can be

corrected if the function is monotonic (Yamasaki, et al., 2014). The use of a number of imaging

modes within electron tomography is outlined in this section, with a particular focus on

spectroscopic analysis of nanoparticles in three dimensions.

68
4.3.1.1. HAADF Tomography

The rapid growth of electron tomography studies in materials science over the last decade has

coincided with the development of HAADF tomography (Midgley & Dunin-Borkowski, 2009;

Midgley & Weyland, 2003). This is due to the extreme reduction in diffraction contrast when

using HAADF STEM in comparison with BF TEM. Thus, electron tomography of crystalline

specimens has been made possible without violating the projection requirement. Due to the Z

contrast of HAADF imaging, it has been used to distinguish between sample volumes of

different composition.

HAADF tomography was first demonstrated by Midgley et al in the study of heterogeneous

catalysts (Midgley & Weyland, 2003) and has seen widespread use for distinguishing heavy

metal nanoparticles on a lighter support material (Friedrich, et al., 2009; Weyland, 2002; Ziese,

et al., 2004). HAADF tomography of individual nanoparticles can be used to separate discrete

elemental phases, for example in well separated core-shell nanoparticles (Gomez-Grana, et al.,

2013; Mourdikoudis, et al., 2015; Rodriguez-Fernandez, et al., 2014). HAADF tomography of

nanoparticles has even been extended to atomic resolution (Chen, et al., 2013; Scott, et al.,

2012), where the application of separating discrete elemental phases is still valid, for example in

a core-shell PbSe/CdSe nanoparticle (Bals, et al., 2011) and an AgAu nanorod (Goris, et al.,

2013a) using the advanced reconstruction algorithms that are discussed in section 4.5.3.

A problem of HAADF STEM is the “cupping” artefact that arises due to the non-linear damping

of intensity (Van den Broek, et al., 2012). This has the effect of making the interior of objects

appear less intense than the exterior and can cause problems when separating regions of

different elements, for example, when segmenting core-shell nanoparticles. Van den Broek et al.

have proposed a method to correct this artefact, but it still remains as a problem in elemental

segregation of HAADF tomograms.

4.3.1.2. EFTEM/EELS Tomography

To date, the most widely used truly spectroscopic imaging technique is EFTEM tomography.

EFTEM was first applied to biological tomography studies (Koster, et al., 1997), using the zero-

loss image, to enhance image contrast. The first EFTEM tomography studies in the context of

materials science were performed by Möbus (Mobus, et al., 2003) and Midgley (Midgley &

69
Weyland, 2003) somewhat simultaneously. These two papers outline the distinct methods in

which EFTEM maps can be produced:

 Single EFTEM image – Taking a single EFTEM image may be sufficient enough to

image free standing particles but there are stringent conditions as discussed by Möbus

(Mobus, et al., 2003).

 Jump ratio – This involves taking two images, one at the maximum of the ionisation

peak and one just before the edge and then dividing the former by the latter. This

method is free from diffraction contrast but is not able to give quantitative information on

composition.

 Elemental composition – A background subtracted image can be formed by acquiring

three images that are below, on and above the associated ionisation edge. The

processed image can then be used to quantitatively study elemental composition,

although these images may still display diffraction contrast that affects the

reconstruction.

As discussed earlier, inelastic electron scattering should be particularly good for observing low-

Z elements. For example, nitrogen doping can be clearly resolved in carbon nanotubes (Florea,

et al., 2012). EFTEM tomography is not just useful for compositional analysis that complements

HAADF tomography. As well as filtering core loss electrons that give distinct elemental

fingerprints it is also possible to filter electrons that have losses associated with the formation of

plasmons. As HAADF imaging is reliant on differences in Z to provide image contrast it is not

possible to image different materials containing the same element. Plasmon loss images can

provide contrast between, for example, carbon nanotubes in nylon (Gass, et al., 2006) or silicon

nanoparticles in silica (Yurtsever, et al., 2006) that would not provide sufficient contrast from

mass-thickness based imaging modes. This method can also provide spectroscopic

measurements for materials of different elements in which HAADF contrast may be low, such as

a silicon/alumina/titanium nitride nanowire (Haberfehlner, et al., 2012).

Electron tomography using full spectral imaging in EELS was first reported in 2009,

demonstrating reconstruction of the core-loss signal of a number of elements (Jarausch, et al.,

2009). Subsequently, further studies have examined the feasibility of EELS tomography (Yedra,

et al., 2012) and compared imaging modes using EELS and EFTEM (Goris, et al., 2011).

70
Extension of 3D elemental reconstructions to a 4D spectrum volume that includes a full energy

loss dimension has been proposed (Goris, et al., 2014b; Yedra, et al., 2014), but it is unclear

whether reconstruction of channels containing no core-loss or low-loss information has any

physical meaning or meets the projection requirement. The full 4D spectrum volume has been

used to track changes in the oxidation state of Ce in 3D, to provide a 3D distribution of oxidation

states in a ceria nanoparticle (Goris, et al., 2014b). Plasmon imaging has also been utilised in

EELS tomography to study the three-dimensional distribution of plasmon modes in an Ag

nanocube (Nicoletti, et al., 2013), although careful consideration of the specimen is needed in

this case.

4.3.1.3. EDX Tomography

Whilst EFTEM tomography has been reported regularly over the past decade, the majority of

reports of EDX tomography have come over the past two years. This is largely due to the

difficulties associated with tilt series acquisition that result from the EDX detector geometry

(Mobus, et al., 2003). Specifically, the detector is usually positioned at the same height as the

sample, or slightly above, and maximum counts are recorded when the sample is tilted to 15-

20°. Thus for most microscopes, in tilting through a large range of angles, as is required for

electron tomography, the number of counts is greatly reduced at angles away from the optimum.

In addition, as discussed in section 2.3, the restricted TEM polepiece means that EDX detectors

typically have a limited solid angle of 0.1 - 0.3 sr and therefore only collect a few percent of

generated X-rays, contributing to long collection times.

The first report of EDX tomography was in the study of nanoparticles in a polycrystalline FeAl

matrix by Möbus (Mobus, et al., 2003). In this report, the difficulties of performing EDX

tomography were clearly illustrated. Images were acquired at tilt intervals of 10° owing to the

long acquisition times required and the shadowing of the detector by the sample holder led to

overall intensity variations over the range of tilt angles, therefore significant artefacts were

present in the weighted-backprojected reconstructions. However, EDX maps taken at these

large intervals have been shown to be useful in a process known as ‘hybrid tomography’

(Mobus, et al., 2010; Saghi, et al., 2007) in which EDX data is combined with HAADF

tomograms in order to provide chemical sensitivity or simply to improve the morphological

characterisation. The development of more advanced reconstruction algorithms, such as the

71
compressed sensing approaches discussed in section 4.5.3, now offer the possibility to reduce

artefacts in reconstructions from projections at 10° angular increments (Leary, et al., 2013).

Other early EDX tomography studies utilised a needle shaped sample to avoid missing

projections associated with X-ray detector shadowing (Kotula, et al., 2007; Yaguchi, et al.,

2008). However, these studies still required long acquisition times (> 50ms per pixel dwell

times) due to the small solid angle of detectors and were undertaken at relatively large tilt

increments (5 or 10°).

Recent advances in EDX detector design have meant that EDX tomography has become a truly

viable alternative to EFTEM tomography with the advantage that the full spectral range is

collected simultaneously. The most pervasive detector design in these studies is the Super-X

detector, introduced in section 2.3.2. The four detectors increase the solid angle of collection to

around 0.7 - 0.9 sr and their symmetrical position around the optic axis means that X-ray counts

do not fall as sharply when tilting the sample. Developments in field emission gun (FEG)

technology and aberration correction have also led to a potential increase in beam current for

the same probe size. Therefore, it is possible to achieve higher counts in EDX at the same

resolution and a higher signal to background than in initial EDX tomography studies.

An early report of EDX tomography using the Super-X detector system investigated the

structure of a high-k dielectric transistor (Lepinay, et al., 2013). Using a needle shaped

specimen it was possible to acquire EDX maps at 2° for a full 180° tilt range and this provided a

reconstruction with largely reduced artefacts compared to reconstructions from far fewer

projections. Subsequently, EDX tomography has been extended to nanoparticle samples

situated on carbon films on TEM supports grids (Genc, et al., 2013; Goris, et al., 2014a;

Liakakos, et al., 2014). In these papers no account was made for the detector shadowing due to

the sample holder or grid. However, there has been some interest in how this shadowing may

be characterised and compensated for in the literature, with a model to determine detector

shadowing put forward recently (Yeoh, et al., 2015). Additionally, an EDX tomography tilt series

has been acquired simultaneously to an EELS tilt series, paving the way for complementary

spectroscopic tomography (Haberfehlner, et al., 2014).

72
Even with advances in detector design, EDX spectrum images will continue to require a higher

overall electron dose than that required in HAADF STEM imaging. When studying beam

sensitive specimens high electron doses result in morphological changes over a tilt series that

invalidate reconstructions. For this reason, reconstruction algorithms such as the discrete

tomography and compressed sensing algorithms discussed in section 4.5.3, that allow

reconstruction from a limited number of projections, will be particularly useful. EDX tomography

is particularly suited to these algorithms as its data is easily separated in to discrete elemental

data.

A clear advantage of EDX spectroscopy in tomographic imaging is its utility at greater

thicknesses than HAADF STEM and particularly EFTEM/EELS. This is particularly useful when

considering that the thickness of standard “thin foil” TEM samples increases by up to (3x) in the

beam direction when tilting. Another advantage of EDX over EELS could be in atomic-resolution

spectroscopy due to contrast reversal and non-locality in EELS (Wang, et al., 2008). This

advantage will become particularly useful if atomic resolution discrete tomography (Van Aert, et

al., 2011) is attempted for EDX signals.

4.4. Alignment of Projections

Invariably, after acquisition the acquired images in the tilt series must be aligned so that they

share a common coordinate system. Firstly, all projections must be aligned spatially so that the

coordinate system of each projection can be described simply by a combination of Euler angles,
j j
so that the origin of each projected coordinate system is the same (x ’,y ’ = 0). Secondly, in

single-axis tomography, the orientation of the tilt axis must be determined so that the

backprojection operation is performed at the correct orientation.

Spatial alignment of images in the tilt series in the most basic sense can be achieved through

simple cross-correlation (Frank & Brandt, 2006). The cross-correlation is almost always

performed between pairs of adjacent tilt images, as the similarity is largest between adjacent

images. Alignment of adjacent images can lead to the accumulation of errors, as small errors

can be propagated and accumulated through the tilt series (Frank, et al., 1987), and for this

reason cross-correlation often uses the central image as the initial reference and proceeds to

the tilt extremes in both directions. It has also been suggested to use an iterative method to

73
determine the projection to use as the initial reference that provides the best alignment (Jones &

Haerting, 2013), rather than simply starting from the central image. Alignment of images of

nanoparticles, or images that do not differ in the objects that they contain, may also be

performed via a variant of centre of mass determination (Scott, et al., 2012), although this

technique has not been widely used.

In bright-field TEM tomography of biological samples, often the image contrast is not good

enough for cross-correlation to provide an accurate result (Frank & Mastronarde, 2006). In

many cases, metal nanoparticles are used as fiducial markers which can be individually tracked

through the tilt series (Frank & Mastronarde, 2006). This not only allows spatial alignment of

images but allows accurate determination of the thickness of the sample and the tilt axis

orientation. Similarly, if image contrast is good enough, “patches” or features of the image itself

can be used as fiducials in much the same way (Brandt & Ziese, 2006).

If image alignment has not used fiducial markers, or a similar technique, to track the tilt axis

then the tilt axis orientation must be subsequently defined. The orientation of the tilt axis can be

found in the most straightforward manner by observing a sum of the tilted images, in which

image intensity forms a streak along the direction perpendicular to the tilt axis (Fig 4.3a). The

“series summation” technique will not work for spherical nanoparticles, or nanoparticles in which

all three dimensions are similar, however, as there will be no elongation of features at different

tilt angles. Determination of tilt axis orientation is also often performed using an arc minimisation

technique (Midgley, 2007). If the tilt axis is misoriented in the reconstruction, the reconstruction

will show the intensity of objects smeared out as arcs along the true tilt axis. A shift in the tilt

axis is therefore present as arcing in the same direction for all slices, whereas a misoriented tilt

axis that has its centre in the same position results in arcing in different directions above and

below the midpoint (Fig 4.3c-e). By performing reconstructions of discrete 2D slices along the tilt

axis it is possible to quickly visualise the presence of arcing and correct the alignment of the tilt

axis accordingly.

74
Figure 4.3. Representation of methods for tilt axis determination, using a sample of Au and Pd

nanoparticles on graphene. a) Projection at 0° tilt displaying the lines (in red) at which

reconstructions are performed to display the arcing in (c-e). b) Summation of the entire tilt series,

displaying streaking in the direction perpendicular to the tilt axis at an angle α to the horizontal.

c-e) Reconstructions of individual slices (i-iii) represented in (a) for c) a tilt axis shifted from the

correct orientation by 40 pixels, d) a tilt axis tilted from the correct orientation by 15° and e) a tilt

axis approximately at the correct orientation. The sample represented here is Au and Pd

nanoparticles on graphene, supplied by Dr Peter Toth (University of Manchester).

75
4.5. Reconstruction

4.5.1. Backprojection

Subsequent to the alignment of a tilt series, the 2D data is then reconstructed into a 3D

tomogram. Reconstruction methods in tomography can be split into two categories, real space

and reciprocal space techniques. In electron tomography, reconstruction techniques almost

exclusively begin with a backprojection operation. In real space, backprojection simply involves

projecting the intensity in a pixel of a projection back along a line at the same angle in which the

projection was taken. The intensity of voxels due to backprojection from a single projection is

defined by Frank and Radermacher (Frank & Radermacher, 2006) as multiplying the projected

pixel intensity by a 3D spread function (𝑙(𝑥 𝑗 − 𝑥 ′𝑗 , 𝑦 𝑗 − 𝑦 ′𝑗 , 𝑧 𝑗 )), related to the point spread

function of incoherent imaging (equation 4.7).

𝑝𝑗𝑏 (𝑥 𝑗 , 𝑦 𝑗 , 𝑧 𝑗 ) = ∬ 𝑝𝑗 (𝑥 𝑗 , 𝑦 𝑗 )𝑙(𝑥 𝑗 − 𝑥 ′𝑗 , 𝑦 𝑗 − 𝑦 ′𝑗 , 𝑧 𝑗 )𝑑𝑥 ′𝑗 𝑑𝑦 ′𝑗 . (4.7)

However, in many situations the spread function is ignored such that 𝑝𝑗𝑏 (𝑥 𝑗 , 𝑦 𝑗 , 𝑧 𝑗 ) = 𝑝𝑗 (𝑥 𝑗 , 𝑦 𝑗 ),

which can result in misrepresented intensity in nanoparticles, although this can be corrected

through measurement and subsequent application of the point spread function (Heidari Mezerji,

et al., 2015). Backprojection from a full set of projections is then performed by simply summing

the intensity from all projections, such that the backprojected intensity, b(x,y,z), is given by

𝑏(𝑥, 𝑦, 𝑧) = ∑𝑗 𝑝𝑗𝑏 (𝑥 𝑗 , 𝑦 𝑗 , 𝑧 𝑗 ). (4.8)

Reconstruction using a simple backprojection operation is illustrated in Fig 4.4., in which a

phantom object is reconstructed from differing numbers of projections. Reconstruction of tilt

series’ of a limited number of projections (> 2° intervals) through backprojection can result in

streaking artefacts (observed to different extents in Fig 4.4(b-d)), although it may be possible to

remove these via an interpolation scheme (Cao, et al., 2010).

76
Figure 4.4. Backprojection of an object along a varying number of directions. a) Image of a two-

phase phantom object to be back-projected, b) backprojection of object (a) along a single

direction, c) backprojection of object (a) along 3 equally-spaced directions and d) backprojection

of object (a) along 10 equally-spaced directions.

In reciprocal space, a projection is related to a reconstruction volume by the projection slice

theorem. This theorem states that the Fourier transform of a projection of an object corresponds

to a central slice through the full three dimensional Fourier transform of the object at the

equivalent tilt angle (Fig 4.5). Considering this theorem it is therefore possible to perform a

reconstruction simply by summing the Fourier transforms of a number of projections, to which

can then be applied a 3D inverse Fourier transform to retrieve the reconstruction (Gilbert,

1972b).

77
 j
Figure 4.5. a) The projected intensity p (x) of an object f(x,z) at an angle of θj. b) An illustration of

the angular sampling of an object in Fourier space under a single tilt acquisition scheme with a

missing wedge. The red line in (b) corresponds to the Fourier transform of the projected
j
intensity p (x) at the same angle θj.

The projection slice theorem also illustrates a problem with the standard backprojection

algorithm. A finite number of projections sample low frequency information at smaller intervals

than that of high frequency information and therefore low frequency information dominates

reconstructions using backprojection alone (Fig 4.6b). In order to compensate for this sampling

a weighting is often applied as a ramp filter in Fourier space, i.e. points in Fourier space are

multiplied by the distance from the tilt axis, that acts to give greater weighting to higher

frequency information. This is commonly known as ‘weighted’ or ‘filtered’ backprojection

(WBP/FBP) and is a very commonly used reconstruction technique, the effect of which is

illustrated in Fig 4.6c. In ‘filtered’ backprojection a low-pass filter is also applied to the Fourier

transform of the reconstructed object in order to limit high frequency noise which would be

enhanced by the ramp filter. Fig 4.5 also displays the missing wedge of information in Fourier

space that occurs as a result of the tilt limits imposed on some single-axis tilt series’. The effect

of this on the reconstructed intensity is a smearing of intensity at the extreme tilt angles and

elongation of the object in the z-direction, equivalent to the beam direction at 0° tilt (Fig 4.6d).

78
Figure 4.6. Reconstructions of the phantom object in (a) from projections at 1° intervals using b)

an unweighted backprojection operator from projections over the full tilt range, c) a weighted

backprojection operator from projections over the full tilt range and d) a weighted backprojection

operator with projections from between ± 70° displaying the missing wedge.

4.5.2. Algebraic/Iterative Reconstruction Techniques

It has become more and more commonplace in electron tomography to improve on an initial

weighted backprojection reconstruction by employing an iterative comparison to the original tilt

series data. This type of reconstruction uses successive reconstructions to produce theoretical

projections that are then compared to the initial acquired projections. The difference in

projections can then be projected back to obtain a difference reconstruction which can be

applied to the previous reconstruction via addition or multiplication.

The first class of algebraic reconstruction techniques, generally known as the algebraic

reconstruction technique (ART) (Gordon, et al., 1970), compared a single projection from the

reconstruction with the corresponding original projection and made corrections only in that

direction, before moving on to the next projection. Whilst this method is simple and fast to

implement computationally it has been shown to produce results that diverge in the presence of

noise.

One of the most widely used classes of reconstruction algorithm is the simultaneous iterative

reconstruction technique (SIRT) (Gilbert, 1972a). Instead of making corrections after the

analysis of each individual projection, SIRT algorithms simultaneously compare all projections in

order to ensure any corrections made do not cause the reconstruction to diverge from the most

likely solution. With the advance of computer processing speeds, and the introduction of multi-

core GPUs, iterative techniques are becoming increasingly fast to implement.

79
4.5.3. Advanced Reconstruction Techniques

Recently, there has been a lot of interest in using prior information to assist in the reconstruction

of electron tomography datasets. This prior information could be that there are only a few

intensity levels expected in the reconstruction (Batenburg, et al., 2009) or that the intensity in

the reconstruction should be found on a regular lattice (Goris, et al., 2013a; Van Aert, et al.,

2011). Most commonly this form of reconstruction is represented by a compressed sensing

basis, i.e. the reconstruction is assumed to be sparse in some basis (Leary, et al., 2013).

Enforcing sparsity is the equivalent of maximising the number of zero voxels in a reconstruction.

For example, a subset of the compressed sensing approach (including discrete tomography and

total variation minimisation) assumes that the reconstruction is made up of a small number of

grey levels and enforces this condition on the reconstruction (Goris, et al., 2013b). Total

variation minimisation assumes sparsity in the gradient domain, i.e. minimizes the number of

pixels in which there is a change in intensity. Discrete tomography can be viewed as an extreme

case of sparsity in the gradient domain, where not only sparsity is enforced but the intensity of

voxels in this domain is limited to a small number of values. These types of reconstruction

should be particularly useful for spectroscopic electron tomography in which only intensity from

a single element is reconstructed as a single grey level could be applied to assign voxels either

containing the element or not.

A compressed sensing approach can also be performed in the image domain, assuming that

the number of voxels containing intensity is sparse. This approach is particularly suitable for

imaging modes in which background pixel intensities are zero, such as EDX tomography. This

type of algorithm has been shown to produce high fidelity reconstructions from a few as 10

projections (Leary, et al., 2013), which will be particularly useful for EDX tomography in which

the maximum allowable dose inhibits the number of projections acquired.

4.6. Resolution in Electron Tomography

The resolution of a reconstruction from an electron tomography tilt series is dependent upon

many factors, including but not limited to: the initial image resolution (discussed in section 2.4),

the reconstruction volume, tilt angle interval, noise characteristics and reconstruction routine

(Midgley & Weyland, 2003). In all cases, the resolution along the tilt axis, which is defined here

80
as the y-axis with a resolution of dy, should be that of the initial projections if the tilt series is

aligned perfectly. In directions perpendicular to the tilt axis it is commonplace to define the

resolution using the Crowther criterion (Crowther, et al., 1970), which gives the resolution of a

backprojected reconstruction perpendicular to the tilt axis, dx and dz, equation 4.9.

𝜋𝐷
𝑑𝑥 = 𝑑𝑧 = , (4.9)
𝑁𝑝

where D is the volume of the reconstruction and Np is the number of projections. In many single

tilt-axis studies, and as displayed in Fig 4.6d, the missing wedge results in elongation of objects

in the z-direction. The resolution in z, dz, is elongated due to the missing wedge by a factor e xz

that is given by (Radermacher, 1988),

𝜃 +sin 𝜃 cos 𝜃 (4.10)

𝑒𝑥𝑧 = √ 𝑚𝑎𝑥 𝑚𝑎𝑥 𝑚𝑎𝑥
,
𝜃 𝑚𝑎𝑥−sin 𝜃 cos 𝜃
𝑚𝑎𝑥 𝑚𝑎𝑥

where θmax is the maximum tilt angle away from 0° at which projections are acquired in radians.

This results in a resolution in z of,

𝑑𝑧 = 𝑑𝑥 𝑒𝑥𝑧 , (4.11)

For a somewhat standard value of θmax of 70°, this results in an elongation in dz by a factor of

1.31.

In practice, equations 4.9 – 4.11 apply to backprojected reconstructions, but are known to

typically overestimate the resolution in real space of reconstructions using iterative techniques

(Heidari Mezerji, et al., 2011). The resolution of reconstructions using SIRT or other more

advanced techniques is determined only through measurement of the final reconstruction (Chen,

et al., 2014; Heidari Mezerji, et al., 2011).

4.8. 3D Imaging Beyond STEM Tomography

In section 4.3.1 it was outlined how STEM tomography is able to provide morphological and

spectroscopic information of samples at the nanometre scale. In addition, the use of both bright

field and weak beam dark field imaging in conjunction with electron tomography can provide 3D

reconstructions visualizing dislocations in materials (Barnard, et al., 2006; Liu, et al., 2014),

although there are stringent conditions to ensure the projection requirement is met. It is also

81
possible to scan a finely focussed precessed beam over the sample and hence to obtain the

three dimensional distribution of crystallographic grains in a sample via tilt series acquisition,

although this requires comparison of obtained diffraction patterns to a standard library (Gemmi

& Oleynikov, 2013; Mugnaioli, et al., 2009; Zhang, et al., 2010).

Three-dimensional information in the TEM does not necessarily require the acquisition of a

tomographic tilt series. A method related to tilt series electron tomography is single particle

reconstruction, an important method in the biological sciences, in which a large number of

images of particles of an identical structure are used as input projections for a 3D reconstruction

(Frank, 2002; van Heel, et al., 2000). This method has been recently applied in the physical

sciences to image the 3D structure of a single nanoparticle in a liquid, taking advantage of the

fact that the nanoparticle rotates to different orientations due to the motion of the fluid

surrounding it (Park, et al., 2015). Other three dimensional imaging methods in the STEM

include scanning confocal electron microscopy, using a large probe-forming angle to reduce the

depth of focus of the probe and extract information through a sample (Frigo, et al., 2002; Nellist

& Wang, 2012), which can also be used spectroscopically in combination with STEM-EELS (Xin,

et al., 2013).

A complementary high resolution 3D imaging technique is atom probe tomography (APT)

(Blavette, et al., 1993; Kelly & Miller, 2007; Seidman, 2007). APT uses an applied voltage to

destructively evaporate atoms from a needle-shaped sample, which are detected at a

positionally sensitive detector, with the position of the atom determined to nanometre resolution

in all dimensions. The time-of-flight of the evaporated atom is used to determine the chemistry,

so that each atom is defined as a particular element or, in some cases, isotope. 3D information

from atom probe tomography has been successfully combined with data from STEM

tomography in a number of cases (Herbig, et al., 2015; Xiong & Weyland, 2014). For example,

Arslan et al undertook correlative atom probe tomography with HAADF-STEM tomography to

provide 3D reconstructions of the same sample volume in an Al-Ag alloy (Arslan, et al., 2008).

A major drawback of all studies in the TEM (and APT studies) is the small sample volumes

analysed (sample thicknesses of a few hundred nanometres as a maximum). The small scale of

volumes analysed raises the question of whether the volume chosen is representative of the

bulk samples used in industrially significant processes. In the case of nanoparticle analysis,

82
while electron tomography fully characterises individual nanoparticles or small-scale

nanoparticle ensembles, questions may arise whether these are representative of the full

nanoparticle population. The use of characterisation techniques that probe a larger sample

volume at a lower resolution, alongside electron tomography, allows verification that the STEM

data is truly representative of the wider sample volume.

Three dimensional imaging in the SEM probes a larger sample volume albeit at a lower

resolution and is possible via sequential sectioning and imaging of one face of the material.

Serial removal of sample sections can be achieved either through use of an ultramicrotome

(Denk & Horstmann, 2004; Zankel, et al., 2009) or a focussed ion beam (FIB) (Uchic, et al.,

2007; Zankel, et al., 2014). The spatial resolution of the two-dimensional slices is limited by the

resolution of the imaging technique in the SEM and the resolution in the slicing direction is

limited by the slice thickness (approximately 10 nm in the FIB) (Mobus & Inkson, 2007; Villinger,

et al., 2012). Three-dimensional information in the SEM can include morphological information

through secondary electron imaging or backscattered electron imaging, chemical information

through EDX spectroscopy and crystallographic information via electron backscatter diffraction

(Goldstein, et al., 2003; Reimer, 1998).

The FIB is also a useful tool for performing sample preparation for electron and atom probe

tomography, allowing preparation of both thin TEM lamellae (Giannuzzi, et al., 1998; Mayer, et

al., 2007) and needle shaped samples for TEM and APT (Thompson, et al., 2007). Additionally,

in contrast to sample preparation via electropolishing, the FIB allows accurate ‘site-specific’

positioning of samples within a larger volume.

At larger length scales X-ray tomography has become a highly versatile tool for 3D imaging of

materials (Maire & Withers, 2014; Stock, 2008). X-ray tomography is a similar technique to TEM

tomography, with a series of projections taken of a sample rotated at different angles using X-

rays as the illuminating radiation rather than electrons. The contrast in images formed by X-rays

is typically due to attenuation of X-rays, via the Beer-Lambert law expressed in equation 2.9 in

section 2.3.3.2, although other imaging mechanisms such as phase contrast (Cloetens, et al.,

1999; Momose, et al., 1996; Wilkins, et al., 1996) provide alternative contrast formation

mechanisms.

83
The collection of attenuated images in X-ray tomography is, in general, far quicker than that in

STEM tomography and the rotation of the sample can be performed with a higher degree of

accuracy in the rotated angle. For these reasons, it is possible to acquire a much greater

number of projections in X-ray tomography than in STEM tomography. In X-ray

nanotomography, in which the resolution is below 1 μm, the time to acquire a single projection is

typically limited to minutes and therefore the total number of projections is typically no more

than a few hundred. In lower resolution studies high speed acquisition has allowed in-situ

studies of processes such as fatigue (Buffiere, et al., 2010) and porosity evolution (Babout, et

al., 2001).

The use of 3D imaging modes at larger length scales alongside STEM studies allows

verification that the STEM sample is representative of a bulk sample and puts any STEM study

in a wider material context (Mobus & Inkson, 2007). For instance, Tariq et al combined X-ray

tomography, FIB-SEM tomography and electron tomography to assess porosity in a silica-

alumina ceramic, which allowed characterisation of pore sizes from 10 nm to 10 μm in

diameter(Tariq, et al., 2011). Multiscale characterisation using a number of different 3D imaging

techniques is a rapidly developing field (Lopez-Haro, et al., 2013; Thiele, et al., 2013; Yazzie, et

al., 2012) with great promise for answering many problems in materials science, geology and

the life sciences.

4.9. Summary

Electron tomography provides a method of transforming the two dimensional information of

transmission electron microscopy into three dimensional information. Through the use of

HAADF, EELS or EDXS imaging modes the distribution of elements can be characterised within

a single nanoparticle in three dimensions. The geometry of the holder within the polepiece of

the STEM and the placement of detectors limits the allowable angles at which an image can be

acquired. The maximum allowable dose also places limitations on the ability to acquire a large

number of images, particularly for the spectroscopic imaging modes. Novel algorithms present

an opportunity to overcome dose limitations, as they allow high fidelity reconstructions from a

smaller number of projections.

84
 5. Investigating STEM-EDX Tomography Methodology

5.1. Introduction

This chapter is concerned with examining the methodology of energy dispersive X-ray

tomography in the STEM and exploring the conditions under which the technique provides an

accurate reconstruction. The inclusion of large solid angle silicon drift detectors within the STEM

has led to a renewed interest in the technique and to a number of studies over the past few

years (Genc, et al., 2013; Goris, et al., 2014; Haberfehlner, et al., 2014; Lepinay, et al., 2013;

Liakakos, et al., 2014). To ensure that STEM-EDX tomography provides accurate

reconstructions it is necessary to understand the technique’s limitations or the necessary

methodology to prevent the inclusion of artefacts.

The key effects to consider are shadowing of the X-ray detectors and absorption of X-rays

within the sample. The extent of detector shadowing varies as a function of the sample tilt angle,

depending on the geometry of the sample holder and EDX detector, and is shown to introduce

artefacts into the reconstruction of a tilt series in a similar manner to the missing wedge effect

outlined in section 4.5.1. In order to compensate for detector shadowing, previous studies have

suggested multiplication by a correction factor (Yeoh, et al., 2015), or utilising a sample holder

in which detector shadowing is reduced to zero or negligible levels (Goris, et al., 2014; Lepinay,

et al., 2013). In this chapter, characterisation of detector shadowing is employed in order to

apply an experimentally determined correction to the acquisition time of each spectrum image.

This chapter also considers the extent of X-ray absorption in a particular AgAu nanoparticle and

presents general levels of X-ray absorption in elemental media. X-ray absorption may cause

intensity variations across an elemental map that affect the validity of its use in tomographic

reconstruction. Similarly, variations in detector shadowing across an image area may cause an

overall intensity variation and the extent of this effect is considered in this chapter.

The AgAu nanoparticles used in this study were synthesized by Alexandra Macedo and Pedro

Camargo (Department of Chemistry at the University of Sao Paulo). The optical profiling of the

TEM holder was conducted by Arne Janssen.

85
The results of this project have been submitted to Ultramicroscopy. The original manuscript for

“STEM-EDX Tomography of Bimetallic Nanoparticles: A Methodological Investigation” is

presented below.

86
Paper: STEM-EDX Tomography of Bimetallic Nanoparticles: A

Methodological Investigation
a a c a a,b
Thomas JA Slater , Arne Janssen , Pedro HC Camargo , M Grace Burke , Nestor J Zaluzec ,
a
Sarah J Haigh

a
School of Materials, University of Manchester, Manchester, M13 9PL, UK

b
Nanoscience and Technology Division, Argonne National Laboratory, Argonne, Illinois 60439,

US

c
Departamento de Química Fundamental, Instituto de Química, Universidade de São Paulo,

São Paulo, Brazil

Corresponding author: Sarah Haigh, sarah.haigh@manchester.ac.uk

Highlights:

 We investigate the methodology of STEM-EDX tomography of nanoparticles

 We present a time-varied acquisition scheme to compensate for detector shadowing
 The ability of STEM-EDX tomography to meet the projection requirement is discussed
Abstract: This paper presents an investigation of the limitations and optimization of energy

dispersive X-ray (EDX) tomography within the scanning transmission electron microscope,

focussing on application of the technique to characterising the 3D elemental distribution of

bimetallic AgAu nanoparticles. The detector collection efficiency when using a standard

tomography holder is characterised using a tomographic data set from a single nanoparticle and

compared to a standard low background double tilt holder. Optical depth profiling is used to

investigate the angles and origin of detector shadowing as a function of specimen field of view.

A novel time-varied acquisition scheme is described to compensate for variations in the intensity

of spectrum images at each sample tilt. Finally, the ability of EDX spectrum images to satisfy the

projection requirement for nanoparticle samples is discussed, with consideration of the effect of

absorption and shadowing variations.

Keywords: Energy dispersive X-ray spectroscopy, electron tomography, bimetallic nanoparticles

1. Introduction

87
The characterisation of nanoparticles has been greatly enhanced over the past decade by the

increasing availability of three-dimensional structural information obtained via electron

tomography [1,2]. This technique involves tilting the sample to different angles along one or

more axes and collecting a series of images at each tilt angle. This data can be used to

reconstruct the three dimensional sample volume via an established tomographic algorithm.

Electron tomography reconstructions using tilt series data sets of transmission electron

microscope (TEM) or scanning transmission electron microscope (STEM) images have allowed

the morphology and distribution of nanoparticles on a substrate to be fully characterised in three

dimensions [3-5]. However, obtaining complementary three dimensional elemental information

is more difficult. Using electron energy loss spectroscopy (EELS) or energy dispersive X-ray

(EDX) spectroscopy in the STEM it is possible to obtain elemental maps that show features

down to the atomic scale [6,7]. This two dimensional spectrum imaging, in which a spectrum is

collected for every pixel, is now routine but, the combination of spectroscopic imaging with

electron tomography, has proved far more challenging experimentally. Practically, the principle

limitation when acquiring both EELS and EDX tomographic spectrum image tilt series are the

long acquisition times associated with even a single spectrum image [8].

New EDX detector geometries in the transmission electron microscope are capable of EDX

spectrum imaging at a wide range of tilt angles and with a significantly improved solid angle for

X-ray collection, greatly increasing the ability to acquire EDX tomography data sets [9,10]. In

particular, the Super-X detector configuration, composed of four separate silicon drift detectors

(SDDs) arranged symmetrically around the optic axis, has proved capable of performing EDX

tomography with much shorter acquisition times at each specimen tilt angle [11,12]. However,

even for these new generation large solid angle detectors, the geometrical percentage of X-rays

detected is low (approximately 6% for 0.7 sr), such that the signal to noise ratio of EDX

spectrum images is often poor. The low signal detection means that very high electron doses

are often required, through the use of high probe currents and/or long acquisition times. In the
9 2
tilt series discussed in this work a total dose of approximately 7x10 electrons/nm was

employed, similar to that reported for a previous study of a single transistor published by

Lepinay et al [12]. This will often limit the application of the technique, as many nanoparticles

are insufficiently robust to withstand this high total dose without significant structural change. An

important aim of experimental procedures in EDX tomography should therefore be to limit the

88
overall electron dose whilst maximising the signal at each angle. Limiting the number of

projections to as few as possible has the advantage of improving the quality of the spectrum

image data for each tilt angle and is therefore highly desirable, although requiring the use of

advanced reconstruction algorithms [13,14] to realise high fidelity reconstruction when few

projections are used.

Historically, one of the key challenges for EDX tomography has been that traditional single

detector EDX systems are often limited to a very narrow range of tilt angles [8]. At other angles

the penumbra of the specimen holder prevents an X-ray signal being detected by ‘shadowing’

the EDX detector [8]. The design of new detector geometries such as the Super-X detector

system [15] means that X-rays may be detected for a wider range of specimen tilt angles (+/-

70°), although some detector shadowing does occur for samples that are deposited on standard

TEM grids or that use traditional high-tilt tomography holders [16]. Shadowing at all tilt angles is

only eliminated through the use of 360° rotation tomography holders [12] and these are not

compatible with all types of specimen. Shadowing of X-ray detectors will cause intensity

variations as a function of tilt angle and will lead to systematic errors in the intensity contribution

to the final reconstruction. Consequently, it is desirable for shadowing-induced intensity

variations to be corrected either as part of the acquisition procedure or by post processing.

This paper investigates a novel methodology for EDX tomography of isolated nanoparticles. As

an example we have used AgAu nanoparticles for which EDX tomography has already been

demonstrated to give important insights into the different elemental distributions as a function of

overall composition [17]. The details of detector shadowing are investigated and a varied-time

acquisition scheme is introduced to compensate changes in detector solid angle. Finally, the

ability of EDX spectra to satisfy the projection requirement for nanoparticle samples is

discussed, with reference to absorption and shadowing variation across a sample.

2. Experimental Methods

High angle annular dark field (HAADF) scanning transmission electron microscope imaging was

performed on an FEI probe-corrected Titan G2 80-200 S/TEM with a high brightness X-FEG

electron source and Super-X energy dispersive silicon drift detectors (SDDs). The microscope

was operated at an accelerating voltage of 200 kV with a beam current of 0.5 nA, a

89
convergence angle of 21 mrad and a HAADF acceptance inner angle of 50 mrad. EDX

spectrum images were acquired in the Bruker Esprit software using the Titan’s Super-X detector

system with a dwell time of 30 μs and a typical image size of 512x512 pixels. This detector has
2
a total solid angle of approximately 0.7 sr distributed over four 30 mm SDDs equally separated

at 45° from the holder tilt axis at an elevation angle of approximately 18° from the horizontal [15]

(Fig 1). Full spectrum image datacubes were exported from Esprit to Gatan’s DigitalMicrograph

software. As individual pixels were too noisy for accurate quantitative analysis, elemental X-ray

counts for each tilt angle were extracted from spectra obtained from a summation over all pixels

in the full spectrum image. Background subtraction was performed via a window method using

energy windows (of the same width as that of the signal window) before and after the peaks of

interest, taking care to ensure that no other peaks are included in these regions [18]. Elemental

maps were extracted at energies of 2.92 keV – 3.06 keV (Ag Lα) and 9.58 keV – 9.82 keV (Au

Lα) and are shown after applying a 5-pixel smoothing (as detailed in section 3.2).

A Fischione 2020 single tilt tomography holder was compared to an FEI high-visibility low-

background double-tilt specimen holder. EDX tomography used an angular increment of 10°

and a tilt range of ±70°. Traditional tilt series data sets were obtained with a constant acquisition

time of 300s for each spectrum image while in our novel time-varied tomographic scheme

acquisition times were varied between 236s and 895s in order to compensate for the calibrated

detector shadowing at the particular specimen tilt. Manual alignment was used to centre the

specimen at each tilt angle and HAADF images were acquired simultaneously with spectrum

images at each tilt angle. Additional intermediate HAADF images were collected every 5° to

assist with subsequent alignment of the data set via cross-correlation of HAADF images. Spatial

image alignment was performed using cross-correlation of HAADF images in FEI’s Inspect3D

software package. Tilt axis alignment was also undertaken in Inspect3D followed by using a

simultaneous reconstruction technique (SIRT) to perform the reconstruction with 20 iterations.

ImageJ [20] and FEI’s Avizo software platforms were used for visualisation of reconstructions.

To investigate the effect of variable intensity data sets a phantom was simulated in MATLAB

software consisting of a simple three-phase image; an 8-bit image of a circle (signal

intensity=193) surrounded by a ring (signal intensity=253) on a background of intensity 1. The

radon function was used to simulate projections between ±70° at 10° intervals. The inverse

90
radon function was then used to perform a filtered backprojection, firstly, on projections of the

same intensity and, subsequently, with projections multiplied by the same factor as the intensity

differences provided by shadowing from a single tilt tomography holder on all four Super-X

detectors (Fig 2c), as well as multiplication factors taken from the variation in intensity from only

two Super-X detectors (Fig 2b).

Three-dimensional optical images of the Fischione 2020 holder were acquired with a Keyence

VK-X210 3D laser scanning confocal microscope. The microscope is equipped with a violet

laser (wavelength 408 nm) and a 16-bit photomultiplier for accurate detection of reflected light

on surfaces. High-resolution three-dimensional (3D) maps of the top and bottom surfaces of the

holder were obtained by stitching 54 single images of the top side and 25 single images of the

bottom side together. The assembled three-dimensional map of each surface was analysed with

the Keyence VK Analyser Software package.

AgAu nanoparticles were chosen as a suitable test system for this study because of their

robustness to the structural changes induced by the electron beam and from the presence of

two elements with characteristic X-rays in different parts of the X-ray spectrum. These particles
-
were synthesised via a galvanic replacement reaction between Ag nanoparticles and AuCl 4 (aq),

as detailed in previous work [21]. This process leads to nanoparticles with a range of

morphologies including hollow donut-like structures like that shown in Fig 2a. At Au contents

less than approximately 20 at% the gold is segregated to the nanoparticle surface while at

higher Au contents the nanoparticle surface is enriched in Ag. A nanoparticle sample of average

composition Ag 78 at% Au 22 at%, as measured by flame absorption spectroscopy, was

deposited from solution onto continuous carbon-coated copper 200-mesh TEM grids (from Agar

Scientific, product code AGS160). Care was taken to acquire tilt-series’ of nanoparticles in the

centre of the grid squares.

3. Results and Discussion

3.1. Calibration of detector shadowing

In order to account for detector shadowing, when using a standard tomography holder, the

extent of shadowing at each angle must first be accurately calibrated. Doing this requires a

sample that will itself give a constant X-ray signal independent of tilt angle. Robust single

91
nanoparticles like that shown in Fig 2a present a suitable specimen from which X-ray counts

associated with the elements present within the nanoparticles should not significantly vary with

specimen tilt angle. Any variations in X-ray peak intensity from spectral images of single

nanoparticles should be due to detector shadowing alone.

Figure 1. Schematic illustrating the geometry of the Super-X detectors in the FEI Titan G2 80-

200 (S)TEM. (a) viewed in the direction of the electron beam and (b) viewed for a 2D cross-

section passing through detectors 2 and 4 at 45° to the long axis of the specimen holder.

Detectors are labelled 1-4 as used in FEI TIA software.

Figure 2. Calibration of Super-X detector shadowing in the Titan G2 for the Fischione 2020

single tilt tomography holder. a) EDX Au and Ag elemental map for a single AgAu nanoparticle

at 0 degrees showing the particle that was used to perform the detector shadowing calibration,

b) background subtracted Au Lα X-ray counts (9.7 keV) at each pair of detectors as a function

of tilt angle, taken from EDX spectrum images acquired for the same AgAu nanoparticle using

an acquisition time of 5 mins, 10° tilt increments and an angular range of ±70°, c) background-

subtracted summed X-ray counts for Au Lα (9.7 keV) and Au Mα (2.1 keV) peaks.

92
In order to characterise the variation in total X-ray count rates as a function of tilt angle,

spectrum images were acquired with a Fischione 2020 single-tilt tomography holder using the

Titan’s Super-X EDX detector system (Fig 1) for a single AgAu nanoparticle. Spectrum images

were acquired for 5 minutes every 10° for a tilt range of ±70° (Fig 2b,c). Detectors mounted on

either side of the holder tilt axis will show similar behaviour and the response of detectors 3 + 4

is symmetrical to that of detectors 1 + 2 (Fig 2b). The response of these pairs of detectors is

similar to that of a standard geometry single SDD detector [16]. The full detector response (Fig

2c) consists of the sum of detectors 1-4. Perhaps surprisingly, this data demonstrates that for

this specimen holder the poorest X-ray detection efficiency occurs at low specimen tilts. At 0°

the Super-X detector collects only 30% of the X-rays measured at a specimen tilt of 60° for the

same nanoparticle, due to the penumbra of the Fischione 2020 tomography holder. The Au Lα

(9.7 keV) and Au Mα (2.1 keV) characteristic X-ray peaks are found to display an almost

identical relationship with specimen tilt angle (Fig 2c), demonstrating that detector shadowing

does not vary with X-ray energy for this energy range. This suggests that the same

compensation of shadowing can be used for all characteristic X-ray peaks above 2.1 keV.

Shadowing of the X-ray detectors is possibly due to either the penumbra of the sample holder or

the bars of the copper TEM grid, or a combination of both on different sides of the holder. In

order to determine the source of shadowing, and the angles at which X-rays are shadowed in tilt

series, we have taken an optical profile of the top and bottom surfaces of the 2020 holder. A

three dimensional image of the surface of the Fischione holder is produced (Fig 3a,c), in which

line profiles of height variations of the holder give the angles at which the holder shadows any

emitted X-rays. The detectors are located at 45° angles azimuthally from the tilt axis (Fig 1a)

and therefore we have extracted line profiles at an azimuthal angle of 45° from both sides (Fig

3b,d). Initially we assume the nanoparticle specimen is located on the carbon film in the centre

of a grid square in the middle of the holder. We also assume that the carbon film of the grid lies

flat on top of the grid, the square holes have a width of 90 μm, the grid bar height is 20 μm and

the grid is oriented at 45° to the tilt axis. We neglect bowing of the carbon film; a phenomenon

which is known to occur for a number of different grids from a range of suppliers and which may

increase with the large dose supplied to the area surrounding the nanoparticle. The line profiles

in Fig 3 reveal a polar angle of 21° from the specimen to the top surface of the holder and 18° to

the bottom surface. In comparison, the 200 mesh copper TEM support will provide no

93
shadowing above the sample and shadowing up to an elevation angle of 24° below the sample.

Thus for the somewhat idealised situation described above, shadowing above the sample

occurs at a polar angle of 21° due to the sample holder and shadowing below the sample

occurs at a polar angle of 24° due to the grid bars (Fig 4). Use of these angles for shadowing

assumes point detectors at 45° azimuthal angles to the tilt axis, but we note that this is an

approximation as the finite detector size will cause shadowing to vary over the width of the

detector.

94
Figure 3. Optical profiles of the top and bottom surfaces of the Fischione 2020 holder. a) 3D

profile of the top surface of the holder with the direction of line profile, b, illustrated. b) Line

profile of plane indicated in a, displaying a polar angle of approximately 21° from the centre of

the grid to the highest point of the top surface of the holder. c) 3D profile of the bottom surface

of the holder with line profile d illustrated. d) Line profile of plane indicated in c, from which a

polar angle of approximately 18° from the centre of the grid to the bottom surface of the holder

is found.

95
Figure 4. Schematic diagram showing the elevation angles from the sample at which the holder

and grid bars will shadow emitted X-rays above and below the specimen in the direction of a

single Super-X detector (45° azimuthal angle).

The angle intersected by each detector can be calculated through prior knowledge of the total

solid angle and the detector areas. Given the Super-X detector system is characterised by a
2
total solid angle of approximately 0.7 sr and a single detector area of 30 mm , the distance of

the detectors from the optic axis is calculated from simple geometry as approximately 12 mm.

Zaluzec’s equation for solid angle determination [22] is:

(𝑟𝑎2 +𝑑 2 −𝑑√𝑟𝑎2 +𝑑 2 ) (1)

Ω = 2𝜋 [ ],
𝑟𝑎2 +𝑑 2

where ra is the radius of a single detector and d is the radial distance of the detector from the

area of interest. Equation 1 can be used to estimate the distance from the sample to the

detector and therefore allows determination of the polar angle between the uppermost and

lowermost parts of the detector. This polar angle is approximately 27° in this case, although this

neglects detector tilt which is known to be present and which means that the angular extent of

the detector is not equal when considering the polar and azimuthal subtending angles. We have

determined the elevation angle of the detector using the minimum in X-ray counts observed for

our tilt series. Fitting Gaussians to the data in Fig 2b gives minimum counts at an average α-tilt

of 18° for each pair of detectors. A polar angle, θ, in the plane of the detectors, can be

translated to a polar angle, α, in the plane normal (perpendicular) to the α-tilt axis by the

equation:

96
 𝑡𝑎𝑛𝜃
𝛼 = tan−1 . (2)
𝑐𝑜𝑠45°

Thus where θ = 27°, 𝛼 =36° and given the elevation angle of the centre of the detector normal to

the α-tilt axis is 18° this predicts that the detector subtends an angle from 𝛼 =0° to 𝛼 =36 °. This

suggests that each pair of detectors should be fully shadowed when the sample holder is tilted

to the angular range of ± 8° to ± 32° (given the holder shadowing shown in Fig 4 and converting

the θ angles (21 and 24°) to 𝛼 angles in the detector plane (28 and 32°)). These values appear

to qualitatively fit the tilt-series data corresponding to normalised detector counts of less than

10%.

The tilt series data is also compared to the detector shadowing model proposed by Yeoh et al

[23] (Fig 5a). We have used a value of δ = 5° for the detector tilt angle and values of d = 10.5

mm and r = 2.9 mm for the distance to the detector and the radius of the detectors respectively.

The data here fits best with their proposed model, for the shadowing angles we have

determined, when a detector elevation angle of 16° is used (Fig 5a). The counts at negative tilt

angles appear to fit well qualitatively to the detector model, although the counts at positive tilt

angles do not. This may be due to a discrepancy between the detection efficiencies of the two

detector pairs. This model also allows consideration of the effect of analysing a nanoparticle

away from the centre of a grid square as shown in Fig 5. If the nanoparticle is away from the

centre of the grid square the tilt series becomes asymmetric, although not in the same manner

as the data acquired, with the counts of tilts in one direction raised whilst the counts of tilts in

the other direction are lowered.

97
Figure 5. Modelling detector shadowing using the model of Yeoh et al [23]. a) Plot of normalised

counts with respect to tilt angle for the positions within a grid square indicated in (b).

Experimental data points for Au Lα (9.7 keV) X-rays (reproduced from Fig 2c) are also shown

(black dots) for reference b) Diagram indicating nanoparticle position within a 90 μm x 90 μm

grid square with black in the centre, blue 10 μm from the edge towards detector 1 and 45 μm

from the edge towards detector 2 and red 10 μm from both edges towards detector 1 and

detector 2.

A tilt series of a similar AgAu nanoparticle using an FEI high-visibility low-background double-tilt

holder was also acquired for comparison of shadowing using a standard specimen holder (Fig

6). In comparison, the non-tomographic holder displays the reversed dependence on tilt angle,

i.e. low tilt angles give the least detector shadowing over the full tilt range of this holder (±30°).

This is due to the low-profile of the top of the holder that is designed specifically not to shadow

the Super-X detectors and is consistent with similar measurements [24]. Whilst this holder

displays a lower variation in counts around 0° tilt, its limited tilt range (±30°) would lead to large

missing wedge artefacts if it were used for tomographic reconstructions.

98
Figure 6. Calibration of Super-X detector shadowing in the Titan G2 for the FEI high-visibility

low-background double tilt holder using a similar particle to that shown in Fig 2a. a) Au Lα

counts at each pair of detectors as a function of tilt angle, b) background-subtracted Au Lα (9.7

keV) and Au Mα (2.1 keV) X-ray counts extracted from EDX spectrum images acquired for the

same AgAu nanoparticle using an acquisition time of 5 mins and 5° tilt increments for angles of

±30°.

3.2. Considering the effect of uncompensated detector shadowing on tomographic

reconstruction

Variations in the intensity of different projections at different tilt angles can lead to artefacts

within tomographic reconstructions [25]. This is demonstrated in Fig 7 for reconstruction of a

simulated two-phase object. Fig 7b shows the object reconstructed using 14 simulated

projections each with the same total intensity (10° tilt intervals over an angular range of ±70°). In

comparison, Fig 7c shows the same object reconstructed from projections where the total

intensity varies in a similar way to that predicted by the shadowing variations revealed in Fig 2c

for the 2020 single-tilt tomography holder, when used with the Super-X detector system. To

further illustrate the effect, the projections were backprojected with intensity variations following

those of only one pair of Super-X detectors (1+2) (Fig 7d), as calibrated in Fig 2b. This

illustrates the effect of using a conventional single side-mounted EDX detector. All

reconstructions show streaking artefacts associated with a missing wedge of projections and

large angular intervals respectively, causing the spherical phantom to become “lemon-shaped”

(with “lumps” in the direction of the missing wedge). However, the varied intensity

reconstructions also show noticeable variations in intensity, where the intensity of the outer ring

is diminished in the plane of the projections with lower intensity (as indicated by the white

99
arrows in Fig 7b,c). When considering this effect in Fourier space, it can be thought of as

weighting more strongly features in central slices away from the slice at k z = 0 (as represented

in Fig 7 e-f), thus reducing contrast for features in the central slices around that at k z = 0. Where

there is sufficient signal within the shadowed regions, these artefacts can be avoided by

compensating for the variation in intensity due to shadowing.

100
Figure 7. Demonstration of artefacts associated with variations in projection intensities. a)

Simulated two-phase object to be reconstructed. b-d) Phantom images reconstructed from 15

simulated projections of the two phase image in (a) for a tilt range of ±70° and angular intervals

of 10°. b) Reconstruction using projections of the same intensity, displaying missing wedge and

streaking artefacts only. c) Reconstruction applying the intensity variations due to shadowing

(measured when employing all 4 Super-X detectors with a single tilt tomography holder, as

shown in Fig 2c). d) Reconstruction applying the intensity variations from only a pair of Super-X

detectors on one axis, as shown in Fig 2b. (e-f) Representation in Fourier space of the

projections used to reconstruct the images in (b-d) respectively. e) Constant intensity

projections. f) Variable intensity projections weighted for the full Super-X detector geometry. g)

Variable intensity projections weighted for one pair of Super-X detectors.

101
3.3. Methods for compensating detector shadowing

We propose two methods by which the shadowing of the Super-X detector system can be

compensated and variations in the count rates at each specimen tilt angle can be reduced. The

first is to acquire spectrum images for a constant time at each tilt angle and then multiply each

spectrum image by a ‘shadowing compensation’ factor that normalises the total counts, as

suggested in the early work of Möbus [8]. This is straightforward for samples in which the total

counts do not change with specimen tilt, such as single nanoparticles, but for samples in which

the total counts change with tilt angle, such as thin films, a calibration sample would be needed.

The other drawback of this approach is that the signal-to-noise ratio across projections will vary

and simply multiplying the lowest intensity projections, or alternatively scaling down the highest

intensities, will retain the low signal-to-noise ratios that are present for these projections.

Therefore, although this approach has the advantage of simplicity, it will not present the most

accurate acquisition scheme. An alternative is to adjust the acquisition time at each tilt angle so

as to achieve a constant total specimen X-ray signal for each projection. This allows the

maximum tolerable electron dose to be optimally distributed over all projections or can be used

to minimise total acquisition time. The disadvantage of this approach is that it requires prior

knowledge of the detector shadowing for a particular sample holder and microscope

combination: obtained using a separate tomographic data set. This calibration can be obtained

either by measuring the time taken to acquire a fixed number of X-ray counts for a specific peak

within summed spectrum images at each tilt angle or by comparing the different X-ray signals

for summed spectrum images obtained with a constant acquisition time as a function of holder

tilt (as illustrated in Figs 2 and 4). We note that as different sample holders have different

geometries and as there are variations in the position of EDX detectors on different instruments,

calibration of the detector shadowing will need to be performed for each sample holder and

microscope combination in order to derive specific time varied acquisition schemes. This

calibration will also differ for different positions within a grid square, as illustrated in Fig 5, so the

accurate position within the grid square should be tracked.

We have acquired EDX spectrum images every 10° using variable tilt-dependent acquisition

times as shown in Fig 8a. The Au Lα, Au Mα and Ag Lα specimen X-ray counts displayed far

smaller variations than for a fixed-time acquisition scheme for the whole range of specimen tilt

102
angles (intensity variations were within ±15 %) demonstrating the success of our time-

dependent acquisition scheme for reducing variations in X-ray signals (Fig 8b). However, even

these remaining small variations in summed X-ray counts are larger than can be explained by

statistical noise fluctuations alone and could be caused by two possible effects. The pixel dwell

time and spectrum image size are constant for all tilt angles and this dictates the time increment

to which the acquisition time can be specified for each tilt angle. For example, a dwell time of 30

μs and an image size of 512x512 pixels results in a discrete minimum time increment of 8s.

Small variations in the X-ray counts are therefore likely to be due to the discrete time

increments enforced in the data acquisition. Proportionally, this will have the largest effect for

spectrum images acquired with the shortest acquisition times, so for our data will cause the

greatest errors at high tilt angles. The accuracy of the shadowing calibration will also affect the

accuracy of the X-ray count rates obtained in the time-dependent data series. Low X-ray signals

in the initial fixed time data are likely to cause larger errors when predicting optimal acquisition

times. For our holder-microscope combination the spectrum images around 0° tilt have the

lowest X-ray signals (approximately 6000 total Au-Lα counts at 0° tilt compared to

approximately 19000 counts at 70°). Our calibration shows the largest deviation at small tilt

angles (especially at -10° as shown in Fig 8b, c) this latter effect is likely to be the more

significant for our data. We also note that for very high count rates the ratios of pixel dwell time

to detector processing time and count rate can result in lost data, thus also potentially

contributing to this variability. The combination of dwell time (approximately a few tens of μs),

detector time constant (approximately 2 μs) and count rates (< 10Kcps) in our experiments did

not reach this regime.

103
 900
800
700
600

Time (s)
500
400
300
200

-80 -60 -40 -20 0 20 40 60 80

Angle ()

Ag L
60000 Au L
Au M
50000

40000
Counts

30000

20000

10000

0
-80 -60 -40 -20 0 20 40 60 80
Angle ()

10000 1.00
0.99
9000
0.98
Proportion of time
Ag L Counts

8000 0.97
0.96
7000 0.95
0.94
6000
0.93
5000 0.92
-80 -60 -40 -20 0 20 40 60 80
Angle ()

Figure 8. a) Total acquisition time at each angle employed in the time-varied acquisition scheme

reported in this study. b) Au Lα, Au Mα and Ag Lα X-ray counts at each specimen tilt angle

under the varied acquisition time scheme. The variation in counts of each elemental peak using

the time-varied acquisition scheme are significantly reduced (variation < ±15%). c) Comparison

of Ag Lα counts at each specimen tilt angle and to the proportion of the calibrated time the

acquisition actually took.

104
The accuracy of the shadowing calibration is poorer where the collection efficiency is low. This

accuracy could be improved by repeating the calibration using longer acquisition times or by

measuring the time required to reach a certain value of specimen X-ray counts.

3.4. Post-acquisition alignment and filtering of EDX tomography data

The choice of tilt increment for the EDX tilt series data set is a compromise determined by

factors such as sample stability, achievable tilt range and X-ray count rates. For a fixed

tolerable electron dose, larger tilt increments, such as 10°, allow longer acquisition times for

each specimen tilt angle, resulting in improved signal to noise ratios within the individual

spectrum images. However, large tilt increments have been shown to limit the fidelity of

tomographic reconstructions when using standard reconstruction algorithms such as the

simultaneous iterative reconstruction technique (SIRT) [13,26]. The use of advanced

reconstruction algorithms may allow high fidelity reconstructions even when using large tilt

intervals of 10° or more [13], but this then poses a problem in the alignment of tilt series’.

Alignment procedures based on cross-correlation are less accurate when aligning images taken

every 10° than for smaller tilt intervals due to the larger differences between subsequent images

[27]. Difficulties with alignment of low signal-to-noise ratio spectrum images can be overcome by

performing image registration using the simultaneously acquired HAADF images and

subsequently applying this registration data to the elemental maps. Supplementary HAADF

images can be acquired at smaller tilt increments to aid registration without significantly

increasing the total acquisition time for the data set. To aid alignment we have acquired HAADF

images at every 5° to assist in the alignment of EDX maps acquired at every 10°.

The majority of EDX tomography studies have employed smoothing functions to the obtained

elemental maps to partially compensate for low signal intensity and a poor signal-to-noise ratio

[11,12]. To investigate the effect of smoothing filters on EDX elemental maps it is desirable to

work with a simulated data set. In this work we have simulated EDX maps using a test object

consisting of a pair of concentric rings; representing a hollow two-phase nanoparticle with the

outer phase having a higher concentration of the elemental being mapped (Fig 9a(i)). To match

the signal to that observed in typical experimental data, the intensity of the image was set to

zero for 30% of pixels selected at random from the whole field of view. An equivalent proportion

of zero value pixels are typically observed in experimental Au Lα elemental maps acquired over

105
300 s for the nanoparticles used in this study (Fig 9c). Noise was then added, equivalent to 20

counts per 100x100 pixels, producing a simulated EDX map (Fig 9a(ii)) which provides a good

match to the experimental data shown in Fig 9c. A 5-pixel smoothing window (Fig 9a(iii)) is

shown to suppress noise and give a signal which closely resembles the original test object.

Lepinay et al report that an edge-preserving smoothing filter suppressed noise well for their test

images [12]. However, using this filter on our test image we have found no significant

improvement from a generalised smoothing filter and also the production of negative intensity

artefacts (Fig 9a(iv) and 9b(iv)). We have therefore employed a 5-pixel smoothing to the

experimentally acquired Ag and Au elemental maps.

106
Figure 9. Producing a simulated EDX elemental map which closely resembles the experimental

data. a) i) Initial test image of object representing a two phase nanoparticle. ii) Simulated

elemental maps in which 30% of data points have been set to zero and Poisson noise added. iii)

Image b processed with 5-pixel smoothing window showing effective suppression of noise in the

image. iv) Image b processed with an edge preserving filter demonstrating the appearance of

negative intensity artefacts within the nanoparticle. All images are shown with the same

intensity scale and are 512x512 pixels. b) Line profiles (i-iv) through the images (a(i-iv))

107
respectively at the positions indicated by the arrows. c) Line profile taken from representative

experimental dataset with the corresponding Au Lα map of an AgAu nanoparticle shown inset.

We note that the signal-to-noise ratio for individual spectra can be improved by binning spectral

images or acquiring smaller-sized spectrum images at lower magnification. This has the

advantage that each pixel contains a more reliable spectrum that can undergo background

subtraction and may also assist with other spectral processing procedures such as multivariate

statistical analysis (MSA) [28,29]. The disadvantage is that binning of spectral images is likely

to degrade spatial resolution and may also reduce the ability to perform advanced

reconstruction algorithms such as compressed sensing [13].

4. Application to nanoparticle samples

The time-dependent acquisition scheme illustrated in Fig 8 has been successfully applied to

produce tomographic reconstructions for AgAu nanoparticles. The projections of Au Lα and Ag

Lα are shown in Fig 10(i-ii) at tilt angles of -60°, 0° and 60°. The spectra from the full spectrum

images at each of these tilt angles is also displayed in Fig 10iii.

108
Figure 10. Projections at a) -60°, b) 0° and c) 60° for i) Au Lα counts and ii) Ag Lα counts. iii)

Spectra from the full spectrum images at a) -60°, b) 0° and c) 60°.

Fig 11 shows orthoslices and a volume render for one of these reconstructions, demonstrating

the segregation of Ag to the surface at a composition of 60 at% Ag and 40 at% Au. A discussion

of the correlation between compositional segregation and catalytic activity for this system can

be found in Slater et al [17]. The reconstructions here represent counts from only one peak in

each case (Au Lα and Ag Lα) but the counting statistics could be improved by using the full

family of peaks. However, care should be taken to avoid overlapping peaks, as is the case for a

number of peaks in the Au M and Ag L families in this case.

109
Figure 11. Example of a tomographic reconstruction using a time-dependent acquisition scheme

performed for a AgAu nanoparticle. a) Orthoslice through the Ag reconstruction normal to the

optic axis, displaying clear segregation of Ag to the surface of the nanoparticle. b) Orthoslice

through the Ag reconstruction parallel to the optic axis. c) Orthoslice through the Au

reconstruction normal to the optic axis. d) Orthoslice through the Au reconstruction parallel to

the optic axis. e) Volume rendering of the Ag and Au reconstructions, Ag volume is green and

Au volume is red.

Previous studies have also successfully applied a standard ‘constant time’ acquisition approach

to successfully reconstruct similar AgAu nanoparticles [30] as well as other nanoparticle

systems [11,31]. However, there has been little discussion of the ability of the EDX signal

intensity to satisfy the projection requirement.

4.1. Projection requirement

The projection requirement of tomography states that the intensity of the signal used to perform

the reconstruction must be a monotonic function of the quantity to be reconstructed [32], and

direct proportionality between intensity and the physical quantity is desirable. A key question for

EDX tomography is therefore whether the characteristic X-ray intensity generated in the STEM

meets this requirement for the constraints of a particular sample and detector geometry.

The intensity of the characteristic X-ray signal for element A is related to the probability of

ionisation. Where nA is the number density of atoms of element A within the sample, QA is the

110
ionisation cross section and t is the sample thickness, the probability of ionisation for an atom of

element A is given by the product nAQAt.

The expected characteristic X-ray generation can then be calculated by consideration of the

competition between X-ray emission and other de-excitation methods, such as Auger electron

emission, defined as the fluorescence yield (ωA) [18]. The intensity of the X-ray signal for a

specific characteristic X-ray of element A (IA) is therefore given by:

Ω (3)
𝐼𝐴 = n𝐴 Q 𝐴 t ω𝐴 𝑎𝐴 𝐷𝑒 ( ) 𝜀𝐴 ,
4𝜋

where aA is the relative transition probability, De is the total number of electrons incident on the

sample, Ω is the detector solid angle and εA is the detector efficiency.

The only quantities in equation 3 which vary as a function of the two dimensional position within

a spectrum image are the number density of atoms of element A (nA) and the thickness of the

sample. The intensity of the characteristic X-rays described in equation 3 therefore fully satisfies

the projection requirement; the detected signal is directly proportional to the mass-thickness of

element A. The exception to this is when crystals are oriented with the electron beam passing

along a major zone-axis, resulting in the electron beam being more tightly bound to atomic

columns due to electron channelling [28]. For this reason, a quantitative analysis should avoid

channelling conditions or exclude projections of crystals close to major zone axes from

reconstructions.

X-rays are produced not only through emission of characteristic X-rays from atoms within the

sample but also through the deceleration of the electron beam in the sample. Bremsstrahlung,

or ‘braking radiation’, is emitted as a continuous spectrum of X-rays that decreases in intensity

with an increase in emitted X-ray energy. A sufficient signal-to-background ratio at each pixel

allows subtraction of the Bremsstrahlung background through either background modelling or a

simple two-window method [18]. If background subtraction is not performed, the contribution of

Bremsstrahlung will affect the intensity of elemental maps produced at each tilt angle and

therefore potentially influence the final tomographic reconstruction. The Bremsstrahlung signal

scales proportionally with atomic number, Z, and therefore could cause several artefacts. For

example, low energy X-rays may contain erroneous intensity in regions of high Z. For

111
reconstruction of high energy X-ray peaks the Bremsstrahlung background is low and the

contribution of Bremsstrahlung X-rays is not likely to significantly affect reconstructions.

4.2. X-Ray Absorption

Equation 3 also neglects the effect that X-ray absorption and fluorescence within the sample will

have on the intensity of characteristic X-rays emitted. Here, we will consider the implications

that absorption will have on the projection requirement. For absorption to negatively affect the

ability of the sample to accurately satisfy the projection requirement, the absorption must vary

across the spectrum image. The absorption of X-rays is described by an exponential attenuation

law [18]:

𝜇
−𝜌𝐿( ) (4)
𝐼 = 𝐼0 𝑒 𝜌 ,

where I/I0 is the fraction of X-rays not absorbed at a particular energy, ρ is the density of the

𝜇
element and ( ) is the mass attenuation coefficient of the element. Equation (4) can be
𝜌

rearranged to calculate a maximum path length through the sample, L, for a fixed amount of

absorption at a specific X-ray energy:

𝐼
𝑙𝑛 𝐼0 (5)
𝐿= 𝜇 .
𝜌( )
𝜌

Fig 12 uses elemental mass attenuation coefficients and standard specimen densities [33] to

predict the maximum allowable X-ray path length for a single element sample as a function of

atomic number in the limit of 1% absorption (I/I0 =0.99) at two different X-ray energies (2 keV

and 10 keV). This figure demonstrates that absorption is less than 1% and can safely be

neglected for characteristic X-ray energies of 10 keV or above for a path length less than 50 nm

and for an atomic number of less than 60. For less energetic X-rays (2 keV), a less stringent

limit of 10% absorption (I/I0 =0.9) yields similar acceptable sample constraints (specimen size

less than 50 nm and atomic number less than 60). The values displayed in Fig 12 provide a

guide to understanding the size of specimens acceptable for EDX tomography and which X-ray

energies are likely to provide elemental maps that most accurately satisfy the projection

requirement.

112
Figure 12. Maximum X-ray path length in a sample as a function of atomic number for a single

element sample when considering a limit of 1% absorption (I/I0 =0.99) at two different X-ray

energies (red, 2 keV and blue, 10 keV) and for a less stringent 10% absorption limit (I/I0 =0.9)

for an X-ray energy of 10 keV (green). Mass attenuation coefficients and the densities of

elements used to calculate this data have been taken from NIST [33].

Taking a 1% absorption limit, pure elemental Ag has a maximum allowable path length through

the specimen of 80 nm for 10 keV X-rays and 7 nm for 2 keV X-rays. The same absorption limit

for pure elemental Au gives a maximum path length through the specimen of 44 nm for 10 keV

X-rays and 5 nm for 2 keV X-rays. In this study we focus on bimetallic AgAu nanoparticles with

diameters of approximately 40 nm (Fig 11), although the hollow morphology of the particles

means that for most particles the maximum projected thickness is in practise typically 10 nm or

less. For the AgAu nanoparticle used in this study (Fig 11), the particle is roughly 50 at% of

each element and the maximum X-ray path length is approximately 5 nm. The attenuation

coefficient of mixtures and compounds can be obtained additively by:

𝜇 𝜇 (6)
= ∑𝑖 𝑤𝑖 ( ) ,
𝜌 𝜌 𝑖

where wi is the weight fraction of the ith atomic constituent [34]. Absorption of 2 keV X-rays

through the nanoparticle of interest (Fig 11) was found to be at maximum 1%, validating the use

113
of X-ray peaks above 2 keV in the reconstruction of this nanoparticle (Au Mα = 2.1 keV).

However, 1 keV X-rays will undergo maximum absorption of approximately 4% within the

nanoparticle which suggests that carbon (0.277 keV) and oxygen (0.525 keV) Kα X-rays will

show significant variations in absorption even in this small sample. The use of a maximum path

length differs from standard absorption correction procedures [35] which integrate the X-ray

path lengths along the electron beam within the sample. The complex geometry of these

particles makes integration challenging and hence we choose to consider the maximum path

length as this will provide an overestimate to X-ray absorption within an image and so a

conservative estimate of thickness.

4.3. Variations in Detector Shadowing Across a Sample Area

Another possible violation of the projection requirement comes from the fact that the extent of

detector shadowing varies as a function of sample position, and consequently it is feasible for

shadowing to vary within a single spectrum image. This situation is illustrated in Fig 13 which

shows a 2D cross-section subtending the sample, the shadowing object and the middle of a

single Super-X detector (top right). The difference in angle, Δθ, at which X-rays are shadowed

for two points at the extremes of the spectrum image (separated by a distance Δl), due to the

presence of an object with height H (typically the holder or grid bars) at a distance I from the

image area is calculated geometrically as:

𝐻 𝐻 (7)
Δ𝜃 = tan−1 ( ) − tan−1 ( ).
𝑙 𝑙+Δ𝑙

114
Figure 13. Diagram illustrating the variation in detector shadowing between two points on the

sample separated by a distance Δl, shadowed by an object of height H at a distance l from the

image area.

Using geometrical calculations an order of magnitude estimate of the variation of shadowing

across an EDX spectrum image area can be estimated as a function of α-tilt. We use the

variation in the unshadowed detector area as an estimate of the variation in X-ray counts across

an image area. This is an approximation that does not take in to account the change in X-ray

flux density across the detector but provides an accurate order of magnitude estimate. The

detector area, A, shadowed by an object casting a shadow of height h on the detector, which is

a function of θT, the polar angle due to tilt of the sample, is given by

𝑟−ℎ(𝜃𝑇 ) (8)
𝐴(𝜃𝑇 ) = 𝑟 2 cos −1 ( ) − (𝑟 − ℎ(𝜃𝑇 ))√2𝑟ℎ(𝜃𝑇 ) − ℎ(𝜃𝑇 )2 ,
𝑟

where r is the radius of the detector.

For the Super-X detector system the variation in shadowing for each pair of detectors (1+2 or

3+4 as shown in Fig 1a) must be considered separately but, within a pair, the shadowing can be

assumed to behave identically, due to their equivalent position with respect to the α-tilt axis. The

proportion of the detector area that is partially shadowed across the image, compared to the

detector area that is entirely unshadowed is termed P12 for detector pair 1+2 and P34 for

detector pair 3+4.

115
We begin by discussing the shadowing resulting from the top side of the sample holder, for the

situation where both pairs of detectors are partially shadowed simultaneously (Fig 14a). From

simple trigonometry the height, h, of the shadow cast upon the detector is given by

ℎ = 𝑑 sec(𝜃𝐸 − 𝜃𝐵 ) sin(𝜃𝑆 + 𝜃𝑇 − 𝜃𝐵 ) sec(𝜃𝑆 + 𝜃𝑇 − 𝛿), (9)

where d is the distance to the centre of the detectors from the sample, θE is the elevation angle

of all detectors (assumed to be the same), θB is the angle to the bottom of the detector pair, θS

is the angle over which the object shadows the detector pair, and δ is the polar tilt angle of the

detector. All angles are defined in the plane of the sample and two Super-X detectors (at a 45°

azimuthal angle to the α-tilt axis) as show in Fig 14a. An α-tilt of the sample can be translated to

a polar tilt angle θT, as 𝜃𝑇 = tan−1 (tan 𝛼 cos 45).

In the range of tilt angles in which detector shadowing from the top of the sample holders occurs

P12 and P34 are defined as

A(−𝜃𝑇 +Δ𝜃)−A(−𝜃𝑇 ) (10)

P12 = between −𝜃𝑇 + 𝜃𝑆 + Δ𝜃 = 𝜃𝑈 and −𝜃𝑇 + 𝜃𝑆 = 𝜃𝐵
A(−𝜃𝑇 +Δ𝜃)

A(𝜃𝑇 −Δ𝜃)−A(𝜃𝑇 ) (11)

and P34 = between 𝜃𝑇 + 𝜃𝑆 + Δ𝜃 = 𝜃𝑈 and 𝜃𝑇 + 𝜃𝑆 = 𝜃𝐵 .
A(𝜃𝑇 −Δ𝜃)

116
Figure 14. Diagram illustrating the geometry of angles used in determination of the extent of

detector shadowing from (a) the top side of the sample holder and (b) the bottom side of sample

holder (grid bars) .

Equations 10 and 11 provide the variation in shadowing when using only one detector pair,

either detectors 1 and 2 or detectors 3 and 4. The variation in shadowing can reach 100%

across the image area when the detector pair is fully shadowed at one side of the area, i.e. P 12

=1 when Δ𝜃 = 𝜃𝑈 − 𝜃𝑆 + 𝜃𝑇 and P34 =1 when Δ𝜃 = 𝜃𝑈 − 𝜃𝑆 − 𝜃𝑇 . For this reason, when using a

detector located only on one side of the tilt axis, care should be taken to avoid acquiring tomographic

data for tilt angles close to this maximum shadowing condition.

However, when both detector pairs are used, as is the case in this study, the shadowing effect is largely

mitigated by the greater effective detector area provided by the complementary detector pair on the

other side of the tilt axis. To calculate the contribution from both detectors we have to multiply the

variation in shadowing of a single detector pair Pij by the angular fraction of the total detector area that

this detector pair is subtending Qij. This value is given by

A(−𝜃𝑇 +Δ𝜃) A(𝜃𝑇 −Δ𝜃) (12)

𝑄12 = and 𝑄34 = .
A(−𝜃𝑇 +Δ𝜃)+A(𝜃𝑇 −Δ𝜃) A(−𝜃𝑇 +Δ𝜃)+A(𝜃𝑇 −Δ𝜃)

117
Thus, the maximum variation in shadowing across the specimen distance Δl when considering

all detectors, P, can then be expressed as a percentage,

P = 100 × (𝑃12 × 𝑄12 − 𝑃34 × 𝑄34 ) (13)

(A(−𝜃𝑇 +Δ𝜃)−A(−𝜃𝑇 ))−(A(𝜃𝑇 −Δ𝜃)−A(𝜃𝑇 )) (14)

or P = 100 × ( ).
A(−𝜃𝑇 +Δ𝜃)+A(𝜃𝑇 −Δ𝜃)

We have determined the corresponding angles approximately from our tilt-series data (Fig 2b)

and from the optical data of the sample holder (Fig 3). Fitting a Gaussian to our experimental

tilt-series data in Fig 2b reveals an average minimum in counts at an α-tilt of approximately 18°,

equivalent to a polar angle of θT = 13°, which we have related to the elevation angle of all four

detectors. We have used a detector tilt of δ = 0-15° and note that this has little effect on our

calculations, with no change in shadowing variation to 1 significant figure. The angles θ U and θB

are determined from the elevation angle, tilt and radius of the detectors. From the optical profile

data (Fig 3) we know that on the top side of the sample holder the height of the shadowing is h

= 600 μm and the distance to the shadowing is l = 1.6 mm, from which we calculate values of θS

= 21° and Δθ = 0.001° when Δl = 100 nm. Note in this and subsequent equations we have also

tacitly assumed that there are no contributions to this shadowing from any collimators, which

are absent in our system. However, for different configurations, should collimators be present,

then their presence will slightly modify the preceding equations which are based purely on the

effective radius of the detector.

Using these values, over a distance of 100 nm the percent of shadowing due to the top side of
-3
the holder reaches a maximum of 3x10 %. P12 and P34 will have opposite signs, meaning the

variation in shadowing is in the opposite direction for each detector pairs and acts to somewhat

compensate the detector shadowing from each detector pair.

Considering shadowing from the underside of the sample the shadowing height, h, on the

detector is given by

ℎ = 𝑑 sec(𝜃𝑈 − 𝜃𝐸 ) sin(𝜃𝑈 + 𝜃𝑆 − 𝜃𝑇 ) sec(𝜃𝑆 − 𝜃𝑇 + 𝛿). (15)

On the underside of the sample grid, our optical measurements have demonstrated that the

sample only causes shadowing for a polar angle of less than 18°, while the middle of a grid

square to the height of the grid bars in the TEM grids used in this study was 24°. This suggests

that the maximum shadowing angle on the underside of the sample is due to the grid bars (θS =

118
24°), as shown in Fig 4. However, bowing of the carbon film across a grid square could lead to

an increase or decrease in this angle. Assuming the carbon film lays flat, shadowing due to grid

bars can be estimated using values of width and depth obtained from suppliers in each

particular case, or through measurement of grid dimensions. In this case, we have used grids

with a hole of width 90 μm and a grid bar depth of 20 μm. Assuming the sample area is located

in the middle of a grid square, the angular difference between points 100 nm from each other

gives Δθ = 0.05°. For the underside of the sample P12 and P34 are calculated in a similar manner

to the top side of the sample holder:

A(−𝜃𝑇 +Δ𝜃)−A(−𝜃𝑇 ) (16)

P12 = between −𝜃𝑇 + 𝜃𝑆 + Δ𝜃 = 𝜃𝑈 and −𝜃𝑇 + 𝜃𝑆 = 𝜃𝐵
A(−𝜃𝑇 +Δ𝜃)

A(𝜃𝑇 −Δ𝜃)−A(𝜃𝑇 ) (17)

and P34 = between 𝜃𝑇 − 𝜃𝑆 = 𝜃𝑈 and 𝜃𝑇 + 𝜃𝑆 − Δ𝜃 = 𝜃𝐵 .
A(𝜃𝑇 −Δ𝜃)

However, unlike for the top side of the specimen holder both detector pairs are never

simultaneously shadowed (Fig 14b) and therefore the variation in shadowing can be given

simply by

(A(−𝜃𝑇 +Δ𝜃)−A(−𝜃𝑇 )) (18)

P = 100 × ( ).
A(−𝜃𝑇 +Δ𝜃)+1

This reaches a maximum of 0.1% and hence can be largely ignored for high spatial resolution

tomographic data sets. Importantly, for larger image areas, images taken close to the grid bars

or for grids with smaller grid areas this variation may be large enough so that projections at a

number of angles should be removed from the tilt series. For instance, if the nanoparticle was

located at a 10 μm distance to the grid bars closest to detectors 1 and 2, as represented by the

red dot in Fig 5b, the shadowing angle due to grid bars would increase to θS = 63°, increasing

Δθ to 0.2° over a 100 nm image area. In this case the variation in shadowing across the image

area would reach approximately 30%. Care should always be taken to ensure that the sample is

situated away from any grid bars when using a standard TEM grid and the use of fine-mesh grid

entirely avoided. In fact, the use of continuous films may be preferable to entirely remove the

contribution of grid bars to shadowing. We further note that where data is acquired from

specimen areas close to the grid bars or from a grid with a smaller mesh size the shadowing

from the top side may also be due to grid bars rather than the sides of the sample holder.

119
5. Conclusions

New designs of large solid angle energy dispersive X-ray detector systems using multiple

detectors have made EDX tomography more readily accessible. However, careful consideration

of sample and acquisition parameters is required to optimize data acquisition schemes and to

ensure that artefacts associated with low signal to noise ratio data do not affect the fidelity of

reconstructions. We have proposed a novel time-varied acquisition scheme that compensates

the effect of detector shadowing for the Titan’s Super-X EDX system, maximizing the available

X-ray signal when using a penumbra-limited single-tilt tomography holder. This compensated

acquisition scheme provides approximately constant X-ray counts in the characteristic peaks

over the full range of specimen tilt angles.

The intensity of reconstructions in EDX tomography, without considering absorption of X-rays, is

directly proportional to the number of atoms per unit volume of the element in question.

Absorption of X-rays limits the size of samples in which this proportionality holds true and, at

some thickness of sample, absorption will result in violation of the projection requirement,

particularly for high atomic number samples and reconstructions of low energy X-ray signals.

We have considered the potential for shadowing variations within a single spectrum image field

of view and determine that these are insignificant and can be largely ignored for small sample

areas (<100 nm) when using all four Super-X EDX detectors.

Acknowledgements

S.J.H. and T.J.A.S. acknowledge funding from multiple research grants including the

Engineering and Physical Sciences Research Council (EPSRC) UK Grants EP/G035954/1 and

EP/J021172/1 and Defence Threat Reduction Agency Grant HDTRA1-12-1-0013. T.J.A.S.

would like to thank the North West Nanoscience Doctoral Training Centre (NOWNano DTC) for

supporting his work. N.J.Z. also acknowledges support from the Electron Microscopy Center at

the Center for Nanoscale Materials of Argonne National Laboratory, a U.S. Department of

Energy, Office of Science, Office of Basic Energy Sciences User Facility under Contract No.

DE-AC02-06CH11357, as well as, a visiting appointment in the School of Materials at the

University of Manchester. The authors wish to acknowledge the support from HM Government

120
(UK) for the provision of the funds for the FEI Titan G2 80-200 S/TEM associated with research

capability of the Nuclear Advanced Manufacturing Research Centre.

References

[1] Midgley, P. A. & Dunin-Borkowski, R. E. Electron tomography and holography in materials

science, Nat. Mater. 8 (2009) 271-280.
[2] Bals, S., Goris, B., Liz-Marzan, L. M. & Van Tendeloo, G. Three-Dimensional
Characterization of Noble-Metal Nanoparticles and their Assemblies by Electron
Tomography, Angew. Chem. Int. Edit. 53 (2014) 10600-10610.
[3] Gontard, L. C., Dunin-Borkowski, R. E. & Ozkaya, D. Three-dimensional shapes and spatial
distributions of Pt and PtCr catalyst nanoparticles on carbon black. J. Microsc. 232
(2008) 248-259.
[4] Sueda, S., Yoshida, K. & Tanaka, N. Quantification of metallic nanoparticle morphology on
TiO2 using HAADF-STEM tomography. Ultramicroscopy 110 (2010) 1120-1127.
[5] Hernandez-Garrido, J. C. et al. The location of gold nanoparticles on titania: A study by high
resolution aberration-corrected electron microscopy and 3D electron tomography.
Cataly. Today. 160 (2011) 165-169.
[6] Klie, R. F., Arslan, I. & Browning, N. D. Atomic resolution electron energy-loss spectroscopy.
J. Electron. Spectrosc. 143 (2005) 105-115.
[7] D'Alfonso, A. J., Freitag, B., Klenov, D. & Allen, L. J. Atomic-resolution chemical mapping
using energy-dispersive x-ray spectroscopy. Phys. Rev. B 81 (2010) 100101.
[8] Mobus, G., Doole, R. C. & Inkson, B. J. Spectroscopic electron tomography. Ultramicroscopy
96 (2003) 433-451.
[9] Zaluzec, N. J. Innovative Instrumentation for Analysis of Nanoparticles: The π Steradian
Detector. Microscopy Today 17 (2009) 56-59.
[10] Phillips, P. J. et al. A New Silicon Drift Detector for High Spatial Resolution STEM-XEDS:
Performance and Applications. Microsc. Microanal. 20 (2014) 1046-1052.
[11] Genc, A. et al. XEDS STEM tomography for 3D chemical characterization of nanoscale
particles. Ultramicroscopy 131 (2013) 24-32.
[12] Lepinay, K., Lorut, F., Pantel, R. & Epicier, T. Chemical 3D tomography of 28 nm high K
metal gate transistor: STEM XEDS experimental method and results. Micron 47 (2013)
43-49.
[13] Leary, R., Saghi, Z., Midgley, P. A. & Holland, D. J. Compressed sensing electron
tomography. Ultramicroscopy 131 (2013) 70-91.
[14] Goris, B., Roelandts, T., Batenburg, K. J., Heidari Mezerji, H. & Bals, S. Advanced
reconstruction algorithms for electron tomography: From comparison to combination.
Ultramicroscopy 127 (2013) 40-47.
[15] von Harrach, H. S. et al. An integrated Silicon Drift Detector System for FEI Schottky Field
Emission Transmission Electron Microscopes. Microsc. Microanal. 15 (2009) 208-209.

121
[16] Slater, T.J.A., Camargo, P. H. C., Burke, M. G., Zaluzec, N. J. & Haigh, S. J. Understanding
the limitations of the Super-X energy dispersive x-ray spectrometer as a function of
specimen tilt angle for tomographic data acquisition in the S/TEM. JPCS 522 (2014)
012025.
[17] Slater, T. J. A. et al. Correlating Catalytic Activity of Ag–Au Nanoparticles with 3D
Compositional Variations. Nano Lett. (2014) 1921-1926.
[18] Williams, D. B. & Carter, C. B., Transmission Electron Microscopy, second ed., Springer,
New York, 2009.
[19] Kremer, J. R., Mastronarde, D. N. & McIntosh, J. R. Computer visualization of three-
dimensional image data using IMOD. J. Struct. Biol. 116 (1996) 71-76.
[20] Rasband, W. S., ImageJ, U. S. National Institutes of Health, Bethesda, Maryland, USA,
1997-2014.
[21] Petri, M. V., Ando, R. A. & Camargo, P. H. C. Tailoring the structure, composition, optical
properties and catalytic activity of Ag-Au nanoparticles by the galvanic replacement
reaction. Chem. Phys. Lett. 531 (2012) 188-192.
[22] Zaluzec, N. J. Analytical Formulae for Calculation of X-Ray Detector Solid Angles in the
Scanning and Scanning/Transmission Analytical Electron Microscope. Microsc.
Microanal. 20 (2014) 1318-1326.
[23] Yeoh, C. S. M. et al. The Dark Side of EDX Tomography: Modeling Detector Shadowing to
Aid 3D Elemental Signal Analysis. Microscopy and Microanalysis FirstView (2015) 1-6.
[24] Schlossmacher, P., Klenov, D., Freitag, B. & von Harrach, H. S. Enhanced Detection
Sensitivity with a New Windowless XEDS System for AEM Based on Silicon Drift
Detector Technology. Microscopy Today 18 (2010) 14-20.
[25] Thomas, J. M. et al. The chemical application of high-resolution electron tomography:
Bright field or dark field? Angew. Chem. Int. Edit. 43 (2004) 6745-6747.
[26] Batenburg, K. J. et al. 3D imaging of nanomaterials by discrete tomography.
Ultramicroscopy 109 (2009) 730-740.
[27] Frank, J. & Brandt, S., Electron Tomography, second ed., Springer, New York, 2008.
[28] Kotula, P., Brewer, L., Michael, J. & Giannuzzi, L. Computed Tomographic Spectral
Imaging: 3D STEM-EDS Spectral Imaging. Microsc. Microanal. 13 (2007) 1324-1325.
[29] Yaguchi, T., Konno, M., Kamino, T. & Watanabe, M. Observation of three-dimensional
elemental distributions of a Si device using a 360 degrees-tilt FIB and the cold field-
emission STEM system. Ultramicroscopy 108 (2008) 1603-1615.
[30] Goris, B., Polavarapu, L., Bals, S., Van Tendeloo, G. & Liz-Marzan, L. M. Monitoring
galvanic replacement through three-dimensional morphological and chemical mapping.
Nano Lett. 14 (2014) 3220-3226.
[31] Liakakos, N. et al. Co-Fe Nanodumbbells: Synthesis, Structure, and Magnetic Properties.
Nano Lett. 14 (2014) 2747-2754.
[32] Hirsch, P. B., Whelan, M. J. & Howie, A. On Production Of X-Rays In Thin Metal Foils.
Philos. Mag. 7 (1962) 2095.

122
[33] Seltzer, S.M. and Hubbell, J.H. (1995), Tables and Graphs of Photon Mass Attenuation
Coefficient and Mass Energy-Absorption Coefficients for Photon Energies 1 keV to 20
MeV for Elements Z = 1 to 92 and Some Dosimetric Materials, Appendix to invited
plenary lecture by J.H. Hubbell ``45 Years (1950-1995) with X-Ray Interactions and
Applications'' presented at the 51st National Meeting of the Japanese Society of
Radiological Technology, April 14-16, 1995, Nagoya, Japan.
[34] Seltzer, S. M. Calculation of Photon Mass Energy-Transfer and Mass Energy-Absorption

Coefficients. Rad. Res. 136 (1993) 147-170.

[35] Philibert, J. A method for calculating the absorption correction in electron-probe

microanalysis. In 3rd International Congress on X-ray Optics and Microanalysis, Pattee,

H. H., Cosslett, V. E. and Engstrom, A. (Eds.), 379.

123
 6. STEM-EDX Tomography of AgAu Nanoparticles

6.1. Introduction

This chapter describes the application of STEM-EDX tomography to investigate the elemental

distribution in AgAu nanoparticles synthesized by the galvanic replacement reaction. In

particular, the catalytic yield of propargylamines, a building block in many anti-apoptotic drugs

(Maruyama, et al., 2002a; Maruyama, et al., 2002b; Tatton, et al., 2002), in a three component

coupling reaction is correlated to changes in elemental distribution within the AgAu nanoparticle

catalyst.

The galvanic replacement reaction allows the synthesis of hollow nanoparticles using a

sacrificial template to determine the morphology of the final nanoparticle (Sun & Xia, 2004; Xia,

et al., 2013). A metallic nanoparticle template undergoes a redox reaction with a metal salt, with

the salt reducing at the surface of the nanoparticle and the metal template oxidising. The use of

an Au metal salt, HAuCl4, with an Ag nanoparticle template is widely reported and results in a

predominantly Au hollow shell structure (Au, et al., 2008; Sun & Xia, 2004; Zhang, et al., 2008).

These previous studies claim that all of the AgAu nanostructures formed possess a fully alloyed

structure. The precise distribution of Ag and Au is difficult to determine due to the very similar

lattice parameter of the two elements.

Hollow nanoparticles, such as those synthesized through galvanic replacement, are of particular

interest to catalysis due to their increased surface area compared to solid nanoparticles. The

galvanic replacement reaction offers a facile method to create hollow structures through a single

reaction step, but the synthesized nanoparticle has a composition and structure dependent on

the precise reaction conditions (Au, et al., 2008; Sun & Xia, 2003). Here, the effect of varying

the concentration of the Au salt is investigated and is shown to result in large changes in

composition, morphology and elemental distribution. The complex elemental distributions

observed are a prime candidate for investigation using STEM-EDX tomography and the three

dimensional elemental reconstructions obtained provide a clear picture of how the elemental

distribution changes with composition.

124
The AgAu nanoparticles used in this study were synthesized by Alexandra Macedo and Pedro

Camargo (University of Sao Paulo), who also performed the catalytic testing of the samples.

XPS analysis of the samples was performed by Sven Schroeder (University of Leeds).

The results of this project were published in Nano Letters in 2014. The original manuscript for

“Correlating Catalytic Activity of Ag-Au Nanoparticles with 3D Compositional Variations” is

presented below. The supplementary information associated with this manuscript is provided

directly following the manuscript.

125
Paper: Correlating Catalytic Activity of Ag-Au Nanoparticles with 3D

Compositional Variations
1 2 1 1,3
Thomas J. A. Slater , Alexandra Macedo , M. Grace Burke , Paul O’Brien , Pedro H. C
2* 1*
Camargo , Sarah J. Haigh

1
School of Materials, University of Manchester, Manchester, M13 9PL, UK

2
Departamento de Química Fundamental, Instituto de Química, Universidade de São Paulo,

São Paulo, Brazil

3
School of Chemistry, University of Manchester, Manchester, M13 9PL, UK

KEYWORDS: Galvanic replacement reaction, hollow bimetallic nanoparticles, heterogeneous

catalysis, scanning transmission electron microscopy, energy dispersive X-ray spectroscopy,

electron tomography

ABSTRACT: Significant elemental segregation is shown to exist within individual hollow silver-

gold (Ag-Au) bimetallic nanoparticles produced from the galvanic reaction between Ag particles
-
and AuCl4 . Three dimensional compositional mapping using energy dispersive X-ray (EDX)

tomography within the scanning transmission electron microscope (STEM) reveals that

nanoparticle surface segregation inverts from Au-rich to Ag-rich as Au content increases.

Maximum Au surface coverage was observed for nanoparticles with approximately 25at% Au

which correlates to the optimal catalytic performance in a three-component coupling reaction

between cyclohexanecarboxyaldehyde, piperidine and phenylacetylene. These results provide

important fundamental insights towards optimizing the properties and reducing the cost of

precious metal catalysts.

The exceptional catalytic properties of Au nanoparticles as compared to the inert nature of the
1
bulk element are well known . The emergence of catalytic activity as particles’ critical

dimensions approach nanometer scale is in part due to the increase in the particle’s surface to
1
volume ratio . Synthetic routes that maximize this property, especially in distinct geometries
2,3
such as hollow shells or cages , have been extensively investigated. Bimetallic noble metal
4
nanoparticles are known to have excellent catalytic performance in a wide range of reactions .

The extent of alloying in these systems is known to influence catalytic activity and selectivity but

126
 5
an understanding of the origin of this effect is lacking . Hollow bimetallic nanoparticles of

accurately tunable size, shape, and composition can be readily prepared by a galvanic
6-8
replacement reaction . The Ag-Au bimetallic system offers the opportunity to both reduce the
9,10
cost of the Au catalysts and enhance catalytic performance . Hollow Ag-Au nanostructures of

various compositions and shapes have been synthesized by galvanic reactions between Au
- 10,11
metal salts, such as AuCl4 , and Ag nanoparticle sacrificial templates . Previous low

resolution TEM imaging and optical spectroscopy observations suggest that this process

initiates with the creation of Ag ions at the surface and the formation of a gold surface layer

which immediately transforms into an Au-Ag alloy because of the fast interdiffusion between
7,12
gold and silver at 100 °C . As the replacement reaction proceeds, dealloying and processes

such as Ostwald ripening induce morphological reconstruction, as well as the formation, and

enlargement, of pinholes in the walls. However, the alloying in individual particles, as well as the

effect of elemental segregation on catalytic properties was previously unknown. Details of

elemental distribution have now been made clear by analytical EDX analysis within the STEM

that has only become possible due to recent advances in EDX detector design.

Ag-Au nanoparticles with controlled compositions and structures (solid vs. hollow interiors) were

synthesized by the galvanic replacement reaction between Ag nanoparticles and different

-
amounts of AuCl4 . The average compositions as determined by flame atomic absorbance

spectrometry were Ag93Au7, Ag82Au18, Ag78Au22 and Ag66Au34. The catalytic activity of the Ag-Au

nanoparticles as a function of composition and structure was investigated for the three-
3
component (A ) coupling reaction between cyclohexanecarboxyaldehyde, piperidine, and

phenylacetylene to form propargylamines (Fig.1a). Propargylamines are useful intermediates in

the synthesis of heterocyclic compounds and represent structural elements of several natural

products and drugs, which has led to their use as intermediates in the total synthesis of such
13,14
molecules . Recently, it has been shown that this transformation can be catalyzed by
15,16
nanostructured Au via a C-H activation mechanism . The yield of propargylamines as a

function of the Au atomic percentage in the Ag-Au nanoparticles is given in Fig.1b. All the

bimetallic and hollow Ag-Au nanoparticles produce higher yields of propargylamine than their

solid and monometallic counterparts (Ag or Au nanoparticles). This observation is in agreement

with the increased surface area in the bimetallic nanoparticles due to the formation of hollow

interiors as shown in the transmission electron microscope (TEM) images (Fig.1c).

127
Figure 1. Ag-Au bimetallic nanoparticles improve propargylamine yield. a, Three component

coupling reaction to form propargylamines. b, Propargylamine yield as a function of the

composition of the nanoparticle catalysts demonstrating a spike in yield for an average

composition of 18at%Au (numerical %yields are noted on graph). c, TEM images of the

synthesized Ag-Au nanoparticles with labels corresponding to the average bimetallic

compositions plotted in b, (i) Ag93Au7, (ii) Ag82Au18, (iii) Ag78Au22 and (iv) Ag66Au34.

More surprising is an exceptionally high peak in yield of 60% at a catalyst composition of 18 at%

Au (Fig.1b). TEM imaging shows that the sizes and morphologies of the nanoparticles are

broadly similar for all bimetallic compositions (Fig.1c), suggesting that the observed peak in the

catalytic results are not explained by size, shape or surface area differences alone.

Nanoscale compositional variations have been observed to strongly affect the catalytic activity
17,18
of platinum (Pt) alloy nanoparticles . Au and Ag form a miscible solid solution over the full

range of compositions, the lattice parameter of Au is just 0.2% larger than that of Ag. Therefore,

structurally sensitive techniques: X-ray diffraction, electron diffraction and high resolution TEM

imaging are unable to distinguish between monometallic, alloyed or composite nanoparticles in

19,20 20
the Ag-Au system. UV-visible spectroscopy and X-ray photoelectric spectroscopy have

been used to elucidate surface composition for bimetallic nanoparticles but cannot provide

information on the local distribution of elements or on the variation of alloying inside the volume

128
of a nanoparticle. High angle annular dark-field (HAADF) scanning transmission electron

microscope (STEM) imaging is able to reveal local compositional variation if there are

sufficiently large differences in the average atomic number of constituent elements. However,

the non-uniform cross section of our particles meant that HAADF STEM imaging of the

nanoparticles in this study was inconclusive (Supplementary Information Fig.S1a).

Through the combination of STEM imaging with electron energy loss (EEL) or EDX

spectroscopy it is possible to obtain chemical information with a spatial resolution that

21,22
approaches 0.1nm for bulk oxides . Using high efficiency EDX spectroscopy, coupled with

aberration corrected STEM imaging, high resolution chemical spectrum images for >30 of the

Ag-Au bimetallic nanoparticles of different compositions were obtained (Fig.2). These images

are the first observation of a reversal in the surface segregation within individual nanoparticles

with increased Au content. Nanoparticles with a low Au content (average composition Ag93Au7)

have an approximately 1nm Au-rich surface layer, whereas those from the sample with the

highest Au concentration (average composition Ag66Au34) exhibited Ag surface segregation.

Nanoparticles of intermediate average composition Ag82Au18 displayed mixed segregation; most

appeared homogeneous in composition but a small fraction (~20%) showed an Au-rich surface

layer. Nanoparticles with intermediate average composition Ag78Au22 were also mostly

homogeneously alloyed except for a small fraction that displayed Ag surface segregation. The

apparent surface segregation behavior for an individual nanoparticle correlates well with its

actual Au composition measured by processing the EDX sum spectra obtained over the entire

field-of-view for each spectrum image (Fig.2e). In order to gain a better understanding of the

surface segregation behavior, line profiles have been extracted perpendicular to the surface of

the elemental maps (Fig.2a-d). Such two dimensional elemental images can be difficult to

interpret because different compositions overlap when projected along the electron beam

direction. For example, Fig.3c shows a line compositional profile extracted from the 2D

elemental maps in Fig.3a,b in which Ag appears to be clearly segregated to the surfaces for the

left-hand side of the nanoparticle ring but not for the right-hand side.

129
Figure 2. STEM EDX mapping showing the inversion of surface segregation behaviour for Ag-

Au nanoparticles as the Au content is increased. EDX maps for representative nanoparticles

from populations with average stoichiometry (a) Ag93Au7, (b) Ag82Au18, (c) Ag78Au22 and (d)

Ag66Au34. The actual composition of each nanoparticle shown inset was measured using the

summed EDX spectrum for the whole field of view of each spectrum image. The chemistry of 20

nanoparticles with a range of compositions was mapped and the structure classified as to

whether the particle appeared to show a gold-rich surface, a silver rich surface or no surface

segregation (alloyed composition). The average nanoparticle composition was then measured

from the summed spectral image and the results are shown in e. The change from gold to silver

surface segregation is found to correlate with the gold content of the nanoparticle.

130
Figure 3. 3D chemical mapping of an Ag-Au nanoparticle with a high Au content showing Ag

surface segregation. Two dimensional EDX maps acquired at zero degrees of tilt showing the

distribution of a, Au and b, Ag. c, Compositional variation extracted right to left from the 2D

maps at the position of the line shown in a and b. 2D EDX maps extracted from the 3D

tomograms showing the distribution of d, Au and e, Ag. f, Compositional variation extracted right

to left from the 2D slices shown in d and e. The line profile extracted from the tomographic data

slice more clearly reveals the silver surface segregation due to the improved resolution in the z

direction. g, Surface visualization of the Au (red) and Ag (green) tomograms revealing the silver

surface segregation (see supplementary movie S1). The average composition measured for the

whole nanoparticle is given inset. The scale bars in a, b, d and e are 10nm.

STEM imaging can be extended to provide three-dimensional information through electron

tomography, in which two-dimensional images are acquired at a range of sample tilts in a

method similar to X-ray tomography. Electron tomography using HAADF STEM imaging has
23
proven a useful tool in the analysis of nanoparticle size, shape and distribution . Energy filtered
24 24,25
transmission electron microscopy (EFTEM) and STEM imaging with EDX spectroscopy

can provide chemical sensitivity to electron tomographic reconstructions but both approaches

are challenging. New EDX detector designs (see Supplementary Information) combined with the

increased currents and small probe size obtainable in aberration corrected STEM instruments

mean EDX tomography is now an increasingly viable tool for nanoparticle characterization.

131
In order to better understand the surface segregation behavior of the nanoparticles, EDX

tomograms were compared for a nanoparticle exhibiting Ag-surface segregation with another

showing evidence of Au-rich surface segregation. The Ag surface-segregated nanoparticle

reveals a fairly uniform surface layer approximately 2 nm thick (Fig.3d, e). The compositional

line profile from the tomogram (Fig.3f) reveals the benefit of the three dimensional data set for

nanoparticle surface analysis, more clearly demonstrating the segregated surface compared to

the profiles obtained from the two dimensional data (Fig.3c). This is due to the improved

resolution in the z-direction of the three dimensional data, removing the contribution of the top

and bottom surfaces from the line profile data. Processing of the EDX sum spectrum revealed

an average composition of Ag60Au40 for the particle shown in Fig.3, close to the average

composition of the whole sample (Ag66Au34). Fig.3f shows a surface visualization of the Au and

Ag tomograms together, which clearly demonstrate the surface segregation of Ag (see

supplementary video S1 for a rotating 3D render). In contrast, the tomogram of a nanoparticle

taken from the low Au sample Ag93Au7 shows clear evidence of a thinner (~1nm) non-uniform

Au surface segregation present on both the inner and outer surfaces of the hollow particle

(Fig.4) (see supplementary video S2 for a rotating 3D render).

Figure 4. EDX Tomogram of nanoparticle with average composition Ag93Au7 with linescan. a,

Au and b, Ag slices through the respective tomograms. Surface segregation of Au can be seen

in the slice through the Au tomogram. c, Linescan through slices a and b showing the

distribution of Ag and Au. Surface segregation of Au is observed although there appears to be a

thin surface of Ag outside of the Au. d, Surface visualization of the Au (red) and Ag (green)

tomograms displaying the surface segregation of Au. The scale bars in a and b are 10nm.

132
The 2D compositional mapping and 3D tomographic data show a clear inversion in the surface

composition for these particles as the Au content is increased. Previous theoretical and lower

spatial resolution experimental studies have suggested that particles are either homogeneously
6,7,26,27
alloyed (due to the lower Ag-Au bond energy compared to either Au-Au or Ag-Ag ) or
28
have Ag segregated to the surface as a result of this element’s lower surface energy .

Our results demonstrate that the Ag and Au distribution in the nanoparticles strongly depends

upon the composition of the particle. At low Au content (<18 at%) galvanic replacement initially

proceeds to give Au substitution for Ag at the surface, forming a non-continuous Au coating. As

more Ag is replaced by Au, there is an increased driving force to minimize the particle’s

intermetallic bond energy and surface energy which produces alloying and finally, at higher

concentrations of Au, Ag surface segregation. The maximum catalytic yield is observed at 18

at% Au, close to the point at which the particles change from Au surface segregation to a

homogeneously alloyed composition (Fig. 2e). This is consistent with the observation that Au
3 29
provides a greater yield than Ag for A coupling reactions . It is feasible that the greater

electronegativity of Au compared to Ag allows electrons to be transferred from Ag to Au,

increasing the electron density of the surface relative to that which would be found in either Ag

or Au monometallic systems and improving C-H bond activation of alkynes.

Our results provide direct 3D observation of an inversion in the segregation of silver and gold for

the first time. This approach is not confined to galvanic replacement reactions but could be

applied to nanomaterials synthesized via many other chemical and physical processing routes.

There are other examples of bimetallic systems which have shown a peak in catalytic

performance for a specific composition. In the absence of suitable local compositional

information a peak in mass specific activity has been attributed solely to a change in the bulk
30
composition of the nanoparticle . Our results demonstrate that segregation in bimetallic

particles is influenced by the synthetic route and is not simple to predict. Furthermore, the

experimental results from this study provide direct experimental evidence that nanoscale

surface segregation can influence nanoparticle properties, providing insights for the improved

design and understanding of nanoparticle synthesis and catalytic activities.

133
ASSOCIATED CONTENT

Supporting Information.

Additional information includes details of synthesis, the catalysis of the three-component

coupling-reaction, STEM imaging and tomography. This includes figures demonstrating the

single/poly-crystalline nature of nanoparticles and EDX spectra. Videos of aligned tilt series and

rotating views of isosurface renders are also available. This material is available free of charge

via the Internet at http://pubs.acs.org.

AUTHOR INFORMATION

Corresponding Author

*E-mail: sarah.haigh@manchester.ac.uk and camargo@iq.usp.br

Author Contributions

PHCC and AM synthesized the nanoparticles, performed preliminary TEM imaging and

measured catalytic data. TJAS and SJH performed the STEM imaging and EDX analysis and

wrote the paper. TJAS performed the 3D reconstructions. All authors contributed to the

interpretation and commented on the manuscript.

Notes

The authors declare no competing financial interest

ACKNOWLEDGMENTS

TJAS, POB and SJH thank the UK Engineering and Physical Sciences Research Council,

NowNANO doctoral training centre for funding support. SJH thanks the USA Defense Threat

Reduction Agency (grant number HDTRA1-12-1-0013) and Gates Foundation for funding

support. PHCC and AM thank FAPESP and CNPq for funding support (grant numbers

2011/06847-0 and 471245/2012-7, respectively). The authors wish to acknowledge the support

from HM Government (UK) for the provision of the funds for the FEI Titan G2 80-200 S/TEM

associated with research capability of the Nuclear Advanced Manufacturing Research Centre.

ABBREVIATIONS

134
EDX, energy dispersive X-ray; STEM, scanning transmission electron microscope; HAADF,

high-angle annular dark field; EFTEM, energy filtered transmission electron microscopy.

REFERENCES

(1) Herzing, A. A.; Kiely, C. J.; Carley, A. F.; Landon, P.; Hutchings, G. J.,

Identification of active gold nanoclusters on iron oxide supports for CO oxidation. Science 2008,

321 (5894), 1331-1335.

(2) Kim, S. W.; Kim, M.; Lee, W. Y.; Hyeon, T. Fabrication of hollow palladium

spheres and their successful application to the recyclable heterogeneous catalyst for Suzuki

coupling reactions. J. Am. Chem. Soc. 2002, 124, 7642-7643.

(3) Lou, X. W.; Archer, L. A.; Yang, Z. C., Hollow Micro-/Nanostructures: Synthesis

and Applications. Adv. Mater. 2008, 20 (21), 3987-4019.

(4) Sankar, M.; Dimitratos, N.; Miedziak, P. J.; Wells, P. P.; Kiely, C. J.; Hutchings,

G. J., Designing bimetallic catalysts for a green and sustainable future. Chem. Soc. Rev. 2012,

41 (24), 8099-8139.

(5) Toshima, N.; Yonezawa, T., Bimetallic nanoparticles - novel materials for

chemical and physical applications. New J. Chem. 1998, 22 (11), 1179-1201.

(6) Petri, M. V.; Ando, R. A.; Camargo, P. H. C., Tailoring the structure,

composition, optical properties and catalytic activity of Ag-Au nanoparticles by the galvanic

replacement reaction. Chem. Phys. Lett. 2012, 531, 188-192.

(7) Sun, Y. G.; Xia, Y. N., Mechanistic study on the replacement reaction between

silver nanostructures and chloroauric acid in aqueous medium. J. Am. Chem. Soc 2004, 126

(12), 3892-3901.

(8) Skrabalak, S. E.; Chen, J.; Sun, Y.; Lu, X.; Au, L.; Cobley, C. M.; Xia, Y., Gold

Nanocages: Synthesis, Properties, and Applications. Acc. Chem. Res. 2008, 41 (12), 1587-

1595.

(9) Kim, Y.; Hong, J. W.; Lee, Y. W.; Kim, M.; Kim, D.; Yun, W. S.; Han, S. W.,

Synthesis of AuPt Heteronanostructures with Enhanced Electrocatalytic Activity toward Oxygen

Reduction. Angew. Chem. Int. Ed. 2010, 49 (52), 10197-10201.

(10) Bracey, C. L.; Ellis, P. R.; Hutchings, G. J., Application of copper-gold alloys in

catalysis: current status and future perspectives. Chem. Soc. Rev. 2009, 38 (8), 2231-2243.

135
 (11) Cobley, C. M.; Xia, Y., Engineering the properties of metal nanostructures via

galvanic replacement reactions. Mater. Sci. Eng. R-Rep. 2010, 70 (3-6), 44-62.

(12) Sun, Y. G.; Xia, Y. N., Alloying and dealloying processes involved in the

preparation of metal nanoshells through a galvanic replacement reaction. Nano Lett. 2003, 3

(11), 1569-1572.

(13) Miura, M.; Enna, M.; Okuro, K.; Nomura, M., Copper-catalyzed reaction of

terminal alkynes with nitrones – selective synthesis of 1-aza-1-buten-3-yne and 2-azetidinone

derivatives. J. Org. Chem. 1995, 60 (16), 4999-5004.

(14) Jenmalm, A.; Berts, W.; Li, Y. L.; Luthman, K.; Csoregh, I.; Hacksell, U.,

Stereoselective epoxidation of phe-gly and phe-phe vinyl isosteres. J. Org. Chem. 1994, 59 (5),

1139-1148.

(15) Gonzalez-Bejar, M.; Peters, K.; Hallett-Tapley, G. L.; Grenier, M.; Scaiano, J. C.,

Rapid one-pot propargylamine synthesis by plasmon mediated catalysis with gold nanoparticles

on ZnO under ambient conditions. Chem. Comm. 2013, 49 (17), 1732-1734.

(16) Kidwai, M.; Bansal, V.; Kumar, A.; Mozumdar, S., The first Au-nanoparticles

catalyzed green synthesis of propargylamines via a three-component coupling reaction of

aldehyde, alkyne and amine. Green Chem. 2007, 9 (7), 742-745.

(17) Cui, C.; Gan, L.; Heggen, M.; Rudi, S.; Strasser, P., Compositional segregation

in shaped Pt alloy nanoparticles and their structural behaviour during electrocatalysis. Nature

Mater. 2013, 12 (8), 765-771.

(18) Stamenkovic, V. R.; Mun, B. S.; Arenz, M.; Mayrhofer, K. J. J.; Lucas, C. A.;

Wang, G.; Ross, P. N.; Markovic, N. M., Trends in electrocatalysis on extended and nanoscale

Pt-bimetallic alloy surfaces. Nature Mater. 2007, 6 (3), 241-247.

(19) Chen, D. H.; Chen, C. J., Formation and characterization of Au-Ag bimetallic

nanoparticles in water-in-oil microemulsions. J. Mater. Chem. 2002, 12 (5), 1557-1562.

(20) Wang, A. Q.; Chang, C. M.; Mou, C. Y., Evolution of catalytic activity of Au-Ag

bimetallic nanoparticles on mesoporous support for CO oxidation. J. Phys. Chem. B 2005, 109

(40), 18860-18867.

(21) Muller, D. A., Structure and bonding at the atomic scale by scanning

transmission electron microscopy. Nature Mater. 2009, 8 (4), 263-270

136
 (22) D'Alfonso, A. J.; Freitag, B.; Klenov, D.; Allen, L. J., Atomic-resolution chemical

mapping using energy-dispersive x-ray spectroscopy. Phys. Rev. B 2010, 81 (10), 100101.

(23) Midgley, P. A.; Dunin-Borkowski, R. E., Electron tomography and holography in

materials science. Nature Mater. 2009, 8 (4), 271-280.

(24) Mobus, G.; Doole, R. C.; Inkson, B. J., Spectroscopic electron tomography.

Ultramicroscopy 2003, 96 (3-4), 433-451.

(25) Genc, A.; Kovarik, L.; Gu, M.; Cheng, H.; Plachinda, P.; Pullan, L.; Freitag, B.;

Wang, C., XEDS STEM tomography for 3D chemical characterization of nanoscale particles.

Ultramicroscopy 2013, 131 (0), 24-32.

(26) Deng, L.; Hu, W.; Deng, H.; Xiao, S.; Tang, J., Au-Ag Bimetallic Nanoparticles:

Surface Segregation and Atomic-Scale Structure. J Phys. Chem. C 2011, 115 (23), 11355-

11363.

(27) Wei, L. Y.; Qi, W. H.; Huang, B. Y.; Wang, M. P., Surface segregation of Au-Ag

bimetallic nanowires. Comp. Mater. Sci. 2013, 69, 374-380.

(28) Gong, H. R., Electronic structures and related properties of Ag-Au bulks and

surfaces. Mat. Chem. Phys. 2010, 123 (1), 326-330.

(29) Huang, B.; Yao, X.; Li, C.-J., Diastereoselective synthesis of alpha-oxyamines

via gold-, silver- and copper-catalyzed, three-component couplings of alpha-oxyaldehydes,

alkynes, and amines in water. Adv. Synth. Catal. 2006, 348 (12-13), 1528-1532.

(30) Zhang, H.; Jin, M. S.; Wang, J. G.; Li, W. Y.; Camargo, P. H. C.; Kim, M. J.;

Yang, D. R.; Xie, Z. X.; Xia, Y. A., Synthesis of Pd-Pt Bimetallic Nanocrystals with a Concave

Structure through a Bromide-Induced Galvanic Replacement Reaction. J. Am. Chem. Soc. 2011,

133 (15), 6078-6089.

137
Supporting Information - Correlating catalytic activity of Ag-

Au nanoparticles with 3D compositional variations

Methods

Analytical grade chemicals AgNO3 (silver nitrate, Sigma-Aldrich, ≥ 99.0% purity),

HAuCl4·3H2O (hydrogen tetrachloroaurate trihydrate, 48% in gold, Synth), PVP (polyvinyl-

pyrrolidone, Sigma- Aldrich, MW 55,000 g/mol), PVP (polyvinyl-pyrrolidone, Sigma-Aldrich,

MW 10,000 g/mol), EG (ethylene glycol, Vetec, 99.5% purity), and MgSO4·XH2O

(magnesium sulfate, ≥ 98%, Vetec) were used as received. C7H11O

(cyclohexanecarboxyaldehyde, 97%, Sigma-Aldrich), C5 H11N (piperidine, 99%, Sigma-

Aldrich), and C8H6 (phenylacetylene, 98%, Sigma-Aldrich) were purified by distillation

before use. All solutions were prepared using distilled water.

Synthesis of Ag and AgAu nanoparticles

For the synthesis of the Ag nanoparticle templates, PVP (10 g, MW 10,000 g/mol) was

dissolved in 75 mL of ethylene glycol at room temperature, and 400 mg AgNO3 was added to

this solution. The resulting suspension was stirred at room temperature until the complete

dissolution of AgNO3. This mixture was heated up to 100 °C at a constant rate of 1 °Cmin-1

and the reaction was then allowed to proceed for 1.5 h. After, 175 mL of distilled water was

added and the reaction mixture was allowed to cool to room temperature. The suspension

containing the Ag nanoparticles was washed three times with water by successive

centrifugation at 7000 rpm and removal of the supernatant. After washing, the resulting Ag

nanoparticles were suspended in 250 mL of water. This suspension was then used for the

synthesis of AgAu nanoparticles.

The synthesis of bimetallic and hollow AgAu nanoparticles used the galvanic reaction

between Ag nanoparticles and AuCl4-(aq). In a typical procedure, 500 mL of a PVP solution

(1 mg/mL, MW 55,000 g/mol) and 27.8 mL of the Ag nanoparticle suspension were added

into a 1000 mL round bottom flask. This system was heated up to 100 °C for 10 min. Then,

138
 -
0.2 mM AuCl4 (aq) solution was added dropwise and the reaction allowed to proceed at

100 °C for another 10 min. 100, 200, 300, or 400 mL aliquots of a 0.2 mM AuCl4-(aq) solution

were used to prepare AgAu nanoparticles with average compositions of Ag93Au7 (Ag 93at%

Au 7at%), Ag82Au18 (Ag 82at% Au 18at%), Ag78Au22 (Ag 78at% Au 22at%), and

Ag66Au34 (Ag 66at% Au 18at%), respectively. The Au and Ag atomic percentages were

measured by flame atomic absorption spectrometry with a Shimadzu spectrophotometer,

model AA-6300, equipped with an air-acetylene flame. After cooling down to room

temperature, each suspension of AgAu nanoparticles was washed three times with water by

successive rounds of centrifugation at 7000 rpm and removal of the supernatant. After

washing, the AgAu nanoparticles were suspended in 10 mL of water for the catalytic tests.

Catalytic activities towards three-component coupling reaction (A3)

The catalytic activities of AgAu nanoparticles towards the three-component coupling reaction

3
(A ) were screened using cyclohexanecarboxyaldehyde, piperidine, and phenylacetylene as

substrates. A 10 mL suspension containing the AgAu nanoparticles was added to a 50 mL

round bottom flask and sonicated under N2 for 5 min. Then, cyclohexanecarboxyaldehyde,

(168 mg, 1.5 mmol), piperidine (85 mg, 1.0 mmol) and phenylacetylene (153 mg, 1.5 mmol)

were added to the suspension containing the AgAu nanoparticles and stirred under a nitrogen

atmosphere. The Au mol% in the reaction corresponded to 1.6, 1.5, 2.0, and 2.5 for Ag93Au7,

Ag82Au18, Ag78Au22, and Ag66Au34 nanoparticles, respectively (relative to piperidine). This

mixture was heated up to 90 oC and the reaction was allowed to proceed for 12 h under

vigorous stirring. After that, the reaction mixture was extracted with dichloromethane (3 x 20

mL), leaving the AgAu nanoparticles in the aqueous phase. The organic layer was dried over

MgSO4·XH2O and the solvent was removed under reduced pressure. The crude product was

purified by column chromatography using silica gel as the stationary phase and an

appropriate eluent (19:1 hexane:ethyl acetate) to yield the corresponding propargylamine.

The catalytic activities of Ag and Au monometallic nanoparticles were also investigated for

comparison. In this case, a similar procedure was performed. For Ag nanoparticles, 25.4 mL

of the suspension after washing was centrifuged and suspended in 10 mL of distilled water

(Ag mol% corresponded to 5 relative to piperidine). For Au, 300 mL of a suspension

139
containing Au nanoparticles were centrifuged and suspended in 10 mL of distilled water (Au

mol% corresponded to 5 relative to piperidine). The Au nanoparticles were prepared by

dissolving 100 mg of HAuCl4·3H2O in 950 mL of distilled water. This solution was heated to

100 °C. A solution of sodium citrate (50 mL, 0.5 wt%) was then added and the reaction was

allowed to proceed for 60 min. The resulting suspension containing the Au nanoparticles was

cooled to room temperature and stored until use. The product, 1-(1-cyclohexyl-3-phenylprop-

1 13
2-yn-1-yl) piperidine, was identified by H and C NMR spectroscopy and GC-MS

1
spectrometry: H NMR (200 MHz, CDCl3): δ 7.54 – 7.19 (m, 5H), 3.10 (d, J = 9.9 Hz, 1H),

2.75 – 2.26 (m, 4H), 2.08 (t, J = 7.5 Hz, 1H), 1.85 – 0.76 (m, 16H) ppm; 13C NMR (50 MHz,

CDCl3): 131.83, 128.29, 127.72, 123.92, 87.88, 86.25, 64.50, 50.86, 39.70, 31.46, 30.55,

-83
26.93, 26.41, 26.23, 24.84 ppm. MS: m/z (%) 198 (M , 100), 153 (3), 128 (9), 115 (65), 103

(4), 91 (7), 77 (5), 69 (4), 55 (11), 41 (38). Analytical thin layer chromatography (TLC) was

carried out by using silica gel 60 F254 pre-coated plates. Visualization was accomplished

with vanillin (vanillin 0.01 g/mL in AcOH/H2SO4 (99:1) solution) or ninhydrin as color reagent

(ninhydrin 0.005 g/mL in EtOH (96%) solution). The gas chromatography analyses were

performed on Shimadzu GC2014 equipment, with flame ionization detection (FID), N2 as

1
carrier gas and equipped with a DB-5 column (30m x 0.25mm x 0.25mm). The H NMR (200

MHz) and 13C NMR (50 MHz) spectra were recorded on Bruker AC 200 spectrometer using

tetramethylsilane (TMS) as the internal standard. Chemical shifts were reported in parts per

million (ppm, δ) downfield from the TMS. Low-resolution mass spectra were recorded on

Shimadzu GC-17A coupled with QP5050A MS, using HP-5MS column (30m x 0.25mm x

0.25mm).

Transmission electron microscope imaging.

Nanoparticles were prepared for (S)TEM imaging by placing a few drops of a suspension

onto a holey carbon coated copper support grid (200 or 300 mesh) and allowing to air dry.

TEM imaging was performed using a JEOL 1010 transmission electron microscope operated

at 80 kV. STEM and EDX spectrum imaging were performed using a probe-side aberration-

corrected FEI Titan G2 80-200 S/TEM operated at 200 kV. STEM images were collected

using a convergence angle of 18 or 26 mrad and a high angle annular dark-field (HAADF)

140
detector with an inner angle of 55 mrad. EDX compositional analysis was performed using

the Super-X detector configuration (4 x 30 mm2 SDDs) with a solid angle of ~0.7 srad and a

beam current of ~1.2 nA. The two dimensional elemental spectrum images shown in Figure 2

were acquired with a total dose of 1.0 μC (5.9x10 electrons), an acquisition time of ≈10

minutes and a dwell time of 15μs per pixel. The compositional analysis for each nanoparticle

was obtained from an EDX sum spectrum (obtained by a summation of all spectra contained

in the whole field-of-view for a spectrum image containing only a single nanoparticle) and

performing quantification for the gold/silver ratio with background fitting and standard Cliff-

Lorimer “k-factors”. For display purposes all spectrum images were processed using a 3-pixel

smoothing window in the Bruker Esprit software.

Tomographic reconstruction.

The three dimensional tomographic reconstructions (HAADF-STEM images and EDX

spectrum images) were simultaneously acquired using the above imaging conditions. Data

was acquired at 5° intervals using a Fischione 2020 single tilt tomography holder and a

specimen tilt range of ±70°. Acquisition times at each angle were varied to compensate for

the tilt induced shadowing of the detector with the acquisition time as a function of tilt angle

given in Fig. S3. Bruker Esprit software was used to acquire and process the EDX spectrum

images.

Compositional maps were extracted and processed using a 3-pixel smoothing window in

Esprit and exported as separate RGB files for the Au and Ag Lα maps. Maps were then

converted to 16-bit unsigned integer files for visualization using Gatan DigitalMicrograph™

and ImageJ software. Maps were formatted and aligned in FEI Inspect3D software with

alignment being performed via cross- correlation calculated from the simultaneously acquired

HAADF images. Tomographic reconstruction was performed using the Inspect3D SIRT

algorithm using 5 iterations. Line profiles of both 2D maps and slices through the 3D data

cubes were processed in ImageJ. Further visualization of the tomographic data was

performed using the Avizo software package.

141
HAADF STEM and EDX

Atomic resolution imaging of the nanoparticles using HAADF STEM has revealed that the

particles have a complex hollow morphology for the whole range of compositions studied in

this work. Nanoparticles were found to have the expected face centered cubic structure with

both single crystal (Fig. S1a,b) and polycrystalline structures found for all compositions. The

extent of surface segregation appeared independent of whether nanoparticles were single

crystal or polycrystalline. Spectra obtained from the four particles shown in Fig. 2 are

displayed in Fig. S3.

Supporting Figure S1. HAADF-STEM images of AgAu nanoparticles with different

compositions. HAADF-STEM images showing typical nanoparticles from populations of

average compositions (a) Ag93Au7 , (b) Ag82Au18 , (c) Ag78Au22, and (d) Ag66Au34 .

Fourier Transforms of the images shown inset illustrate that the nanoparticles are largely

crystalline. Distinguishing surface segregation of Au or Ag from the HAADF images is non-

trivial. (e)-(h) the corresponding HAADF images (a)-(d) with overlay shading illustrating the

different crystallographic regions and presence of both single crystal and polycrystalline

nanoparticles. The spots in the inset Fourier transforms illustrate which spots

correspond to the different regions.

142
Supporting Figure S2. Summed EDX spectra extracted from the EDX spectral data cubes

used to obtain the compositional images in Fig 2 (a-d). The change in

composition through the nanoparticles is illustrated through the change in the relative

heights of the Au and Ag peaks for each nanoparticle. The Cu peak is due to the copper

grid on which the particles are deposited. The Fe peak is due to the pole-piece of the

microscope and the O peak is confirmed to not be due to oxidation of nanoparticles as

mapping of O matches the HAADF image rather than being localized at the surface. The

lack of a Cl peak in the spectra of nanoparticles confirms no AgCl present in the formed

nanoparticles.

We have used a scanning transmission electron microscope fitted with a high brightness

source, aberration corrected probe optics and symmetrical arrangement of large angle EDX

spectrometersS1 in order to demonstrate EDX tomographic reconstruction with a resolution

close to 1 nm. However, even for this microscope a Fischione 2020 high tilt tomography

holder suffers from significant detector shadowing due to the sides of the holder blocking the

path of X-rays emitted from the sample to the detector. Fig. S3a shows that shadowing

changes the detected sample counts by a factor of three when using standard TEM grids with

a commercial high-tilt tomography holder.

143
Supporting Figure S3. Time-varied acquisition scheme for recording tomographic STEM

EDX data. a, X-ray sample counts as a function of specimen tilt angle for the same

particle using a uniform acquisition time of 300 seconds for each tilt angle. b, Acquisition

time as a function of tilt angle necessary to correct for measured variation in detector

counts shown in (a). c, X-ray sample counts as a function of specimen tilt angle obtained

when the calculated non-uniform acquisition times (given in b) are applied to tomographic

data acquisition.

To overcome this limitation we have employed a novel acquisition scheme under which the

time to acquire each EDX map is varied in order to compensate for detector shadowing at

each angle. The time-varied approach can be justified considering the equation describing

the intensity of the EDX signal for element A, IA, (assuming no absorption or fluorescence

S2,S3
takes place within the sample) :

𝐼𝐴 = 𝑁𝐴 𝑄𝐴 𝜔𝐴 𝑎𝐴 𝑡𝐷𝑒 (Ω/4π)𝜀𝐴

where NA is the number density of element A, QA is the ionization cross-section, ωA is the

fluorescence yield, aA is the relative transition probability of the peak in question, t is the

thickness of the sample, De is the electron dose, Ω/4π is the proportion of the whole 4π solid

angle that the detector subtends and εA is the detector efficiency. Shadowing of element A

can be included as a factor in detector efficiency εA. To reconstruct the number density of

element A (NA) accurately then IA should not vary as a function of any other parameter as a

function of specimen tilt angle. This requirement is clearly violated through variation in

detector shadowing as a function of tilt angle, i.e. εA varies with specimen tilt angle. To

counteract this variation one of the other parameters in equation 1 could be altered to render

the product of all parameters other than NA constant over all tilt angles. The only parameter

144
that can be altered independently from the sample to normalize the equation is De, the

electron dose, which can be varied straightforwardly by altering the acquisition times of each

spectral image. This approach assumes that the entire nanoparticle is shadowed equally at

each tilt angle; a reasonable approximation considering the small size of the nanoparticles

relative to the detector. Using the calibrated acquisition times shown in Fig.S3b signal counts

for Au and Ag X-ray lines of a single nanoparticle are observed to remain constant (±10%).

An example of the HAADF STEM images, Au and Ag elemental maps for one of the

tomographic data sets obtained is shown in Fig.S4. HAADF-STEM images of the AgAu

particle displayed in Fig.3 acquired before and after tilt-series acquisition are shown in Fig.S4

along with an atomic resolution HAADF-STEM image of the same particle.

Supporting Figure S4. HAADF-STEM images of AgAu nanoparticles. a, before and b, after

tilt series acquisition showing little change in structure of the nanoparticle. c, atomic

resolution HAADF-STEM image (with inset FFT) acquired after tilt-series acquisition showing

an indistinguishable Ag/Au FCC lattice.

X-ray Photoelectron Spectroscopy (XPS)

XPS analysis was performed for (i) the pure Ag precursor nanoparticles and for

each of the synthesised nanoparticle ensembles having mean compositions of (ii)

Ag93Au7, (iii) Ag82Au18, (iv) Ag78Au22 and (v) Ag66Au34, as measured by Flame

Absorbance Spectroscopy (FAS). XP spectra were recorded with a Kratos Axis Ultra

instrument, employing a monochromatic Al Kα source (1486.69 eV), a hemispherical

analyser with a hybrid (electrostatic and magnetic) lens system, charge

neutralization by filament-generated magnetically channelled low-energy electrons, and a

delay line detector (DLD). Samples were prepared by depositing 50 µL of

145
nanoparticle suspension into aluminium lids for differential scanning calorimetry

investigations (DSC Consumables Inc., #84005), followed by evaporation of the solvent.

Experiments were performed while operating the X-ray source with a power of 180 W (15

kV and 12 mA), with a pressure of 10−8 mbar during analysis. The instrument was

operated in CAE (constant analysis energy) mode, with a pass energy of 20 eV for high

resolution scans of the photoemission from individual core levels, with a

calibrated intensity/energy response and transmission function.S4 High resolution

spectra were measured within the spectral range of interest with 0.1 eV steps and 600 ms

dwell time per data point. Repeats were carried out to check for radiation damage. Survey

spectra for elemental analysis were obtained using an analyser pass energy of 80 eV, with

0.5 eV steps and 300 ms dwell time per data point.

Analysis of the data was carried out with Casa XPS software.S5 A Shirley background was

S5,S6
used in all curve-fitting along with a GL(80) lineshape (80% Gaussian, 20% Lorentzian

S5
using the Gaussian/Lorentzian product form). Samples were referenced to the

photoemission from aliphatic hydrocarbons at 285 eV.S7,S8 Repeatability of the peak

positions was within ±0.1 eV. The results of the analysis of the Ag and Au components in

the spectra are summarised in Supplementary Table S1. The high resolution Au 4f and Ag 3d

photoemission lines are displayed in Supplementary Figure S5.

146
 Elemental

analysis Surface
Bulk BE FWHM Area BE FWHM Area
Surface Au 4f7/2 Ag 3d5/2
Ag 0 0 - -(eV) -(eV) - (eV) 0.85
367.9 (eV) 1368
(at%
Ag93Au 7 (FAS)
7 Au) (XPS)
19 enrichment
2.7 83.21 0.85 (a.u.)
213 367.9 0.80 (a.u.)
1007
Ag82Au 1 18 35 1.9 83.44 0.78 626 367.9 0.80 1739
(Surface/Bulk
Ag78Au 2 22 48 2.2 83.69 0.85 4025 368.0 0.85 4998
8 66Au 3 34
Ag 48 1.4
) 83.66 0.81 740 368.1 0.90 1017
2
4

Supporting Table S1. Comparison of the elemental analysis by FAS (overall bulk

composition) and quantitative XPS (surface and near-surface composition). Enrichment in

Au in the surface region of the particles is evident from the ratio between the Au

concentrations detected by XPS and FAS. The results of the chemical shift analysis via curve

fitting analysis of the high resolution Au 4f and Ag 3d photoemission signals are summarised

in the two main column on the right side of the table. For each photoemission line the core

level binding energy (BE), full width at half maximum (FWHM) and area of the Gauss-Lorentz

peak are reported.

XPS data confirmed the expected presence of Au and Ag. Table S1 compares the

quantification for the Au/Ag ratio obtained via XPS to that measured using FAS. XPS is highly

surface sensitive, with the great majority of the signal coming from within a few nm of the

surface (the inelastic mean free paths for Au 4f and Ag 3d electrons are approximately

S9
1.6 nm and 1.5 nm, respectively). These results show a greater Au content than was

measured for the bulk, demonstrating that a significant amount of deposited Au remains at

or at least near the surface for mean compositions up to 34 at% Au. This Au enrichment is

most pronounced for the lowest Au composition (7 at%) and smallest for the highest Au

composition (34 at%). We did not obtain data for mean compositions above 34 at% Au,

preventing an observation of a transition to Ag surface segregation which, according to

our STEM EDX measurements, occurs for individual particles above ~40 at% Au (Fig. S2e),

but which due to the spread in composition is not likely to be observed for the

ensemble population until significantly higher Au contents.

147
Supporting Figure S5. Au 4f (left) and Ag 3d (right) photoemission spectra of the

AgxAu100-x nanoparticle preparations and best fit models obtained using 80%

Gaussian/20% Lorentzian lineshapes and a Shirley background. The spectra are stacked

according to Au content, from 0 at% (top) to 34 at% (bottom). The FAS-determined overall

composition of the nanoparticles is indicated within the figure for each spectrum.

The core level binding energies (BEs) associated with the Au 4f7/2 and Ag 3d5/2 emission

lines of the nanoparticles indicate strongly that Au and Ag are present in the metallic state,

demonstrating that the change in catalytic performance is not related to a change in

oxidation state of either element. This agrees with STEM-EDX results in which no

enhanced oxygen signal was observed for the surfaces of the nanoparticles. The

S4
observed Au 4f7/2 BEs are significantly lower than the bulk Au metal value (83.96 eV ),

which is in good agreement with previously reported values for Au, Ag and their alloys in

S10-S12
nanoparticulate form. The Au 4f7/2 BEs are lowest for the samples with the

smallest Au content, indicating that the observed catalytic activity is associated with a

significant electronic perturbation of the Au component. All observed Au 4f7/2 BEs as well as

the overall smaller variations in the Ag 3d5/2 BEs are in line with previously reported

S11,S13,S14 Overall,
chemical shifts associated with Au/Ag alloying and de-alloying.

expected variations in core level BEs due to alloying are of the same magnitude as surface

core level shiftsS15 and the shifts associated with nanoparticle size and morphology.S10

148
Supporting References

S1 von Harrach, H. S. et al. An integrated Silicon Drift Detector System for FEI Schottky

Field Emission Transmission Electron Microscopes. Microscopy and Microanalysis 15, 208-

209 (2009).

S2 Watanabe, M. & Williams, D. B. The quantitative analysis of thin specimens: a

review of progress from the Cliff-Lorimer to the new zeta-factor methods. Journal of

Microscopy- Oxford 221, 89-109 (2006).

S3 Cliff, G. & Lorimer, G. W. QUANTITATIVE-ANALYSIS OF THIN SPECIMENS.

Journal of Microscopy-Oxford 103, 203-207 (1975).

S4 Seah, M. P.; Gilmore, L. S.; Beamson, G. XPS: Binding energy calibration of

electron spectrometers 5 - Re-evaluation of the reference energies. Surf. Interface Anal. 1998,

26, 642-649.

S5 Fairley, N.; Carrick, A. The CASA Cookbook: Recipes for XPS Data Processing, pt.

1; Acolyte Science: Knutsford, Cheshire, 2005.

S6 Briggs, D.; Seah, M. P.; Sherwood, P. M. A. In Practical Surface Analysis, Volume

1: Auger and X-ray Photoelectron Spectroscopy, 2 ed.; Briggs, D., Seah, M. P., Eds.; Wiley:

Chichester, 1990.

S7 The XPS of Polymers Database; 1 ed.; Surface Spectra Ltd: Manchester, UK, 2000.

S8 Stevens, J. S.; Schroeder, S. L. M. XPS Analysis of Saccharides and Their X-ray

Induced Degradation. Surf. Interface Anal. 2009, 41, 453-462.

S9 Tanuma, S.; Powell, C. J.; Penn, D. R. Calculations of Electron Inelastic Mean Free

Paths II.Data for 27 Elements over the 50-2000 eV range. Surf. Interface Anal. 1991, 17,

911-926.

S10 Radnik, J.; Mohr, C.; Claus, P. On the origin of binding energy shifts of core levels of

supported gold nanoparticles and dependence of pretreatment and material synthesis. Phys.

Chem. Chem. Phys. 2003, 5, 172-177.

149
S11 Bzowski, A.; Kuhn, M.; Sham, T. K.; Rodriguez, J. A.; Hrbek, J. Electronic structure

of Au-Ag bimetallics: Surface alloying on Ru(001). Phys. Rev. B 1999, 59, 13379-13393.

S12 Mason, M. G. Electronic Structure of Supported Small Metal Clusters. Phys. Rev. B

1983, 27, 748-762.

S13 Tyson, C. C.; Bzowski, A.; Kristof, P.; Kuhn, M.; Sammy-Naiken, R.; Sham, T. K.

Charge Redistribution in Au-Ag Alloys from a Local Perspective. Phys. Rev. B 1992, 45,

8924-8928.

S14 Watson, R. E.; Hudis, J.; Perlman, M. L. Charge Flow and d Compensation in Gold

Alloys.Phys. Rev. B 1971, 4, 4139-4144.

S15 Egelhoff, J. Core-Level Binding-Energy Shifts at Surfaces and in Solids. Surf. Sci.

Rep. 1987, 6, 253-415.

150
 7. Summary and Future Work

Energy dispersive X-ray spectroscopy allows elemental mapping of nanoparticles and

quantification of the composition of individual nanoparticles or ensembles of nanoparticles.

Quantification of EDX spectra is a straightforward procedure if nanoparticles are small enough

that absorption and fluorescence have a negligible effect. This thesis demonstrates that when

X-ray absorption is not negligible, X-ray absorption correction can be applied for well-defined

nanoparticle geometries, as is demonstrated in chapter 3 for spherical particles. The absorption

correction factor derived here is applicable to all approximately homogeneous spherical

nanoparticles and can be used to provide an estimate of X-ray absorption in approximately

spherical nanoparticles. In the future, this type of absorption correction could be easily

automated so that absorption correction is applied to spectrum images of spherical particles,

taking only inputs on the parameters defining the elements present and detector geometry.

However, when considering nanoparticles of a complex geometry it may be necessary to

provide a three dimensional reconstruction of the nanoparticle in order to calculate X-ray

absorption accurately. For nanoparticles of approximately homogeneous composition X-ray

absorption correction could be applied through acquisition of a HAADF tomogram and

calculation of X-ray path length at each voxel.

Nanoparticles of inhomogeneous elemental distribution may require a STEM-EDX or EELS

tomogram to attempt to quantify composition at each voxel. Although there has been recent

progress in the area of STEM-EDX tomography, quantitative results have not yet been

demonstrated. Currently, EDX tomography can be used to provide qualitative data visualising

the distribution of single elements within a nanoparticle or collection of nanoparticles, as

demonstrated in chapter 6 and (Genc, et al., 2013; Goris, et al., 2014). Accurate quantification

of STEM-EDX tomograms will require higher counts per pixel in spectrum image projections

than those presented here and are also complicated by artefacts introduced by the absorption

of X-rays. Absorption of X-rays within the sample will cause variations in elemental maps away

from the true elemental distribution, although the use of four detectors equally spaced around

the optic axis compensates for this somewhat if none of the detectors are shadowed. STEM-

EDX is often referred to as semi-quantitative, partly due to the inaccuracy of absorption

correction in many samples. In the future, STEM-EDX tomography has the potential to calculate

151
the absorption of X-rays accurately through a sample, so that a more accurate quantification

can be achieved in two or three dimensions. Iteratively calculating the absorption of X-rays

through a STEM-EDX reconstruction and applying this to the projections acquired should

eventually converge to a solution of the true composition, neglecting the effect of X-ray

fluorescence.

X-ray absorption in the sample holder rather than the sample results in shadowing of the X-ray

detectors, which this thesis has demonstrated can be an important effect in the accurate

reconstruction of an EDX tomography tilt series. Furthermore, it is demonstrated that detector

shadowing can be characterised through acquisition of tilt series’ of spherical particles, in

addition to the modelling of detector shadowing shown in previous literature (Yeoh, et al., 2015).

Here, the variations in intensity caused by variations in detector shadowing are compensated

for by varying the acquisition time of each projection, so that X-ray count rate is maintained

(Chapters 5 and 6). This thesis also demonstrated that variations in shadowing across an image

may be large if using detectors on only one side of the tilt axis, but when using detectors on

either side of the tilt axis can be largely ignored. Novel positioning of detectors, such as an

annular detector above the sample holder (Zaluzec, 2009), or one parallel to the tilt-axis have

the potential to reduce this effect in single detector systems.

Ultimately, beam damage of the sample may still limit the X-ray counts at each pixel in STEM-

EDX tomography as the cross-section for X-ray production is much lower than the cross-

sections associated with image formation in EELS or HAADF-STEM. Use of advanced

reconstruction algorithms will allow good fidelity reconstructions from few projections (Goris, et

al., 2013; Leary, et al., 2013), which will allow a larger per image dose if the total dose is limited

due to beam damage of the sample. A larger dose per image will be particularly beneficial for

EDX tomography, as STEM-EDX spectrum imaging suffers from much poorer signal to noise

than imaging via EELS or HAADF-STEM. Another way to limit damage may be to use cryo-

holders that keep the sample at a temperature below -170 °C (Baker & Rubinstein, 2010; Unger,

2001), although these holders typically have high sides that shadow EDX detectors to a large

degree.

The optimal conditions for STEM-EDX tomography are highly dependent on the sample

analysed, but it is possible to make some recommendations on the two-dimensional spectrum

152
images acquired. In all studies it is important to maximise the counts at each pixel (single pixel

spectra are displayed in Appendix A) to improve the signal to noise ratio. For this reason,

maximising pixel size without degrading resolution is highly recommended for each spectrum

image, either through set up at acquisition or through binning of pixels post-acquisition. It is

advisable to calculate an estimate of resolution before acquisition, through consideration of the

factors discussed in section 2.4, so that pixel size can be set close to the beam diameter. The

maximum dose per each tilt angle should then be used, which is dependent on the number of

projections acquired.

The number of projections necessary is entirely dependent on the reconstruction algorithm used

but here it was found that larger than 5° tilt increments restricts the fidelity of reconstructions

using SIRT. The SIRT algorithm appears to enhance the inherent noise present in STEM-EDX

data when using successive iterations and choice of the number of iterations should be made

carefully, further discussion of the number of SIRT iterations is presented in Appendix A.

EDX spectrum imaging and tomography has been applied here to the study of AgAu

nanoparticles synthesized by the galvanic replacement reaction. A reversal in the surface

segregation of Au and Ag is observed with increasing Au content and the composition at the

point at which Au surface segregation ceases appears to correlate with that at which catalytic

yield in a three component coupling reaction is at maximum. STEM-EDX tomography appears

to be an excellent method to investigate the 3D compositional variations in this type of bimetallic

nanoparticle and future work could apply the approach developed here to other bimetallic

systems containing heavy elements such as Au and Pt. It is also feasible in the future that

STEM-EDX tomography could be applied to in-situ studies of nanoparticles in the STEM,

whether this is in-situ heating (Gontard, et al., 2014), mechanical testing (Dehm, et al., 2012; Yu,

et al., 2015) or the observation of reactions in liquids and gases (de Jonge & Ross, 2011;

Sharma, 2005; Wang, et al., 2015). In particular, the development of in-situ heating (Lewis, et

al., 2014a) and liquid cell (Lewis, et al., 2014b) holders with low-profile designs specifically for

STEM-EDX analysis will greatly help in this work.

153
7.1. Correlative Tomography

The previous chapters have detailed the application of EDX tomography to freestanding

nanoparticles, in which the length scale of interest is 1-100 nm, ideally suited to investigation via

the STEM-EDX technique. However, the extension of electron tomography to probe nanoscale

features in bulk material samples raises the questions of whether the volume analysed in

electron tomography is representative of the bulk, or whether the volume chosen is located at

the correct position of interest.

Many material studies have used a range of three dimensional imaging techniques in order to

investigate structural features that span multiple length scales from the atomic to the

macroscopic (Li, et al., 2015; Lopez-Haro, et al., 2013; Tariq, et al., 2011). However, only

recently has it been possible to apply a range of techniques on the same sample volume at

multiple length scales, which can then be spatially correlated to ensure the high resolution

techniques are acquired at a precisely known location (Burnett, et al., 2014a; Burnett, et al.,

2014b; Merkle, et al., 2014).

Over the past six months I have undertaken initial studies on the correlation of STEM-EDX

tomography data with that of X-ray nanotomography. The sample I have studied is that of a

cavitated grain boundary in a AISI-type 316H stainless steel extracted from within a weld region

of an AGR power station steam header (Burnett, et al., 2014a). Nanoparticles precipitated at the

grain boundary are characterised with STEM-EDX tomography and are located precisely at a

position defined through correlation of the different datasets obtained.

In order to perform high-resolution X-ray nanotomography the location of a cavitated grain

boundary was identified using X-ray microtomography data undertaken by Tim Burnett and

Robert Bradley (University of Manchester) (Burnett, et al., 2014a). The X-ray microtomography
2
was performed on a matchstick sample of length of 0.74 mm and cross-section of 0.36 mm .

The presence of a cavitated grain boundary was confirmed through extraction of a sub-surface

cross-section at the specified site. Subsequently, a ~120 μm high pillar of 30 μm diameter was

isolated from the matchstick sample through annular milling in a plasma-FIB (Fig 7.1a),

undertaken by Tim Burnett. The plasma-FIB provides a rapid method of machining volumes at

the scales of tens of micrometres, providing milling rates up to fifty times greater than

154
conventional Ga+ ion beam milling (Delobbe, et al., 2014). The pillar was attached to a pin

sample holder for further analysis after lift-out (Fig 7.1b).

Figure 7.1. a) SEM image within the plasma-FIB dual beam instrument showing the pillar

sample after annular milling. b) The 30 μm diameter pillar attached to a pin for further analysis.

c) An SEM image of the cross-sectional slab cut from the nano-CT pillar before liftout. d) An ion-

beam image of one polished face of the slab attached to a TEM grid. e) Ga ion-beam image of

the “block” cut from the cross-sectional slab and attached to the lift-out needle. f) SEM image of

the final STEM needle.

X-ray nanotomography reveals the presence of three grain boundaries within the pillar, two

decorated with cavities and one visible only through small intensity differences due to large

intergranular carbides (Fig 7.2). The X-ray nanotomography was performed by Robert Bradley

(University of Manchester). The three grain boundaries meet at a triple junction in the middle of

the pillar, which can be visualised in three dimensions through manual tracking of intensity

variations due to the intergranular carbides (Fig 7.2). The location and morphology of individual

cavities along the cavitated grain boundary are fully determinable through visualisation of the X-

ray CT data. The large carbides that are visible as slight changes in intensity along the grain

boundary are also present along the cavitated grain boundary but their morphology cannot be

determined due to the resolution limits of the X-ray CT.

155
Figure 7.2. X-ray nanotomography of a pillar of stainless steel 316. Volume visualisation

displaying an isosurface rendering of the outer surface of the pillar, traces along the three grain

boundaries and the surface of the segmented cavities. Also displayed is a slice through the X-

ray nanotomography reconstruction displaying the three grain boundaries and the intensity

changes associated with cavities and large intergranular carbides.

Visualisation of the X-ray CT data allowed identification of a preferred orientation for extraction

of a cross-sectional “slab” from which to remove needle shaped samples for high resolution 3D

STEM tomography. This orientation was chosen as favourable because in this geometry the full

range of X-ray CT slices were seen to contain all three grain boundaries relaxing the

requirement for precise vertical positioning of the slab. A slab approximately 5μm thick was

lifted out from the pillar (Fig 7.1c) and attached to a “pronged” half-moon TEM grid (Fig 7.1d).

The sides of the slab were polished to allow the acquisition of SEM images, which could be

manually correlated to slices through the X-ray CT data. The removal of the slab sample was

undertaken by Remco Geurts (FEI, Eindhoven).

A block with length 22 μm and a square 2 μm x 2 μm cross section was removed from the slab

(Fig 7.1e) using the FIB-SEM and attached to the top of an on-axis tomography pin. Each of the

four faces of the block were polished to allow SEM images to be taken of the grain boundary at

156
each face (Fig 7.5a), in order to identify the location of cavities and grain boundaries within the

block. This was needed in order to determine the optimal location for subsequent extraction of a

suitable STEM tomography needle sample of length greater than 1 μm and diameter 200 – 300

nm using annular milling in the FIB-SEM. Extraction of the block and the annular milling steps

outlined were performed by Giacomo Bertali (University of Manchester).

STEM-EDX tomography was employed here in order to provide straightforward identification of

different nanoscale secondary phases. Previous lower resolution imaging of this material

(Burnett, et al., 2014a) has identified the presence of secondary phases of ferrite, M 23C6

carbides and G-phase decorating similar cavitated boundaries within a different portion of the

same austenitic steel sample. STEM-EDX elemental mapping of the needle shaped sample

prepared here reveals the same secondary phases, sandwiched between two austenitic

domains (Fig 7.3). The prepared needle fully encapsulates a small (~ 50 nm diameter) carbide

situated at an austenite-ferrite interface, allowing a full morphological description of an individual

precipitate.

Figure 7.3. STEM-EDX elemental maps of the STEM needle displaying the intensity of X-ray

counts associated with the a) Fe Kα, b) Cr Kα, c) Ni Kα, d) Mo Lα, e) Nb Kα, f) O Kα, g) C Kα

and h) Si Kα peaks. The austenite phase at top and bottom is associated with Fe, Ni and to a

lesser extent Mo and Mn, the ferrite phase is predominantly Fe, the M23C6 carbides contain Cr,

Mn, Mo, C and O, the G-phase precipitate contains Ni, Mo and Si. The surface of the needle

has high carbon and silicon intensities that are associated with organic contamination.

The reconstruction of STEM-EDX tomography involves the reconstruction of the X-ray intensity

associated with individual elements, such that there is a separate reconstruction of the intensity

distribution of each element. Here, we have reconstructed the distribution of Fe, Cr and Ni in

three dimensions (Fig 7.4 a-c), as these elements allow segmentation of the identified phases.

157
The regions in which there is a high Fe intensity with an associated Ni signal is associated with

the austenite phase and that of a high Fe intensity without an associated Ni intensity is

associated with ferrite. Similarly, a high intensity in the Cr reconstruction corresponds to the

carbide precipitates and a high intensity in the Ni reconstruction with no associated Fe intensity

is associated to the G-phase. Through segmentation of the elemental data using these

constraints the individual phases are segmented in three dimensions (Fig 7.4 d-f).

Figure 7.4. Visualisation of the three dimensional reconstruction of elemental distributions and

the associated segmented phases. Volume visualisations of a) the Fe Kα signal (red), b) the Cr

Kα signal (yellow) and c) the Ni Kα signal (blue). Surface visualisations of the segmented

phases showing the d) ferrite phases (green), e) carbide precipitates (purple) and f) austenite

(orange) and G-phase (cyan).

Spatial correlation of the EDX tomography and X-ray nanotomography volumes is not possible

from the two volumes alone, as features present in the EDX tomography volume are not visible

in the X-ray CT volume due to the limited spatial resolution of X-ray CT. Instead, volume

registration is achieved through correlation of the position of SEM images acquired through the

158
FIB preparation steps with the X-ray CT volume. Registration of the X-ray CT volume and SEM

images is achieved through a manual feature matching, taking advantage of the cavities

decorating the grain boundary (Fig 7.5).

An SEM image of one face of the extracted cross-sectional slab is initially matched to a slice

through the X-ray CT volume to determine the approximate position of the block. SEM images

of each face of the block are then manually positioned to match slices in the X-ray CT data.

Cavities present on each of the four faces display a distinctive morphology that allows

correlation to slices in the X-ray CT data. In this way, the location of the extracted block can be

determined to an accuracy of less than 100 nm in all dimensions, positioning the SEM

coordinates into that of the X-ray CT coordinate system. The viewing directions of the four faces

which we number from (i) to (iv), are thus also defined with respect to the X-ray CT coordinates.

159
Figure 7.5. Images displaying the faces of the extracted block and the correlation between a)

SEM images and b) slices through the CT data of faces i-iv. c) Faces (i) and (ii) displayed in

three dimensions from the CT data with the X-ray CT coordinate system indicated.

Correlation of the position of the STEM needle within the X-ray CT volume relies primarily on

the correlation of the block faces. Knowledge that the needle was fabricated in one of the

160
corners of the block, and that it does not intersect any cavities, positions the needle in an area
2
in the defined X and Y dimensions of less than 1 μm in the material coordinate system.

Correlation of the XY position of the needle with that of a top-down view SEM image further

increases the positional accuracy to within a few hundred nanometres (Fig 7.6b). The position

of the needle in the Z direction of the CT volume, that of the long axis of the needle, is harder to

determine and is achieved here through positioning the centre of the precipitates at a Z-

coordinate in the materials coordinate system that corresponds to the location of precipitates in

the correlated SEM image of face (iii) (Fig 7.6a). This assumes that the precipitates are located

at the same Z-coordinates at the XY position of the needle as they are found at the XY

coordinates of face (iii).

Figure 7.6. a) Visualisation of the STEM-EDX pillar with respect to the SEM image of face (iii)

(flipped) and b-c) with respect to the top-down SEM image taken after annular milling with (c)

and without (b) segmented cavities. The coordinate systems indicated all correspond to the X-

ray CT coordinates.

The rotational transformation of the STEM reconstruction coordinates is determined through

correlation of an SEM image of the needle to the STEM-EDX 3D data. An SEM image of the

needle with a viewing direction in the same orientation as that of block face (i) is compared to

the final reconstruction. The SEM image of the needle shows bright regions associated with the

higher Z of the Nb containing G-phase precipitates. The viewing direction of block face (i)

specifies a direction in the material coordinate system at which the orientation of the segmented

161
G-phase can be matched to that in the SEM image of the needle, this defines the rotational

transformation about the Z direction (long axis of the needle).

Figure 7.7. Correlation of the orientation of the STEM-EDX needle embedded in to the micro-CT

data (a) with that from the SEM image of the final pillar (b) at the same orientation in the SEM

as in Fig 5a(i).

The steps outlined above represent the first steps towards an automated workflow for spatially

correlated studies combining STEM tomography with larger scale techniques such as X-ray

nanotomography. This type of correlative tomography is easily extendable to further techniques

such as atom probe tomography and FIB-SEM slice and view in the future. Correlative

tomography also has great promise for use in a number of industrially relevant material systems.

For instance, the location of individual nanoparticles that display a particular fluorescence

characteristic in dye-sensitized solar cells could be identified in 3D through confocal light

microscopy. Three dimensional imaging through X-ray nanotomography could reveal the

morphology of the site of the nanoparticle in the solar cell and further pinpoint the location of the

nanoparticle for extraction of a needle shaped sample via FIB-SEM. The 3D morphology and

chemistry of this identified single nanoparticle could then be obtained through STEM

tomography or APT.

Correlative tomography also has the potential to analyse in-situ processes at multiple length

scales in the future. In the instance of cracking of an engineering material due to mechanical

stress or a corrosive environment, it is already possible to investigate this type of process in-situ

162
at the macroscale via X-ray CT. If the in-situ X-ray CT is stopped at an important point in the

process of failure of the material, a correlative tomography approach will allow extraction of the

crack-tip for further micro- or nanoscale analysis via FIB-SEM or STEM tomography. Correlative

tomography offers an exciting opportunity to understand the links between nano- and

microscale features and the macroscale properties of materials.

163
8. References

ALLEN, L.J., FINDLAY, S.D., OXLEY, M.P. & ROSSOUW, C.J. (2003). Lattice-Resolution Contrast
From A Focused Coherent Electron Probe. Part I. Ultramicroscopy 96(1), 47-63.

AMENDOLA, V. & MENEGHETTI, M. (2009). Size Evaluation Of Gold Nanoparticles By UV-Vis

Spectroscopy. Journal Of Physical Chemistry C 113(11), 4277-4285.

ANDERHALT, R. & SANDBORG, A. (2011). Windowless Silicon Drift Detectors. Microscopy And
Microanalysis 17(S2), 620-621.

ARMIJO, J.S. (1968). Intergranular Corrosion Of Nonsensitized Austenitic Stainless Steels.

Corrosion 24(1), 24.

ARMSTRONG, J.T. & BUSECK, P.R. (1975). Quantitative Chemical Analysis Of Individual
Microparticles Using The Electron Microprobe. Theoretical. Analytical Chemistry 47(13), 2178-
2192.

ARSLAN, I., TONG, J.R. & MIDGLEY, P.A. (2006). Reducing The Missing Wedge: High-Resolution
Dual Axis Tomography Of Inorganic Materials. Ultramicroscopy 106(11–12), 994-1000.

ARSLAN, I., MARQUIS, E.A., HOMER, M., HEKMATY, M.A. & BARTELT, N.C. (2008).
Towards Better 3-D Reconstructions By Combining Electron Tomography And Atom-Probe
Tomography. Ultramicroscopy 108(12), 1579-1585.

AU, L., LU, X. & XIA, Y. (2008). A Comparative Study Of Galvanic Replacement Reactions
Involving Ag Nanocubes And Aucl2- Or Aucl4. Advanced Materials 20(13), 2517.

BABOUT, L., MAIRE, E., BUFFIERE, J.Y. & FOUGERES, R. (2001). Characterization By X-Ray
Computed Tomography Of Decohesion, Porosity Growth And Coalescence In Model Metal
Matrix Composites. Acta Materialia 49(11), 2055-2063.

BAER, D.R., GASPAR, D.J., NACHIMUTHU, P., TECHANE, S.D. & CASTNER, D.G. (2010).
Application Of Surface Chemical Analysis Tools For Characterization Of Nanoparticles.
Analytical And Bioanalytical Chemistry 396(3), 983-1002.

BAKER, L.A. & RUBINSTEIN, J.L. (2010). Radiation Damage In Electron Cryomicroscopy. Methods
In Enzymology, Vol 481: Cryo-Em, Part A - Sample Preparation And Data Collection 481, 371-
388.

BALDI, A., NARAYAN, T.C., KOH, A.L. & DIONNE, J.A. (2014). In Situ Detection Of Hydrogen-
Induced Phase Transitions In Individual Palladium Nanocrystals. Nature Materials 13(12), 1143-
1148.

164
BALS, S., CASAVOLA, M., VAN HUIS, M.A., VAN AERT, S., BATENBURG, K.J., VAN TENDELOO, G. &
VANMAEKELBERGH, D. (2011). Three-Dimensional Atomic Imaging Of Colloidal Core-Shell
Nanocrystals. Nano Letters 11(8), 3420-3424.

BARNARD, J.S., SHARP, J., TONG, J.R. & MIDGLEY, P.A. (2006). High-Resolution Three-
Dimensional Imaging Of Dislocations. Science 313(5785), 319.

BATENBURG, K.J., BALS, S., SIJBERS, J., KUBEL, C., MIDGLEY, P.A., HERNANDEZ, J.C., KAISER, U.,
ENCINA, E.R., CORONADO, E.A. & VAN TENDELOO, G. (2009). 3D Imaging Of Nanomaterials By
Discrete Tomography. Ultramicroscopy 109(6), 730-740.

BATSON, P.E. (1993). Simultaneous STEM Imaging And Electron Energy-Loss Spectroscopy
With Atomic-Column Sensitivity. 366(6457), 727-728.

BELIC, D., CHANTRY, R.L., LI, Z.Y. & BROWN, S.A. (2011). Ag-Au Nanoclusters: Structure And
Phase Segregation. Applied Physics Letters 99(17), 171914.

BELL, D.C.. & ERDMAN, N. (2013). Low Voltage Electron Microscopy: Principles And
Applications. Chichester, UK: Wiley.

BETHE, H. (1930). Zur Theorie Des Durchgangs Schneller Korpuskularstrahlen Durch Materie.
Annalen Der Physik 397(3), 325-400.

BLAVETTE, D., BOSTEL, A., SARRAU, J.M., DECONIHOUT, B. & MENAND, A. (1993). An
Atom-Probe For 3-Dimensional Tomography. Nature 363(6428), 432-435.

BLELOCH, A. & RAMASSE, Q. (2011). Lens Aberrations: Diagnosis And Correction. In Aberration-
Corrected Analytical Transmission Electron Microscopy, Pp. 55-87. John Wiley & Sons, Ltd.

BOOTZ, A., VOGEL, V., SCHUBERT, D. & KREUTER, J. (2004). Comparison Of Scanning
Electron Microscopy, Dynamic Light Scattering And Analytical Ultracentrifugation For The Sizing
Of Poly(Butyl Cyanoacrylate) Nanoparticles. European Journal Of Pharmaceutics And
Biopharmaceutics 57(2), 369-375.

BRAIDY, N., JAKUBEK, Z.J., SIMARD, B. & BOTTON, G.A. (2008). Quantitative Energy Dispersive X-
Ray Microanalysis Of Electron Beam-Sensitive Alloyed Nanoparticles. Microscopy And
Microanalysis 14(2), 166-175.

BRANDT, S.S. & ZIESE, U. (2006). Automatic TEM Image Alignment By Trifocal Geometry.
Journal Of Microscopy-Oxford 222, 1-14.

BROWN, L.M. (1981). Scanning-Transmission Electron-Microscopy - Microanalysis For The

Microelectronic Age. Journal Of Physics F-Metal Physics 11(1), 1-26.

BRYDSON, R. & HONDOW, N. (2011). Electron Energy Loss Spectrometry And Energy
Dispersive X-Ray Analysis. In Aberration-Corrected Analytical Transmission Electron
Microscopy, Brydson, R. (Ed.), Pp. 163-210. John Wiley & Sons, Ltd.

165
BUFFIERE, J.Y., MAIRE, E., ADRIEN, J., MASSE, J.P. & BOLLER, E. (2010). In Situ
Experiments With X Ray Tomography: An Attractive Tool For Experimental Mechanics.
Experimental Mechanics 50(3), 289-305.

BURNETT, T.L., GEURTS, R., JAZAERI, H., NORTHOVER, S.M., MCdONALD, S.A., HAIGH, S.J.,

BOUCHARD, P.J. & W ITHERS, P.J. (2014a). Multiscale 3D Analysis Of Creep Cavities In AISI Type

316 Stainless Steel. Materials Science And Technology 31(5), 522-534.

BURNETT, T.L., MCdONALD, S.A., GHOLINIA, A., GEURTS, R., JANUS, M., SLATER, T., HAIGH, S.J.,
ORNEK, C., ALMUAILI, F., ENGELBERG, D.L., THOMPSON, G.E. & WITHERS, P.J. (2014b). Correlative
Tomography. Scientific Reports 4, 4711.

CAO, M., ZHANG, H.B., LU, Y., NISHI, R. & TAKAOKA, A. (2010). Formation And Reduction Of
Streak Artefacts In Electron Tomography. Journal Of Microscopy 239(1), 66-71.

CASSIDY, C., SINGH, V., GRAMMATIKOPOULOS, P., DJURABEKOVA, F., NORDLUND, K. & SOWWAN, M.
(2013). Inoculation Of Silicon Nanoparticles With Silver Atoms. Scientific Reports 3, 7.

CAZAUX, J. (1995). Correlations Between Ionization Radiation-Damage And Charging Effects In

Transmission Electron-Microscopy. Ultramicroscopy 60(3), 411-425.

CHANTRY, R.L., SIRIWATCHARAPIBOON, W., HORSWELL, S.L., LOGSDAIL, A.J., JOHNSTON, R.L. & LI,
Z.Y. (2012). Overgrowth Of Rhodium On Gold Nanorods. Journal Of Physical Chemistry C
116(18), 10312-10317.

CHEN, C.C., ZHU, C., WHITE, E.R., CHIU, C.Y., SCOTT, M.C., REGAN, B.C., MARKS, L.D., HUANG, Y.
& MIAO, J.W. (2013). Three-Dimensional Imaging Of Dislocations In A Nanoparticle At Atomic
Resolution. Nature 496(7443), 74.

CHEN, D., FRIEDRICH, H. & DE WITH, G. (2014a). On Resolution In Electron Tomography Of

Beam Sensitive Materials. Journal Of Physical Chemistry C 118(2), 1248-1257.

CHEN, Y.Q., SLATER, T.J.A., LEWIS, E.A., FRANCIS, E.M., BURKE, M.G., PREUSS, M. & HAIGH, S.J.
(2014b). Measurement Of Size-Dependent Composition Variations For Gamma Prime (Γ′)
Precipitates In An Advanced Nickel-Based Superalloy. Ultramicroscopy 144, 1-8.

CHO, S.J., IDROBO, J.C., OLAMIT, J., LIU, K., BROWNING, N.D. & KAUZLARICH, S.M. (2005). Growth
Mechanisms And Oxidation Resistance Of Gold-Coated Iron Nanoparticles. Chemistry Of
Materials 17(12), 3181-3186.

CLIFF, G. & LORIMER, G.W. (1975). Quantitative-Analysis Of Thin Specimens. Journal Of

Microscopy-Oxford 103(MAR), 203-207.

CLOETENS, P., LUDWIG, W., BARUCHEL, J., VAN DYCK, D., VAN LANDUYT, J., GUIGAY,
J.P. & SCHLENKER, M. (1999). Holotomography: Quantitative Phase Tomography With

166
Micrometer Resolution Using Hard Synchrotron Radiation X Rays. Applied Physics Letters
75(19), 2912-2914.

COLLIEX, C., CRAVEN, A.J. & WILSON, C.J. (1977). Fresnel Fringes In Stem. Ultramicroscopy 2(4),
327-335.

COLLIEX, C. & MORY, C. (1984). Quantitative Aspects Of Scanning Transmission Electron

Microscopy. In Quantitative Electron Microscopy, Chapman, J. N. And Craven, A. (Eds.), Pp.
149. Glasgow: CRC.

CORTIE, M.B. & MCDONAGH, A.M. (2011). Synthesis And Optical Properties Of Hybrid And
Alloy Plasmonic Nanoparticles. Chemical Reviews 111(6), 3713-3735.

CRAVEN, A. (2011). Details Of STEM. In Aberration-Corrected Analytical Transmission Electron

Microscopy, Pp. 111-161. John Wiley & Sons, Ltd.

CROWTHER, R.A., DEROSIER, D.J. & KLUG, A. (1970). Reconstruction Of 3 Dimensional Structure
From Projections And Its Application To Electron Microscopy. Proceedings Of The Royal
Society Of London Series A-Mathematical And Physical Sciences 317(1530), 319.

CUENYA, B.R. (2013). Metal Nanoparticle Catalysts Beginning To Shape-Up. Accounts Of

Chemical Research 46(8), 1682-1691.

CUI, C., GAN, L., LI, H.-H., YU, S.-H., HEGGEN, M. & STRASSER, P. (2012). Octahedral Ptni
Nanoparticle Catalysts: Exceptional Oxygen Reduction Activity By Tuning The Alloy Particle
Surface Composition. Nano Letters 12(11), 5885-5889.

CUI, C., GAN, L., HEGGEN, M., RUDI, S. & STRASSER, P. (2013). Compositional Segregation In
Shaped Pt Alloy Nanoparticles And Their Structural Behaviour During Electrocatalysis. Nature
Materials 12(8), 765-771.

D'ALFONSO, A.J., FREITAG, B., KLENOV, D. & ALLEN, L.J. (2010). Atomic-Resolution Chemical
Mapping Using Energy-Dispersive X-Ray Spectroscopy. Physical Review B 81(10), 100101.

DAHMEN, T., BAUDOIN, J.-P., LUPINI, A.R., KUEBEL, C., SLUSALLEK, P. & DE JONGE, N. (2014).
Combined Scanning Transmission Electron Microscopy Tilt- And Focal Series. Microscopy And
Microanalysis 20(2), 548-560.

DE JONGE, N. & ROSS, F.M. (2011). Electron Microscopy Of Specimens In Liquid. Nature
Nanotechnology 6(11), 695-704.

DEANS, S.R. (2007). The Radon Transform And Some Of Its Applications. Mineola, NY: Dover
Publications Inc.

167
DEFRISE, M., VIERGEVER, M., TODD-POKROPEK, A. (1988). Possible Criteria for Choosing

the Number of Iterations in Some Iterative Reconstruction Methods. In Mathematics and

Computer Science in Medical Imaging 39, Pp. 293-303. Springer Berlin Heidelberg.

DEHM, G., LEGROS, M. & KIENER, D. (2012). In-Situ TEM Straining Experiments: Recent
Progress In Stages And Small-Scale Mechanics. In In-Situ Electron Microscopy, Pp. 227-254.
Wiley-VCH Verlag GMBH.

DENK, W. & HORSTMANN, H. (2004). Serial Block-Face Scanning Electron Microscopy To

Reconstruct Three-Dimensional Tissue Nanostructure. Plos Biology 2(11), 1900-1909.

DOROFEEV, G.A., STRELETSKII, A.N., POVSTUGAR, I.V., PROTASOV, A.V. & ELSUKOV,
E.P. (2012). Determination Of Nanoparticle Sizes By X-Ray Diffraction. Colloid Journal 74(6),
675-685.

DROUIN, D., COUTURE, A.R., JOLY, D., TASTET, X., AIMEZ, V. & GAUVIN, R. (2007). CASINO V2.42
- A Fast And Easy-To-Use Modeling Tool For Scanning Electron Microscopy And Microanalysis
Users. Scanning 29(3), 92-101.

DUNCUMB, P. (1962). Enhanced X-Ray Emission From Extinction Contours In A Single-Crystal

Gold Film. Philosophical Magazine 7(84), 2101.

DUNN, D.N. & HULL, R. (1999). Reconstruction Of Three-Dimensional Chemistry And Geometry
Using Focused Ion Beam Microscopy. Applied Physics Letters 75(21), 3414-3416.

EGERTON, R.F. (2009). Electron Energy-Loss Spectroscopy In The TEM. Reports On Progress
In Physics 72(1), 016502.

EGERTON, R.F. (2011a). Physics Of Electron Scattering. In Electron Energy-Loss Spectroscopy

In The Electron Microscope, Pp. 111-229. Springer US.

EGERTON, R.F. (2011b). Quantitative Analysis Of Energy-Loss Data. In Electron Energy-Loss

Spectroscopy In The Electron Microscope, Pp. 231-291. Springer US.

EGERTON, R.F. (2013). Control Of Radiation Damage In The TEM. Ultramicroscopy 127, 100-
108.

EGERTON, R.F., LI, P. & MALAC, M. (2004). Radiation Damage In The TEM And SEM. Micron
35(6), 399-409.

EGERTON, R.F., MClEOD, R., WANG, F. & MALAC, M. (2010). Basic Questions Related To
Electron-Induced Sputtering In The TEM. Ultramicroscopy 110(8), 991-997.

EGERTON, R.F., WANG, F. & CROZIER, P.A. (2006). Beam-Induced Damage To Thin Specimens
In An Intense Electron Probe. Microscopy And Microanalysis 12(01), 65-71.

168
EGGERT, T., BOSLAU, O., GOLDSTRASS, P. & KEMMER, J. (2004). Silicon Drift Detectors With
Enlarged Sensitive Areas. X-Ray Spectrometry 33(4), 246-252.

FERRER, D., BLOM, D.A., ALLARD, L.F., MEJIA, S., PEREZ-TIJERINA, E. & JOSE-YACAMAN, M. (2008).
Atomic Structure Of Three-Layer Au/Pd Nanoparticles Revealed By Aberration-Corrected
Scanning Transmission Electron Microscopy. Journal Of Materials Chemistry 18(21), 2442-2446.

FERRANDO, R., JELLINEK, J. & JOHNSTON, R.L. (2008). Nanoalloys: From Theory To
Applications Of Alloy Clusters And Nanoparticles. Chemical Reviews 108(3), 845-910.

FERTIG, J. & ROSE, H. (1981). Resolution And Contrast Of Crystalline Objects In High-Resolution
Scanning-Transmission Electron-Microscopy. Optik 59(5), 407-429.

FINDLAY, S.D., ALLEN, L.J., OXLEY, M.P. & ROSSOUW, C.J. (2003). Lattice-Resolution Contrast
From A Focused Coherent Electron Probe. Part II. Ultramicroscopy 96(1), 65-81.

FIORINI, C., LONGONI, A., PEROTTI, F., LABANTI, C., ROSSI, E., LECHNER, P., SOLTAU, H. &
STRUDER, L. (2002). A Monolithic Array Of Silicon Drift Detectors For High-Resolution Gamma-
Ray Imaging. IEEE Transactions On Nuclear Science 49(3), 995-1000.

FLADISCHER, S. (2013). Application of new EDXS quantification methods in TEM for organic
semiconductor devices. PhD thesis, University of Graz.

FLOREA, I., ERSEN, O., ARENAL, R., IHIAWAKRIM, D., MESSAOUDI, C., CHIZARI, K., JANOWSKA, I. &
CUONG, P.-H. (2012). 3D Analysis Of The Morphology And Spatial Distribution Of Nitrogen In
Nitrogen-Doped Carbon Nanotubes By Energy-Filtered Transmission Electron Microscopy
Tomography. Journal Of The American Chemical Society 134(23), 9672-9680.

FRANK, J., MCeWEN, B.F., RADERMACHER, M., TURNER, J.N. & RIEDER, C.L. (1987). 3-
Dimensional Tomographic Reconstruction In High-Voltage Electron-Microscopy. Journal Of
Electron Microscopy Technique 6(2), 193-205.

FRANK, J. (2002). Single-Particle Imaging Of Macromolecules By Cryo-Electron Microscopy.

Annual Review Of Biophysics And Biomolecular Structure 31, 303-319.

FRANK, J. & BRANDT, S. (2006). Markerless Alignment In Electron Tomography. In Electron

Tomography, Pp. 187-215. Springer New York.

FRANK, J. & HAWKES, P. (2006). The Electron Microscope As A Structure Projector. In Electron
Tomography, Pp. 83-111. Springer New York.

FRANK, J. & MASTRONARDE, D. (2006). Fiducial Marker And Hybrid Alignment Methods For
Single- And Double-Axis Tomography. In Electron Tomography, Pp. 163-185. Springer New
York.

FRANK, J. & RADERMACHER, M. (2006). Weighted Back-Projection Methods. In Electron

Tomography, Pp. 245-273. Springer New York.

169
FRIEDRICH, H., DE JONGH, P.E., VERKLEIJ, A.J. & DE JONG, K.P. (2009). Electron Tomography For
Heterogeneous Catalysts And Related Nanostructured Materials. Chemical Reviews 109(5),
1613-1629.

FRIGO, S.P., LEVINE, Z.H. & ZALUZEC, N.J. (2002). Submicron Imaging Of Buried Integrated
Circuit Structures Using Scanning Confocal Electron Microscopy. Applied Physics Letters
81(11), 2112-2114.

GAN, L., HEGGEN, M., O'MALLEY, R., THEOBALD, B. & STRASSER, P. (2013). Understanding And
Controlling Nanoporosity Formation For Improving The Stability Of Bimetallic Fuel Cell Catalysts.
Nano Letters 13(3), 1131-1138.

GASS, M.H., KOZIOL, K.K.K., WINDLE, A.H. & MIDGLEY, P.A. (2006). Four-Dimensional Spectral
Tomography Of Carbonaceous Nanocomposites. Nano Letters 6(3), 376-379.

GENC, A., KOVARIK, L., GU, M., CHENG, H., PLACHINDA, P., PULLAN, L., FREITAG, B. & WANG, C.
(2013). XEDS STEM Tomography For 3D Chemical Characterization Of Nanoscale Particles.
Ultramicroscopy 131, 24-32.

GEMMI, M. & OLEYNIKOV, P. (2013). Scanning Reciprocal Space For Solving Unknown
Structures: Energy Filtered Diffraction Tomography And Rotation Diffraction Tomography
Methods. Zeitschrift Fur Kristallographie 228(1), 51-58.

GIANNUZZI, L.A., DROWN, J.L., BROWN, S.R., IRWIN, R.B. & STEVIE, F. (1998).
Applications Of The FIB Lift-Out Technique For TEM Specimen Preparation. Microscopy
Research And Technique 41(4), 285-290.

GILBERT, P. (1972a). Iterative Methods For 3-Dimensional Reconstruction Of An Object From

Projections. Journal Of Theoretical Biology 36(1), 105.

GILBERT, P.F.C. (1972b). Reconstruction Of A 3-Dimensional Structure From Projections And Its
Application To Electron-Microscopy .2. Direct Methods. Proceedings Of The Royal Society Of
London Series B-Biological Sciences 182(1066), 89.

GOLDSTEIN, J.I., COSTLEY, J.L., LORIMER, G.W. & REED, R.J.B. (1977). Quantitative X-Ray
Analysis In The Electron Microscope. In Scanning Electron Microscopy, Johari, O. (Ed.), Pp.
315. Chicago, IL: IITRI.

GOLDSTEIN, J., NEWBURY, D.E., JOY, D.C., LYMAN, C.E., ECHLIN, P., LIFSHIN, E.,
SAWYER, L. & MICHAEL, J.R. (2003). Scanning Electron Microscopy And X-Ray Microanalysis.
New York: Springer US.

GOMEZ-GRANA, S., GORIS, B., ALTANTZIS, T., FERNANDEZ-LOPEZ, C., CARBO-ARGIBAY, E.,
GUERRERO-MARTINEZ, A., ALMORA-BARRIOS, N., LOPEZ, N., PASTORIZA-SANTOS, I., PEREZ-JUSTE,
J., BALS, S., VAN TENDELOO, G. & LIZ-MARZAN, L.M. (2013). Au@Ag Nanoparticles: Halides
Stabilize {100} Facets. Journal Of Physical Chemistry Letters 4(13), 2209-2216.

170
GONTARD, L.C., DUNIN-BORKOWSKI, R.E., FERNANDEZ, A., OZKAYA, D. & KASAMA, T. (2014).
Tomographic Heating Holder For In Situ TEM: Study Of Pt/C And Ptpd/Al2O3 Catalysts As A
Function Of Temperature. Microscopy And Microanalysis 20(3), 982-990.

GORDON, R., BENDER, R. & HERMAN, G.T. (1970). Algebraic Reconstruction Techniques (ART)
For 3-Dimensional Electron Microscopy And X-Ray Photography. Journal Of Theoretical Biology
29(3), 471.

GORIS, B., BALS, S., VAN DEN BROEK, W., VERBEECK, J. & VAN TENDELOO, G. (2011). Exploring
Different Inelastic Projection Mechanisms For Electron Tomography. Ultramicroscopy 111(8),
1262-1267.

GORIS, B., DE BACKER, A., VAN AERT, S., GOMEZ-GRANA, S., LIZ-MARZAN, L.M., VAN TENDELOO, G.
& BALS, S. (2013a). Three-Dimensional Elemental Mapping At The Atomic Scale In Bimetallic
Nanocrystals. Nano Letters 13(9), 4236-4241.

GORIS, B., POLAVARAPU, L., BALS, S., VAN TENDELOO, G. & LIZ-MARZAN, L.M. (2014a). Monitoring
Galvanic Replacement Through Three-Dimensional Morphological And Chemical Mapping.
Nano Letters 14(6), 3220-3226.

GORIS, B., ROELANDTS, T., BATENBURG, K.J., HEIDARI MEZERJI, H. & BALS, S. (2013b). Advanced
Reconstruction Algorithms For Electron Tomography: From Comparison To Combination.
Ultramicroscopy 127, 40-47.

GORIS, B., TURNER, S., BALS, S. & VAN TENDELOO, G. (2014b). Three-Dimensional Valency
Mapping In Ceria Nanocrystals. Acs Nano 8(10), 10878-10884.

GROBELNY, J., DELRIO, F.W., PRADEEP, N., KIM, D.-I., HACKLEY, V.A. & COOK, R.F.
(2011). Size Measurement Of Nanoparticles Using Atomic Force Microscopy. Characterization
Of Nanoparticles Intended For Drug Delivery 697, 71-82.

HABERFEHLNER, G., BAYLE-GUILLEMAUD, P., AUDOIT, G., LAFOND, D., MOREL, P.H., JOUSSEAUME,
V., ERNST, T. & BLEUET, P. (2012). Four-Dimensional Spectral Low-Loss Energy-Filtered
Transmission Electron Tomography Of Silicon Nanowire-Based Capacitors. Applied Physics
Letters 101(6), 063108.

HABERFEHLNER, G., ORTHACKER, A., ALBU, M., LI, J. & KOTHLEITNER, G. (2014). Nanoscale Voxel
Spectroscopy By Simultaneous EELS And EDS Tomography. Nanoscale 6(23), 14563-14569.

HAIDER, M., ROSE, H., UHLEMANN, S., SCHWAN, E., KABIUS, B. & URBAN, K. (1998). A Spherical-
Aberration-Corrected 200 Kv Transmission Electron Microscope. Ultramicroscopy 75(1), 53-60.

HAISS, W., THANH, N.T.K., AVEYARD, J. & FERNIG, D.G. (2007). Determination Of Size And
Concentration Of Gold Nanoparticles From UV-Vis Spectra. Analytical Chemistry 79(11), 4215-
4221.

171
HEIDARI MEZERJI, H., VAN DEN BROEK, W. & BALS, S. (2013). Quantitative Electron Tomography:
The Effect Of The Three-Dimensional Point Spread Function. Ultramicroscopy 135, 1-5.

HEIDARI MEZERJI, H., VAN DEN BROEK, W. & BALS, S. (2011). A Practical Method To Determine
The Effective Resolution In Incoherent Experimental Electron Tomography. Ultramicroscopy
111(5), 330-336.

HENRY, C.R. (2005). Morphology Of Supported Nanoparticles. Progress In Surface Science

80(3-4), 92-116.

HERBIG, M., CHOI, P. & RAABE, D. (2015). Combining Structural And Chemical Information At
The Nanometer Scale By Correlative Transmission Electron Microscopy And Atom Probe
Tomography. Ultramicroscopy 153, 32-39.

HIGGINS, J.P., ARRIVO, S.M., THURAU, G., GREEN, R.L., BOWEN, W., LANGE, A.,
TEMPLETON, A.C., THOMAS, D.L. & REED, R.A. (2003). Spectroscopic Approach For On-Line
Monitoring Of Particle Size During The Processing Of Pharmaceutical Nanoparticles. Analytical
Chemistry 75(8), 1777-1785.

HILLHOUSE, H.W. & BEARD, M.C. (2009). Solar Cells From Colloidal Nanocrystals:
Fundamentals, Materials, Devices, And Economics. Current Opinion In Colloid & Interface
Science 14(4), 245-259.

HIRSCH, P., HOWIE, A., NICHOLSON, R., PASHLEY, D. & WHELAN, M. (1977). Electron Microscopy
Of Thin Crystals. Huntington, New York: Krieger.

HOWIE, A. (1979). Image-Contrast And Localized Signal Selection Techniques. Journal Of

Microscopy-Oxford 117(SEP), 11-23.

HUTTER, E. & FENDLER, J.H. (2004). Exploitation Of Localized Surface Plasmon Resonance.
Advanced Materials 16(19), 1685-1706.

HWANG, B.J., SARMA, L.S., CHEN, J.M., CHEN, C.H., SHIH, S.C., WANG, G.R., LIU, D.G.,
LEE, J.F. & TANG, M.T. (2005). Structural Models And Atomic Distribution Of Bimetallic
Nanoparticles As Investigated By X-Ray Absorption Spectroscopy. Journal Of The American
Chemical Society 127(31), 11140-11145.

HSIEH, J. (2009). Computed Tomography: Principles, Design, Artifacts, And Recent Advances.
Wiley.

JAIN,P. K., HUANG, X., EL-SAYED, I.H., & EL-SAYED, M.A. (2008). Noble Metals on the
Nanoscale: Optical and Photothermal Properties and Some Applications in Imaging, Sensing,
Biology, and Medicine. Accounts of Chemical Research 41(12), 1578-1586.

172
JARAUSCH, K., THOMAS, P., LEONARD, D.N., TWESTEN, R. & BOOTH, C.R. (2009). Four-
Dimensional STEM-EELS: Enabling Nano-Scale Chemical Tomography. Ultramicroscopy
109(4), 326-337.

JESSON, D.E. & PENNYCOOK, S.J. (1995). Incoherent Imaging Of Crystals Using Thermally
Scattered Electrons. Proceedings Of The Royal Society-Mathematical And Physical Sciences
449(1936), 273-293.

JIANG, H.L., AKITA, T., ISHIDA, T., HARUTA, M. & XU, Q. (2011). Synergistic Catalysis Of
Au@Ag Core-Shell Nanoparticles Stabilized On Metal-Organic Framework. Journal Of The
American Chemical Society 133(5), 1304-1306.

JONES, S.D. & HAERTING, M. (2013). A New Correlation Based Alignment Technique For Use In
Electron Tomography. Ultramicroscopy 135, 56-63.

JOSHI, A. & STEIN, D.F. (1972). Chemistry Of Grain-Boundaries And Its Relation To
Intergranular Corrosion Of Austenitic Stainless-Steel. Corrosion 28(9), 321.

KANO, S., TADA, T. & MAJIMA, Y. (2015). Nanoparticle Characterization Based On STM And
STS. Chemical Society Reviews 44(4), 970-987.

KATO, M., KAWASE, N., KANEKO, T., TOH, S., MATSUMURA, S. & JINNAI, H. (2008). Maximum
Diameter Of The Rod-Shaped Specimen For Transmission Electron Microtomography Without
The "Missing Wedge". Ultramicroscopy 108(3), 221-229.

KELLY, T.F. & MILLER, M.K. (2007). Invited Review Article: Atom Probe Tomography. Review
Of Scientific Instruments 78(3), 031101.

KIM, J., CHOI, H., NAHM, C. & PARK, B. (2012). Surface-Plasmon Resonance For
Photoluminescence And Solar-Cell Applications. Electronic Materials Letters 8(4), 351-364.

KIMOTO, K., ASAKA, T., NAGAI, T., SAITO, M., MATSUI, Y. & ISHIZUKA, K. (2007). Element-Selective
Imaging Of Atomic Columns In A Crystal Using STEM And EELS. Nature 450(7170), 702-704.

KIRKLAND, E.J. (2010). Advanced Computing In Electron Microscopy, Second Ed. Advanced
Computing In Electron Microscopy, Second Ed.

KNOLL, M. & RUSKA, E. (1932). Das Elektronenmikroskop. Zeitschrift Für Physik 78(5-6), 318-
339.

KOGUCHI, M., KAKIBAYASHI, H., TSUNETA, R., YAMAOKA, M., NIINO, T., TANAKA, N., KASE, K. &
IWAKI, M. (2001). Three-Dimensional STEM For Observing Nanostructures. Journal Of Electron
Microscopy 50(3), 235-241.

KOSTER, A.J., GRIMM, R., TYPKE, D., HEGERL, R., STOSCHEK, A., WALZ, J. & BAUMEISTER, W.
(1997). Perspectives Of Molecular And Cellular Electron Tomography. J. Struct. Biol. 120(3),
276-308.

173
KOTHLEITNER, G., NEISH, M.J., LUGG, N.R., FINDLAY, S.D., GROGGER, W., HOFER, F. & ALLEN, L.J.
(2014). Quantitative Elemental Mapping At Atomic Resolution Using X-Ray Spectroscopy.
Physical Review Letters 112(8).

KOTULA, P., BREWER, L., MICHAEL, J. & GIANNUZZI, L. (2007). Computed Tomographic Spectral
Imaging: 3D STEM-EDS Spectral Imaging. Microscopy And Microanalysis 13(S2), 1324-1325.

KOTULA, P., MICHAEL, J. & ROHDE, M. (2008). Results From Two Four-Channel Si-Drift Detectors
On An SEM: Conventional And Annular Geometries. Microscopy And Microanalysis 14(S2),
116-117.

KRAKOW, W., JOSEYACAMAN, M. & ARAGON, J.L. (1994). Observation Of Quasi-Melting At The
Atomic-Level In Au Nanoclusters. Physical Review B 49(15), 10591-10596.

KRAMERS, H.A. (1923). On The Theory Of X-Ray Absorption And Of The Continuous X- Ray
Spectrum. Philosophical Magazine 46(275), 836-871.

KRAUSE, M.O. (1979). Atomic Radiative And Radiationless Yields For K And L Shells. Journal Of
Physical And Chemical Reference Data 8(2), 307-327.

KRIVANEK, O.L., DELLBY, N. & LUPINI, A.R. (1999). Towards Sub-Angstrom Electron Beams.
Ultramicroscopy 78(1-4), 1-11.

KRIVANEK, O.L., LOVEJOY, T.C., DELLBY, N., AOKI, T., CARPENTER, R.W., REZ, P., SOIGNARD, E.,
ZHU, J., BATSON, P.E., LAGOS, M.J., EGERTON, R.F. & CROZIER, P.A. (2014). Vibrational
Spectroscopy In The Electron Microscope. Nature 514(7521), 209.

KRIVANEK, O.L., LOVEJOY, T.C., DELLBY, N. & CARPENTER, R.W. (2013). Monochromated STEM
With A 30 Mev-Wide, Atom-Sized Electron Probe. Microscopy 62(1), 3-21.

KYOTANI, T., TSAI, L.F. & TOMITA, A. (1997). Formation Of Platinum Nanorods And Nanoparticles
In Uniform Carbon Nanotubes Prepared By A Template Carbonization Method. Chemical
Communications (7), 701-702.

LAOUFI, I., SAINT-LAGER, M.C., LAZZARI, R., JUPILLE, J., ROBACH, O., GARAUDEE, S.,
CABAILH, G., DOLLE, P., CRUGUEL, H. & BAILLY, A. (2011). Size And Catalytic Activity Of
Supported Gold Nanoparticles: An In Operando Study During CO Oxidation. Journal Of
Physical Chemistry C 115(11), 4673-4679.

LEARY, R., DE LA PENA, F., BARNARD, J.S., LUO, Y., ARMBRUESTER, M., THOMAS, J.M. & MIDGLEY,
P.A. (2013a). Revealing The Atomic Structure Of Intermetallic Gapd2 Nanocatalysts By Using
Aberration-Corrected Scanning Transmission Electron Microscopy. Chemcatchem 5(9), 2599-
2609.

LEARY, R., SAGHI, Z., MIDGLEY, P.A. & HOLLAND, D.J. (2013b). Compressed Sensing Electron
Tomography. Ultramicroscopy 131, 70-91.

174
LEE, J.-H., JANG, J.-T., CHOI, J.-S., MOON, S.H., NOH, S.-H., KIM, J.-W., KIM, J.-G., KIM, I.-S.,
PARK, K.I. & CHEON, J. (2011). Exchange-Coupled Magnetic Nanoparticles For Efficient Heat
Induction. Nature Nanotechnology 6(7), 418-422.

LEPINAY, K., LORUT, F., PANTEL, R. & EPICIER, T. (2013). Chemical 3D Tomography Of 28 Nm
High K Metal Gate Transistor: STEM XEDS Experimental Method And Results. Micron 47, 43-
49.

LEWIS, E., SLATER, T., PRESTAT, E., MACEDO, A., O'BRIEN, P., CAMARGO, P. & HAIGH, S. (2014a).
Real-Time Imaging And Elemental Mapping Of Agau Nanoparticle Transformations. Nanoscale
6, 13598-13605.

LEWIS, E.A., HAIGH, S.J., SLATER, T.J.A., HE, Z., KULZICK, M.A., BURKE, M.G. & ZALUZEC, N.J.
(2014b). Real-Time Imaging And Local Elemental Analysis Of Nanostructures In Liquids.
Chemical Communications 50, 10019-10022.

LI, R.F., WU, G.H., JIANG, L.T. & SUN, D.L. (2015). Characterization Of Multi-Scale Porous
Structure Of Fly Ash/Phosphate Geopolymer Hollow Sphere Structures: From Submillimeter To
Nano-Scale. Micron 68, 54-58.

LIAKAKOS, N., GATEL, C., BLON, T., ALTANTZIS, T., LENTIJO-MOZO, S., GARCIA-MARCELOT, C.,
LACROIX, L.-M., RESPAUD, M., BALS, S., VAN TENDELOO, G. & SOULANTICA, K. (2014). Co-Fe
Nanodumbbells: Synthesis, Structure, And Magnetic Properties. Nano Letters 14(5), 2747-2754.

LIAO, Y. & MARKS, L.D. (2013). Reduction Of Electron Channeling In EDS Using Precession.
Ultramicroscopy 126, 19-22.

LINK, S., WANG, Z.L. & EL-SAYED, M.A. (1999). Alloy Formation Of Gold-Silver Nanoparticles
And The Dependence Of The Plasmon Absorption On Their Composition. Journal Of Physical
Chemistry B 103(18), 3529-3533.

LIU, G.S., HOUSE, S.D., KACHER, J., TANAKA, M., HIGASHIDA, K. & ROBERTSON, I.M.
(2014). Electron Tomography Of Dislocation Structures. Materials Characterization 87, 1-11.

LIU, J.Y. (2005). Scanning Transmission Electron Microscopy and its application to the study of
nanoparticles and nanoparticle systems. Journal of Electron Microscopy 54(3), 251-278.

LIU, Z., MA, L., ZHANG, J., HONGSIRIKARN, K. & GOODWIN, J.G., JR. (2013). Pt Alloy
Electrocatalysts for Proton Exchange Membrane Fuel Cells: A Review. Catalysis Reviews-
Science and Engineering 55(3), 255-288.

LIZ-MARZAN, L.M. (2006). Tailoring surface plasmons through the morphology and assembly
of metal nanoparticles. Langmuir 22(1), 32-41.

175
LOPEZ-HARO, M., JIU, T., BAYLE-GUILLEMAUD, P., JOUNEAU, P.-H. & CHANDEZON, F. (2013).
Multiscale tomographic analysis of polymer-nanoparticle hybrid materials for solar cells.
Nanoscale 5(22), 10945-10955.

LOUIS, A. K., NATTERER, F. (1983). Mathematical Problems of Computerized-Tomography.

Proceedings of the IEEE 71 (3), 379-389.

MAIRE, E. & WITHERS, P.J. (2014). Quantitative X-ray tomography. International Materials
Reviews 59(1), 1-43.

MARKS, L.D. (1994). EXPERIMENTAL STUDIES OF SMALL-PARTICLE STRUCTURES.

Reports on Progress in Physics 57(6), 603-649.

MARUYAMA, W., AKAO, Y., CARRILLO, M.C., KITANI, K., YOUDIUM, M.B.H. & NAOI, M. (2002a).
Neuroprotection by propargylamines in Parkinson's disease - Suppression of apoptosis and
induction of prosurvival genes. Neurotoxicology and Teratology 24(5), 675-682.

MARUYAMA, W., BOULTON, A.A., YOUDIM, M.B.H. & NAOI, M. (2002b). Anti-apoptotic function of
propargylamines. Catecholamine Research: from Molecular Insights to Clinical Medicine 53,
241-244.

MAYER, J., GIANNUZZI, L.A., KAMINO, T. & MICHAEL, J. (2007). TEM sample preparation
and FIB-induced damage. Mrs Bulletin 32(5), 400-407.

MAYORAL, A., DEEPAK, F.L., ESPARZA, R., CASILLAS, G., MAGEN, C., PEREZ-TIJERINA, E. & JOSE-
YACAMAN, M. (2012). On the structure of bimetallic noble metal nanoparticles as revealed by
aberration corrected scanning transmission electron microscopy (STEM). Micron 43(4), 557-564.

MERKLE, A., LECHNER, L., STEINBACH, A., GELB, J., KIENLE, M., PHANEUF, M., UNRAU, D., SINGH, S.
& CHAWLA, N. (2014). Automated correlative tomography using XRM and FIB-SEM to span
length scales and modalities in 3D materials. In Microscopy and Analysis, pp. S10-S13. Wiley.

MIDGLEY, P.A. & WEYLAND, M. (2003). 3D electron microscopy in the physical sciences: the
development of Z-contrast and EFTEM tomography. Ultramicroscopy 96(3-4), 413-431.

MIDGLEY, P.A. & WEYLAND M. (2007). Electron Tomography. In Nanocharacterisation.

HUTCHISON, J. & KIRKLAND, A. (Eds.). The Royal Society of Chemistry.

MIDGLEY, P.A. & DUNIN-BORKOWSKI, R.E. (2009). Electron tomography and holography in
materials science. Nature Materials 8(4), 271-280.

MISZTA, K., BRESCIA, R., PRATO, M., BERTONI, G., MARRAS, S., XIE, Y., GHOSH, S., KIM, M.R. &
MANNA, L. (2014). Hollow and Concave Nanoparticles via Preferential Oxidation of the Core in
Colloidal Core/Shell Nanocrystals. Journal of the American Chemical Society 136(25), 9061-
9069.

176
MOBUS, G., DOOLE, R.C. & INKSON, B.J. (2003). Spectroscopic electron tomography.
Ultramicroscopy 96(3-4), 433-451.

MOBUS, G. & INKSON, B.J. (2007). Nanoscale tomography in materials science. Materials Today
10(12), 18-25.

MOBUS, G., SAGHI, Z., GUAN, W., GNANAVEL, T., XU, X. & PENG, Y. (2010). Hybrid Tomography
for structural and chemical 3D imaging on the nanoscale. In Electron Microscopy and Analysis
Group Conference 2009, Baker, R. T. (Ed.), Bristol: Iop Publishing Ltd.

MOMOSE, A., TAKEDA, T., ITAI, Y. & HIRANO, K. (1996). Phase-contrast X-ray computed
tomography for observing biological soft tissues. Nature Medicine 2(4), 473-475.

MOURDIKOUDIS, S., CHIREA, M., ZANAGA, D., ALTANTZIS, T., MITRAKAS, M., BALS, S., LIZ-MARZAN,
L.M., PEREZ-JUSTE, J. & PASTORIZA-SANTOS, I. (2015). Governing the morphology of Pt-Au
heteronanocrystals with improved electrocatalytic performance. Nanoscale 7(19), 8739-8747.

MUGNAIOLI, E., GORELIK, T. & KOLB, U. (2009). "Ab initio" structure solution from electron
diffraction data obtained by a combination of automated diffraction tomography and precession
technique. Ultramicroscopy 109(6), 758-765.

MUKERJEE, S. & SRINIVASAN, S. (1993). Enhanced electrocatalysis of oxygen reduction on

platinum alloys in proton-exchange membrane fuel-cells. Journal of Electroanalytical Chemistry
357(1-2), 201-224.

MULLER, D.A. & SILCOX, J. (1995). Radiation damage of Ni3Al by 100 kev electrons.
Philosophical Magazine A 71(6), 1375-1387.

MYERS, B., MCILWRATH, K., WU, J., LI, S., INADA, H. & DRAVID, V. (2010). Performance
Characterization of Hitachi HD-2300A STEM with Dual-EDS Configuration. Microscopy and
Microanalysis 16(S2), 942-943.

NARAYANAN, R. & EL-SAYED, M.A. (2004). Shape-dependent catalytic activity of platinum

nanoparticles in colloidal solution. Nano Letters 4(7), 1343-1348.

NELLIST, P.D. (2011). Theory and Simulations of STEM Imaging. In Aberration-Corrected

Analytical Transmission Electron Microscopy, pp. 89-110. John Wiley & Sons, Ltd.

NELLIST, P.D. & WANG, P. (2012). Optical Sectioning and Confocal Imaging and Analysis in
the Transmission Electron Microscope. Annual Review of Materials Research 42, 125-143.

NEWBURY, D.E. (2006). The New X-ray Mapping: X-ray Spectrum Imaging above 100 kHz
Output Count Rate with the Silicon Drift Detector. Microscopy and Microanalysis 12(1), 26-35.

NICOLETTI, O., DE LA PENA, F., LEARY, R.K., HOLLAND, D.J., DUCATI, C. & MIDGLEY, P.A. (2013).
Three-dimensional imaging of localized surface plasmon resonances of metal nanoparticles.
Nature 502(7469), 80.

177
PARK, J., ELMLUND, H., ERCIUS, P., YUK, J.M., LIMMER, D.T., CHEN, Q., KIM, K., HAN,
S.H., WEITZ, D.A., ZETTL, A. & ALIVISATOS, A.P. (2015). Nanoparticle imaging. 3D structure
of individual nanocrystals in solution by electron microscopy. Science 349(6245), 290-295.

PATTERSON, A.L. (1939). The Scherrer formula for x-ray particle size determination. Physical
Review 56(10), 978-982.

PENCZEK, P., MARKO, M., BUTTLE, K. & FRANK, J. (1995). Double-tilt electron tomography.
Ultramicroscopy 60(3), 393-410.

PENNYCOOK, S.J. & JESSON, D.E. (1990). High-resolution incoherent imaging of crystals.
Physical Review Letters 64(8), 938-941.

PHILIBERT, J. A method for calculating the absorption correction in electron-probe microanalysis.

In 3rd International Congress on X-ray Optics and Microanalysis, Pattee, H. H., Cosslett, V. E. &
Engstrom, A. (Eds.), pp. 379. Academic.

POTAPOV, P.L. & SCHRYVERS, D. (2004). Measuring the absolute position of EELS ionisation
edges in a TEM. Ultramicroscopy 99(1), 73-85.

POWELL, C.J. (1976). Cross sections for ionization of inner-shell electrons by electrons. Reviews
of Modern Physics 48(1), 33.

POWELL, C.J. & JABLONSKI, A. (2010). NIST Electron Inelastic-Mean-Free-Path Database.

Gaithersburg, MD: National Institute of Standards and Technology.

PRODAN, E., RADLOFF, C., HALAS, N.J. & NORDLANDER, P. (2003). A hybridization model
for the plasmon response of complex nanostructures. Science 302(5644), 419-422.

RADERMACHER, M. (1988). 3-dimensional reconstruction of single particles from random and

nonrandom tilt series. Journal of Electron Microscopy Technique 9(4), 359-394.

RAETHER, H. (1988). Surface-plasmons on smooth and rough surfaces and on gratings.

Springer Tracts in Modern Physics 111, 1-133.

RAYLEIGH. (1879). XXXI. Investigations in optics, with special reference to the spectroscope.
Philosophical Magazine Series 5 8(49), 261-274.

REED, S.J.B. (1982). The single-scattering model and spatial-resolution in x-ray-analysis of thin
foils. Ultramicroscopy 7(4), 405-409.

REIMER, L. (1995). Energy-Filtering Transmission Electron Microscopy. Springer Berlin

Heidelberg.

REIMER, L. (1998). Scanning Electron Microscopy. Springer Berlin Heidelberg.

178
RODRIGUEZ-FERNANDEZ, D., ALTANTZIS, T., HEIDARI, H., BALS, S. & LIZ-MARZAN, L.M. (2014). A
protecting group approach toward synthesis of Au-silica Janus nanostars. Chemical
Communications 50(1), 79-81.

RODRIGUEZ-GONZALEZ, B., BURROWS, A., WATANABE, M., KIELY, C.J. & LIZ-MARZAN, L.M. (2005).
Multishell bimetallic auag nanoparticles: synthesis, structure and optical properties. Journal of
Materials Chemistry 15(17), 1755-1759.

ROSSELL, M.D., RAMASSE, Q.M., FINDLAY, S.D., RECHBERGER, F., ERNI, R. & NIEDERBERGER, M.
(2012). Direct Imaging of Dopant Clustering in Metal-Oxide Nanoparticles. Acs Nano 6(8), 7077-
7083.

RUOFF, R.S., LORENTS, D.C., CHAN, B., MALHOTRA, R. & SUBRAMONEY, S. (1993). Single-crystal
metals encapsulated in carbon nanoparticles. Science 259(5093), 346-348.

SAGHI, Z., XU, X., PENG, Y., INKSON, B. & MOBUS, G. (2007). Three-dimensional chemical
analysis of tungsten probes by energy dispersive x-ray nanotomography. Applied Physics
Letters 91(25), 251906.

SAKAMOTO, T., CHENG, Z.H., TAKAHASHI, M., OWARI, M. & NIHEI, Y. (1998). Development of an
ion and electron dual focused beam apparatus for three-dimensional microanalysis. Japanese
Journal of Applied Physics Part 1-Regular Papers Short Notes & Review Papers 37(4A), 2051-
2056.

SASAKI, K., NAOHARA, H., CHOI, Y., CAI, Y., CHEN, W.-F., LIU, P. & ADZIC, R.R. (2012). Highly
stable Pt monolayer on pdau nanoparticle electrocatalysts for the oxygen reduction reaction.
Nature Communications 3, 1115.

SCOTT, M.C., CHEN, C.C., MECKLENBURG, M., ZHU, C., XU, R., ERCIUS, P., DAHMEN, U., REGAN,
B.C. & MIAO, J.W. (2012). Electron tomography at 2.4-angstrom resolution. Nature 483(7390),
444-491.

SEIDMAN, D.N. (2007). Three-dimensional atom-probe tomography: Advances and

applications. Annual Review of Materials Research 37, 127-158.

SERP, P. & PHILIPPOT, K. (2013). Nanomaterials in Catalysis. Wiley-VCH Verlag.

SHAO, M.H., HUANG, T., LIU, P., ZHANG, J., SASAKI, K., VUKMIROVIC, M.B. & ADZIC, R.R.
(2006). Palladium monolayer and palladium alloy electrocatalysts for oxygen reduction.
Langmuir 22(25), 10409-10415.

SHARMA, R. (2005). An environmental transmission electron microscope for in situ synthesis and
characterization of nanomaterials. Journal of Materials Research 20(7), 1695-1707.

SHERIDAN, P.J. (1989). Determination of experimental and theoretical kasi factors for a 200-kv
analytical electron-microscope. Journal of Electron Microscopy Technique 11(1), 41-61.

179
SILCOX, J. (1998). Core-loss EELS. Current Opinion in Solid State & Materials Science 3(4),
336-342.

SONDHEIMER, E.H. (2001). The mean free path of electrons in metals. Advances in Physics 50(6),
499-537.

SOURMAIL, T. (2001). Precipitation in creep resistant austenitic stainless steels. Materials

Science and Technology 17(1), 1-14.

SPENCE, J.C.H., ZUO, J.M. & LYNCH, J. (1989). On the holz contribution to stem lattice images
formed using high-angle dark-field detectors. Ultramicroscopy 31(2), 233-240.

STOCK, S.R. (2008). Recent advances in X-ray microtomography applied to materials.

International Materials Reviews 53(3), 129-181.

SUENAGA, K. & KOSHINO, M. (2010). Atom-by-atom spectroscopy at graphene edge. Nature

468(7327), 1088-1090.

SUN, Y.G. & XIA, Y.N. (2003). Alloying and dealloying processes involved in the preparation of
metal nanoshells through a galvanic replacement reaction. Nano Letters 3(11), 1569-1572.

SUN, Y.G. & XIA, Y.N. (2004). Mechanistic study on the replacement reaction between silver
nanostructures and chloroauric acid in aqueous medium. Journal of the American Chemical
Society 126(12), 3892-3901.

SUZUKI, S. (2015). Development of a Novel Surface Elemental Analysis Methodology: X-ray-

Aided Noncontact Atomic Force Microscopy (XANAM). Bulletin of the Chemical Society of
Japan 88(2), 240-250.

TAFTO, J. & LEHMPFUHL, G. (1982). Direction dependence in electron-energy loss spectroscopy

from single-crystals. Ultramicroscopy 7(3), 287-294.

TARIQ, F., HASWELL, R., LEE, P.D. & mCcOMB, D.W. (2011). Characterization of hierarchical pore
structures in ceramics using multiscale tomography. Acta Materialia 59(5), 2109-2120.

TATTON, W.G., CHALMERS-REDMAN, R.M.E., JU, W.J.H., MAMMEN, M., CARLILE, G.W., PONG, A.W.
& TATTON, N.A. (2002). Propargylamines induce antiapoptotic new protein synthesis in serum-
and nerve growth factor (NGF)-withdrawn, NGF-differentiated PC-12 cells. Journal of
Pharmacology and Experimental Therapeutics 301(2), 753-764.

TEDMON, C.S., VERMILYE.D.A. & ROSOLOWS. J.H. (1971). Intergranular corrosion of

austenitic stainless steel. Journal of the Electrochemical Society 118(2), 192.

THIELE, E.W. (1939). Relation between catalytic activity and size of particle. Industrial and
Engineering Chemistry 31, 916-920.

180
THIELE, S., FURSTENHAUPT, T., BANHAM, D., HUTZENLAUB, T., BIRSS, V., ZIEGLER, C.
& ZENGERLE, R. (2013). Multiscale tomography of nanoporous carbon-supported noble metal
catalyst layers. Journal of Power Sources 228, 185-192.

THOMPSON, K., LAWRENCE, D., LARSON, D.J., OLSON, J.D., KELLY, T.F. & GORMAN, B.
(2007). In situ site-specific specimen preparation for atom probe tomography. Ultramicroscopy
107(2-3), 131-139.

TOSHIMA, N. & YONEZAWA, T. (1998). Bimetallic nanoparticles - novel materials for chemical
and physical applications. New Journal of Chemistry 22(11), 1179-1201.

TRAN, D.T., JONES, I.P., PREECE, J.A., JOHNSTON, R.L., DEPLANCHE, K. & MACASKIE, L.E. (2012).
Configuration of microbially synthesized Pd-Au nanoparticles studied by STEM-based
techniques. Nanotechnology 23(5), 055701.

TREACY, M.M.J. (2011). Z Dependence of Electron Scattering by Single Atoms into Annular
Dark-Field Detectors. Microscopy and Microanalysis 17(6), 847-858.

UCHIC, M.D., HOLZER, L., INKSON, B.J., PRINCIPE, E.L. & MUNROE, P. (2007). Three-
dimensional microstructural characterization using focused ion beam tomography. Mrs Bulletin
32(5), 408-416.

UNGER, V.M. (2001). Electron cryomicroscopy methods. Current Opinion in Structural Biology
11(5), 548-554.

VAN AERT, S., BATENBURG, K.J., ROSSELL, M.D., ERNI, R. & VAN TENDELOO, G. (2011). Three-
dimensional atomic imaging of crystalline nanoparticles. Nature 470(7334), 374-377.

VAN DEN BROEK, W., ROSENAUER, A., GORIS, B., MARTINEZ, G.T., BALS, S., VAN AERT, S. & VAN
DYCK, D. (2012). Correction of non-linear thickness effects in HAADF STEM electron
tomography. Ultramicroscopy 116, 8-12.

VAN HEEL, M., GOWEN, B., MATADEEN, R., ORLOVA, E.V., FINN, R., PAPE, T., COHEN, D.,
STARK, H., SCHMIDT, R., SCHATZ, M. & PATWARDHAN, A. (2000). Single-particle electron
cryo-microscopy: towards atomic resolution. Quarterly Reviews of Biophysics 33(4), 307-369.

VARELA, M., FINDLAY, S.D., LUPINI, A.R., CHRISTEN, H.M., BORISEVICH, A.Y., DELLBY, N., KRIVANEK,
O.L., NELLIST, P.D., OXLEY, M.P., ALLEN, L.J. & PENNYCOOK, S.J. (2004). Spectroscopic imaging
of single atoms within a bulk solid. Physical Review Letters 92(9), 095502 .

VILLINGER, C., GREGORIUS, H., KRANZ, C., HOEHN, K., MUENZBERG, C., VON WICHERT,
G., MIZAIKOFF, B., WANNER, G. & WALTHER, P. (2012). FIB/SEM tomography with TEM-like
resolution for 3D imaging of high-pressure frozen cells. Histochemistry and Cell Biology 138(4),
549-556.

181
VON HARRACH, H., KLENOV, D., FREITAG, B., SCHLOSSMACHER, P., COLLINS, P. & FRASER, H.
(2010). Comparison of the Detection Limits of EDS and EELS in S/TEM. Microscopy and
Microanalysis 16(S2), 1312-1313.

VON HARRACH, H.S., DONA, P., FREITAG, B., SOLTAU, H., NICULAE, A. & ROHDE, M. (2009). An
integrated Silicon Drift Detector System for FEI Schottky Field Emission Transmission Electron
Microscopes. Microscopy and Microanalysis 15, 208-209.

WANG, C.M., LIAO, H.-G. & ROSS, F.M. (2015). Observation of materials processes in liquids by
electron microscopy. MRS Bulletin 40(1), 46-52.

WANG, C.M., GENC, A., CHENG, H., PULLAN, L., BAER, D.R. & BRUEMMER, S.M. (2014). In-Situ
TEM visualization of vacancy injection and chemical partition during oxidation of Ni-Cr
nanoparticles. Scientific Reports 4, 3683.

WANG, D. & LI, Y. (2011). Bimetallic Nanocrystals: Liquid-Phase Synthesis and Catalytic
Applications. Advanced Materials 23(9), 1044-1060.

WANG, D., XIN, H.L., HOVDEN, R., WANG, H., YU, Y., MULLER, D.A., dIsALVO, F.J. & ABRUNA, H.D.
(2013). Structurally ordered intermetallic platinum-cobalt core-shell nanoparticles with enhanced
activity and stability as oxygen reduction electrocatalysts. Nature Materials 12(1), 81-87.

WANG, D., YU, Y., XIN, H.L., HOVDEN, R., ERCIUS, P., MUNDY, J.A., CHEN, H., RICHARD, J.H.,
MULLER, D.A., dIsALVO, F.J. & ABRUNA, H.D. (2012). Tuning Oxygen Reduction Reaction Activity
via Controllable Dealloying: A Model Study of Ordered Cu3Pt/C Intermetallic Nanocatalysts.
Nano Letters 12(10), 5230-5238.

WANG, P., D'ALFONSO, A.J., FINDLAY, S.D., ALLEN, L.J. & BLELOCH, A.L. (2008). Contrast
Reversal in Atomic-Resolution Chemical Mapping. Phys. Rev. Lett. 101(23), 236102.

WATANABE, M. (2011). X-Ray Energy-Dispersive Spectrometry in Scanning Transmission

Electron Microscopes. In Scanning Transmission Electron Microscopy, Pennycook, S. J. &
Nellist, P. D. (Eds.), pp. 291-351. Springer New York.

WATANABE, M. & WILLIAMS, D.B. (2006). The quantitative analysis of thin specimens: a review of
progress from the Cliff-Lorimer to the new zeta-factor methods. Journal of Microscopy-Oxford
221(2), 89-109.

WEYLAND, M. (2002). Electron tomography of catalysts. Topics in Catalysis 21(4), 175-183.

WILKINS, S.W., GUREYEV, T.E., GAO, D., POGANY, A. & STEVENSON, A.W. (1996). Phase-
contrast imaging using polychromatic hard X-rays. Nature 384(6607), 335-338.

WILLETS, K.A. & VAN DUYNE, R.P. (2007). Localized surface plasmon resonance
spectroscopy and sensing. In Annual Review of Physical Chemistry, pp. 267-297. Palo Alto:
Annual Reviews.

182
WILLIAMS, D.B. & GOLDSTEIN, J.I. (1991). Quantitative x-ray-microanalysis in the analytical
electron-microscope. Electron Probe Quantitation, 371-398.

WILLIAMS, D., GOLDSTEIN, J., NEWBURY, D. & LUND, M.W. (1995). Current Trends in Si(Li)
Detector Windows for Light Element Analysis. In X-Ray Spectrometry in Electron Beam
Instruments, pp. 21-31. Springer US.

WILLIAMS, D.B., PAPWORTH, A.J. & WATANABE, M. (2002). High resolution X‐ray mapping
in the STEM. Journal of Electron Microscopy 51(S1), S113-S126.

WILLIAMS, D.B. & CARTER, C.B. (2009a). Imaging. In Transmission Electron Microscopy, New
York: Springer US.

WILLIAMS, D.B. & CARTER, C.B. (2009b). Basics. In Transmission Electron Microscopy, New
York: Springer US.

WILLIAMS, D.B. & CARTER, C.B. (2009c). Spectrometry. In Transmission Electron Microscopy,
New York: Springer US.

WILSON, A.R., SUN, K., CHI, M., WHITE, R.M., lEbEAU, J.M., LAMB, H.H. & WILEY, B.J. (2013).
From Core-Shell to Alloys: The Preparation and Characterization of Solution-Synthesized aupd
Nanoparticle Catalysts. Journal of Physical Chemistry C 117(34), 17557-17566.

WROGGEN, U. (1997). Optical Properties of Semiconductor Quantum Dots. Berlin: Springer

Berlin Heidelberg.

XIA, X., WANG, Y., RUDITSKIY, A. & XIA, Y. (2013). 25th Anniversary Article: Galvanic
Replacement: A Simple and Versatile Route to Hollow Nanostructures with Tunable and Well-
Controlled Properties. Advanced Materials 25(44), 6313-6333.

XIA, Y., YANG, H. & CAMPBELL, C.T. (2013). Nanoparticles for Catalysis. Accounts of
Chemical Research 46(8), 1671-1672.

XIN, H.L.L., DWYER, C., MULLER, D.A., ZHENG, H.M. & ERCIUS, P. (2013). Scanning
Confocal Electron Energy-Loss Microscopy Using Valence-Loss Signals. Microscopy and
Microanalysis 19(4), 1036-1049.

XIONG, X. & WEYLAND, M. (2014). Microstructural Characterization of an Al-Li-Mg-Cu Alloy by

Correlative Electron Tomography and Atom Probe Tomography. Microscopy and Microanalysis
20(4), 1022-1028.

YAGUCHI, T., KONNO, M., KAMINO, T. & WATANABE, M. (2008). Observation of three-dimensional
elemental distributions of a Si device using a 360 degrees-tilt FIB and the cold field-emission
STEM system. Ultramicroscopy 108(12), 1603-1615.

YAMAMOTO, Y., BRADY, M.P., LU, Z.P., LIU, C.T., TAKEYAMA, M., MAZIASZ, P.J. & PINT,
B.A. (2007). Alumina-forming austenitic stainless steels strengthened by laves phase and MC

183
carbide precipitates. Metallurgical and Materials Transactions a-Physical Metallurgy and
Materials Science 38A(11), 2737-2746.

YAMASAKI, J., MUTOH, M., OHTA, S., YUASA, S., ARAI, S., SASAKI, K. & TANAKA, N. (2014). Analysis
of nonlinear intensity attenuation in bright-field TEM images for correct 3D reconstruction of the
density in micron-sized materials. Microscopy 63(5), 345-355.

YAZZIE, K.E., WILLIAMS, J.J., PHILLIPS, N.C., DE CARLO, F. & CHAWLA, N. (2012). Multiscale
microstructural characterization of Sn-rich alloys by three dimensional (3D) X-ray synchrotron
tomography and focused ion beam (FIB) tomography. Materials Characterization 70, 33-41.

YEDRA, L., ELJARRAT, A., ARENAL, R., PELLICER, E., CABO, M., LOPEZ-ORTEGA, A., ESTRADER, M.,
SORT, J., DOLORS BARO, M., ESTRADE, S. & PEIRO, F. (2012). EEL spectroscopic tomography:
Towards a new dimension in nanomaterials analysis. Ultramicroscopy 122, 12-18.

YEDRA, L., ELJARRAT, A., REBLED, J.M., LOPEZ-CONESA, L., DIX, N., SANCHEZ, F., ESTRADE, S. &
PEIRO, F. (2014). EELS tomography in multiferroic nanocomposites: from spectrum images to
the spectrum volume. Nanoscale 6(12), 6646-6650.

YEOH, C.S.M., ROSSOUW, D., SAGHI, Z., BURDET, P., LEARY, R.K. & MIDGLEY, P.A. (2015). The
Dark Side of EDX Tomography: Modeling Detector Shadowing to Aid 3D Elemental Signal
Analysis. Microscopy and Microanalysis 21(3), 759-764

YU, K. & CHEN, J. (2009). Enhancing Solar Cell Efficiencies through 1-D Nanostructures.
Nanoscale Research Letters 4(1), 1-10.

YU, Q., LEGROS, M. & MINOR, A.M. (2015). In situ TEM nanomechanics. Mrs Bulletin 40(1), 62-
68.

YURTSEVER, A., WEYLAND, M. & MULLER, D.A. (2006). Three-dimensional imaging of

nonspherical silicon nanoparticles embedded in silicon oxide by plasmon tomography. Appl.
Phys. Lett. 89(15), 151920.

ZALUZEC, N.J. (2004). XEDS Systems for the Next Generation Analytical Electron Microscope.
Microscopy and Microanalysis 10(S2), 122-123.

ZALUZEC, N.J. (2009). Innovative Instrumentation for Analysis of Nanoparticles: The π Steradian
Detector. Microscopy Today 17(4), 56-59.

ZANKEL, A., KRAUS, B., POELT, P., SCHAFFER, M. & INGOLIC, E. (2009). Ultramicrotomy in
the ESEM, a versatile method for materials and life sciences. Journal of Microscopy-Oxford
233(1), 140-148.

ZANKEL, A., WAGNER, J. & POELT, P. (2014). Serial sectioning methods for 3D investigations
in materials science. Micron 62, 66-78.

184
ZHANG, D., OLEYNIKOV, P., HOVMOLLER, S. & ZOU, X. (2010). Collecting 3D electron
diffraction data by the rotation method. Zeitschrift Fur Kristallographie 225(2-3), 94-102.

ZHANG, Q., XIE, J., LEE, J.Y., ZHANG, J. & BOOTHROYD, C. (2008). Synthesis of Ag@agau metal
core/alloy shell bimetallic nanoparticles with tunable shell compositions by a galvanic
replacement reaction. Small 4(8), 1067-1071.

ZIESE, U., DE JONG, K.P. & KOSTER, A.J. (2004). Electron tomography: a tool for 3D structural
probing of heterogeneous catalysts at the nanometer scale. Applied Catalysis a-General 260(1),
71-74.

ZREIBA, N.A. & KELLY, T.F. (1988). Absorption and fluorescence corrections of characteristic x-
rays from thin spheres. X-Ray Spectrometry 17(6), 229-238.

185
Appendix A - Conditions for successful STEM-EDX tomography

Optimising individual projections

The STEM-EDX tilt series’ used in both chapters 5 and 6 suffer from extremely low count

statistics, for the reasons described in section 2.3. An example single pixel spectrum is

presented in Figure A1 to demonstrate the low counts that can be expected in the case of small

nanoparticle samples.

Figure A1. EDX spectrum from the single pixel indicated in the inset image, taken from a

spectrum image at 0° tilt of the high-Au content AgAu nanoparticle described in chapter 6. A

summed map of Au counts is shown inset.

For the small (less than 30 nm thickness) nanoparticles used in chapters 5 and 6, the conditions

used here resulted in single pixel spectra with between 0 – 5 counts for both the Au and Ag

peaks analysed. While the results in chapter 6 still demonstrate that this number of counts can

be utilised to reconstruct with some information present, it is certainly favourable to try to

increase counts, as could be performed by binning pixels in this case. In order to achieve

maximum counts per pixel, it is advisable to attempt to make the pixel size as large as possible,

without degrading resolution in the 2D spectrum images.

186
In the case of the stainless steel sample described in chapter 7, where the sample can tolerate

a larger dose, single pixel counts per X-ray peak ranged from 0 – 60. A single pixel spectrum

from this tilt series is displayed in Figure A2.

Figure A2. EDX spectrum from the single pixel indicated in the inset image, taken from a

spectrum image at 0° tilt of the stainless steel SS316 needle sample described in chapter 7. A

summed map of Fe counts is shown inset.

Reconstruction algorithms and number of iterations

In chapters 5 and 6 a SIRT algorithm with 5 iterations was used to reconstruct the STEM-EDX

tilt series’. The choice of reconstruction algorithm and in the case of SIRT, the number of

iterations, is not straightforward and here has been carried out through qualitative inspection of

the reconstructions at a number of different conditions. Reconstructions of the tilt series’ used in

chapter 6 of an AgAu nanoparticle are shown in Figure A3 (for Au) and Figure A4 (for Ag). Here,

a SIRT algorithm is preferred over weighted backprojection due to an improvement in the

sharpness of the reconstruction and reduction in intensity of streaking artefacts and of the

missing wedge.

187
Figure A3. XY slices of reconstructions of the Au maps of a STEM-EDX tilt series of one of the

AgAu nanoparticles described in chapter 6, from a tilt series acquired with 5° angular

increments over ±70°.

188
Figure A4. XY slices of reconstructions of the Ag maps of a STEM-EDX tilt series of one of the

AgAu nanoparticles described in chapter 6, from a tilt series acquired with 5° angular

increments over ±70°.

The number of SIRT iterations was chosen based on the point at which high frequency

information became apparent in the reconstructions. As mentioned throughout this thesis,

STEM-EDX almost always has very low signal levels and SIRT is known to enhance high

frequency information, including noise, as the number of iterations increase (Louis & Natterer,

1983; Defrise, 1988). Therefore, for the low signal-to-noise data obtained with STEM-EDX

tomography it may be the case that fewer iterations are used than for similarly acquired HAADF

tilt series'. Here, while the sharpness of the exterior of the nanoparticle improves up to 5

iterations in the Au reconstructions (Figure A3), subsequent iterations appear to act only to

resolve small features within the nanoparticle, which can be associated with noise in the initial

projections.

189