University of Southern California Dissertations and Theses

Real time surface analysis of complex oxide thin films during pulsed laser deposition

(USC Thesis Other)

Real time surface analysis of complex oxide thin films during pulsed laser deposition

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content Copyright 2021 Thomas Orvis

Real Time Surface Analysis of Complex Oxide Thin Films
During Pulsed Laser Deposition

by

Thomas Orvis

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
MATERIALS SCIENCE

May 2021

ii

For the Squid,
for the Hare,
and for the Sea.
iii

Acknowledgements

If I were to give half the acknowledgments that were owed, this already sizable document
would double in length. That said, I'll do my best not to (inevitably) forget anyone.
First off: an inordinate amount of this burden (and so many others) was carried by my
partner, Eli — for which I will forever be in your debt. Thank you. Your presence and
voice has been responsible for so many of the good decisions I've made in my life, and so
much of the good that has come from them, that I can't imagine where I would be, or
what I would do, without you. We did it.
This dew drop world -
is but a drop of dew,
and yet, and yet…
- Kobayashi Issa
The sheer volume of patience exhibited by my advisor, Dr. Ravichandran, has bordered
on the obscene. I feel incredibly lucky that our paths crossed as they did, because I owe
so much of who I have become as a scientist to your mentor- and friendship. Getting this
degree has been both more and less difficult than I originally anticipated. One of my
favorite (terrible) jokes about grad school is: if you have never thought about quitting
grad school, then you have probably never been in grad school. However, I never actually
seriously considered leaving, because your staunch support has always been, in a word,
unwavering. Even when things got bad, and they did, I could rely on your application of
gentle and persistent pressure to keep me moving in the right direction. After my first
year, someone asked me how it had been so far. I replied, realizing how true it was as I
said it, that it felt like becoming the person I had been aspiring to be. That's never been
more true than now. If, somehow, you've yet to realize it, I'll use this opportunity to say:
I have the utmost appreciation and respect for your intelligence, wisdom, kindness, and
passion for science. Thank you for everything you have taught me, and everything you've
done to help me get this far. Thank you.
Mythili Surendran, my first real mentee. Beyond the simple fact that without all of your
help I would never have gotten anything done, collaborating with you has consistently
elevated my understanding and appreciation of the work we do. The skill sets we each
brought to the table were quite different, but naturally complementary. I certainly
benefited enormously from your keen analytical perspective, and like to think that it
helped me improve my ability to do the same. I have a deep respect for your obstinate
persistence in overcoming obstacles, which is probably why we works so well together,
because otherwise I doubt you would have put up with my continuous refusal to answer
iv

questions I thought you should be able to answer yourself, or my dogged insistence that
you should know how to take everything apart and fix it. Most of all, I am grateful for
your continued presence as such an excellent friend.
I’m lucky to have had such charming and lovely lab mates, and I truly appreciate all of
your support. Harish, I’m excited to see where you go from here, and have the utmost
faith that you’ll do great things. Don’t let the probe boss you around too much.
Working with John Curulli and Matt Mecklenburg at CNI has been a great experience,
and their support helped me reach the finish line. Thank you.
Lillyan, Eric, Laws, and Philippe from Staib Instruments truly helped make this a reality
with their passionate support for the project. I can’t wait to see what else they come up
with.
All of our collaborators were a pleasure to work with, especially Tengfei, Arashdeep, and
Rohan from Washington University in St. Louis, and Alex and Shin from NIST. Thanks
for your hard work.
To everyone: thank you.

v

Table of Contents

Dedication ii

Acknowledgements iii

List of Figures viii

List of Tables xxiii

Abstract xxv

1 Introduction 1
1.1 Background and Motivation ............................ 2
1.1.1 Thin Films ................................ . 3
1.1.2 Complex Oxides .............................. 5
1.1.3 Potential and Challenges ......................... 10
1.2 Understanding the Growth Process ....................... 15
1.2.1 Pulsed Laser Deposition ......................... 15
1.2.2 In Situ Structural Characterization ................... 16
1.2.3 In Situ Compositional Characterization ................ 18
1.2.4 Dynamic Thin Film Growth Phenomena ................ 22
1.3 Summary ................................ ..... 25

2 M aterial Synthesis and Characterization
2.1 Introduction ................................ ... 26
2.2 Synthesis ................................ ..... 27
2.2.1 Substrate ................................ . 27
2.2.2 Pulsed Laser Deposition ......................... 39
2.2.3 Growth Parameters ............................ 42
2.3 In Situ Characterization ............................. 51
2.3.1 Reflection High Energy Electron Diffraction ............. 51
vi

2.3.2 Auger Electron Microscopy ....................... 54
2.4 Microscopy ................................ .... 63
2.4.1 Atomic Force Microscopy ........................ 63
2.4.2 Scanning Electron Microscopy ...................... 64
2.4.3 Transmission Electron Microscopy ................... 66
2.5 Structural Characterization .......................... 68
2.5.1 Thin Film X-Ray Diffraction ..................... 69
2.6 Compositional Characterization ......................... 71
2.6.1 Energy Dispersive X-Ray Spectroscopy ................ 71
2.6.2 Electron Energy Loss Spectroscopy .................. 73
2.7 Analysis ................................ ...... 74
2.7.1 Density-Functional Theory ...................... 74
2.7.2 Modeling and Simulations ....................... 75
2.7.3 Spectral Processing ........................... 80

3 In Situ Auger Electron Spectroscopy
3.1 Introduction ................................ ... 86
3.2 The AugerProbe ................................ . 87
3.2.1 Incorporation ............................... 87
3.2.2 Use ................................ ..... 89
3.3 Capabilities ................................ .... 92
3.3.1 Observables ................................ 92
3.3.2 Limits ................................ ... 95
3.4 Spectral Analysis ................................ . 98
3.5 Monitoring Composition in a LaMnO
3
/CaTiO
3
Superlattice ........ 101
3.5.1 Experimental Design .......................... 102
3.5.2 Data and Analysis ........................... 103
3.6 Real Time Growth Observation ........................ 107
3.7 Summary ................................ .... 111

4 In Situ Compositional Quantification
4.1 Introduction ................................ .. 112
4.2 Film Sensitivity to Growth Parameters ................... 113
4.2.1 Laser Fluence and Spot Size ...................... 114
4.2.2 Background Pressure .......................... 116
4.2.3 Summary ................................ 118
4.3 Structural Correlation to Composition .................... 119
4.4 Homoepitaxial SrTiO
3
Growth-Parameter Series .............. 120
4.4.1 X-Ray Diffraction ........................... 121
4.4.2 Auger Spectra .............................. 122
vii

4.5 Direct Observation of Parameter Dependence ................ 124
4.6 Parameter-Free Escape Depth Model ..................... 126
4.6.1 Principle and Description ....................... 126
4.6.2 Sensitivity to Termination ....................... 127
4.7 Termination Analysis ............................. 129
4.8 Quantification ................................ . 131
4.9 Summary ................................ .... 132

5 Direct Observation and Control of Surface Termination
5.1 Introduction ................................ .. 133
5.2 Compositional and Structural Sensitivity of Properties ........... 134
5.3 Dynamic Thin Film Growth Phenomena ................... 137
5.4 Surface Termination .............................. 138
5.4.1 LaAlO
3
/SrTiO
3
............................. 138
5.4.2 SrRuO
3
................................ . 139
5.5 Surface Energy Calculations .......................... 141
5.6 Controlling Surface Termination in Homoepitaxial SrTiO
3
......... 144
5.6.1 Experimental Design .......................... 144
5.6.2 Data and Analysis ........................... 148
5.7 Extension to SrZrO
3
.............................. 151
5.8 Summary ................................ .... 153

6 Future Work and Conclusions
6.1 Future Work ................................ .. 154
6.2 Conclusions ................................ ... 156

References 159
Appendix A: M ATLAB Scripts 192

viii

List of Figures

Figure 1.1 (a) Example of ancient Egyptian application of gold leaf on a model collar
of Hapiankhtifica ca. 1981–1802 B.C. and (b) a gilded glass earring ca. 100
B.C. (c) A tumbaga necklace of the Tairona people from pre-Columbian
Colombia, ca. 10
th
– 14
th
century A.D. The high copper content accounts
for the rosy tonality. Images distributed under CC0 by The Metropolitan
Museum of Art. ......................................................................................... 2
Figure 1.2 An illustration of thin film deposition, with a new material arriving at the
atomically-stepped surface of a single crystal or epitaxial thin film. Image:
the author. ................................................................................................ 3
Figure 1.3 (a) Cross-sectional TEM of a finFET, illustrating the precisely layered
structure. The image is approximately 80 nm across. (b) Cross-sectional
TEM of stacked gate all-around nanowire transistors. Image (a): (Biring,
2014); (b): (Courtland, 2016). ................................................................... 4
Figure 1.4 An illustration of the proposed MESO transistor architecture, which
operates via spin-orbit transduction coupled with electromagnetic
switching. Image: (Manipatruni et al, 2019). ............................................ 4
Figure 1.5 Crystal structures of La
0.7
Sr
0.3
MnO
3
(left) and YBa
2
Cu
3
O
6.9
(right). Note
that due to the distribution of La and Sr in the material, they are
represented equivalently in the La
0.7
Sr
0.3
MnO
3
structure. Image: the author.
.............................................................................................................. 5
Figure 1.6 The periodic table showing which elements may be used for the A, B, and
X-site of the ABX
3
perovskite structure with 100% occupation. Image: the
author, based on data from (Schlom et al, 2008). ..................................... 6
Figure 1.7 Ideal cubic perovskite structure ABX
3
, of space group Pm3m, with A
cations occupying corner sites, B cation occupying cube center, and X
anions occupying face centers, forming a BX
6
octahedra. Image: the author.
.............................................................................................................. 6
ix

Figure 1.8 The original calculation for the band structure of SrTiO
3
, plotted as energy
versus wave vector for all lines of symmetry in the Brillouin zone. Image:
(Kahn et al, 1964). .................................................................................... 7
Figure 1.9 A demonstration of the geometry of bond formation for SrTiO
3
in bonding
and antibonding states. Image: (Dylla et al, 2019).................................... 7
Figure 1.10 The BaTiO
3
structure in the cubic 𝑃𝑚 3𝑚 phase above 393 K (left) and
tetragonal 𝑃 4𝑚𝑚 phase above 263 K (right). Note the distortion of the
tetragonal phase (exaggerated for clarity), with the relative shifts of the
central oxygen atoms highlighted with a dashed line. Image: the author. . 8
Figure 1.11 An illustration of the principle of broken symmetry: at high energy there
is one available state, but at low energy, even though the solution is still
symmetric, the outcome will not be. Image: the author. .......................... 10
Figure 1.12 Spontaneous magnetization with the two possible solutions (±𝑚 0) (solid
line) and in an applied field (dashed line). Image: (Schwabl, 2006). ...... 12`
Figure 1.13 Helmholtz free energy as a function of magnetization at 𝑇 > 𝑇𝐶 , 𝑇 = 𝑇𝐶 ,
and 𝑇 < 𝑇𝐶 , showing the symmetric split of solution states indicative of
broken symmetry. Image: (Schwabl, 2006). .............................................. 13
Figure 1.14 An illustration of the properties that can arise from different forms of
broken symmetry, as well as the relationship between their electronic
degrees of freedom. Image: the author, inspired by (Hwang et al, 2012). . 14
Figure 1.15 The Rutherford backscattering spectrum from the as-grown YBa
2
Cu
3
O
6.9

film on carbon, showing remarkable consistency with the simulation and
demonstrating the efficacy of pulsed laser deposition for compositional
transfer of complex stoichiometry oxides. Image: (Dijkkamp et al, 1987). ..
............................................................................................................. 15
Figure 1.16 Photograph of the pulsed laser deposition system used in this work during
the ablation of a target. The electron beam, heater and substrate, target,
and laser path are indicated. Image: the author. .................................... 15
Figure 1.17 Cross-sectional low-magnification high angle annular dark field image of a
(Ba,Ca)(ZrTi)O
3
/SrRuO
3
/GdScO
3
sample (left), magnification of the
(Ba,Ca)(ZrTi)O
3
/SrRuO
3
interface (center) and SrRuO
3
/GdScO
3
interface
(left) overlaid with atomic models illustrating octahedral tilt. Image: (Liu
et al, 2019). .............................................................................................. 16
Figure 1.18 Typical before (top) and after (bottom) images of substrates (small
rectangles, ~3 x 5 mm and 5 x 5 mm, top to bottom) mounted to resistive
heaters with silver paint. The colors seen on the heater in the bottom image
x

are the result of thin-film interference, rather than the color of the
deposition itself. Image: the author. ......................................................... 17
Figure 1.19 (right) An example of a reflection high energy electron diffraction pattern
from a SrTiO
3
single crystal substrate surface at 800℃ in 10
-4
mbar O
2
in
our pulsed laser deposition chamber. The direct beam can be seen above
the diffraction pattern, with the horizontal shadow below it from the edge
of the heater. The large ring around the direct beam is from scatter through
the glass on which the phosphor is applied, as well as the viewport behind
it. Note the vertical habit of the diffraction pattern, indicating a smooth
2D surface, and the Kikuchi lines from inelastic collision processes. Also
note the doubly-spaced diffraction spots below the specular spot, which
demonstrate surface reconstruction. Image: the author. ........................... 17
Figure 1.20 An example RHEED oscillations from the deposition of 16 unit cells of
LaAlO
3
on SrTiO
3
. The sixteen clear oscillations in intensity are the result
of the surface roughening (decrease) and then smoothing again (increase).
The decay in intensity over the growth is not uncommon and has many
possible causes. Image: the author. .......................................................... 18
Figure 1.21 An illustration of the Auger process, in which incoming energy (E
Incoming
)
knocks out a core-level electron. A higher-level electron falls to take its
place, releasing energy as it does so, which proceeds to knock out an even
higher energy electron, which is then the Auger electron. An important
distinction of this process is the Auger electron energy’s independence from
E
Incoming
. ................................................................................................... 19
Figure 1.22 Characteristic Auger transition energies as a function of atomic number.
The shells associated with the transition are labeled and coded by color.
Image: the author, with data compiled from (Ferguson, 1989). ............... 20
Figure 1.23 Charging of an SrTiO
3
substrate at room temperature (left), compared to
the same substrate at 750℃ using the same imaging conditions. Note the
distortion of the direct beam and resulting diffraction pattern in the image
taken at room temperature. Image: the author. ....................................... 20
Figure 1.24 An illustration of possible differences in film growth modes depending on
kinetics and surface energy. If the deposition has insufficient surface
diffusion or low self-nucleation energy then island growth can dominate
(top), rather than developing layer-by-layer (bottom) into a smooth film.
Image: the author. .................................................................................... 23
Figure 1.25 An example of the growth parameter-dependence of BiFeO
3
, with marked
growths data-mined from the literature (left) and their resulting properties
xi

(right). The congregation of parameters shows the Goldilocks zone effect.
Image: (Young et al, 2018). ...................................................................... 23
Figure 1.26 Deposition of GaN on diamond with growth conditions such that the
formation of nanowires is energetically preferable to film formation. This is
an example of utilization of growth parameters for self-assembly. Image:
(Schuster et al, 2012). .............................................................................. 24
Figure 2.1 An illustration of the cubic SrTiO
3
unit cell with the corresponding atomic
coordinates (top), as compared to the octahedral distortion in CaZrO
3

(bottom left), and the position of the octahedra within the primitive cell
(dark dashed line, bottom right), shown within four cells demarcated with
a thin dashed line. Image: the author. ..................................................... 27
Figure 2.2 Examples of the planes corresponding to the identified Miller indices within
a body-centered cubic unit cell. Image: the author. ................................. 28
Figure 2.3 Illustration of the structural layering in (001) SrTiO
3
. Image: the author.
............................................................................................................. 29
Figure 2.4 The BaTiO
3
phase diagram as a function of TiO
2
mole percent. Image:
(Belruss et al, 1971). ................................................................................ 30
Figure 2.5 Illustration of the range of single crystal substrate lattice constants
available for the deposition of epitaxial complex oxide perovskite thin films.
Image: the author. .................................................................................... 31
Figure 2.6 Atomic force micrographs of single crystal substrates: as-received polished
SrTiO
3
(left), etched and annealed TiO
2
terminated SrTiO
3
(center), and
annealed NdGaO
3
(right). The height scales are 5, 5, and 3 nm (left to
right). Image: the author. ........................................................................ 33
Figure 2.7 The typical annealing time/temperature recipe used for single crystal
substrates. Image: the author. .................................................................. 33
Figure 2.8 A 10 x 5 mm (left) single crystal substrate of SrTiO
3
, and a 10 x 10 mm
(right) single crystal substrate of SrTiO
3
doped with 2% Nb, making it
conductive and, as a result, black in color. Image: the author. ................ 34
Figure 2.9 Atomic force micrographs of poorly etched SrTiO
3
surfaces showing
insufficient etching and remnant SrO (left), excessive etching forming etch-
pits (center), and excessive etching attacking dislocation faults caused when
breaking the crystal (right). The image height scales are 5 nm, 20 nm, and
20 nm (left to right). Image: the author. .................................................. 35
Figure 2.10 An atomic force micrograph of a perfectly etched and annealed TiO
2

terminated SrTiO
3
substrate. The height scale is 4 nm. Image: the author.
............................................................................................................. 35
xii

Figure 2.11 A picture of the deposition chamber interior used in this work, showing
the AugerProbe, electron gun, sample heater, and target. Image: the author.
............................................................................................................. 39
Figure 2.12 An illustration of the laser path used to control the size, shape, and energy
of the beam. Image: the author. ............................................................... 39
Figure 2.13 A picture of the PLD chamber overlayed with examples of the many
growth parameters which must be considered during the deposition process.
Image: the author. .................................................................................... 40
Figure 2.14 Photographs of plume formation from laser-ablated targets. The difference
in shape between the two images arises from the different points in time
the pictures were taken during plume formation, as well as the difference
in laser energy and background pressure. Image: the author. .................. 41
Figure 2.15 Qualitative profile of electron temperature (T
e
) for the three initial stages
before ablation: (1) – absorption of radiation, (2) – propagation of the
thermal wave, (3) – electron-lattice relaxation, indicated by temperature
equilibrium between the electrons and the lattice (T
i
). Image: (Phipps,
2007). ....................................................................................................... 42
Figure 2.16 SEM images of microprocessing conducted with a ns pulsed 266 nm laser
(left), and a 100 fs pulsed 780 nm laser (right), on glass, demonstrating the
different ablative methods for different pulse-widths, where the ns pulse
results in more heat being transferred to the surface. Image: (Lucas et al,
2012). ....................................................................................................... 43
Figure 2.17 The relationship between pressure and mean free path (𝜆 ) for the gas
temperatures indicated, assuming the kinetic diameter for O
2
of
approximately 346 pm. The different pressure regimes are demarcated for
the 75 mm target-substrate distance used in our chamber. Image: the
author. ..................................................................................................... 45
Figure 2.18 Illustration of the interfacial energies at play when determining the
behavior of nucleation and growth in a thermodynamically stable system,
where 𝛾𝑓𝑠 , 𝛾𝑠𝑣 , and 𝛾𝑓𝑣 are the interfacial energies between the film and
substrate, substrate and vapor, and film and vapor, respectively. Image:
(Ohring, 2001). ......................................................................................... 47
Figure 2.19 Illustration of the step-flow growth mode, in which surface roughness is
not changed despite deposition due to high surface diffusivity which allows
transport to the low-energy step edge locations. Image: the author. ........ 48
Figure 2.20 RHEED oscillations observed during homoepitaxial deposition of SrTiO
3
.
Note the sustained layer-by-layer growth until lost, with 38 oscillations
xiii

observed in 1500 s, corresponding to approximately 15 nm of deposition.
Image: the author. .................................................................................... 50
Figure 2.21 A picture of SrTiO
3
(001) surface diffraction with RHEED, collected with
a 5 kV electron source at 10
-4
mbar O
2
. Image: the author. ..................... 51
Figure 2.22 An illustration of the Ewald sphere construction for a 2D surface (left),
with the resulting reciprocal lattice rods, image distributed under CC:BY –
Ponor. (Right) an image of the diffraction pattern formed with RHEED
when the surface is 3-dimensional. Shown is a CaZrO
3
thin film with a 35
kV beam at 10
-4
mbar O
2
. Image: the author. .......................................... 52
Figure 2.23 Contrasting diffraction patterns from CaTiO
3
(left) and LaMnO
3
(right),
showcasing the different structures and thus different patterns observed.
Image: the author. .................................................................................... 52
Figure 2.24 A demonstration of the principle behind RHEED oscillations, showing the
higher intensity observed from a smooth surface, corresponding to a plotted
peak, compared to the lower intensity observed from a rough surface,
corresponding to a plotted trough. Image: the author. ............................. 53
Figure 2.25 RHEED oscillations from 3 different growths (left) showing consistent
long-term oscillations (top), quickly lost oscillations (center), and odd
intensity behavior (bottom). Overlapping RHEED oscillations (right)
during the deposition of CaZrO
3
, showing two clear signals and an envelope.
Image: the author. .................................................................................... 53
Figure 2.26 Diagrammatic representation of the K and L energy levels. Image: the
author. ..................................................................................................... 56
Figure 2.27 Comparison of the models used to approximate contributions to the
inelastic mean free path from electron-electron scattering (Eq. 2.14),
plasmon scattering (Eq. 2.15), and the model developed by Seah and Dench
from experimental data (Eq. 2.16). Image: the author. ............................ 59
Figure 2.28 A diagram demonstrating the basic operating principle of the AugerProbe.
Incoming electrons are energy-filtered by the collimator lens, then exposed
to the retarding field which allows the derivative signal to be determined
by lock in amplification. Image: the author. ............................................ 60
Figure 2.29 Auger spectra from 20 eV to 2 keV, showing the presence of characteristic
peaks in the E*N(E) signal, as well as the dN/dE signal. Image: the author.
............................................................................................................. 61
Figure 2.30 An illustration of the relationship between direct RFA signal and its
derivative. Image: (Winklehner, 2013). .................................................... 62
xiv

Figure 2.31 The principle of atomic force microscopy, illustrated as a cantilever being
monitored with a photodiode observing a reflected laser (right), and (left)
a 3D plot of an atomic force micrograph of a SrTiO
3
surface showing atomic
steps. Visible peaks are undesirable dirt. Image: the author. ................... 63
Figure 2.32 Plotted height (right) for the line shown in the micrograph (left). The
observed steps are 1 unit cell high (4 Å), showing the successful etch and
termination control of SrTiO
3
. Image: the author. ................................... 64
Figure 2.33 An illustration of the operating principle for scanning electron microscopes.
Image: (Marturi, 2013). ............................................................................ 65
Figure 2.34 A scanning electron micrograph of an ablated polycrystalline CaZrO
3

target surface showing the ablated region (dark, smooth) in contrast with
the unablated region (light, rough). Image: the author. ........................... 65
Figure 2.35 Illustration of the lens configurations for imaging (left) versus diffraction
(right) modes in a standard TEM. Image distributed under CC:BY – Eric
Kvaalen. ................................................................................................... 66
Figure 2.36 A sample image of self-assembled polymer nanoparticles taken with a
Fischer Scientific Talos CryoTEM. Image: the author. ............................ 67
Figure 2.37 Illustration of the principle of Bragg’s law, in which constructive
interference occurs when the difference in path length is equal to an integer
value of the wavelength. Image distributed under public domain. ........... 68
Figure 2.38 Schematic for XRD instrument configuration (left), showing 𝜃 and 2𝜃 . The
Ewald sphere construction illustrating the diffraction condition, in which
the change in wave vector between the incident and scattered X-ray equals
a reciprocal lattice vector. Image: the author. .......................................... 69
Figure 2.39 Illustration of the geometry used to generate a rocking curve (left) with an
example rocking curve taken from off-stoichiometry homoepitaxial SrTiO
3

(right), which caused the visible shoulder. Image: the author. ................. 69
Figure 2.40 X-ray diffraction of a BaZrO
3
thin film 97 nm thick grown on a (101)
DyScO
3
substrate. The quality of the film is illustrated by the Pendellösung
fringes indicating a smooth surface and interface. The BaZrO
3
rocking
curve, shown inset, has a narrow full width at half maximum, supporting
the claim of film quality. Image: the author............................................. 70
Figure 2.41 An example energy dispersive X-ray spectrum from BaZrO
3
. Image: the
author. ..................................................................................................... 71
Figure 2.42 Energy dispersive X-ray spectroscopy map of an ablated CaZrO
3
target,
with the electron micrograph shown in the bottom left. The colored images
are elemental maps for the labeled elemental peaks, and indicate that there
xv

was no preferential ablation of this target, as the ablated and unablated
regions have the same composition. Image: the author. ........................... 72
Figure 2.43 (a) Image showing an LaAlO
3
capping layer, SrRuO
3
layer and SrTiO
3

substrate. (b) Elemental maps for Ti L edge, O K edge and La M edge, for
the region highlighted as white box in (a). Each elemental map is
normalized within itself. (c) Extracted EEL spectra for Ti L edge, O K edge
and La M edge, where the color of each spectrum corresponds to the region
of same color highlighted as boxes in (a). Image: taken by collaborators. 73
Figure 2.44 Illustration of the impact stacking sequence has on escape-depth signal, for
two arbitrary materials with equal arbitrary thickness, such that 𝜆 = 1,
where Depth = 1 = 𝜆 . The deviation in signal is immediately visible, even
for only 2 layers of each material generating signal. Image: the author. .. 76
Figure 2.45 Demonstration of the peak-to-peak measurement for an O
KLL
peak plotted
in N(E): the dN/dE signal has a clear trough which can be subtracted from
the clearly defined peak. Image: the author. ............................................ 82
Figure 2.46 An example of a situation where a clear peak in the dN/dE signal is not
observed, because the energy-width of the Sr
MNN
scan is too narrow, even
though the peak can clearly be seen. Image: the author. ......................... 83
Figure 2.47 Incremental observation of the deposition of LaAlO
3
on SrTiO
3
, 1 unit cell
at a time, as indicated, up to 7 unit cells of LaAlO
3
on SrTiO
3
. Note the
shift in signal, both to lower energy from the Sr
MNN
peak at around 85 eV
to the La
NVV
peak around 76 eV, and the loss of the Sr
MNN
peak around 60
eV. Image: the author. ............................................................................. 84
Figure 2.48 Illustration of the peak area selected for an AUC measurement, with a
given start and end point. Image: the author. .......................................... 84
Figure 3.1 The Auger probe in 55 mm position for spectral acquisition, 30° off the
normal axis. Image: the author. ............................................................... 87
Figure 3.2 Electron diffraction pattern of a poorly aligned SrTiO
3
substrate viewed
on a phosphor screen damaged after excessive deposition-buildup. Image:
the author. ............................................................................................... 88
Figure 3.3 Comparison of the diffraction pattern formed from a SrTiO
3
substrate with
a 10 keV beam (left) and a 35 keV beam (right). Note the higher order
diffraction spots on the lower energy diffraction pattern. Image: the author.
............................................................................................................. 90
Figure 3.4 The portion of the AugerProbe outside of the growth chamber, with arrows
identifying the two (of three) visible axes used to control probe axial tilt.
Image: the author. .................................................................................... 91
xvi

Figure 3.5 Auger spectra of 24 elements, intensity adjusted for visibility and
comparison, plotted as E*N(E) (top) and N(E) (top, inset), as well as
dN/dE (bottom). Numbers correspond to spectra and peak identifications
(for peaks marked with arrows) found in Table 8. Image: the author. ... 93
Figure 3.6 Normalized O, Ga, and Nd peak-to-peak intensities of spectra collected
with 10 second total dwell time as a function of chamber pressure, and
(inset) the O
KLL
N(E) spectra with denoted pressure, intensity shifted
linearly for clarity. Approximately 50% of the oxygen Auger signal is lost
at 5x10
-3
mbar and effectively all of it is lost by 5x10
-2
mbar, as illustrated
by the decreasing oxygen peak intensity shown in the inset. The shift in
oxygen peak energy is very likely due to systematic errors arising from the
instrument conditions. The dashed line is a guide to the eye. Image: the
author. ..................................................................................................... 95
Figure 3.7 Peak shape and position comparison for spectra of characteristic O, Nd,
and Ga peaks at increasing background partial pressures of oxygen. For all
three elements, the top spectra were collected at 5x10
-7
mbar, with each
successive row below it increasing by an order of magnitude until the
bottom spectra, which were taken at 5x10
-3
mbar. The dashed line connects
the maximum position for each peak, showing that the energy shift observed
is consistent across all three elements. Image: the author. ....................... 96
Figure 3.8 The Mn
LMM
and La
MNN
peaks collected from LaMnO
3
thin films of
increasing thicknesses between depositions, identified in the legend. Note
the overlap between the La
MNN
peak located at approximately 636 eV and
the Mn
LMM
satellite peak located at approximately 643 eV. Spectra shown
are 5 summed scans. Image: the author. .................................................. 97
Figure 3.9 (right) The La
NVV
peak collected from LaMnO
3
thin films of increasing
thicknesses between depositions, identified in the legend. Despite the
appearance of low signal intensity due to the large secondary electron
background, the peak can be consistently measured. Image: the author. . 97
Figure 3.10 The O
KLL
peak plotted as N(E). The top panel shows the accompanying
dN/dE signal with circles indicating the data points from which the peak-
to-peak value is calculated by subtracting the local minimum from the local
maximum. The bottom panel is a visual representation of the trapezoidal
numerical integration method, with the spectra split into trapezoids of equal
width. Image: the author. ........................................................................ 98
Figure 3.11 Examples of spectra plotted as E*N(E) and dN/dE, with visual examples
of the area under the curve and peak-to-peak methods of quantifying
xvii

intensity, La and Mn peak-overlap, and narrow scan summing. Image: the
author. ................................................................................................... 100
Figure 3.12 Electron diffraction data from the deposition of a CaTiO
3
/LaMnO
3

superlattice on a NdGaO
3
single crystal substrate. (a) RHEED oscillations
during growth, with four oscillations per material, per layer, and distinction
between the two materials indicated by background color, with the lower
intensity oscillations from the LaMnO
3
and the higher intensity oscillations
from the CaTiO
3
. (b-d) Electron diffraction patterns demonstrating
smoothness of the as-grown film after growth (b), after deposition of four
atomic layers of CaTiO
3
(c), and after deposition of four atomic layers of
LaMnO
3
(d). Image: the author. ........................................................... 102
Figure 3.13 Cartoon of the stacking sequence of deposition and Auger spectra
acquisition for the CaTiO
3/
LaMnO
3
superlattice. The alternating colors
correspond to the regions deposited between stopping for spectroscopy, as
do the numbers. Image: the author. ....................................................... 103
Figure 3.14 A plot of normalized (to the oxygen signal) elemental Auger spectra signals
during the growth of a CaTiO
3
/LaMnO
3
superlattice. Note that the A-site
intensity (Ca and La) data is offset from the B-site intensity (Ti and Mn)
data for clarity. Each data point is from an area under the curve calculated
from scans of the characteristic peak for that specific element, with 10
seconds total dwell time, normalized to the oxygen signal from that specific
collection. The Auger spectra were collected in situ between depositions of
known thickness, as calibrated by RHEED and previous growths. Lines are
exponential fits to the data in approximation of escape depth as outlined
in the text. Image: the author. ............................................................... 104
Figure 3.15 Fits to the normalized intensity data of CaTiO
3
/LaMnO
3
superlattice
layers during deposition. Comparing the fits (long dash) to the calculated
escape depths (short dash) for these elements reveals that the escape depth
is proportional to the Auger electron energy, as expected, and has
agreement within one unit cell (0.4 nm) for all elements. Image: the author.
........................................................................................................... 106
Figure 3.16 Illustration of the Sr
MNN
peak when plotted as E*N(E). Image: the author.
........................................................................................................... 107
Figure 3.17 Demonstration of the pulse-probe method’s ability to observe RHEED
oscillations despite the long recovery time. Shown here with 3 pulses
between each scan. Image: the author. ................................................... 108
xviii

Figure 3.18 Comparison between the high energy Ti
LMM
peak located at 381 eV and
the Ti
MVV
peak located at 36 eV, with energy scale and intensity adjusted
for demonstration. Image: the author. ................................................... 109
Figure 3.19 Measured intensity of RHEED specular spot (noted as RHEED intensity)
and AES data simultaneously collected using the pulse-probe method
outlined in the text for the growth of homoepitaxial SrTiO
3
. The Auger
lines used to monitor Sr and Ti are Sr
MNN
, and Ti
LMM
transitions in (a),
and to monitor O, Sr and Ti are O
KLL
, Sr
MNN
, and Ti
LMM
transitions in (b).
Laser pulses result in a roughening of the surface (a rapid decrease in
RHEED intensity) followed by recovery during Auger acquisition (gradual
increase in RHEED intensity). Depending on the duration of the spectral
collection and the number of pulses between scans, the shape of the
resulting oscillations will vary with either long scans and long recovery time
as shown in (a), or short scans and thus short recovery as shown in (b).
Likewise, the sensitivity of the scans to compositional changes is dependent
on the number of data points and their integration time, resulting in a
trade-off between the quality of RHEED and AES data. Image: the author.
........................................................................................................... 110
Figure 4.1 Laser spot sizes on thermal paper for the ‘R2’ mask at different lens
positions. Relative to the spot with best focus, labeled with an arrow,
increasing sizes correspond to 5 mm increments of lens transit away from
the chamber, and decreasing sizes are 5 mm increments towards the
chamber. Ruler present for scale: 1 mm increments. Image: the author. 115
Figure 4.2 A simple Monte Carlo simulation of the pressure-dependent flight paths
between target and substrate (bottom) and final particle distribution on
the substrate surface (top) illustrates a significant deviation in deposition
composition for Sr and Ti arising solely from the difference in atomic mass.
Comparing high (left) versus low (right) pressure shows the dramatic
influence of growth parameters on the resulting composition. Image: the
author. ................................................................................................... 116
Figure 4.3 Simple Monte Carlo simulation showing the difference in projected flight
paths for Sr and Ti when the species are elemental (left) versus oxidized
(right). Note that the pressure in both simulations is the same, and
therefore the dispersion difference is purely the result of less disparity in
the species’ mass. Image: the author. ..................................................... 117
Figure 4.4 Illustration of octahedral tilt accompanying oxygen vacancies in cubic
perovskites, thereby causing lattice expansion. Image: (Marrocchelli et al,
2015). ..................................................................................................... 119
xix

Figure 4.5 Excellent SrTiO
3
surface showing atomic steps after etch treatment.
Vertical scale is 5 nm. Image: the author. .............................................. 120
Figure 4.6 XRD and AES E*N(E) of (001) oriented homoepitaxial STO thin films.
The primary substrate (002) peak at 2θ = 46.51° (lightly dashed line) is
used as a reference to determine c-axis expansion of the thin films by
comparing its position to the lower 2θ (002) peak position (heavily dashed
line). A clear trend can be observed in which the medium laser fluence films
(1.8 J cm
-2
) have the least lattice expansion, while the low laser fluence
films (0.8 J cm
-2
) have the greatest, and higher pressure exacerbates the
low laser fluence trend. Comparing the Sr
MNN
and Ti
MVV
Auger spectra
shows a similar visible trend, in which higher laser fluence increases the
intensity of the Ti
MNN
peak, while lower laser fluence and higher pressure
increase the intensity of the Sr
MNN
peak. Image: the author. ................. 122
Figure 4.7 Auger signal intensity quantified as the ratio of the summed Sr
MNN
peak-
to-peak value to that of Ti
MVV
, for nine homoepitaxial SrTiO
3
thin films
grown with the indicated laser fluences and pressures (bottom), as well as
their c-axis lattice constants as derived from thin film XRD (top).
Horizontal dashed lines indicate the Sr/Ti ratio and c-axis lattice constant
of the single crystal SrTiO
3
substrate. Image: the author. ..................... 123
Figure 4.8 Sr/Ti ratio of the P2P calculated from summed Auger spectra collected
after the deposition of every two unit cells of homoepitaxial SrTiO
3
, as a
function of the number of deposited layers. The laser fluence was changed
periodically during the deposition, as indicated by the shape and color of
the markers. A single deposition of two unit cells of TiO
2
(equivalent to
two half unit cells of SrTiO
3
) was performed after depositing 70 unit cells
of SrTiO
3
, demarcated as a star. The original composition of the SrTiO
3

substrate, as determined with AES, is indicated with a thin dashed gray
line. The results of a parameter-free escape depth model used to explain the
compositional variance as a function of deposition is shown as thick dashed
lines. Image: the author. ........................................................................ 124
Figure 4.9 The data from the growth originally shown in Figure 4.8, in comparison to
the parameter-free escape depth model that accounts for termination. The
termination model is clearly a better fit to the data, with χ
2
= 0.011, versus
χ
2
= 0.370 for the bulk model. Image: the author. ................................. 130
Figure 4.10 Normalized Sr/Ti Auger signal versus c-axis lattice constant for the
homoepitaxial SrTiO
3
thin films presented in Figure 4.6 and 4.7, also
plotted as excess Sr quantified from the parameter free escape depth model
outlined in the text. For comparison, composition and c-axis data have been
xx

plotted from the literature,
385,389
showing agreement with our reported
trends. Image: the author. ...................................................................... 131
Figure 5.1 An illustration of the polar/nonpolar LaAlO
3
/SrTiO
3
interface, showing
the resulting formation of a 2-dimensional electron gas. Image: the author.
........................................................................................................... 135
Figure 5.2 The Ruddlesden-Popper structure of Sr
3
Ru
2
O
7
. Image: the author. ..... 137
Figure 5.3 Ionic model behavior of the LaAlO
3
/SrTiO
3
interface for either a
TiO
2
/LaO
+
interface (top) or a SrO/AlO
2
-
interface (bottom), with
compensating surface gas formation. Image: the author. ........................ 138
Figure 5.4 RHEED intensity oscillations for SrRuO
3
films grown by pulsed laser
deposition, showing an extra half-oscillation when growing on a TiO
2

terminated surface versus a SrO terminated surface. Image: (Rijnders et al,
2004). ..................................................................................................... 139
Figure 5.5 (a) The slab models of SrO-terminated and TiO
2
-terminated SrTiO
3
. (b)
Variation of surface energy of SrTiO
3
with SrO and TiO
2
termination as a
function of the chemical potential of SrO (µ
SrO
). Image: collaboration. . 141
Figure 5.6 SrO- and RuO
2
-terminated slab models of SrRuO
3
. (b) Variation in surface
energy of SrRuO
3
with SrO and RuO
2
termination as a function of the
chemical potential of SrO. Image: collaboration. .................................... 141
Figure 5.7 Layer-resolved, spin-polarized density of states (DOS) of SrRuO
3
on
SrTiO
3
substrates. The SrRuO
3
film is terminated by SrO surface. The DOS
corresponding to SrRuO
3
layer is shaded in grey color and that of SrTiO
3

are shaded blue. The SrTiO
3
or SrRuO
3
layer corresponding to the DOS is
shown in the atomic model on the right. Image: collaboration. ............ 142
Figure 5.8 Layer-resolved, spin-polarized density of states (DOS) of SrRuO
3
on
SrTiO
3
substrates. The SrRuO
3
film is terminated by RuO
2
surface. Image:
collaboration. ......................................................................................... 142
Figure 5.9 Model configuration of SrO-terminated and RuO
2
-terminated SrRuO
3

films (left side) on an SrTiO
3
substrate, as well as TiO
2
-terminated and
SrO-terminated SrTiO
3
films sandwiching an SrRuO
3
layer (right side).
Comparison of the chemical potential-dependent surface energy of the
SrRuO
3
- (top-center) and SrTiO
3
-capped (bottom-center) structures.
Image: collaboration. .............................................................................. 143
Figure 5.10 Phase diagram describing stability of SrTiO
3
(a) and SrRuO
3
(b) as a
function of the chemical potential of Sr (µ
Sr
) and O (µ
O
). The phase
boundaries are indicated by Sr-rich or Ti-rich (Ru-rich) conditions during
growth of SrTiO
3
(SrRuO
3
). Image: collaboration. ................................ 143
xxi

Figure 5.11 Illustration of the heterostructure grown with termination switching.
Image: the author. .................................................................................. 144
Figure 5.12 RHEED oscillations for the ¼ unit cell growth conducted as a ‘dummy’
deposition to determine growth rate. Image: the author. ....................... 145
Figure 5.133 Illustration of the second heterostructure grown with termination
switching. Image: the author. ................................................................. 145
Figure 5.14 (a) Wide field-of-view HAADF image showing the LaAlO
3
capping layer,
SrRuO
3
layer and SrTiO
3
substrate. Scale bar corresponds to 2 nm. (b)
Elemental maps for Ti L edge, O K edge and La M edge, for the region
highlighted as white box in (a). Each elemental map is normalized within
itself. (d) Extracted EEL spectra for Ti L edge, O K edge and La M edge,
where the color of each spectrum corresponds to the region of same color
highlighted as boxes in (a). Image: collaborators. .................................. 146
Figure 5.15 (a) Atomic resolution HAADF image showing the oxide heterostructure
with LaAlO
3
capping layer, SrRuO
3
layer and SrTiO
3
substrate. Scale bar
corresponds to 1 nm. (b) A part of the HAADF image chosen for intensity
analysis. (c) The A and B-site HAADF intensity profiles for the regions
highlighted in (b). Image: collaborators. ................................................ 147
Figure 5.16 Auger electron spectra signal intensity calculated from areas beneath the
curves for spectra collected at known thickness intervals during the
deposition of homoepitaxial SrTiO
3
on a TiO
2
-terminated SrTiO
3
substrate,
with termination switching controlled by the deposition of SrRuO
3
. The
Auger lines used to monitor Sr, Ru, and Ti are Sr
MNN
, Ru
MNN
,

and Ti
LMM

transitions and their intensities clearly track the surface termination of the
film both before and after the switching event. The dashed lines are modeled
signal intensity for the structures, described in the text, with thick lines
corresponding to the assumed switching events shown in the structure
beneath the plot, and thin lines corresponding to the same growth without
termination switching. Marker size is proportional to error. Image: the
author. ................................................................................................... 148
Figure 5.17 Auger electron spectra signal intensity calculated from areas beneath the
curves for spectra collected in real time, during the deposition of
homoepitaxial SrTiO
3
on a TiO
2
-terminated STO substrate, with
termination switching controlled by the deposition of SrRuO
3
or TiO
2
. The
Auger lines used to monitor Ti and Sr are the Ti
MVV
and Sr
MNN
transitions,
and their intensities clearly track the surface termination of the film both
before and after all switching events. The dashed lines are modeled signal
intensity for the structures, described in the text, with thick lines
xxii

corresponding to the assumed switching events shown in the structure
beneath the plot, and thin lines corresponding to the same growth without
termination switching. Marker size is proportional to error. Image: the
author. ................................................................................................... 149
Figure 5.18 A repetition of the growth shown in Figure 5.17, with improved spectral-
acquisition methods to observe Ru from the single deposited monolayer.
Image: the author. .................................................................................. 150
Figure 5.19 Auger electron spectra signal intensity calculated from areas beneath the
curves for spectra collected at known thickness intervals during the
deposition of homoepitaxial STO on a TiO
2
-terminated STO substrate
followed by epitaxial SrZrO
3
with termination switching controlled by the
deposition of SrRuO
3
. The Auger lines used to monitor Sr, Zr, and Ti are
Sr
MNN
, Zr
MNN
,

and Ti
MVV
transitions and their intensity-ratios track the
surface termination of the film both before and after the switching event.
The dashed lines are modeled signal intensity for the structures, described
in the text. Image: the author. ............................................................... 151
Figure 6.1 Example Auger spectra from the full energy range available, collected in
our PLD chamber with the AugerProbe. Image: the author. ................. 157

xxiii

List of Tables

Table 1 Examples of the diverse properties exhibited by complex oxide perovskites.
.............................................................................................................. 9
Table 2 Reported Auger peak locations in dN/dE mode. Data from Ferguson, 1989. ..... 21
Table 3 Miller index notations and their corresponding definitions. .................... 28
Table 4 Single crystal substrates used for epitaxial growth of complex oxide
perovskites, listing the available orientations, crystal structure, a and c
lattice constants, growth method, and source. If the orientation or growth
method are not listed, it indicates the information was not provided by the
supplier. For orthorhombic structures, the lattice constants listed are for
the pseudo-cubic unit cell. For growth methods, ‘Cz’ refers to Czochralski,
‘Vern’ refers to Verneuil, and ‘TSSG’ refers to top seeded solution growth.
For the sources, ⸙ means it was available from Crystec, ⸸ was from
Powerway, and ⸹ was from MTI; ※ is (Mazur et al, 1997), and ‡ is (Helden
et al, 2019). .............................................................................................. 32
Table 5 The relationship between the designations for the first three electron shells
in comparison to the optical nomenclature. ............................................. 55
Table 6 Final states for all KLL transitions, adapted from (Ferguson, 1989). ..... 57
Table 7 Outline of steps required for operation of the AugerProbe. ...................... 89
Table 8 Identification of peaks denoted by arrows for the spectra shown and
numbered in Figure 3.5, sorted by reference number (column 1). Column 2
is the energy at which the peak was observed, and column 3 is the element
responsible for the peak, with questionable identifications noted. Column 4
is the atomic number, and column 5 is the identity of the transition
responsible for the peak. The reference numbers marked with the ○ symbol
are from single crystal substrates, the ⸸ symbol indicates they are from
PLD-grown thin films, and the § symbol indicates they are from metal foil.
............................................................................................................. 94
xxiv

Table 9 Escape depths calculated from fits made to deposition data shown in Fig.
3.15 following Eq. (3.2), compared to escape depths calculated for the same
elements following Eq. (3.3). .................................................................. 106
Table 10 Predicted Auger signal intensity for Ti
MVV
and Sr
MNN
peaks as a function
of layer using the parameter-free escape depth model with the ‘bulk’ case
(left), and the same model applied with 5% excess strontium in the
highlighted cells for 2 unit cells (middle), and 4 unit cells (right). Note the
decrease in signal increase as a result of the exponential decay of the escape
depth. ..................................................................................................... 126
Table 11 Predicted Auger signal intensity for Ti
MVV
and Sr
MNN
peaks as a function
of layer using the parameter-free escape depth model with the ‘termination
case for TiO
2
termination (left) and SrO termination (right). ............... 128

xxv

Abstract

Real Time Surface Analysis of Complex Oxide Thin Films During Pulsed
Laser Deposition

by

Thomas Orvis
Doctor of Philosophy in Materials Science

University of Southern California

Dr. Jayakanth Ravichandran, Chair

Complex oxides play host to myriad unconventional physical properties ranging from
colossal magneto-resistance to high-temperature superconductivity and serve a
foundational role in numerous emerging technologies. When these materials are
dimensionally confined as thin films and heterostructures, broken symmetry leads to the
emergence of new metastable phases bearing functional properties and phenomena such
as two dimensional electron gases and quantum phase transitions. The ability to engineer
thin films and interfaces with sufficient precision to observe and study these phenomena
requires a comprehensive understanding of the deposition and growth process, and the
limitations thereof – this is the primary obstacle to the continued development of this
field and the next-generation devices it promises. Structural characterization of films
during the growth process has advanced considerably over the past several decades, with
in situ monitoring of structure and thickness with atomic layer resolution during
deposition now commonplace. This progress has been foundational to current
achievements in the quality and control of thin film deposition techniques, but is likewise
limited in scope. The compositional and chemical characterization of thin films is still
largely limited to pseudo in situ and ex situ techniques which fail to fully unravel the
complexities of growth dynamics during the deposition process. Although surface
composition characterization techniques are prolific, they are all too often cumbersome in
design or sensitive during application. Their incorporation into the harsh oxygen-rich
environment required for complex oxide deposition, with techniques such as pulsed laser
deposition, is therefore a difficult task. Recent developments in Auger electron
spectroscopy probe-design, however, have allowed its incorporation into a pulsed laser
deposition system, making the real time in situ observation of atomic and chemical
xxvi

composition possible for the first time. We have performed a series of experiments to
elucidate its capabilities and sensitivity down to the sub-monolayer scale. Through this
process we have demonstrated the ability to qualitatively and quantitively observe real
time changes in composition and surface termination of the prototypical complex oxide
perovskite SrTiO
3
during the deposition process. These achievements open the door to a
more robust understanding of the complexities inherent to the thin film deposition process,
including temporal phenomena such as termination-switching and dynamic layer
rearrangement. A more complete view of the deposition process will serve to deepen the
community’s understanding of these material systems, and play a significant role in
enabling research – comparable to that of in situ structural characterization decades ago.
The next generation of fundamental science and thin film technology will be built on our
ability to observe and understand these processes in action, and throughout this process
in situ and real time compositional characterization will play a pivotal role.

1

Chapter 1
Introduction

Human history is a story about the relationship between understanding the world and
changing it. When a stone was hefted as a hammer for the first time it became a tool, and
in short time all stone became a material. Three hundred and fifty thousand years ago
people were traveling up to 50 km to procure iron-rich stone for pigment or obsidian for
tools.
1
They understood stone in a way that modern humans never will, because they
relied on that knowledge for survival. As time progressed and the nature of that knowledge
changed and grew, so too did the sophistication of our technology, art, and thought.
When constrained by our biology, the human inclination is to reach for a tool. Improving
the collective understanding has improved our technology, which has improved our
understanding, and so on. When limited by the scope of our vision we invented telescopes
to bring the distant near, and microscopes to make the small large. As we developed the
techniques to capture images and share them, the distance between ideas became smaller,
even as the ideas themselves became larger. The resulting feedback loop has defined the
path of human history: science has led the way, devising new methods and tools to
visualize things for the first time, or in a new way. Technology and society follow behind,
using these tools to understand and engineer the world on a smaller and more complex
scale.
This work is on thin films and surface analysis, which is just another example of our
species’ attempts to visualize, understand, and engineer the world on yet a smaller scale.
Compared to thin films, human-scale material is the ‘bulk’: with surface to volume ratio
so low that the only thing that really matters is the volume, and all three of dimensions
defining its shape. However, as you make that material thinner and thinner, eventually
the only dimension that becomes relevant is its thickness – because in the other two
dimensions it may as well be infinite. Thin films are special because of their lower
dimensionality: as the volume is decreased the surface becomes louder and louder until,
finally, it is all that you can hear.
2
1.1 Background and M otivation

Humans have been creating thin films for over 5000 years, beginning with ancient
Egyptians beating gold foil for gilding,
2
with examples shown in Figure 1.1. Simply
through mechanical means (and, I imagine, a great deal of patience) they were able to
make gold leaf only 300 nm thick; for reference, that is about 200 times thinner than a
human hair. The Inca developed a depletion-gilding technique using a gold-copper alloy
called tumbaga.
3
An object fashioned from the alloy would be thermally oxidized and the
copper-oxide dissolved from the surface. This made the surface appear gold while the
volume remained copper-rich.

Figure 1.1 (a) Example of ancient Egyptian application of gold leaf on a model collar of Hapiankhtifica ca. 1981–
1802 B.C. and (b) a gilded glass earring ca. 100 B.C. (c) A tumbaga necklace of the Tairona people from pre-Columbian
Colombia, ca. 10
th
– 14
th
century A.D. The high copper content accounts for the rosy tonality. Images distributed under
CC0 by The Metropolitan Museum of Art.
Over the following years the craft improved, along with the invention of various forms of
more-or-less toxic vapor deposition for mirrors and other applications. With electricity
came electroplating, and coatings became a crucial component of corrosion and wear
protection. However, it was silicon that shifted the paradigm for thin films.
With the advent of microelectronics, applications for thin films found a new niche and
their value became more than just decorative, protective, or reflective. Suddenly, the film
was the object instead of simply defining it. In modern semiconductor manufacturing a
silicon wafer can undergo more than 250 different steps of depositing various films,
3

patterning, then etching them away, all to control the thickness and precision of the
interfaces at the heart of our devices.

3
1.1.1 Thin Films
A surface in vacuum is in a high energy state because it is a field of broken bonds. To
minimize energy it will rearrange itself, causing structural distortions and electronic
reconstruction. When an interface is formed (through deposition, for example, illustrated
in Figure 1.3) this same process occurs in both materials as they sort out the lowest
energy arrangement. An interface is a transition between two states, and is therefore both
of them – and neither. The structures that form, and the material properties that arise
from their formation, are often not possible in either material by itself. Understanding the
structural and electronic consequences of interfacial reconstruction is the foundation of
modern atomic-scale thin film engineering.

Figure 1.2 An illustration of thin film deposition, with a new material arriving at the atomically-stepped surface of
a single crystal or epitaxial thin film. Image: the author.
Designing and synthesizing thin films and layered heterostructures is the forefront of
current technology. This is exemplified in modern transistor architecture as new designs
become more and more complex. FinFETs,
5
transistors built on-edge to save space while
remaining operational (like fins), are in common use today after over a decade of process
development,
6,7
shown in Figure 1.3 (a). Nanowires are a proposed next step, shown in
Figure 1.3 (b), which effectively bury the transistor channel with an all-around gate, with
claims that they are the solution to the 5 nm node.
8-10
Another area of active research is
alternative materials, such as germanium for replacing silicon channels to improve electron
mobility,
11,12
or the III-V materials such as GaN/SiC (for high frequency and high power
applications) that helped enable 5G wireless technology.
13-15
These few examples illustrate
some of the capabilities and future trends of the so-called ‘mean-field electron system’
materials, which simply means that their electronic behavior can be fairly approximated
with mean-field theory, that reduces the many-bodied problem of electron-electron
interactions in a solid to a single averaged effect.
16
Correlated materials, in contrast, are
not well approximated by mean-field theory, and require consideration of electron-electron
effects to accurately describe their behavior.
Despite the incredible advances being made at the forefront of the electronics industry,
however, traditional silicon technology is still quickly marching towards its fundamental
4
limits, forcing a collective reconsideration of how we will design and engineer devices in a
post-Moore’s law environment.
19-21
There are numerous approaches being explored as
candidates for the future beyond complementary metal-oxide semiconductor (CMOS)
transistors, ranging from alternative architectures with nanoscale thermal management,
to non-thermal equilibrium systems and non-von Neumann architectures.
22
One of the
particularly compelling candidates is the magnetoelectric spin-orbit (MESO) transistor.
23

This novel design utilizes a multiferroic material as a magnetoelectric, such as BiFeO
3
,
coupled to a spin-charge converter, such as the LaAlO
3
/SrTiO
3
interface. As a scalable
device it has been shown, theoretically, to have superior switching energy, lower switching
voltage, enhanced logic density, and ultralow standby power, with the proposed device
architecture illustrated in Figure 1.4. The operational principle is that charge and voltage
are the inputs, which are converted to magnetism by the magnetoelectric effect and then
back to charge by the spin-orbit effect. However, MESO transistors still need improved
room-temperature electromagnetic coupling and, on the flip side, stronger spin-charge
conversion to function as intended.
24

Figure 1.3 (a) Cross-sectional TEM of a finFET, illustrating the precisely layered structure. The image is
approximately 80 nm across. (b) Cross-sectional TEM of stacked gate all-around nanowire transistors. Image (a):
(Biring, 2014); (b): (Courtland, 2016).
Figure 1.4 An illustration of the proposed MESO transistor architecture, which
operates via spin-orbit transduction coupled with electromagnetic switching. Image:
(Manipatruni et al, 2019).
5
Regardless of the specific approach taken for ‘Beyond CMOS,’ it is generally agreed that
improved functionality will be fundamental to the effort.
22-25
What this means is that
improved performance is going to rely on materials with functional properties that extend
past the mean-field and into the realm of correlation effects, such as the previously
mentioned electromagnetism and spin-orbit coupling in MESO transistors. Among the
diverse materials being explored, complex oxides have emerged as a compelling candidate
due to their correlated electron systems resulting in enhanced and exotic functionality.
24-
27

1.1.2 Complex Oxides
Complex oxides are defined, simply, as those materials containing oxygen and two or more
other elements, or one other element with multiple oxidation states. They are a rich and
varied field encompassing everything from magnetite (Fe
3
O
4
) and emeralds
(Be
3
Al
2
(SiO
3
)
6
), to the colossal magneto resistor lanthanum strontium manganate (La
1-
x
Sr
x
MnO
3
) and high-T
C
superconductor yttrium barium copper oxide (YBa
2
Cu
3
O
7-x
),
28-33

with crystal structures shown in Figure 1.5. What makes this class of materials notable
is its broad range of useful electronic, magnetic, and optical properties arising from
strongly correlated d or f electron orbitals; this simply means that the behavior of these
systems cannot be accurately described without accounting for electron-electron
interactions. While undoubtedly more complicated than non-correlated mean-field
systems, this complex behavior is responsible for equally complex properties with a long
history, and promising future, of
applications.
Perovskite is the name of the naturally
occurring mineral CaTiO
3
, but is used
more generally in reference to the family
of compounds that share its crystal
structure.
34
The general chemical
formula is ABX
3
, where A and B are two
different cations and X is an anion, often
oxygen, bound to both cations. The
cations and anion can be chosen from
discrete groups in the periodic table,
35

shown in Figure 1.6, but the selection
is large enough that the number of
options is daunting. The idealized
perovskite structure is cubic, with space
group Pm3
̅
m, shown in Figure 1.7, and
Figure 1.5 Crystal structures of La0.7Sr0.3MnO3 (left) and
YBa2Cu3O6.9 (right). Note that due to the distribution of La
and Sr in the material, they are represented equivalently in the
La0.7Sr0.3MnO3 structure. Image: the author.
6
is adopted by, among others, BaZrO
3
and the flagship perovskite semiconductor SrTiO
3
.
36

In the cubic unit cell, the corners are occupied by A cations, the body-center is occupied
by the B cation, and face-centers hold X anions, forming BX
6
octahedra.
Because most complex oxide perovskites are wide-bandgap insulators they may not seem
like the natural choice for electronic applications.
37
Of course this is accurate, and why
the foundation of modern electronics is silicon: the requirements for controlling its
fabrication and conductive properties have been – relatively – low hurdles, thereby
allowing on-chip transistor manufacturing and the
advent of modern computing. However, the silicon road
of the past sixty years has been propelled by
miniaturization, squeezing more transistors onto a chip
through lithographic scaling.
20-22,38
Although reaching
our current status has required inspiring innovation,
the approaching limits are dictated by the fundamental
laws of physics.
39
As a result, the cost of
miniaturization is increasing beyond profitability, and
the future of computing has shifted to reenvision the
solution-space.
40,41
The demand for single-thread
performance is declining, pushing the paradigm
towards edge-computing,
42,43
where overall latency in a
network is decreased through conducting more
Figure 1.6 The periodic table showing which elements may be used for the A, B, and X-site of the ABX3
perovskite structure with 100% occupation. Image: the author, based on data from (Schlom et al, 2008).
Figure 1.7 Ideal cubic perovskite
structure ABX3, of space group Pm3
̅
m, with
A cations occupying corner sites, B cation
occupying cube center, and X anions
occupying face centers, forming a BX6
octahedra. Image: the author.
7
computation closer to the end
user by taking advantage of
the distributed nature of the
‘Internet of Things’. A
requirement of this scale of
redistribution is increasing the
capabilities of devices with
integrated multifunctional
properties.
44-46
However,
silicon and III-V technology
struggles with functional
properties, whereas complex
oxide perovskites are
abundantly multifunctional materials. Rather than trying to replace traditional
semiconductors, which are the best materials for what they do, their integration with
complex oxides can bring expanded application to existing high-performance architectures,
harnessing the emergent phenomena of correlated systems.
23-27,35,41,45

The electronic structure of complex oxide perovskites is the origin of its functional
properties which is, therefore, dictated by the structure and positions of the elements
comprising it. Using SrTiO
3
as an example, we have a cubic unit cell with a Ti
4+
cation
at its center, surrounded on all six cubic faces by O
2-
anions, and Sr
2+
cations occupying
the cubic corners. The calculated band structure of SrTiO
3
was first published in 1964,
47

shown in Figure 1.8, and has seen considerable review since.
48-51
The crux of the story,
however, has essentially remained the same: the valence and conduction bands are
composed of strong hybridization of titanium d-states and oxygen p-states, either in
bonding or anti-bonding covalent interactions.
52
The bonding state forms the valence
band, while the anti-bonding states form
mutually orthonormal t
2g
hybrids which
create the conduction band. The
significance of this structure is found in its
symmetry, which begs to be broken: d/p-
hybridization results in x-y-z t
2g
orbitals,
but only two of the three directions can be
strong 𝜋 -hybridizations, while the third
must be a weak 𝛿 -interaction. This
‘asymmetrical symmetry’ is illustrated in
Figure 1.9 and demonstrates why these
materials are considered correlated, as well
as being responsible for the high dielectric
constant of SrTiO
3
.
Figure 1.8 The original calculation for the band structure of SrTiO3,
plotted as energy versus wave vector for all lines of symmetry in the Brillouin
zone. Image: (Kahn et al, 1964).
Figure 1.9 A demonstration of the geometry of bond
formation for SrTiO3 in bonding and antibonding states.
Image: (Dylla et al, 2019).
8
Expanding on this premise, consider BaTiO
3
,
34
which undergoes a phase transformation
from trigonal to orthorhombic at 183 K, then to tetragonal at 263 K, and finally to cubic
at 393 K, all due to displacement of the Ti cation. As the cubic Pm3
̅
m phase cools below
393 K one axis elongates slightly while the others contract, forming a tetragonal structure
with space group P4mm, illustrated in Figure 1.10. This subtle shift causes a change in
the octahedral dimensions causing the equatorial oxygens to shift, forming electric dipoles
that are the source of ferroelectricity in tetragonal BaTiO
3
.
These examples illustrate how the perovskite structure itself can be responsible for
material properties, and it takes little imagination to see how these same combinations of
structure and electronic configuration can give rise to other phenomena such as
magnetism, colossal magnetoresistance, and transparent conductivity. Beyond SrTiO
3
and
BaTiO
3
there are numerous perovskite oxides, and effectively infinite solid solutions that
mix them together to tune their properties. Some examples of the property and
compositional diversity shown in the perovskite oxide family are shown in Table 1.

Figure 1.10 The BaTiO3 structure in the cubic 𝑃𝑚 3
̅
𝑚 phase above 393 K (left) and tetragonal 𝑃 4𝑚𝑚 phase above
263 K (right). Note the distortion of the tetragonal phase (exaggerated for clarity), with the relative shifts of the central
oxygen atoms highlighted with a dashed line. Image: the author.

9
Table 1 Examples of the diverse properties exhibited by complex oxide perovskites.
Property or Application M aterial
Ionic Conductor
53
Doped LaInO
3

Proton Conductor
54
BaZrO
3

Solid Oxide Fuel Cells
55
BaCe
1-x
YO
3

Colossal Dielectric
56
CaCu
3
Ti
4
O
12

Ferroelectric
34
BaTiO
3

Piezoelectric
57
PbZrO
3
-PbTiO
3

Ferrielectric
58
Pb
2
MnWO
6

Relaxor Ferroelectric
59
Pb(Mg
1/3
Nb
2/3
)O
3

Ferromagnetic
60
La
0.7
Sr
0.3
MnO
3

Paramagnetic
61
SrTiO
3

Antiferromagnetic
62
LaFeO
3

Ferrimagnetic
63
CaCu
3
Fe
4
O
12
Spin Glass
64
Ca
2
MgOsO
6
Multiferroic
65
BiFeO
3

Metallic
66
SrRuO
3

Metal-Insulator Transition
67
NaOsO
3

Superconductivity
68
BaPb
0.75
Bi
0.25
O
3

High TC Superconductivity
32
YBa
2
Cu
3
O
6.95
Spin Polarization
69
CaMnO
3

Magnetoresistance
70
La
0.7
Ca
0.3
MnO
3

Zero Thermal Expansion
71
0.6(PbTiO
3
)-0.3(Bi(Zn
0.5
Ti
0.5
)O
3

Field-Controlled Transparency
72
Pb
1-x
La
x
(Zr
1-y
Ti
y
)
1-x/4
O
3

Electro-optic Phase Modulation
73
LiNbO
3

Moving forward, the inherent complexities of these systems, though difficult to control,
must be viewed as an opportunity rather than a burden. Improving our ability to
understand the interplay between structure, properties, and processing will allow us to
create electronic devices designed with truly functional materials. Exciting applications
on the near horizon are taking advantage of the high electron density and functional
properties of complex oxide perovskites to create new devices with fundamentally different
behavior than has been used traditionally, such as phase-transition transistors,
23,74

negative capacitors,
75
and photovoltaic cells based on correlation effects.
24,25,76
Considering
that we are still only on the near-edge of understanding these materials, we will
undoubtedly continue being surprised by the properties and applications that emerge.

10
1.1.3 Potential and Challenges
Complex oxide perovskite thin films and heterostructures show great promise for
applications, with active pursuit of implementation for resistive and ferroelectric random
access memory,
77,78
ferroelectric tunnel junctions,
79
logic-in-memory non-von-Neumann
architectures,
80,81
nano-neurons for neuromorphic computing,
82
quantum computing,
83

magnetic tunnel junctions and ordered insulators for spintronics,
84,85
on-chip integrated
photonics,
86
and photovoltaics.
24,25,76
Many of the functional properties found in complex
oxides arise due to broken symmetry of the order parameter, resulting in the phase
transitions observed in these materials.
87

For instance, the phase transition from a
paramagnetic to ferromagnetic state occurring at
the Curie temperature (T
C
) is associated with the
order parameter of magnetization. An example is
La
0.7
Sr
0.3
MnO
3
with a T
C
of 378 K, above which
the spins are randomly oriented and thus
symmetric with zero magnetization.
60
On cooling,
however, the spins align, spontaneously breaking
the symmetry. Another example is the previously-
mentioned paraelectric-ferroelectric transition
observed in BaTiO
3
below 393 K, for which the
order parameter is polarization. Cubic BaTiO
3
has
zero polarization (symmetric), but that symmetry
is lost when it becomes tetragonal, resulting in
ferroelectricity.
34
Slightly more complex is the
paramagnet-antiferromagnet transition which
occurs at the Néel temperature (T
N
). For LaFeO
3

T
N
is 740 K, and the order parameter is the
staggered magnetization of two sublattices.
62,88,89

Symmetry is a great unifier in natural philosophy,
forcing obedience on the shapes, interactions, and
evolutions which can naturally occur.
90
In
symmetry physics, a fundamental idea is that the
states of a system do not need to have the same
symmetries as the theory which describes the system, which defines spontaneous broken
symmetry. The result is that a physical system in an initial symmetric state ends up in
an asymmetric state. A simple illustration of this phenomenon is shown in Figure 1.11,
where at high energy levels the settled state is at the symmetric bottom of a trough, but
as the energy decreases the rules governing outcomes remain symmetric but force an
asymmetric outcome: one of two equivalent new lowest energy settled states, chosen
spontaneously. The significance of broken symmetries is quite profound, manifest in
Figure 1.11 An illustration of the principle of
broken symmetry: at high energy there is one
available state, but at low energy, even though the
solution is still symmetric, the outcome will not
be. Image: the author.
11
different ways throughout the field of physics, but of particular significance here because
broken symmetry governs phase transformations and provides the definition for most
phases of matter.
An order parameter can generally be defined as a quantity which is zero above the critical
point and finite below it, characterizing the transitional changes which occur. These can
be structural, such as lattice displacements, or otherwise, such as the component of total
magnetic moment. To demonstrate, consider an overview of the Ising model description
of ferromagnetism. The Hamiltonian in an external magnetic field is given by Equation
1.1, where 𝜎 are the Pauli spin matrices, 𝐽 ( 𝑙 − 𝑙 `) is the exchange interaction between
lattice sites (𝑙 and 𝑙 `), and the applied magnetic field is represented by the −ℎ∑ 𝜎 𝑙 𝑙 term.
89

ℋ = −
1
2
∑ 𝐽 ( 𝑙 − 𝑙 `) 𝜎 𝑙 𝜎 𝑙 `
− ℎ∑ 𝜎 𝑙 𝑙 𝑙 ,𝑙 `
(1.1)
This equation is stating that the exchange interaction is translationally invariant, and
dependent only on the distance between lattice sites. Simplifying this expression for an
entire crystal requires approximation, therefore let us introduce an average value for the
magnetization per lattice site (possible because of the translational invariance) given by
Equation 1.2, and an abbreviation for the vanishing exchange interaction, given by
Equation 1.3. The result is an approximation for the molecular field, 𝐽 ̃
𝑚 .
𝑚 = 〈𝜎 𝑙 〉 (1.2)
𝐽 ̃
≡ 𝐽 ̃
( 0)≡ ∑ 𝐽 ( 𝑙 )
𝑙 (1.3)
Again, relying on the translational invariance of the Hamiltonian, the Fourier transform
of the exchange coupling can be taken, shown in Equation 1.4, thereby allowing the
formulation of the density matrix, simplified as Equation 1.5.
𝐽 ̃
( 𝒌 )= ∑ 𝐽 ( 𝑙 ) 𝑒 −𝑖 𝒌 ∙𝒙 𝑙 𝑙 (1.4)
𝜌 ∝ ∏ 𝑒 𝜎 𝑙 ( ℎ+𝐽 ̃
𝑚 ) /𝑘𝑇
𝑙 (1.5)
Using these approximations, the original Hamiltonian can be simplified to a mean field
approximation, resulting in a mean field density matrix and partition function. These
approximations result in the equation of state in the molecular field approximation for
magnetization shown in Equation 1.6, where 𝛽 is the thermodynamic beta (
1
𝑘 𝐵 𝑇 ) .
12
𝑚 = tanh( 𝛽 ( 𝐽 ̃
𝑚 + ℎ) ) (1.6)
This is striking particularly because of its comparison to a paramagnet, which is equivalent
without the molecular field, 𝐽 ̃
𝑚 . The difference arises because in the case of a ferromagnet
the applied field ℎ is amplified by the internal molecular field. If we define the critical
temperature (or Curie temperature) with Equation 1.7, and 𝜏 with Equation 1.8, we can
look at the behavior of the magnetization around this point.
𝑇 𝐶 =
𝐽 ̃
𝑘 (1.7)
𝜏 =
𝑇 −𝑇 𝐶 𝑇 𝐶 (1.8)
When the applied field is zero, the magnetization can be expanded in a Taylor series,
shown in Equation 1.9, with solutions of 𝑚 = 0 for 𝑇 > 𝑇 𝐶 , and 𝑚 = ±𝑚 0
, 𝑚 0
=
√3( −𝜏 )
1/2
for 𝑇 < 𝑇 𝐶 .
𝑚 = tanh𝛽 𝐽 ̃
𝑚 =
𝑇 𝐶 𝑇 𝑚 −
1
3
(
𝑇 𝐶 𝑇 𝑚 )
3
+ ⋯ (1.9)
This demonstrates that the magnetization is, indeed, the order parameter of the
ferromagnetic phase transformation, as shown in Figure 1.12. Furthermore, when ℎ and
𝜏 are nonzero, the expansion leads to Equation 1.10, showing that an applied magnetic
field produces a finite magnetization even
above T
C
.
ℎ
𝑘 𝑇 𝐶 = 𝜏𝑚 +
1
3
𝑚 3
(1.10)
However, this also reveals that for a given
applied field ℎ the value of the
magnetization below T
C
is not unique, as
shown by the three solutions for ℎ = 0. To
determine which solutions are stable
requires calculation of the free energy.
Continuing with the prior mean field
approximation the Helmholtz free energy
can be calculated as Equation 1.11, and
plotted as shown in Figure 1.13, which
Figure 1.12 Spontaneous magnetization with the two
possible solutions (±𝑚 0
) (solid line) and in an applied field
(dashed line). Image: (Schwabl, 2006).
13
clearly illustrates the origin of the simple description previously shown in Figure 1.11.
𝐹 ( 𝑇 , 𝑚 )=
1
2
𝑇 𝐶 𝑚 2
− 𝑇 log{2 cosh(
𝑇 𝐶 𝑚 +
ℎ
𝑘 𝑇 ) } +
𝑚 ℎ
𝑘 (1.11)
Why this concept matters to thin films is
actually quite simple: interfaces are, by their
very nature, a point of broken symmetry.
The theory governing a crystal lattice
assumes it is infinite in every direction, and
therefore has inversion symmetry.
87
As soon
as it is not infinite (the case when an
interface is rudely interjected onto it) that
symmetry is broken. For perovskites,
breaking this symmetry can, all by itself,
lead to spontaneous electric polarization,
offering a perfect point of origin for the
investigation of novel phases and quantum
phenomena.
85
As another way of viewing it,
the magnetization order parameter
described above can be considered instead
as time reversal symmetry. In the
paramagnetic state moving forward and
backward in time you will follow the same
path. However, the path is not the same – the symmetry is broken – in magnetic materials
exhibiting hysteresis. Another example is gauge symmetry, in which the order parameter
is the macroscopic wavefunction, and the broken symmetry is the gauge invariance with
respect to its phase. This means that certain measurements will be constant for a
particular system, reflected in the range of allowed solutions. The formation of Cooper
pairs superconductivity breaks gauge symmetry.
91

To better describe superconductivity, consider two electrons with opposing spin. Within
a solid they are two of, effectively, 𝑁 → ∞ electrons. However, if these two electrons form
a bound state with spin zero, they no longer behave as fermions and can be treated as a
boson. The two electrons then share a wavefunction, and the system energy is equivalent
to 𝑁 − 2. This reduction in energy is statistically insignificant unless it occurs en masse
in a system with sufficiently low energy to provide it significance. The formation of pairs
results in their adoption of complex wavefunctions. Because the wavefunction is the
system order parameter and is broken by the complex form, this emergence signals the
formation of the superconducting state and the spontaneous breaking of gauge invariance.
Figure 1.13 Helmholtz free energy as a function of
magnetization at 𝑇 > 𝑇 𝐶 , 𝑇 = 𝑇 𝐶 , and 𝑇 < 𝑇 𝐶 , showing the
symmetric split of solution states indicative of broken
symmetry. Image: (Schwabl, 2006).
14
The interplay between the symmetries outlined here, and their corresponding electronic
degrees of freedom, shown in Figure 1.14, produces various emergent phenomena
characterized by their correlated behavior. Designing thin films and heterostructures with
deliberation and awareness of these properties provides an inherent advantage. Oxide
interfaces, approached in this manner, can be used to engineer thin films and
heterostructures exhibiting novel properties and functionality not possible in the bulk.

Figure 1.14 An illustration of the properties that can arise from different forms of broken symmetry, as well as the
relationship between their electronic degrees of freedom. Image: the author, inspired by (Hwang et al, 2012).
The challenges associated with this process are, however, quite formidable. Foremost
among them are the previously-mentioned processing/structure/property relationships of
these materials. This is essentially another way of saying that the electronic structure of
these materials (which dictates their manifest properties) is a function of their atomic
structure, which is determined by the energetics of their formation (aka processing).
Understanding the subtleties of these interrelationships can allow the active design and
synthesis of films with functional properties tuned for specific applications. However, we
are still in the process of discovering the scope of our ignorance, and all signs suggest it is
deep.
15
1.2 Understanding the Growth Process

The first step to understanding the underlying relationships in complex oxides is
developing a greater comprehension of structure and composition at the atomic scale,
which is sensitive to the dynamics of thin film growth itself. However, growth parameters
play an inherently complex role in the energetics of film formation,
92
the quality of the
film,
93
and even its composition,
94-96
making this process as challenging as it is important.
To achieve this goal we need better, more robust, in situ thin film characterization tools
to monitor the dynamics of these processes as they occur.

1.2.1 Pulsed Laser Deposition
Pulsed laser deposition (PLD) has been the method of choice for depositing complex oxides
since the late 1980s, when it was used to grow a thin film of YBa
2
Cu
3
O
6.95
(YBCO) that
was shown to be compositionally the same as the target,
33
shown in Figure 1.15. Prior
to this shocking discovery, the laser ablation and deposition process was known to exist
but considered inferior to other thin film growth techniques because it is, truth be told,
inferior for most materials. However, the
demonstration of PLD for YBCO brought
attention to this technique because no one had
been able to deposit superconducting films of
equivalent quality using other methods. It was
Figure 1.15 The Rutherford backscattering spectrum
from the as-grown YBa2Cu3O6.9 film on carbon, showing
remarkable consistency with the simulation and
demonstrating the efficacy of pulsed laser deposition for
compositional transfer of complex stoichiometry oxides.
Image: (Dijkkamp et al, 1987).
Figure 1.16 Photograph of the pulsed laser
deposition system used in this work during the
ablation of a target. The electron beam, heater and
substrate, target, and laser path are indicated. Image:
the author.
16
quickly shown that the compositional transfer exhibited with YBCO was also possible
with other complex stoichiometry materials, and it has since been utilized to grow high
quality epitaxial thin films, heterostructures, and superlattices of a diversity of
materials.
35,97-101

The basic principle of PLD may seem rudimentary: we shoot a target, with a laser, which
evaporates the material and then condenses that vapor on a substrate, illustrated in
Figure 1.16. However, by controlling the laser energy,
102
spot size,
103
and pulse rate;
104

chamber pressure;
105
target composition,
106
density,
107
and distance to the substrate;
108

substrate temperature,
109
surface quality,
110
and crystallographic orientation;
111
it is
possible to deposit thin films of the highest quality. That may seem like a lot to keep
track of, and it is, but it is not significantly more complicated than any other thin film
growth technique. Using the tools available to monitor deposition, careful PLD has
produced thin films and heterostructures of remarkable quality and precision,
112-114
as
shown in Figure 1.17.

Figure 1.17 Cross-sectional low-magnification high angle annular dark field image of a
(Ba,Ca)(ZrTi)O3/SrRuO3/GdScO3 sample (left), magnification of the (Ba,Ca)(ZrTi)O3/SrRuO3 interface (center) and
SrRuO3/GdScO3 interface (left) overlaid with atomic models illustrating octahedral tilt. Image: (Liu et al, 2019).

1.2.2 In Situ Structural Characterization
An unfortunately unintuitive aspect of the thin film deposition process is the lack of
human-scale feedback. Typical substrates used in PLD are on the order of a few mm
2
,
and the films grown are typically thinner than visible with human vision. The substrate
oftentimes leaves the vacuum chamber looking much the same as it did when it went in,
shown in Figure 1.18. In order to “see” the deposited film requires some method of
characterization that depends on the desired properties and nature of the deposition. It
should be clear that this is not particularly efficient, and some form of feedback and
process control during deposition would greatly improve the results. Luckily, the past
17
several decades has produced significant advances in in situ thin film characterization
techniques.
115

The requirements of in situ characterization are fairly
straightforward: the method must be usable in the
growth environment. The primary limitations are
typically either that the growth environment will not
support the characterization technique, or the technique
cannot be incorporated into the growth environment.
Many thin film deposition systems have characterization
techniques that claim to be in situ, but cannot be used
on samples in the same conditions as used for growth. A
common example is a deposition system which has the
capability of transporting samples in vacuum to a
characterization chamber,
116-118
or can characterize films
at a different pressure or temperature than used for
growth.
119,120
I will refer to these types of systems as
”pseudo in situ.” To clarify, these methods have
tremendous practical application, but must be
distinguished in this work from the in situ
characterization of surfaces that have not changed
environment since their formation.
Reflection high energy electron diffraction (RHEED) was
initially developed as a surface characterization
technique akin to low energy electron diffraction
(LEED),
121,122
with an example diffraction pattern
shown in Figure 1.19. Serving solely as a means of
observing structural changes of surfaces, it was quite
useful.
123
However, it became an essential addition to
epitaxial growth when it was discovered with GaAs that
the intensity of the specular spot would oscillate as a
function of surface roughness during deposition, making
Figure 1.18 Typical before (top) and
after (bottom) images of substrates (small
rectangles, ~3 x 5 mm and 5 x 5 mm, top
to bottom) mounted to resistive heaters
with silver paint. The colors seen on the
heater in the bottom image are the result
of thin-film interference, rather than the
color of the deposition itself. Image: the
author.
Figure 1.19 (right) An example of a reflection high energy electron
diffraction pattern from a SrTiO3 single crystal substrate surface at 800℃
in 10
-4
mbar O2 in our pulsed laser deposition chamber. The direct beam
can be seen above the diffraction pattern, with the horizontal shadow below
it from the edge of the heater. The large ring around the direct beam is
from scatter through the glass on which the phosphor is applied, as well as
the viewport behind it. Note the vertical habit of the diffraction pattern,
indicating a smooth 2D surface, and the Kikuchi lines from inelastic
collision processes. Also note the doubly-spaced diffraction spots below the
specular spot, which demonstrate surface reconstruction. Image: the
author.

18
it possible to observe the layer-
by-layer growth of thin films in
real time,
124
illustrated in
Figure 1.20 with complex
oxides. This function allowed a
degree of control previously
impossible, and ignited a
renaissance in the study of thin
film heterostructures.
In 1997 a RHEED system was
developed for use with the high
pressure PLD chamber
environment by incorporating
a differential pumping
system.
125
This brought PLD
into the modern era of thin film growth, and improved the quality and precision of
reported thin films and heterostructures dramatically.
115
Using the structural information
obtained from RHEED the standards of complex oxide thin film growth have improved
substantially, and in some cases are on par with films grown by molecular beam epitaxy.
126

Unfortunately, RHEED is only providing half the story of thin film surface evolution
during the deposition process. Attempts have been made to use RHEED as a more
complex in situ composition characterization tool, but the methods have proven unwieldy
and not been popularized.
127,128
If a surface analysis technique were developed that was
as robust for compositional characterization as RHEED is for structural, it would have
the potential to advance the field of complex oxide thin film deposition in a manner
equivalent to that demonstrated by RHEED. Thankfully, this has been achieved.

1.2.3 In Situ Compositional Characterization
Compositional analysis techniques are quite powerful and typically operate by measuring
the characteristic energy of emitted electrons or photons, such as X-ray Photoelectron
Spectroscopy (XPS), Auger Electron Spectroscopy (AES), Energy Dispersive X-ray
Spectroscopy (EDS), and X-ray Fluorescence Spectroscopy (XRF), though some, such as
Secondary Ion Mass Spectrometry (SIMS), observe ions directly. However, the limitations
of these techniques make them largely unsuitable for most in situ measurements.
Depending on the requirements of the characterization and deposition methods, several
systems have been incorporated into growth chambers for truly in situ observation.
120,129-
132
Unfortunately, the same factors which delayed the incorporation of RHEED in PLD
Figure 1.20 An example RHEED oscillations from the deposition of 16
unit cells of LaAlO3 on SrTiO3. The sixteen clear oscillations in intensity
are the result of the surface roughening (decrease) and then smoothing
again (increase). The decay in intensity over the growth is not uncommon
and has many possible causes. Image: the author.
19
have applied to compositional characterization, and only recently has this been addressed
with the introduction of Staib Instruments’ AugerProbe.
133

Auger electron spectroscopy is a highly surface-sensitive technique of observing
characteristic electron emissions (Auger electrons) to identify the presence of specific
elements and/or their chemical states. It has been in use as a surface analysis tool for over
half a century,
134
but tends to be eclipsed by XPS, which also detects Auger electrons as
well as normal photoelectrons. Photoelectrons are not visible in AES because an electron
gun is used as the ionizing radiation, and electron-electron interactions prevent the
photoelectrons from maintaining characteristic energies. The Auger electrons, however,
are the result of a 3-step process in which
they are ejected by the energy released when
a higher energy electron fills a core level hole,
shown in Figure 1.21.
Traditionally AES has been limited to ultra-
high vacuum (UHV) because the relatively
low energy of Auger electrons made them
susceptible to scattering, which prevented
effective data collection with standard
cylindrical mirror analyzers.
135
However, a
recently developed Auger probe has been
designed which overcomes this problem
through differential pumping and optics
modified by incorporating a collimator lens
coupled to an electrostatic retarding field.
The result is an Auger electron probe
capable of operation at above 5 x 10
-3

mbar,
136
a large working distance to
accommodate atomic flux for growth, good
sensitivity, and a robust design resistant to
material deposition.
137

The ability to acquire real time in situ compositional information about thin films during
their deposition has the potential to answer fundamental questions about the dynamics of
the thin film growth process. However, fundamental limitations to AES will always be a
factor in its application and must be accounted for. One of the primary limitations is the
Z limit defined by the process itself: 3 electrons are required to generate an Auger electron,
so the lightest element that can be seen with the method is lithium. There is no
fundamental limit to a maximum Z observable with AES, but for the highest Z values the
only transitions that can be seen reliably at less than 2 keV are low energy N– peaks,
138

Figure 1.21 An illustration of the Auger process, in
which incoming energy (EIncoming) knocks out a core-level
electron. A higher-level electron falls to take its place,
releasing energy as it does so, which proceeds to knock
out an even higher energy electron, which is then the
Auger electron. An important distinction of this process
is the Auger electron energy’s independence from
EIncoming.
20
illustrated in Table II and shown in
Figure 1.22. One of the most common
limitations for AES is charging of
insulating samples. The sample must be
under active ionization by high energy
electrons, and accumulation of surface
charge can occur if it cannot be
conducted away, distorting the position
and visibility of Auger peaks.
139
In a
PLD system charging is often not an
issue because even though the
substrates are frequently insulating at
room temperature, they are sufficiently
conductive at the temperatures used for
growth, illustrated in Figure 1.23.
Furthermore, the electron source for
AES is the same as that used for
RHEED, and surface charging equally
distorts the diffraction pattern, making
this a problem that was initially
addressed over twenty years ago.

Figure 1.22 Characteristic Auger transition energies as a
function of atomic number. The shells associated with the
transition are labeled and coded by color. Image: the author,
with data compiled from (Ferguson, 1989).
Figure 1.23 Charging of an SrTiO3 substrate at room temperature (left), compared
to the same substrate at 750℃ using the same imaging conditions. Note the distortion
of the direct beam and resulting diffraction pattern in the image taken at room
temperature. Image: the author.
21
Table 2 Reported Auger peak locations in dN/dE mode. Data from Ferguson, 1989.

22
1.2.4 Dynamic Thin Film Growth Phenomena
The intricate surface energetics of substrates and films can be the source of surprising
phenomena. Taking a step back, sometimes it is surprising that we are able to grow
epitaxial thin films at all, especially of materials with complex stoichiometries. Yet, we
are able to for the same fundamental reason behind these phenomena: surface energy and
kinetics.
To clarify, let’s return to the example of growing epitaxial thin films of materials with
complex stoichiometries. If you ablate a plume of strontium, titanium, and oxygen from
a target and send it in the direction of a piece of SrTiO
3
, what would you expect to
happen? Clearly the answer to that question depends on a lot of things, but the exercise
remains the same: why should this vapor form more SrTiO
3
? If we want it to form SrTiO
3
,
how do we make that happen? If we don’t want it to form SrTiO
3
, how do we prevent it
from happening?
Answering these questions is the basic decision-making process behind film growth, and
serves to illustrate the role of kinetics in deposition. The short answer is always the same:
the minimization of energy. However, even though the answer is simple, it doesn’t mean
the question is. Returning to SrTiO
3
: when the incoming vapor hits the surface, what
happens next depends on how much energy is available to the particles. If they are too
cold, they will be stuck in place because they will not be able to overcome the energy
barrier keeping them in their current position. This means the strontium and titanium
will be dropped in random places, and won’t be able to form a crystalline layer, much less
an epitaxial one, and the result will be amorphous or, at best, polycrystalline. However,
if they are too hot, they’ll be just as likely to float away as stay on the surface because
they’ll have so much energy that there won’t be sufficient impetus to keep them there, or
they will damage the substrate and growing film by smashing into them. As if that’s not
enough, these values are species-dependent, because strontium and titanium have different
masses and electronic structures. Clearly, then, we’re looking for a Goldilocks zone which,
depending on the material, can be very narrow or very wide,
140
shown in Figure 1.24. If
the temperature of the substrate is in an acceptable range, and the incoming adatoms
have the appropriate amount of energy, it’s possible to grow a crystalline thin film, though
still not guaranteed.
23
Once the atoms are on the surface, they have to figure out how to get organized. In an
ideal scenario, they have enough energy to hop around the surface until they can find a
proper lattice position and minimize their energy. However, there are any number of things
that can interfere with this fairytale, the most obvious being either not enough energy for
surface diffusion, or available positions with lower energy that aren’t where you want the
species to go, illustrated in Figure 1.25. The latter scenario is more complicated,
especially for heteroepitaxial films, because the complex relationship between the surface
energy of the substrate and the competing surface energy of the various species can have
unexpected results. For instance, what happens if the lowest energy position for a species
is a bond with itself, as opposed to with the substrate? This is a situation where excess
energy for surface diffusion can be a negative, as it allows the species to clump into islands
and hills, rather than spreading out evenly on the substrate.
104
The resulting surface would
be rough, and even when the valleys are filled in the film would have a high defect-density.
However, with clever design this can also be deliberately taken advantage of to create
novel film structures, through a self-assembly process,
141-144
shown in Figure 1.26.
Figure 1.25 An example of the growth parameter-dependence of BiFeO3, with marked growths data-mined from the
literature (left) and their resulting properties (right). The congregation of parameters shows the Goldilocks zone effect.
Image: (Young et al, 2018).
Figure 1.24 An illustration of possible differences in film growth modes depending on kinetics and surface energy.
If the deposition has insufficient surface diffusion or low self-nucleation energy then island growth can dominate (top),
rather than developing layer-by-layer (bottom) into a smooth film. Image: the author.
24
So, assuming we have solved all the problems and our films are stacking up epitaxially
like eggs in a carton: what happens if the lowest energy structure now is epitaxial growth,
but the act of growth changes the lowest energy structure? This is just one example of a
possible scenario which can, and does, occur during the growth process.
119,120,145
These
types of dynamic thin film growth events can dramatically alter the resulting structure
without being observed, and may never be discovered if not deliberately investigated.
146

As a result, understanding these types of phenomena is a crucial part of understanding
the larger growth process, and in situ compositional characterization is the best tool for
the job.

Figure 1.26 Deposition of GaN on diamond with growth conditions such that the formation of nanowires is
energetically preferable to film formation. This is an example of utilization of growth parameters for self-assembly.
Image: (Schuster et al, 2012).

25
1.3 Summary

Continued technological progress demands more functionality from electronics without
sacrificing performance. Complex oxide perovskite thin films and heterostructures are a
promising avenue to achieve these goals, with active investigation and development
currently underway on a wide array of fronts. However, the primary limitation to
development is the inherent complexity of this correlated system, especially in comparison
to traditional mean field semiconductors. Improving our understanding of the
process/property/structure relationships of these materials requires a more complete suite
of tools for monitoring dynamic thin film growth events, thereby illuminating a more
comprehensive view of the entire system. The development and application of new
techniques will underwrite the next generation of emergent quantum thin film phenomena
for electronic, photonic, and energy applications.
Chapter two proceeds with a thorough introduction and overview of the material synthesis
methods and characterization techniques used in this work. Special attention is given to
the PLD growth process. Chapter three discusses the implementation and operation of
the AugerProbe, its capabilities and limits, and determination of its sensitivity. Chapter
four outlines methods used to quantify film composition with AES, and the unexpected
observation of termination effects during deposition. Chapter five discusses the methods
used to observe dynamic thin film events in real time during pulsed laser deposition.

26

Chapter 2
M aterials Synthesis and
Characterization

2.1 Introduction

This body of work is about the development and novel application of a recent innovation
by Staib Instruments: the AugerProbe.
133
As such, there should be no doubt as to my
perspective on the merits of surface characterization. However, as is always the case, the
introduction of a new characterization technique is meaningless without verification of its
capabilities by the application of the old. Ensuring the validity of every experimental step
that was taken required the judicious use of the giants beneath our feet: reflection high
energy diffraction, atomic force microscopy, X-ray diffraction and reflection, scanning and
transmission electron microscopy, elemental dispersive X-ray spectroscopy, electron
energy loss spectroscopy, and the extensive knowledge base required to develop models
and simulations. The basic principles of these techniques are outlined in this chapter, with
the addition of relevant examples where appropriate.
This chapter begins with an in-depth examination of pulsed laser deposition, from
substrate selection and preparation to the dynamics of thin film growth. The traditional
in situ thin film characterization technique, reflection high energy electron diffraction, is
discussed, as well as the primary focus of this work: in situ Auger electron spectroscopy.
Atomic force and electron microscopies are introduced, with examples of their application
to this work. X-ray diffraction of thin films for structural characterization is explained,
and compositional characterization by energy dispersive X-ray spectroscopy and electron
energy loss spectroscopy are demonstrated. Lastly, the methods developed and
implemented for analysis and modeling are covered.

27
2.2 Synthesis

Epitaxial thin film growth techniques can produce remarkably precise thin films and
heterostructures. These feats of engineering exemplify the scientific progress that makes
the modern world possible. However, the cost of such accomplishment is not insignificant,
and complex destinations require complex paths to get there. This section outlines the
methods required for epitaxial thin film synthesis using our PLD system, beginning with
substrate preparation and concluding with a description of the theory governing the
deposition process.

2.2.1 Substrate
The substrate is the rock on which the house is built. Without the understanding and
control of the quality of its surface, there is scant understanding or control of the quality
of the film deposited on it. Especially at the forefront of epitaxial thin film and
heterostructure engineering, where the success or failure of devices has been shown to
depend on a single atomic layer,
37
the substrate surface must be as close to atomically
perfect as it can be made. Furthermore, this degree of quality must be reliably, and
consistently, achievable.
Crystallography
As a brief aside, an overview of the
topics in crystallography most relevant
to thin films will be presented here.
Crystallography is the study of the
arrangement of atoms into crystalline
solids, where a crystal is an ordered
structure with symmetric patterns
repeated along three axes in space. The
smallest group of atoms that repeats in
this manner, and represents the
complete symmetry and structure of the
overall crystal, is the unit cell.
147
The
translational position of the unit cell,
such that the repetition defines the
larger crystal, is the lattice. The lengths
of the principal axes (𝑎 , 𝑏 , 𝑐 ), and angles
between them (𝛼 , 𝛽 , 𝛾 ), of the unit cell
Figure 2.1 An illustration of the cubic SrTiO3 unit cell with
the corresponding atomic coordinates (top), as compared to the
octahedral distortion in CaZrO3 (bottom left), and the position
of the octahedra within the primitive cell (dark dashed line,
bottom right), shown within four cells demarcated with a thin
dashed line. Image: the author.
28
are the lattice constants, and are used to define the positions of the atoms. These concepts
are demonstrated in Figure 2.1, which shows the cubic SrTiO
3
unit cell and
corresponding positions, contrasted against the orthorhombic CaZrO
3
structure. The
primary distinction between SrTiO
3
and CaZrO
3
is that SrTiO
3
has a cubic unit cell (𝑎 =
𝑏 = 𝑐 , 𝛼 = 𝛽 = 𝛾 = 90°) while CaZrO
3
is orthorhombic (𝑎 ≠ 𝑏 ≠ 𝑐 , 𝛼 = 𝛽 = 𝛾 = 90°). This
means that crystallographic directions are trivial for SrTiO
3
without a frame of reference
because each face of the cube is identical, but define the orientation of CaZrO
3
because
each pair of faces is unique. Note that the CaZrO
3
octahedra shown on the bottom left in
Figure 2.1, within the distorted prism of Ca, is not an equivalent unit cell to that of
SrTiO
3
because the structure requires two different octahedral distortions, making the
unit cell larger as shown on the bottom right of Figure 2.1.
Crystallographic directions are identified with Miller indices, a notation system for
indexing planes within the crystal structure. They are written with three integers ℎ, 𝑘 , and
𝑙 , using several specific notations depending on the subject, shown in Table 3, where
𝑎 1
, 𝑎 2
, and 𝑎 3
are the basis vectors of the lattice.
Table 3 Miller index notations and their corresponding definitions.
Notation Definition
[ℎ 𝑘𝑙 ] A direction in the basis of the direct lattice.
〈ℎ 𝑘𝑙 〉
All directions that are equivalent to
[ℎ 𝑘𝑙 ] by symmetry.
( ℎ 𝑘𝑙 ) A plane that intercepts the three points
𝑎 1
ℎ
,
𝑎 2
𝑘 , and
𝑎 3
𝑙 .
{ ℎ 𝑘𝑙 }
All planes that are equivalent to
( ℎ 𝑘𝑙 ) by symmetry.

Figure 2.2 Examples of the planes corresponding to the identified Miller indices within a body-centered cubic unit
cell. Image: the author.
29
The significance of crystallographic directions in the context of thin film growth is in their
definition of the surface. Depending on how you cut the unit cell, the resulting surface
configuration can look completely different, illustrated in Figure 2.2. More than just
appearance, the surface configuration will determine the energy and possible
reconstructions of that surface as it tries to minimize its energy. Furthermore, it defines
the structure required for epitaxial growth, as the film will try to continue the surface
structure in the growth direction. Building on the prior discussion of the structure-
properties relationship in complex oxide perovskites from §1.1.2, it is clear that the surface
will have properties based on its structure, and therefore form interfaces with different
properties that depend on that initial structure.
The orientation which receives
the most attention in SrTiO
3
is
(001), or, in other words, with a
surface plane that intercepts the
three points (0,0,1). Viewing
SrTiO
3
along this axis, shown in
Figure 2.3, the structure can be
seen as alternating layers of SrO
and TiO
2
. This is significant
because it allows control of the
(001) surface using chemical
etching methods,
148
outlined
below, which can selectively
form a surface ‘terminated’ with only TiO
2
. This provides even greater control over the
properties of the interfaces grown on this surface, as there are significant differences in
the energy and electronic structure of SrO versus TiO
2
termination.
149

Only when the atomic structure and its orientation are known can thin film growth be
repeatably successful, which is why the silicon wafer manufacturing process has been so
thoroughly investigated by the microelectronics industry. Modern semiconductor
fabrication facilities use 300 mm diameter, crystallographically-oriented, single-crystal
silicon wafers that are 99.9999999% pure (9N) or higher, and have been the standard since
2002.
150
Recognize that these wafers are cut from giant cylindrical single crystals, called
boules, and then consider that the standard wafer in 1960 was 25 mm in diameter: the
progress has been significant.
139

Crystal Growth
There are several methods for growing single crystals large enough to be used for
substrates, the choice of which typically depends on the material. The process used for
silicon, for instance, is typically the Czochralski method, which carefully pulls a precisely-
Figure 2.3 Illustration of the structural layering in (001) SrTiO3.
Image: the author.
30
oriented, rotating, seed crystal out of a vat of molten silicon with tightly controlled
temperature gradients. The design of the pull is such that the seed crystal grows into a
large cylindrical ingot as it is being extracted from the molten silicon, while maintaining
the original crystal orientation. The boule is then sliced and polished into dimensionally-
precise wafers.
SrTiO
3
, on the other hand, is often synthesized with the Verneuil method, for which the
process is somewhat different. Powdered precursors are fed to an oxyhydrogen flame which
melts them onto a pedestal supporting the growing crystal. The pedestal is slowly moved
away from the flame while keeping a molten tip to continue the growth. This method is
capable of producing gem-quality single crystals in boules typically up to approximately
10 cm long and 30 mm in diameter.
152
With carefully controlled growing conditions and
feedstock, the crystals synthesized with this method can be very high quality.
For some materials, the phase diagram in the region of the liquidus line prohibits their
growth directly from the molten form. This can be for various reasons, such as because
the desired phase is metastable and must be reached through a different thermodynamic
pathway or, inversely, that an undesired phase is metastable and must be avoided. BaTiO
3

illustrates this problem because the melting point is ~1900 K, but it undergoes a hexagonal
to cubic phase transition at ~1700 K, shown in the BaTiO
3
phase diagram in Figure 2.4.
This prevents the synthesis of viable single crystal substrates with the Czochralski method
because on cooling the phase transition introduces thermal strains and defects that inhibit
the ultimate quality.
153
To overcome this problem a different synthesis technique is used,
called top seeded solution growth, which works for systems requiring a non-congruent
melt.
154
For BaTiO
3
, the
approach uses the phase
diagram to reduce the
temperature of the liquidus
line by adding excess TiO
2
. A
seed crystal is then pulled
from the melt, much like the
Czochralski method.
However, because the solution
is not the same composition
as the crystal, the diffusion-
limited growth rate will often
be slower.
Figure 2.4 The BaTiO3 phase diagram as a function of TiO2 mole percent.
Image: (Belruss et al, 1971).
31
SrTiO
3
is a very versatile substrate, and the conditions for single crystal growth are well
established, making it the standard for complex oxide film growth. Another popular
complex oxide substrate is (LaAlO
3
)
0.3
(Sr
2
TaAlO
6
)
0.7
(LSAT), which is an example of a
perovskite oxide which can be grown with the Czochralski method, improving production
capability.
155
There are many other single crystal complex oxide substrates available,
listed in Table 4, with enough variation in lattice parameter for many applications,
illustrated in Figure 2.5. However, there is a continued effort to reliably produce single
crystal substrates of new materials in order to expand the range of lattice parameter and
material properties available for the growth of thin films and heterostructures.
Annealing
Substrates are typically purchased polished on one or both sides to an extremely fine
finish, with R
a
roughness on the order of nm or less. However, the polishing process is
mechanical, and the surface is left in a disordered state, regardless of orientation. To
improve the quality of the polished surface the common practice for complex oxide
substrates is to anneal them in flowing oxygen. This serves to allow the surface to
restructure itself after the polishing process, and aims to restore oxygen stoichiometry. If
a vacuum furnace is used, it may reduce the substrate and change its properties, as is the
case with reduced-SrTiO
3
, which conducts due to the formation of oxygen vacancies.
156

Depending on the objective of the growth, this may be a desired initial state for the
substrate. However, to ensure the quality of the substrate surface and limit atmospheric
contamination, the annealing environment is controlled by flowing oxygen. Flowrate
during the annealing process is typically held at 100 standard cubic centimeters per minute
with a mass flow controller and negative-pressure outlet.

Figure 2.5 Illustration of the range of single crystal substrate lattice constants available for the deposition of epitaxial
complex oxide perovskite thin films. Image: the author.
32
Table 4 Single crystal substrates used for epitaxial growth of complex oxide perovskites, listing the available
orientations, crystal structure, a and c lattice constants, growth method, and source. If the orientation or growth method
are not listed, it indicates the information was not provided by the supplier. For orthorhombic structures, the lattice
constants listed are for the pseudo-cubic unit cell. For growth methods, ‘Cz’ refers to Czochralski, ‘Vern’ refers to
Verneuil, and ‘TSSG’ refers to top seeded solution growth. For the sources, ⸙ means it was available from Crystec, ⸸
was from Powerway, and ⸹ was from MTI; ※ is (Mazur et al, 1997), and ‡ is (Helden et al, 2019).
Material Orientation Structure a c Method Source
DyScO3 (110) Orth. 3.96 3.95 Cz ⸙
GdScO3 (110) Orth. 3.97 3.97 Cz ⸙
LaAlO3 (100),(110),(111) Cubic 3.82 - Vern, Cz ⸙
(LaAlO3)0.3-(Sr2AlTaO6)0.7 (100),(110),(111) Cubic 3.87 - Cz ⸙
NdGaO3 (110),(100),(001) Orth. 3.85 3.86 Cz ⸙
SrLaAlO4 (100),(001),(110) Tetrag. 3.75 12.63 Cz ⸙
SrLaGaO4 (100),(001),(110) Tetrag. 3.84 12.68 Cz ⸙
SrTiO3 (100),(110),(111) Cubic 3.91 - Vern ⸙
YAlO3 (100),(110),(111) Orth. 3.67 3.71 Cz ⸙
LaGaO3 (110) Orth. 3.89 3.90 Cz ※
SmScO3 (110) Orth. 3.98 3.99 Cz ‡
NdScO3 - Orth. 3.99 4.00 - ⸸
PrScO3 - Orth. 4.01 4.03 - ⸸
TbScO3 (110) Orth. 3.96 3.96 Cz ⸹
KTaO3 (100),(110),(111) Cubic 3.99 - TSSG ⸹
BaTiO3 (100),(001) Tetrag. 3.99 4.04 TSSG ⸹
NdCaAlO4 (001),(100) Tetrag. 3.69 12.12 Cz ⸹
La0.95Sr0.05Ga0.95Mg0.05O3-x (001),(110) Orth. 3.89 3.90 Cz ⸹

33
The two variables considered in substrate annealing are temperature and time. To improve
the quality of the substrates the desired outcome is to provide the surface atoms with
enough energy to rearrange themselves in a smooth, low-energy, configuration, and enough
time to do so. If the temperature is too low the surface diffusion will also be too low to
observe the desired change in a reasonable timeframe. If the temperature is too high
unwanted side effects can occur, such as bulk-diffusion of strontium to the surface of
SrTiO
3
to form SrO crystals.
157-159

Examples of the various surface states observed before and after annealing are shown with
atomic force micrographs in Figure 2.6. In a diffusion-controlled scenario such as high-
temperature annealing, temperature and time are effectively the same parameter, and
therefore the objective is simply to find a balance between the two which achieves the
desired outcome in the least amount of time possible. The annealing recipe used for
complex oxide substrates which was consistently the most successful is shown in Figure
2.7. The furnace was heated to 1100℃ over 3 hours, held at 1100℃ for 3 hours, and cooled
to room temperature over 3 hours, all under 100 standard cubic centimeters per minute
of flowing oxygen.

Figure 2.7 The typical annealing time/temperature recipe
used for single crystal substrates. Image: the author.
Figure 2.6 Atomic force micrographs of single crystal substrates: as-received polished SrTiO3 (left), etched and
annealed TiO2 terminated SrTiO3 (center), and annealed NdGaO3 (right). The height scales are 5, 5, and 3 nm (left
to right). Image: the author.
34
SrTiO
3

The versatility of STO as a substrate is owed to the qualities which have established it as
the prototypical complex oxide semiconductor: relative ease of high quality synthesis,
lattice constant, dopability, termination control, and material properties. The synthesis
and material properties have been discussed previously. The cubic lattice constant of
SrTiO
3
(3.905 Å) is sufficiently moderate to allow epitaxial film growth of a wide range
of materials and, when coupled with its well-studied properties, makes it a first choice for
many heterostructures. The dopability of SrTiO
3
broadens its appeal, as Nb
5+
substitution
on the Ti
4+
site results in conductivity
proportional to the degree of doping.
160.161
While
it can be doped with any number of elements to
make it n-type (Nb, La) or p-type (Sc, In),
162-164

single crystal substrates of Nb-doped SrTiO
3
with
excellent quality are readily available, illustrated
in Figure 2.8, making it one of the few
commercially available conducting perovskite
oxide substrates. Lastly, the surface termination
can be manipulated with a chemical etching
process, providing an additional degree of control
over the quality and properties of the resulting
heterostructures.
Originally reported in 1998 as a technique for
improving the perfection and reproducibility of
the single-crystal SrTiO
3
surface,
148
using a
selective hydrofluoric acid (HF) etching technique
to consistently produce a TiO
2
terminated surface
became the standard for growth on these substrates. The importance of this technique
increased exponentially, however, when the termination-dependent LaAlO
3
/SrTiO
3

conducting interface was reported in 2004.
37
The principle of this technique is that when
exposed to water, the mixed-termination SrTiO
3
surface hydroxylates, forming Sr(OH)
2
-
complexes from the ionic SrO, while the covalent TiO
2
remains chemically stable. On
exposure to a buffered HF acid solution the hydroxide-complex is preferentially dissolved,
leaving a TiO
2
surface which is then cleaned and recrystallized with a final anneal, as
described above.
The original etching recipe called for a 10 minute ultrasonic soak in deionized water,
followed by a 30 s dip in a 7:1 buffered (with NH
4
F) HF solution, referred to as ‘buffered
oxide etch’ (BOE). However, the use of HF is highly undesirable due to its extreme
toxicity, prompting interest in the use of alternative recipes which could produce the same
Figure 2.8 A 10 x 5 mm (left) single crystal
substrate of SrTiO3, and a 10 x 10 mm (right)
single crystal substrate of SrTiO3 doped with 2%
Nb, making it conductive and, as a result, black in
color. Image: the author.
35
result. Many recipes were reported,
165,166
all claiming equivalent results when using BOE,
buffered HCl, aqua regia, or even just deionized water (DI).
167
When attempting to
replicate the reported studies, however, my results were unfortunately mixed, as shown in
Figure 2.9 with atomic force micrographs of prepared SrTiO
3
surfaces using various
methods.

Figure 2.9 Atomic force micrographs of poorly etched SrTiO3 surfaces showing insufficient etching and remnant SrO
(left), excessive etching forming etch-pits (center), and excessive etching attacking dislocation faults caused when
breaking the crystal (right). The image height scales are 5 nm, 20 nm, and 20 nm (left to right). Image: the author.

After various trials a recipe for
selectively etching SrTiO
3
surfaces
was developed that not only
produced atomically flat TiO
2

terminated substrates, but did so
with excellent consistency. Though
more labor-intensive than some
alternative reports, the method
developed is safer and produces less
surface defects because it relies on
longer etching times with a weaker
BOE solution of 50:1, as opposed to
the originally-reported 7:1 solution.
This has the additional benefit of
improving the quality of multiple
substrates etched simultaneously
due to the relaxed dependence on
specific timing. Results of the
improved recipe can be seen in
Figure 2.10.
Figure 2.10 An atomic force micrograph of a perfectly etched
and annealed TiO2 terminated SrTiO3 substrate. The height scale is
4 nm. Image: the author.
36
Before reporting the recipe, it is important to reiterate the extreme toxicity of HF.
168-170

Prior to any work with HF being conducted, it is imperative that the operators are fully
aware of its risks, and that the workplace is well equipped with appropriate PPE,
emergency shower/eyewash stations, and a readily-available calcium gluconate gel.
Furthermore, work with HF should never be conducted alone. Lastly, due to the nature
of HF, and its tendency to attack ceramics (such as glass) and many plastics, HF-safe
containers, tweezers, and disposal systems must be in place before any work is conducted.

STO Buffered Oxide Etch Recipe

Beginning with a new commercially polished (001) STO substrate:
1. 3 minutes – Ultrasonic bath in Acetone
2. 3 minutes – Ultrasonic bath in IPA
3. 30 minutes – Ultrasonic bath in DI
4. Prepare the following workspace, with sufficient PPE in place and using HF-safe
plastic beakers and tweezers:
a. 1 L DI for rinsing/refilling beakers
b. 300 ml DI for rinsing tweezers
c. 15 ml 50:1 BOE
d. 40 ml DI for rinsing the substrate
e. 3x 40 ml DI for final rinse procedure
f. A clearly visible timer/stopwatch
5. Start the ultrasonic bath containing the beaker with 15 ml 50:1 BOE
6. Start the timer
7. Transfer the substrate from the DI bath to the BOE beaker
a. Rinse tweezers in the tweezer bath
8. Wait for ~1 minute
a. While waiting:
b. Dump the DI bath and rinse the container 3 times with DI water
c. Refill the DI bath and set it aside
9. When the timer is nearing ~55 s, begin removing the substrate from the BOE
10. Transfer the substrate to the 40 ml DI rinse, not in the ultrasonic bath
a. Rinse tweezers in the tweezer bath
11. Wait for ~1 minute
12. When the timer is nearing the 2 minute mark (total elapsed time), be ready to
transfer the substrate
37
13. At the 2 minute mark (total elapsed time), transfer the substrate from the DI
rinse to the BOE beaker, in the ultrasonic bath
a. Rinse tweezers in the tweezer bath
14. Wait for ~1 minute
a. While waiting:
b. Dump the DI rinse beaker and rinse it 3 times with DI water
c. Refill the DI rinse beaker
15. When the timer is nearing ~2:55 s (total elapsed time), begin removing the
substrate from the BOE
16. Transfer the substrate to the 40 ml DI rinse, not in the ultrasonic bath
a. Rinse tweezers in the tweezer bath
17. Wait for ~1 minute
18. When the timer is nearing the 4 minute mark (total elapsed time), be ready to
transfer the substrate
19. At the 4 minute mark (total elapsed time), transfer the substrate from the DI
rinse to the BOE beaker, in the ultrasonic bath
a. Rinse tweezers in the tweezer bath
20. Wait for ~1 minute
a. While waiting:
b. Dump the DI rinse beaker and rinse it 3 times with DI water
c. Refill the DI rinse beaker
21. When the timer is nearing ~4:55 s (total elapsed time), begin removing the
substrate from the BOE
22. Transfer the substrate to the 40 ml DI rinse, not in the ultrasonic bath
a. Rinse tweezers in the tweezer bath
23. Wait for ~1 minute
a. While waiting:
b. Dump the BOE beaker into the hazardous waste container
c. Rinse the BOE beaker with the tweezer rinse DI 3 times, dumping the waste
water into the hazardous waste container
d. Rinse the BOE beaker 3 times with DI water
e. Hang the BOE beaker to dry
24. The timer is no longer important, and from here on timing is simply based on how
long it takes to complete each step
25. Transfer the substrate to the first extra 40 ml DI rinse, in the ultrasonic bath
a. Dump the DI rinse beaker and rinse it 3 times with DI water
b. Refill the DI rinse beaker
26. Transfer the substrate to the original 40 ml DI rinse, in the ultrasonic bath
38
a. Dump the first extra DI rinse beaker and rinse it 3 times with DI water
b. Hang the first extra DI rinse beaker to dry
27. Transfer the substrate to the second extra 40 ml DI rinse, in the ultrasonic
bath
a. Dump the DI rinse beaker and rinse it 3 times with DI water
b. Refill the DI rinse beaker
28. Transfer the substrate to the original 40 ml DI rinse, in the ultrasonic bath
a. Dump the second extra DI rinse beaker and rinse it 3 times with DI water
b. Hang the second extra DI rinse beaker to dry
29. Transfer the substrate to the third extra 40 ml DI rinse, in the ultrasonic bath
a. Dump the DI rinse beaker and rinse it 3 times with DI water
b. Hang the DI rinse beaker to dry
30. Transfer the substrate to the original DI bath
a. Dump the third extra DI rinse beaker and rinse it 3 times with DI water
b. Hang the third extra DI rinse beaker to dry
31. Finish cleaning up the workspace
32. Transfer the substrate to an IPA bath
33. Remove the substrate from the IPA and blow-dry with dry-N
2

34. Place the substrate in a clean alumina crucible and transfer to the annealing
furnace
35. Anneal the substrate under 100 SCCM flowing O
2
for the following durations:
a. Heat to 1100℃ over 3 hours
b. Hold at 1100℃ over 3 hours
c. Cool to room temperature over 3 hours
36. Verify surface quality with AFM

39
2.2.2 Pulsed Laser Deposition
PLD is more complex than meets the eye. Though in appearance simple, controlling the
growth process can turn this rudimentary practice into a technique capable of atomic
precision.
171
The basic premise is that the very short pulse duration (~10
-8
s) of the UV
laser creates a localized plasma when it ablates the target, forming a plume of ejected
material that interacts with the background gas in a pseudo-diffusive manner. When the
plume reaches the high temperature substrate it
impinges on the surface and, depending on the
local kinetics, deposits either epitaxially,
amorphously, or as a polycrystalline film. A
picture of the PLD chamber used for the
experiments in this work is shown in Figure
2.11.
The type of laser used in this work is a common
choice for PLD: a 248 nm (5 eV) KrF excimer
laser. Excimer is short for ‘excited dimer,’ and
refers to the gas species used.
172
Because noble
gases (in this case Kr) are highly inert they do
not readily form compounds, but when forced
into an excited state within the laser will form
an excited complex with F. This compound will
release the excess energy through spontaneous
or stimulated emission, and therefore ‘lase.’
The energy of the laser is significant because it
is high enough to overcome the bandgap of most complex oxides, which is necessary for
proper plume creation and ablation. The short pulse duration is likewise important
because the large energy input in a short timeframe limits target heating and encourages
plume formation.
The laser path, shown in Figure 2.12,
passes through a mask and attenuators,
is twice reflected by mirrors, then
focused with a lens onto the target. By
adjusting the position of the lens, the
size of the spot created by the laser can
be changed, thereby changing the
fluence. However, there is only one
position where the spot will be in focus,
and positions not in focus have the
Figure 2.11 A picture of the deposition chamber
interior used in this work, showing the AugerProbe,
electron gun, sample heater, and target. Image: the
author.
Figure 2.12 An illustration of the laser path used to control
the size, shape, and energy of the beam. Image: the author.
40
potential to apply a non-uniform pulse energy and thus change the way the material is
ablated.
The target is most commonly the same material as the desired film, though manipulation
of growth parameters can be used to grow thin films of different composition than the
target.
173,174
Likewise, different targets can be used sequentially to grow thin films of solid
solutions or new compounds.
175
The target is typically either polycrystalline or single
crystal, with the ideal scenario most often being a single crystal target due to the purity
and consistent structure.
176
Polycrystalline targets have the advantage of being relatively
cheap to synthesize with raw powders, a bench press, and a box furnace.
177
However, these
techniques of making polycrystalline targets are rife with opportunities to introduce
contamination through the crucibles and dies, and can have difficulty achieving a desirable
density. Alternate techniques such as hot-pressing and spark plasma sintering can achieve
much better results, but need more complex equipment and have their own contamination
issues.
178

The heater in our chamber is the surface on which the substrate is mounted with silver
paint, to promote thermal contact and keep it suspended. The heater is then inserted into
the chamber where it perches upside-down above the target. It has a simple resistive
element which can reach higher than
850℃. The heater mount has x, y, z
translation and tilt and azimuth
adjustment, to align the substrate to
the electron beam for RHEED.
The many axes of PLD growth
parameters makes untangling their
influences on one another
difficult.
2,94,96,106,179-192
To demonstrate
the parameters which must be
considered in planning a growth, a
sample are listed in Figure 2.13. Their
influence extends beyond the quality of
the film to the composition, structure,
and therefore properties. This is
because they not only determine the
composition of the plume (and therefore
film), they also dictate the kinetics of
the growth itself. During ablation of the
target material the incident laser pulse
arrives with a very short duration
Figure 2.13 A picture of the PLD chamber overlayed with
examples of the many growth parameters which must be
considered during the deposition process. Image: the author.
41
(typically ~10
-8
s) and high energy fluence (typically >1 J cm
-2
). This creates a local
plasma, vaporizing the target material and ejecting a plume, shown in Figure 2.14.
193,194

The plume propagates towards the substrate, interacting with itself and the background
gas while broadening. Arriving at the substrate, the ablated material condenses on the
surface as dictated by the thermodynamics of nucleation and growth. This process can be
controlled by manipulating the incident plume energy through the laser fluence and
background gas, and the kinetics of nucleation by adjusting the substrate temperature.
The manner in which the film grows can have a significant impact on the resulting
material properties as well, because the kinetics of nucleation and growth will determine
the structure and defect density of the deposited film.

Figure 2.14 Photographs of plume formation from laser-ablated targets. The difference in shape between
the two images arises from the different points in time the pictures were taken during plume formation, as
well as the difference in laser energy and background pressure. Image: the author.
42
2.2.3 Growth Parameters
As discussed in §2.2.2, and illustrated in Figure 2.13, the extent of control one has over
the growth process is limited by the scope of their understanding. The complexity of PLD,
and its reliance on microscopic phenomena, can make it feel almost too complex, or even
random, at times. However, the complexity of PLD is also responsible for its remarkable
flexibility, and developing methods to understand and control this technique despite its
difficulties is crucial to the important role it plays in the field of complex oxide thin films.
When confronting the array of growth parameters which can be manipulated, it may seem
intuitive to approach the problem from the perspective of changing specific parameters
for specific results in the growth process. However, the inherent interrelationships between
parameters makes this approach, often, a dead end. This is because the solution space is
frequently nonlinear and therefore the objective may require the alteration of several
parameters, rather than just one. The relationship between growth parameters and film
properties is indirect. The correct approach, then, is from the perspective of the dynamic
growth process itself. This means the objective must be to control the manner in which
the film grows, rather than the results of the growth, because the relationship between
parameters and the dynamics of the growth process is direct.
Ablation
Let us address the operation of pulsed laser deposition as a whole, starting at the
beginning: with the arrival of the laser pulse. The pulse can be classified as either long or
short, depending on how its energy is
propagated within the material. If the pulse
length 𝜏 is greater than 𝑑 /𝑐 𝑠 , the thickness of
the layer heated by the thermal wave of
electrons divided by the speed of sound in the
material, then the arriving energy will be
unloaded by the lattice as phonons, and the
only ablation which occurs will be surface
evaporation. However, if the pulse length is
short (𝜏 < 𝑑 /𝑐 𝑠 ), the excess energy cannot be
unloaded by the lattice and the result becomes
a function of electron heat transfer, illustrated
in Figure 2.15.
The electric-field amplitude of an
electromagnetic wave is given by Equation
2.1, where Φ is the power density, 𝘀 0
is the
Figure 2.15 Qualitative profile of electron
temperature (Te) for the three initial stages before
ablation: (1) – absorption of radiation, (2) – propagation
of the thermal wave, (3) – electron-lattice relaxation,
indicated by temperature equilibrium between the
electrons and the lattice (Ti). Image: (Phipps, 2007).
43
permittivity of free space, 𝑐 is the speed of light, and 𝑛 is the refractive index.
𝐸 = (
2Φ
𝑐𝑛 𝜀 0
)
1/2
(2.1)
A material with a refractive index of 2.4, such as SrTiO
3
, with a 20 ns pulse length and
fluence of 1 J cm
-2
(standard operating parameters for our laser), would be subjected to a
field strength of ~2.5 x 10
5
V cm
-1
. Note that the dielectric breakdown strength of SrTiO
3
single crystals is ~4 x 10
5
V cm
-1
.
196,197
However, because the pulse duration extends
beyond the time it takes for electron-phonon coupling, the maintained electric field
prevents Coulombic explosion.
94,98
This is advantageous for PLD because ultrafast lasers
(with ps or fs pulse-widths) often produce nanoparticles, which is not conducive to
epitaxial thin film growth.
198

The difference in surface treatment by different pulse-widths
is illustrated in Figure 2.16.
The thickness of the layer affected by the pulse is determined by the absorption coefficient
of the material for that specific wavelength, following a typical Beer-Lambert law profile,
and the pulse length depends on the laser design. The excited electrons are either freed
and escape as photoelectrons, which is a nonthermal process, or they remain and transfer
their energy to the lattice via electron-phonon coupling, which occurs on the order of
picoseconds, and begins the thermal process within the optical absorption depth of the
material 1/𝛼 , with 𝛼 the optical
absorption coefficient. The
length of thermal diffusion 𝑙 𝑇 is
given by Equation 2.2, where 𝐷
is the thermal diffusivity
coefficient.
199
If 𝑙 𝑇 is less than
the absorption depth, which is
similar to 𝜏 < 𝑑 /𝑐 𝑠 , then the
bulk will be heated to 1/𝛼
regardless of 𝜏 . The satisfaction
of this condition guarantees
congruent evaporation,
200
which
is clearly an important factor of
complex oxide deposition. This
is also why the use of fast UV
lasers is standard for complex
oxides: because longer pulse
durations allow too much
thermal diffusion.
Figure 2.16 SEM images of microprocessing conducted with a ns
pulsed 266 nm laser (left), and a 100 fs pulsed 780 nm laser (right), on
glass, demonstrating the different ablative methods for different pulse-
widths, where the ns pulse results in more heat being transferred to the
surface. Image: (Lucas et al, 2012).
44
𝑙 𝑇 = 2√𝐷𝜏 (2.2)
Based on the topics thus far discussed, the laser wavelength will play a significant role in
the ablation process. Higher energy may increase the degree of ionization with minimal
heating effects, but also have a lower optical absorption depth and will therefore have a
decreased ablation yield. Lower energy may require a much higher power (shorter pulse)
to overcome the heating effects, which can then result in a greater amount of nano or
microparticles, which has also been associated with lower laser energy even without the
shorter pulse length.
201
Furthermore, the target material must be considered, as the
function of ablation relies on ionization. Complex oxides tend to have substantial
bandgaps on the order of a few eV and, as a result, wavelengths below that threshold will
require multiple phonon ionization events, which can significantly decrease the ionized
yield and limit the ablative power. The absorption of the laser by the plume being
generated must also be considered, because longer wavelengths have higher absorption,
and when the plume shields the target it significantly decreases the laser efficiency for
ablation.
202
For these reasons, deliberately or by trial-and-error, the community has
generally settled on excimer lasers with wavelengths in the range of 200 to 400 nm (6.2 to
3.1 eV) and pulse-widths on the order of 10 to 20 ns. Excimer lasers have the additional
advantage of a ‘top hat’ energy profile, as opposed to the Gaussian profile of, for instance,
Nd:YAG lasers.
The degree of ionization immediately after the end of the laser pulse has been
experimentally observed to be between 0.1 and 1, with initial plasma temperatures in
excess of 10 000 K.
202,203
Though the surface is quickly shielded from the full duration of
the pulse by the plasma itself, preventing further ablation, the plasma remains in close
thermal contact with the surface. The high degree of ionization means the plasma is
localized at the surface at pressures up to 10
9
Pa, depending on optical absorption of the
target material.
203
The thermal coupling of the plasma and surface which results from this
process causes additional absorption of the plasma energy by the target, resulting in
thermal equilibration and a lower degree of ionization, but additional thermal evaporation.
In fact, experiments have demonstrated that a great deal of the photonic energy is simply
absorbed by the target, with one report finding that this was the case for nearly a third
of the total photonic input.
204
Regardless, after pulse-duration-imposed equilibration, the
high pressure region will expand adiabatically at velocities up to 30 000 m s
-1
.
94,205

Plume Propagation
Determining the manner in which the plasma plume expands into the vacuum or
background gas has been studied extensively with models and time-resolved plume
imaging.
94,98,206-216
Though the specifics of any given plume will depend on the species,
ablation process, and background gas, the general model can be approximated as a drifted
45
Maxwellian with a velocity normal to the surface given by Equation 2.3, where 𝑣 is the
thermal velocity and 𝑣 ̅ is the center-of-mass velocity. While the particle density is still
very high there is still a great deal of collision within the plume, resulting in ion-electron
recombination and electron transfer between species. This ends after the first few mm of
expansion and the plasma has become generally collisionless.
98
As the plasma continues
expanding it cools adiabatically to ~3000 to 5000 K, with species’ kinetic energies from
~1 to several 100s of eV.
217

𝑃 ( 𝑣 ) ~𝑣 3
𝑒 −𝑚 ( 𝑣 −𝑣̅)
2
2𝑘 𝐵 𝑇 (2.3)
Again, because the specifics of plume expansion are dependent on the material, pressure,
and ablation parameters, decisions should be made on a case-by-case basis. The influence
of background gas and pressure will severely alter the manner in which the plume
propagates, as well as the species it contains.
95
For instance, oxidation of the plume can
alter the shape and kinetics of the plume, as well as the composition of the film itself.
218

Figure 2.17 The relationship between pressure and mean free path (𝜆 ) for the gas temperatures indicated, assuming
the kinetic diameter for O2 of approximately 346 pm. The different pressure regimes are demarcated for the 75 mm
target-substrate distance used in our chamber. Image: the author.
46
In the interest of a general approach to overcome the abundance of growth-dependent
specifics, three different pressure regimes have been identified for different plume
behavior.
94,219
They are: ‘vacuum-like,’ with no interaction between the plume and the
background gas; ‘transition,’ with moderate interaction; and ‘diffusion-like,’ where the
interaction between plume and background gas yields broad angular distribution of species
and very slow plume propagation. The regime can be identified by simply calculating the
mean free path (𝜆 ) for the given background pressure with Equation 2.4, where 𝑇 is the
gas temperature and 𝑑 𝑚 =
𝑑 𝑔𝑎𝑠
+𝑑 𝑠𝑝𝑒𝑐𝑖𝑒𝑠 2
.
199

𝜆 =
𝑘 𝐵 𝑇 √2∙𝑃 ∙𝜋 ∙𝑑 𝑚 2
(2.4)
When 𝜆 is greater than the target-substrate distance it is considered the ‘vacuum-like’
regime. Statistically, this simply means that the probability of an ablated particle colliding
with the gas is less than ~60%, given by Equation 2.5, where 𝑑 𝑇𝑆
is the target-substrate
distance. For the target-substrate distance in our chamber, 75 mm, the pressure required
for various mean free paths is plotted in Figure 2.17. Intraplume collisions will continue
to occur as the plume expands, but the greatest influence on its dispersion is the mass of
the individual ablated species, which can cause quite dissimilar behavior for different
elements. The low collision-rate with the background gas will limit lateral plume
expansion, making its center retain the greatest particle density and thus cause a distinct
angular thickness-variance of the deposited material.
210
This regime will also, naturally,
limit reaction between the ablated species and background gas, preventing, for instance,
oxidation. The kinetic energy of the arriving species will be quite high and must be
accounted for in the growth kinetics. However, with such high energies the plume has
been shown to actually rebound from the heater and substrate, mitigating the free
expansion of the plasma plume.
210
𝑃 ( 𝜆 )= 1 − 𝑒 −𝑑 𝑇𝑆
𝜆 (2.5)
When the pressure results in a mean free path in the range 0.1 ∙ 𝑑 𝑇𝑆
< 𝜆 < 𝑑 𝑇𝑆
, it is
considered the ‘transition’ regime, in which some collisions occur between the plume and
background gas.
219
The likelihood of collision can be determined by integrating Equation
2.5 over 𝑑 𝑇𝑆
, with less than 1 in 10,000 particles arriving without collision by the time
𝜆 = 0.1 ∙ 𝑑 𝑇𝑆
. This is a common pressure range for deposition because it is possible to exert
some control over the growth kinetics by adjusting the pressure, and the oxidation of the
plume can improve the film composition. However, this is also where the greatest deviation
in elemental scattering of the plume species is observed. This is purely due to the difference
in mass between the species, but the result can be quite dramatic and has been observed
numerous times experimentally,
189,214-216,220,221
as well as modeled.
222
A simple Monte
47
Carlo simulation of this phenomenon is shown in §4.2.2. Despite the potential for off-
stoichiometry caused by the preferential scattering of lighter species, the benefits gained
within this pressure regime are still significant. Oxidizing the plume without severely
diminishing the kinetic energy of the arriving species can improve the composition as well
as properties of deposited thin films.
205
When 𝜆 is less than 10% of the target-substrate distance it is considered the ‘diffusion-
like’ regime. As shown in Figure 2.17, with a 𝑑 𝑇𝑆
of 75 mm at 1100 K this corresponds to
pressure above ~4 x 10
-2
mbar. Plume propagation within this regime is dictated by
confinement due to the background gas. The breaking effect of plume-gas interactions
reduces plume species divergence, resulting in more homogenous thickness and
composition of deposited materials,
210,223
but the decrease in kinetic energy of the arriving
species can alter the film growth behavior. Additionally, the high pressure can result in
unexpected phenomena such as local density-fluctuations caused by temperature gradients
or excited species,
94
which are difficult to identify without rigorous investigation.
Film Growth
The process of film growth in PLD is, ideally, vapor-phase growth.
199
That is, the plume
transporting the ablated species can be treated as a vapor, and the deposition process will
therefore be governed by the thermodynamics of nucleation and growth.
2,179
This is as
opposed to the formation and deposition of droplets or nano/microparticles, which are not
conducive to epitaxial growth though they may have their uses.
198,201,224
Regardless, the
fundamental benefit to this approach is that we can begin by introducing the system in
thermodynamic equilibrium.

Figure 2.18 Illustration of the interfacial energies at play when determining the behavior of nucleation and growth
in a thermodynamically stable system, where 𝛾 𝑓𝑠
, 𝛾 𝑠𝑣
, and 𝛾 𝑓𝑣
are the interfacial energies between the film and substrate,
substrate and vapor, and film and vapor, respectively. Image: (Ohring, 2001).
48
In a thermodynamically stable system local fluctuations around equilibrium are the
driving force for change. With the introduction of a supersaturated vapor to a solid,
fluctuations will allow the phase transition required to form nuclei at the vapor-solid
interface. The probability of nucleation is determined by the activation energy, and they
will continue to form until a critical density is achieved. The basic concept of this model
is to balance the interfacial energies of the substrate, film, and vapor, illustrated in Figure
2.18.
The interfacial energies at play will determine the manner in which the film nucleates and
crystallizes. For instance, when the interfacial energy between the film and substrate, 𝛾 𝑓𝑠
,
plus the energy between the film and vapor, 𝛾 𝑓𝑣
, is lower than the substrate-vapor energy,
𝛾 𝑠𝑣
, growth will be ‘step-flow,’ where it is energetically preferrable to ‘wet’ the surface
with individual layers (𝛾 𝑠𝑣
≥ 𝛾 𝑓𝑠
+ 𝛾 𝑓𝑣
). On the other hand, if 𝛾 𝑠 𝑣 < 𝛾 𝑓𝑠
+ 𝛾 𝑓𝑣
, then
wetting the surface is not energetically stable and so-called ‘island growth’ occurs. An
intermediate growth mode is possible, ‘layer-island,’ in which the film starts out in layer-
by-layer mode but accumulates biaxial strain from lattice mismatch with the substrate.
At a critical thickness the strain is relieved by the formation of dislocations, which triggers
island formation.
These scenarios are a tidy approach to understanding film growth in thermodynamic
equilibrium, but unfortunately in PLD the deposited film is not in equilibrium, and kinetic
effects must be considered. Although the pulsed nature of this growth technique allows
some time for surface diffusion and reconstruction between pulses, the very high
supersaturation associated with the plume arrival results in a large nucleation rate. The
consequence is that kinetic effects largely determine the resulting growth modes.
Using the homoepitaxial deposition of SrTiO
3
as an example, the interfacial energy
between substrate and vapor is the same as that between the film and vapor, and only
2D growth modes should be possible.
225,226
However, beyond the ideal layer-by-layer mode
described above, there exists a ‘step-flow’ growth mode which is also 2D, illustrated in
Figure 2.19. If the arriving adatoms have sufficient surface diffusion they will freely
migrate to atomic step-edges, as that is the lowest energy location. The problem which
arises from step-flow growth is that the surface step density remains constant, which limits
the characterization of surface roughness used for RHEED. Without that information the
ability to monitor the growth rate is hampered.

Figure 2.19 Illustration of the step-flow growth mode, in which surface roughness is not changed despite deposition
due to high surface diffusivity which allows transport to the low-energy step edge locations. Image: the author.
49

The behavior of the incoming adatoms once they reach the substrate will be determined
by their kinetic properties and the energetic parameters of the surface. Imagining the
deposition process, an arriving adatom will travel the surface looking for other adatoms
to nucleate with, existing nuclei to join, or a step edge to attach to. If it is located on a
step or terrace, it might reach the edge and fall, or remain above, depending on the energy
barrier associated with that transition. It may even desorb, returning to the vapor. The
most important parameter dictating the outcome is likely the surface diffusion coefficient,
𝐷 𝑆 , given by Equation 2.6, where 𝐸 𝐴 is the activation energy for diffusion, 𝑣 is the attempt
frequency, and 𝑎 is the characteristic jump distance. The net distance traveled determines
what behavior is possible, and is given by the diffusion length 𝑙 𝐷 , in Equation 2.7, where
𝜏 is the residence time before the adatom reevaporates.
𝐷 𝑆 = 𝑣 𝑎 2
𝑒 (
−𝐸 𝐴 𝑘 𝐵 𝑇 )
(2.6)
𝑙 𝐷 = √𝐷 𝑆 𝜏 (2.7)
From Equation 2.6, the temperature is clearly an important consideration among growth
parameters as it will determine the diffusivity. If 𝑙 𝐷 is longer than the average terrace
width the result is step-flow growth, as the adatom mobility is sufficient to treat the step
edges as an energy sink, limiting terrace nucleation. The step edges will march together
without modifying the net surface structure. If, on the other hand, 𝑙 𝐷 is smaller than the
average terrace width nucleation will occur until a saturation density is reached. After
saturation, the probability of joining existing nuclei becomes higher than forming new,
and they will grow and coalesce as layer-by-layer growth.
In Ideal layer-by-layer growth, the diffusing adatoms will have enough energy to reach
the step or terrace edges and overcome the energy barrier to fall down, thereby creating
a perfectly smooth film layer before nucleation is initiated again. However, if they either
do not have the ability to reach the edges, or the barrier to descent is too great, nucleation
will occur on existing terraces and form a second layer before the first layer has finished
depositing. In reality, layer-by-layer deposition is commonly somewhere between these
two extremes, which is observed as the gradual decay in RHEED oscillation intensity as
growth progresses and the surface roughens, shown in Figure 2.20. The principles of
RHEED are described in §2.3.
50

Figure 2.20 RHEED oscillations observed during homoepitaxial deposition of SrTiO3. Note the sustained layer-by-
layer growth until lost, with 38 oscillations observed in 1500 s, corresponding to approximately 15 nm of deposition.
Image: the author.

In summary, the pulsed laser deposition process, from ablation to film growth, has many
opportunities for things to go awry. However, despite its complexity, a great deal of effort
has been made to understand the nuances of this technique, resulting in a strong
conceptual grasp of its many moving parts. The physics specific to PLD makes it excellent
for the growth of complex stoichiometry compound thin films and provides unique control
over aspects of the deposition process not possible with other methods.
227-231
With the
continued development of analysis techniques, the ability to understand and control the
composition and structural outcome of growth parameter manipulation will continue to
improve.
94,100,232-234

51
2.3 In Situ Characterization

In situ characterization is a field that will not lose significance as long as people are still
growing thin films. Simply adding the capability to observe structural and thickness
information from the deposition as it was occurring, RHEED completely changed the way
thin films were grown and revolutionized their quality, precision, and consistency.
115

Various characterization techniques are constantly being applied to new problems and in
new ways, broadening the collective comprehension one small step at a time and improving
our synthesis methods.
194
Understanding more about the dynamics of the thin film growth
process will never be a hindrance, so it stands to reason that in situ characterization
techniques will continue to improve – expanding the pinhole through which we observe
these complex and fascinating phenomena.

2.3.1 RHEED
Reflection high energy electron diffraction is a simple concept with enormous depth.
236

The basic premise is to reflect a low-angle electron beam off of a single crystal surface,
thereby forming a diffraction pattern of the 2D lattice, as illustrated in Figure 2.14. The
electron source requires sufficient acceleration to limit unwanted scattering from the
background gas. In our system, the electron acceleration is variable, up to 35 kV, which
permits in situ RHEED at relatively high pressures above 1 mbar.
The diffraction portion of RHEED obeys Bragg’s law
with the subtle difference that because the surface is
2D the diffracted pattern is also 2D. However, when
treated as 3D the reciprocal lattice points become
reciprocal lattice rods, and the diffraction pattern is
located where those rods intersect the Ewald sphere,
shown in Figure 2.21. This has the additional effect
of elongating the diffraction spots when the surface
is very smooth, and forming a 3D lattice when the
surface is very rough, illustrated in Figure 2.22
(left).
While the diffraction pattern holds a tremendous
amount of information, including surface structure,
reconstructions, and quality,
237-244
only some of that
information can be gathered qualitatively. For
instance, the diffraction patterns of different
Figure 2.21 A picture of SrTiO3 (001)
surface diffraction with RHEED, collected with
a 5 kV electron source at 10
-4
mbar O2. Image:
the author.
52
structures will be visibly unique, as will the diffraction patterns of different crystal
orientations of the same structure, with the former shown in Figure 2.22 (right).
Reconstructions can be observed as intermediate diffraction spots, and Kikuchi lines can
be seen as lines connecting zones, shown in Figure 1.19 and 2.21
The reflection portion of RHEED is used for monitoring layer-by-layer deposition.
124,125

By observing the intensity of the 0
th
order diffraction spot, called the specular spot, the
surface smoothness can be seen to oscillate as material is deposited. Low intensity is, of
course, indicative of a rough surface while high intensity corresponds to a smooth surface.
The principle of RHEED oscillations is illustrated in Figure 2.23, and example
oscillations are shown in Figure 2.24.
Figure 2.22 An illustration of the Ewald sphere construction for a 2D surface (left), with the resulting reciprocal
lattice rods, image distributed under CC:BY – Ponor. (Right) an image of the diffraction pattern formed with RHEED
when the surface is 3-dimensional. Shown is a CaZrO3 thin film with a 35 kV beam at 10
-4
mbar O2. Image: the author.
Figure 2.23 Contrasting diffraction patterns from CaTiO3 (left) and LaMnO3 (right), showcasing the different
structures and thus different patterns observed. Image: the author.
53
In practice, there are many factors that can contribute to nonideal RHEED oscillations
illustrated in Figure 2.25. While the intensity of the specular reflection from the clean
substrate is often higher than that observed after deposition has begun, the intensity can
decay to a point where oscillations are either so weak as to barely show up in the noise,
or are completely gone. This can occur for several reasons, the simplest being that the
growth quality is poor, and the surface is not recovering between pulses in a way that
creates layer-by-layer growth. Another possibility is that the growth quality is not poor,
but during the between-pulse recovery period surface diffusion allows the incoming
adatoms to move to the lowest energy location at the step edges. This is called step-flow
growth, and because the surface smoothness does not change, it will not demonstrate
RHEED oscillations, discussed further in §2.2.3.
100,227,245
Other events that can cause
confusion in the RHEED signal are lower frequency oscillations such as beats, envelopes,
or overlapping oscillations, seen in Figure 2.20 and 2.25.
236,245

Figure 2.25 RHEED oscillations from 3 different growths (left) showing consistent long-term oscillations (top),
quickly lost oscillations (center), and odd intensity behavior (bottom). Overlapping RHEED oscillations (right) during
the deposition of CaZrO3, showing two clear signals and an envelope. Image: the author.

Figure 2.24 A demonstration of the principle behind RHEED oscillations, showing the higher intensity observed
from a smooth surface, corresponding to a plotted peak, compared to the lower intensity observed from a rough surface,
corresponding to a plotted trough. Image: the author.
54
2.3.2 Auger Electron Spectroscopy
The generation of Auger electrons occurs through a fairly well-understood process.
138
The
generation of a hole in an inner electron shell prompts a higher energy electron to take its
place, releasing energy in the process, which removes an outer shell electron, the Auger
electron. The energy of the Auger electron will depend on the energies of the first two
shell electrons, and is therefore characteristic to the element it originated from. The energy
of the Auger electron can be calculated with Equation 2.8, where the electron energy levels
are K, L
1
, and L
2
, the Auger electron would be designated KL
1
L
2
.
𝐸 𝐾 𝐿 1
𝐿 2
= 𝐸 𝐾 − 𝐸 𝐿 1
− 𝐸 𝐿 2
(2.8)
Note that the energy of the Auger electron, unlike in XPS, is completely independent of
the original excitation source. As a result, Auger electrons will have the same energy in
XPS as they do in AES. Furthermore, because an electron beam is used as the excitation
source in AES, electron-electron interactions prevent the formation of characteristic
photoelectrons. This increases the background signal, but guarantees that the only peaks
visible are from Auger electrons.
Generation
The letters used to designate the atomic orbitals are from the quantum numbers: 𝑛 , 𝑙 , 𝑚 𝑙 ,
and 𝑚 𝑠 . These correspond to the principal quantum number 𝑛 , designating the shell, the
azimuthal quantum number (angular momentum) 𝑙 , designating the subshell, the
magnetic quantum number (projection of angular momentum) 𝑚 𝑙 , which indicates the
orientation of the subshell, and the spin quantum number 𝑚 𝑠 , indicating the spin. In X-
ray spectroscopy the nomenclature designates the principal quantum number with a letter,
from K (𝑛 = 1), L (𝑛 = 2), M (𝑛 = 3), etc., and the subscript is simply 𝑙 + 𝑚 𝑙 . The
relationship between the designations for the first three electron shells in comparison to
the optical nomenclature is shown in Table 5.

55
Table 5 The relationship between the designations for the first three electron shells in comparison to the optical
nomenclature.
X-Ray
Nomenclature
Optical
Nomenclature
Principal (𝑛 ) Azimuthal (𝑙 ) Magnetic (𝑚 𝑙 ) Spin (𝑚 𝑠 )
K 1s 1 0 0 ±½
L
1
2s
1/2
2 0 0 ±½
L
2
2p
1/2
2 1 0 ±½
L
3
2p
3/2
2 1 ±1 ±½
M
1
3s
1/2
3 0 0 ±½
M
2
3p
1/2
3 1 0 ±½
M
3
3p
3/2
3 1 ±1 ±½
M
4
3d
1/2
3 2 0 ±½
M
5
3d
3/2
3 2 ±1 ±½
M
6
3d
5/2
3 2 ±2 ±½

This means, then, that 𝐸 𝐾 𝐿 1
𝐿 2
transition described in Equation 2.8 is the result of the
ejection of a 1s electron, prompting a higher energy 2s
1/2
electron to take its place. The
difference in energy is then used to remove the 2p
1/2
electron, which becomes the Auger
electron with energy 𝐸 𝐾 𝐿 1
𝐿 2
. It might be assumed that the ejection of the L
2
electron occurs
by a radiative process, but this is actually not the case. The reason is simply that a
radiative process must always involve a change in both 𝑛 and 𝑙 , which is not the case with
a KLL transition. Therefore, the KLL Auger electrons which have been observed must be
generated without the forbidden radiative process, which implies that it should be treated
as a simple two-electron Coulombic readjustment to the generation of the original K ‘hole’.
In the case of a solid, the expression for the Auger energy given in Equation 2.8 is not
entirely accurate, as it does not account for the work function of the material, which must
be overcome to remove the electron. This would imply that the actual definition of the
Auger energy would be given by Equation 2.9, where 𝜑 is the work function and 𝑍 is the
atomic number, indicating the energy of that specific orbital.
𝐸 𝐴 ( 𝑍 )= 𝐸 𝐾 ( 𝑍 )− 𝐸 𝐿 1
( 𝑍 )− 𝐸 𝐿 2
( 𝑍 )− 𝜑 (2.9)
It might be tempting to assume that 𝐸 𝐿 1
≈ 𝐸 𝐿 2,3
for most materials, which it normally is,
but the energy of the orbitals generating the Auger electron must account for the presence
of the positively charged holes used to generate it. This alters the resulting energy level
by a measurable amount, on the order of 10s of eV. A better approximation for the orbital
energy is then: 𝐸 𝐿 2,3
( 𝑍 + ∆) , where ∆ is an empirical constant (~1) representing the
additional kinetic energy acquired as the electron escapes the multiply charged ion.
246

This brings the expression for the Auger energy to Equation 2.10.
247-249
56
𝐸 𝐴 ( 𝑍 )= 𝐸 𝐾 ( 𝑍 )− 𝐸 𝐿 1
( 𝑍 )− 𝐸 𝐿 2,3
( 𝑍 + ∆)− 𝜑 (2.10)
The shifts in orbital energies occurring during Auger generation mean that the Auger
energies cannot be calculated directly from the energy levels generated with traditional
X-ray techniques.
250-253
This arises from a number of processes including quantum
mechanical exchange, relativistic effects on the inner shells, and Coulombic repulsion
between the two holes; all of which are grouped under the general term ‘relaxation
effects’.
254-258

Figure 2.26 Diagrammatic representation of the K and L energy levels. Image: the author.
In addition to the energy of the transition observed, the transitions which are possible
should be known. Looking only at the states of the K and L shells, illustrated in Figure
2.26, it appears that six transitions should be possible: KL
1
L
1
, KL
1
L
2
, KL
1
L
3
, KL
2
L
2
,
KL
2
L
3
, and KL
3
L
3
. This is the case for high-Z elements (> 80), where spin-orbit coupling
creates definite states due to j-j coupling between the electrons and nucleons.
16
For low-
Z elements (< 20), however, L-S coupling occurs which makes the 1s – 2p2p:
3
P transition
forbidden. As a result, only five transitions are possible. In the intermediate Z-regime
alternate couplings between j-j and L-S are possible, which increases the number of
possible transitions to nine,
259
listed in Table 6.

57
Table 6 Final states for all KLL transitions, adapted from (Ferguson, 1989).
Transition Final State Final Configuration
KL
1
L
1

1
S
0
2s
0
2p
6
KL
1
L
2

1
P
1
,
3
P
0
2s
1
2p
5

KL
1
L
3

3
P
2
,
3
P
2
2s
1
2p
5

KL
2
L
2

1
S
0
2s
2
2p
4

KL
2
L
3

1
D
2
2s
2
2p
4

KL
3
L
3

3
P
0
,
3
P
2
2s
2
2p
4

The initial hole creation happens in a timeframe on the order of ~10
-16
s, and
recombination occurs in ~10
-15
s. By Heisenberg’s uncertainty principle, Equation 2.11,
relating the uncertainty in energy (∆𝐸 ) and time (∆𝑡 ), where ℎ is Planck’s constant, this
will result in energy uncertainty of around 4 eV. What if, instead of a KLL transition, the
electron filling the initial hole comes from the same principal quantum number, such as
L
1
L
1
M, or even M
1
M
2
M
3
? These are called a Coster-Kronig transitions,
260
and due to the
significant wavefunction overlap, the recombination time is much faster, on the order of
~10
-16
s. This results in a corresponding increase in energy uncertainty, and therefore these
transitions are broad and weak, occupying the background. However, Coster-Kronig
transitions do affect the intensity of other Auger transitions by influencing their
probability of occurrence. For instance, ionization of L
1
can cause fast Coster-Kronig
L
1
L
2
Y transitions, which reduces the probability of the L
1
XY Auger process due to
competition from the now-possible L
1
L
2
X Auger transition.
261
∆𝐸 ∆𝑡 = ℎ (2.11)
Since the Auger electrons being observed originate in a solid, it might be assumed that
the population of the energy bands would be reflected in the Auger spectra, as it does
with soft X-rays. However, this is not the case and Auger spectra appear to originate from
free atoms.
262
The reason this occurs arises from correlation effects.
263-265
If the local
Coulombic interactions are larger than the one-electron band width, the spectrum consists
of a narrow signal associated with a broad background.
266
However, if the Coulomb
interaction is small, the spectra will resemble the convoluted band density of states.
267

Escape Depth
In any attempt at surface analysis, knowing the depth from which your information is
generated is clearly extremely important. The depth from which electrons are generated
58
for Auger analysis, the ‘escape depth,’ is determined by their ability to leave the material.
The inelastic mean free path (𝜆 ) encapsulates this concept, and is defined as the depth
which attenuates the number of escaping electrons by 𝑒 −1
, illustrated with Equation 2.12,
where 𝑑𝑁 𝑑𝑧 ⁄ is the rate of attenuation for 𝑁 electrons for depth 𝑧 . Alternately, it can be
expressed with Equation 2.13, which can also be viewed as the probability that a given
electron will escape from depth 𝑧 , given its 𝜆 .

𝑑𝑁
𝑑𝑧
= −𝜆𝑁 (2.12)
𝑁 = 𝑁 0
𝑒 −𝑧 𝜆 (2.13)
The scattering processes which determine 𝜆 are various, and include excitation of phonons,
interband transitions, single-electron excitations, and plasmon excitations. These are
collectively referred to as ‘ZAF effects’ in quantitative models for Auger and, for instance,
energy dispersive X-ray spectroscopy. These phenomena are largely dependent on the
specific energy of the electron and the material it is passing through. Higher atomic
number atoms will more readily block or deflect passing Auger electrons due to their
higher electron density (‘Z’). If there are energy transitions available that correspond to
the energy of the Auger electron, then absorption will occur with greater frequency than
in an energetic neighborhood of forbidden transitions (‘A’). Lastly, the absorption of an
Auger electron’s energy can result in fluorescence or even the emission of another Auger
electron (‘F’). Complications which arise from the material-specific impact of ZAF effects
on signal strength make the use of standards necessary for quantitative analysis.
Looking at the possible interactions which can occur: phonon energy is small and can be
ignored, interband transitions are typically less than 15 eV, but single-electron interactions
are frequent at low energies. The escape depth defined by single electron scattering 𝜆 𝑒 is
given by Equation 2.14, where 𝑎 𝐵 is the Bohr radius, 𝐾 = arctan𝑦 + 𝑦 ( 1 + 𝑦 2
) ⁄ , 𝑦 =
( 𝜋 𝑎 𝐵 𝑟 𝑠 ⁄ )
1/2
, 𝑎 𝐵 𝑟 𝑠 is the radius of a sphere with a volume equal to the volume per electron,
and 𝑥 = ( 𝐸 𝐸 𝐹 ⁄ )
1/2
, with 𝐸 the electron energy and 𝐸 𝐹 the Fermi energy. This implies that
𝜆 𝑒 varies as 1 ( 𝐸 − 𝐸 𝐹 )
2
⁄ at or below the Fermi energy.
268
𝜆 𝑒 =
32𝑎 𝐵 𝐾𝑦
𝑥 2
( 𝑥 2
−1)
2
(2.14)
Above the Fermi energy the dominant scattering process becomes the production of
plasmons, which scatter electrons as 𝜆 𝑃𝐿
, given by Equation 2.15.
269
In this case, 𝑦 𝑃𝐿
=
ℏ𝜔 𝐵 𝐸 𝐹 ⁄ , where ℏ𝜔 𝐵 is the characteristic plasmon energy. This implies that 𝜆 𝑃 𝐿 becomes
infinite at and below the plasmon formation energy ( 𝐸 𝐹 + ℏ𝜔 𝐵 ) , and approximates a linear
function of 𝐸 at much higher energies. Experimental data does not show a discontinuity
59
at ( 𝐸 𝐹 + ℏ𝜔 𝐵 ) , however, which means that 𝜆 𝑒 is likely still dominant beyond 𝐸 𝐹 until
being overshadowed by 𝜆 𝑃𝐿
.
𝜆 𝑃𝐿
= 2𝑎 𝐵 𝐸 (
ln[( 1+𝑦 𝑃𝐿
)
1
2 −1]
𝑥 −( 𝑥 2
−𝑦 𝑃𝐿
)
1/2
)
−1
ℏ𝜔 𝐵 (2.15)
The observed experimental behavior of 𝜆 with Auger energy suggests that the intermixing
of the two models provides a reasonable general approach, and that the overall shape of
the result is consistent,
270-273
shown in Figure 2.27. Following this observation, a
thorough survey of escape depth data was conducted by Seah and Dench.
274
Fitting the
data provided Equation 2.16 for inorganic compounds, also plotted in Figure 2.27 in
comparison to the previous models, with the result in monolayers (𝜆 𝑚 ). 𝐸 is the electron
energy and 𝑎 is the monolayer thickness in nm.
𝜆 𝑚 =
2170
𝐸 2
+ 0.72( 𝑎𝐸 )
1/2
(2.16)

Figure 2.27 Comparison of the models used to approximate contributions to the inelastic mean free path from
electron-electron scattering (Eq. 2.14), plasmon scattering (Eq. 2.15), and the model developed by Seah and Dench from
experimental data (Eq. 2.16). Image: the author.

60

Figure 2.28 A diagram demonstrating the basic operating principle of the AugerProbe. Incoming electrons are
energy-filtered by the collimator lens, then exposed to the retarding field which allows the derivative signal to be
determined by lock in amplification. Image: the author.

Detection
The operation of the AugerProbe is based on a robust design suitable to the requirements
for incorporation into a PLD chamber.
275-277
A schematic showing the basic design
principle is shown in Figure 2.28.
133
Essentially, incoming electrons are filtered with a
collimator lens tuned to selectively guide a specific energy through a pinhole. They then
pass through an AC retarding field before impinging on a scintillator. The scintillator
generates photons from the electrons, which are passed through a photomultiplier and
then detected. The direct signal is delivered by the detector, which can be viewed or
passed through a lock in amplifier tuned to the frequency of the AC retarding field. As a
result, the lock in signal is the doubly-filtered energy signal. The probe collects signal for
a predetermined period at individual energies dictated by step-size, thereby forming a
spectrum of intensity, shown in Figure 2.29.
61

Figure 2.29 Auger spectra from 20 eV to 2 keV, showing the presence of characteristic peaks in the E*N(E) signal,
as well as the dN/dE signal. Image: the author.

When the probe is collecting a spectrum, it is detecting signal in one of several ways,
depending on its manner of operation. The most common method is essentially just
counting electrons for a given energy with an electron energy analyzer, such as a
cylindrical mirror analyzer. The benefit of this type of analyzer is its high efficiency and
resolution, but it requires high or ultra-high vacuum to function properly. An alternative
technique uses a phase sensitive detector to pick out the signal from a noisy background.
This type of modulated signal detection is used by retarding field analyzers (RFAs), and
is the method implemented by the AugerProbe.
The background of secondary electrons can be removed, and the signal enhanced by
differentiating the energy spectrum after introducing a modulation voltage. A sinusoidal
modulation with frequency 𝜔 /2𝜋 and amplitude 𝐸 𝑚 is superimposed on the applied
voltage 𝐸 0
as it is ramped up the energy spectrum. The modulation then appears
superimposed on the detected signal, but when fed to a phase-sensitive lock-in amplifier
tuned to the modulating frequency, the amplitude of the detected signal is 𝐸 𝑚 𝐼 ′
( 𝐸 0
) .
278-
282
The collected current can be expressed using the generalized Taylor expansion as
Equation 2.17, where 𝐼 ′
refers to differentiation with respect to 𝐸 .
62
𝐼 ( 𝐸 0
+ 𝐸 𝑚 sin( 𝜔𝑡 ) )= 𝐼 ( 𝐸 0
)+ 𝐸 𝑚 sin ( 𝜔𝑡 )𝐼 ′
( 𝐸 0
)+

𝐸 𝑚 2
2!
sin
2
( 𝜔𝑡 ) 𝐼 ′′
( 𝐸 0
)+
𝐸 𝑚 3
3!
sin
3
( 𝜔𝑡 ) 𝐼 ′′′
( 𝐸 0
)+ ⋯ (2.17)
Using the relationships between sin
2
( 𝜔𝑡 ) and cos ( 2𝜔𝑡 ) leads to Equation 2.18, which
reduces to Equation 2.19 when ignoring terms of higher order than 𝐸 𝑚 2
.
283-285
𝐼 ( 𝐸 )= 𝐼 ( 𝐸 0
)+ 𝐸 𝑚 sin ( 𝜔𝑡 )𝐼 ′
( 𝐸 0
)+
𝐸 𝑚 2
2!
1−cos( 2𝜔𝑡 )
2
𝐼 ′′
( 𝐸 0
)+ ⋯ (2.18)
𝐼 ( 𝐸 )= 𝐼 ( 𝐸 0
)+ 𝐸 𝑚 sin ( 𝜔𝑡 )𝐼 ′
( 𝐸 0
)−
1
4
𝐸 𝑚 2
𝐼 ′′
( 𝐸 0
) cos ( 2𝜔𝑡 ) (2.19)
It turns out that the second harmonic term, characterized by cos ( 2𝜔𝑡 ) , can be used by a
lock-in amplifier to generate data proportional to 𝑑𝑁 ( 𝐸 ) 𝑑𝐸 ⁄ and equal to
( 𝐸 𝑚 2
4 ⁄ ) 𝑑𝑁 ( 𝐸 ) 𝑑𝐸 ⁄ .
286
This allowed the use of retarding field analyzers for Auger analysis,
and is what allows the AugerProbe to operate at the high pressures used for PLD. The
principle is illustrated in Figure 2.30. RFAs are effective in high pressure because the
generation of signal, though dampened by the pressure, is still significant enough to detect
spectra against the noise. The higher efficiency electron energy analyzers, on the other
hand, lose viability with substantial noise, making their high pressure application much
more challenging.

Figure 2.30 An illustration of the relationship between direct RFA signal and its derivative.
Image: (Winklehner, 2013).
63
2.4 M icroscopy

Microscopy is uniquely beneficial to science because it communicates through the universal
medium of images. Visualization is often crucial to understanding, and why so much
import has been given to the successful implementation of new techniques. Used here,
these microscopy techniques have helped illuminate the hidden workings of sample
preparation and thin film growth.

2.4.1 Atomic Force M icroscopy
Developed in the mid-1980s at IBM, the atomic force microscope (AFM) is a simpler
version of the scanning tunneling microscope and was invented by the same people (Gerd
Binnig and Heinrich Rohrer).
287-290
The principle of operation in tapping mode, shown in
Figure 2.31 (right), is to oscillate a piezoelectric cantilever with a sharp tip at its
resonant frequency (typically ~10
5
Hz), then raster that tip across a surface. As it
encounters variations in topography the tip will be deflected according to Hooke’s law,
and the magnitude of that deflection can be observed in the frequency of oscillation.
Measuring the deviation from resonant frequency generates 3D topographical maps of the
surface with sub-nm resolution, shown in Figure 2.31 (left).
There are many different imaging modes that can be used with AFM, depending on the
type of force being measured between the sample and the tip. For the complex oxide thin
films developed in this work, however, the AFM was used exclusively in tapping mode to
determine surface morphology and roughness. It is an especially useful technique for
Figure 2.31 The principle of atomic force microscopy, illustrated as a cantilever being monitored with
a photodiode observing a reflected laser (right), and (left) a 3D plot of an atomic force micrograph of a
SrTiO3 surface showing atomic steps. Visible peaks are undesirable dirt. Image: the author.
64
checking the quality of substrates after etching and annealing because it requires no
additional preparation, can be executed in atmosphere, and has sufficient resolution to
observe 4 Å atomic steps without significant difficulty, shown in Figure 2.32.

Figure 2.32 Plotted height (right) for the line shown in the micrograph (left). The observed steps are 1 unit cell
high (4 Å), showing the successful etch and termination control of SrTiO3. Image: the author.

2.4.2 Scanning Electron M icroscopy
A scanning electron microscope (SEM) uses a focused beam of high energy electrons
(typically ~5 keV to ~30 keV) to produce images by scanning a surface. An illustration
of the principle of operation is provided in Figure 2.33. The interaction of the electron
beam with the material being imaged generates a variety of responses, each providing
different information about the sample.
291
For instance, the most common imaging mode
detects secondary electrons, which provide topographic information about the sample,
while detecting backscattered electrons will instead provide greater elemental contrast and
can also provide different depth of field depending on the detector location. Another
commonly collected signal is X-ray spectra for EDS, which provides compositional
information about the surface, discussed in §2.6.1.
Because the SEM operates by detecting the signal generated by the interaction of the
beam with the sample, the imaging resolution is not limited by typical diffraction limits
like other techniques such as transmission electron microscopy. Instead, the ultimate
resolution is dependent on the electron optics dictating the size of the beam rastering the
65
surface and the electron interaction volume.
This typically results in an ultimate resolution
on the order of ~10
-8
m.
The primary imaging provided by SEM is
topographical, which makes the applications
for thin films on the order of 10 nm thick very
limited, as the roughness of the film should be
well below the imaging limit of the instrument.
Additionally, sample charging of complex
oxides can be problematic for imaging.
However, the SEM has been useful for
characterizing PLD targets for morphology and
composition, shown in Figure 2.34.

Figure 2.33 An illustration of the operating
principle for scanning electron microscopes. Image:
(Marturi, 2013).
Figure 2.34 A scanning electron micrograph of an ablated polycrystalline CaZrO3 target surface showing the ablated
region (dark, smooth) in contrast with the unablated region (light, rough). Image: the author.
66
2.4.3 Transmission Electron M icroscopy
A transmission electron microscope (TEM) uses high energy electrons, typically ~200 keV,
to image samples by passing the beam through the sample. As a result, the samples must
be sufficiently thin to interact with the electron beam without hindering it. Thicker
samples are possible to image, but the resolution becomes limited as a result of the
increasing number of interactions between the electron beam and the sample. Ideally,
there will be a limited number of elastic collisions within the sample per electron, making
the optimal sample thickness around 100 nm, depending on the accelerating voltage and
instrument parameters, etc.
292
Ultimate resolution achievable with TEM is below 50 pm
with aberration-correction, allowing the imaging of individual atoms and the bonds
between them.
293-295

After the electron beam is generated it is focused, aligned, and passed through the sample.
The transmitted electrons are collected, focused, and projected onto a phosphor or CCD
detector for viewing. Depending on the desired application, the electromagnetic lenses will
adopt different configurations to illuminate the sample in the desired manner and collect
the transmitted electrons as required, as shown in Figure 2.35.
Figure 2.35 Illustration of the lens configurations for imaging (left) versus diffraction
(right) modes in a standard TEM. Image distributed under CC:BY – Eric Kvaalen.
67
There is incredible diversity in the possible imaging modes available with a TEM. This is
in part due to the fact that their operation utilizes electron diffraction, which is an
additional layer of information that can be harnessed for analysis. For instance, within an
image it is possible to generate a diffraction pattern from a specific spot by limiting the
probe to that location in a technique known as selected area electron diffraction.
Additionally, much like SEM, it is possible to generate an image by rastering a very small
probe across a large area in a technique called scanning transmission electron microscopy
(STEM). This allows the generation of spectral data from very limited regions, which can
be used for EDS and electron energy loss spectroscopy, discussed in §2.6, or other
spectroscopic techniques. The result is compositional mapping similar to that of SEM, but
with higher resolution and different information as a result of the difference in energies
and means of operation. An example TEM image taken with a Fischer Scientific Talos
CryoTEM is shown in Figure 2.36.
For this work, collaborators conducted STEM imaging using the Nion UltraSTEM
TM
200
microscope (operating at 200 kV) at Oak Ridge National Laboratory. The microscope is
equipped with a cold-field emission electron gun and a fifth-order aberration corrector.

Figure 2.36 A sample image of self-assembled polymer nanoparticles taken with
a Fischer Scientific Talos CryoTEM. Image: the author.
68
2.5 Structural Characterization

X-ray diffraction (XRD) crystallography earned a well-deserved Nobel prize in 1915 for
the Bragg father and son team. Their primary contribution was Bragg’s law, Equation
2.20 and illustrated in Figure 2.37, which describes the constructive interference
observed from crystallographic lattice planes, and is the foundation of the X-ray
diffraction techniques used to determine crystal structure. This incredibly powerful tool
revolutionized modern science with the ability to reveal fundamental laws of physics and
chemistry in the organization of atoms into crystal structures. Since inception, the
applications of X-ray diffraction have expanded to encompass all manner of materials,
with a few notable examples including DNA, viruses, and the application of inelastic
scattering techniques.
296-298

𝑛𝜆 = 2𝑑 sin ( 𝜃 ) (2.20)
The basic premise of XRD is as follows: X-rays of known wavelength are generated and
directed towards a sample, and a detector monitors the collected X-ray intensity as a
function of the angle between the source and detector, shown in Figure 2.38 (left). X-
rays will elastically scatter from the atoms within the investigated material and, if the
Bragg condition is satisfied, they will form an interference pattern as a result of their
interaction with the periodic atomic lattice, which is the diffraction pattern. As with other
diffraction techniques, such as RHEED discussed above, the detector will observe the
constructive interference peak if conditions are met that can be understood with an Ewald
sphere diagram, shown in Figure 2.38 (right). The source and detector are located on the
Figure 2.37 Illustration of the principle of Bragg’s law, in which constructive interference occurs when
the difference in path length is equal to an integer value of the wavelength. Image distributed under public
domain.
69
perimeter of a circle with respect to one another and the sample, which is in the center.
When the change of wave vector between the incident and scattered X-ray is equal to a
reciprocal lattice vector, the constructive diffraction peak can be observed by the detector.

2.5.1 Thin Film X-Ray Diffraction
Using careful alignment techniques with a suitably high resolution system, it is possible
to use X-ray diffraction to gather a great deal of information about the structure of thin
films. Thin film XRD begins with the alignment of the substrate to the equipment using
an iterative process to maximize the signal. The X-ray source should be monochromatized
and parallel to improve resolution. The initial signal observed is the rocking curve, a
measure of the distribution of crystal plane orientations, or mosaicity, by scanning 𝜔
Figure 2.38 Schematic for XRD instrument configuration (left), showing 𝜃 and 2𝜃 . The Ewald sphere construction
illustrating the diffraction condition, in which the change in wave vector between the incident and scattered X-ray equals
a reciprocal lattice vector. Image: the author.
Figure 2.39 Illustration of the geometry used to generate a rocking curve (left) with an example
rocking curve taken from off-stoichiometry homoepitaxial SrTiO3 (right), which caused the visible
shoulder. Image: the author.
70
around a fixed 2𝜃 value, shown with an example in Figure 2.39. If the single crystal
were completely perfect and stationary – impossible even at 0 K due to zero-point energy
– the rocking curve would be infinitely thin indicating only one precise Bragg geometry
represented by the crystal. However, the combination of lattice vibrations and
imperfections in the crystal broaden the distribution of spacing, which is measured by the
rocking curve: checking for satisfaction of the Bragg condition near the ideal scenario. The
rocking curve is typically measured by full width at half-maximum, with larger values
indicating a less perfect crystal.
Once the substrate is aligned to the equipment, the film itself can be measured. This is
typically done by a standard 𝜃 − 2𝜃 scan over the region where film peaks are expected.
Because the single crystal and diffractometer limit the angles available for diffraction, not
all peaks can be observed. If the film is epitaxial then it should share a peak in the same
family as the substrate surface orientation. For example, XRD of BaZrO
3
grown on
DyScO
3
is shown in Figure 2.28, with the BaZrO
3
film (001) direction epitaxial to the
(101) DyScO
3
surface direction, which is why both peaks are visible. However, the out-of-
plane lattice spacing is not the same for both materials, and why the peaks do not overlap.
The oscillations result from constructive and deconstructive interference between
reflections from the film and substrate surfaces. They are called Pendellösung fringes and
their spacing is proportional to the film thickness.
Building on these concepts, high resolution thin film XRD can be used to gather a great
deal of information about a sample. High instrumental precision and source coherency
make it possible to probe the reciprocal space of the sample along specific directions,
generating additional information such as 2D reciprocal space maps.
Figure 2.40 X-ray diffraction of a BaZrO3 thin film 97 nm thick grown on a (101) DyScO3 substrate. The quality
of the film is illustrated by the Pendellösung fringes indicating a smooth surface and interface. The BaZrO3 rocking
curve, shown inset, has a narrow full width at half maximum, supporting the claim of film quality. Image: the author.
71
2.6 Compositional Characterization

Compositional characterization is the foundational motivation for this work, and
specifically the lack of accessible in situ techniques. However, there are of course numerous
techniques that work splendidly with sufficient preparation. There are also many
techniques that work moderately well with moderate preparation.

2.6.1 Energy Dispersive X-Ray Spectroscopy
One of the latter methods is energy dispersive X-ray spectroscopy, commonly found in
electron microscopes. The principle of this technique is that sufficiently high energy
electrons (though any particle should do) remove electrons from their orbitals in the
material being investigated, and higher energy electrons drop to take their place, releasing
the difference in energy as a photon in the process.
299
The energy of this photon is
measured, and its characteristic energy forms part of the atomic structure-fingerprint
identifying the element it originated from. The spectra generated, shown in Figure 2.41,
contain peaks associated with these transitions, which are used to identify the elements
present. In scanning electron microscopy systems, the spectra can be correlated to the
image and form elemental maps, shown in
Figure 2.42.
EDS requires standards for quantification,
300
but
is often touted as quantitative with the use of
internal, ‘factory,’ standards. The resulting
quantification is excellent if the sample is
consistent with the standards, however this is
often only the case with geological and metal
samples. Furthermore, charging in oxide samples
will alter the electron-volume interaction,
thereby changing the resulting spectra.
Regardless, most spectroscopic techniques must
be aware of ZAF effects, and EDS can be
particularly susceptible to them without the
appropriate standards. In theory, the spectra emitted from an excited sample should be
proportional to the quantity of elements present. However, ZAF effects are sample-
dependent discrepancies in this expectation caused by the atomic number effect (‘Z’),
absorption effect (‘A’), and fluorescence excitation effect (‘F’). Due to differences in atomic
size and interaction cross-section, the probability of a particular element’s interaction with
an incident electron is dependent on the atomic number, which is the atomic number
Figure 2.41 An example energy dispersive X-
ray spectrum from BaZrO3. Image: the author.
72
effect. This means that the intensity of X-rays from high Z elements will be larger than
low Z elements. Once an X-ray is emitted, if another element present has an available
transition state with the same energy as the X-Ray, it will decrease the amount of that
X-ray observed, due to the absorption effect. Similarly, if the X-ray emitted has a higher
energy than the transition of another element, the absorption of that X-ray can cause
reemission of the lower energy photon from the second element, further skewing the results
from the ideal. Because these are all element-element interactions, complex materials will
exhibit more overlap between them, making quantification more challenging. Regardless,
despite the unfortunate challenges associated with quantification, EDS does an excellent
job of identifying the presence of elements and their locations, even if it doesn’t know how
much of them there exactly are.

Figure 2.42 Energy dispersive X-ray spectroscopy map of an ablated CaZrO3 target, with the electron
micrograph shown in the bottom left. The colored images are elemental maps for the labeled elemental peaks,
and indicate that there was no preferential ablation of this target, as the ablated and unablated regions have the
same composition. Image: the author.
73
2.6.2 Electron Energy Loss Spectroscopy
In electron energy loss spectroscopy (EELS), conducted with a TEM as it was in this
work, the sample is exposed to the beam of electrons with known energy. As they pass
through the sample some will be scattered inelastically. Measuring the post-scattering
electron energy spectrum, it is possible to determine the causes of energy loss and use that
information to characterize the sample. Much like Auger electrons, discussed in §2.3.2,
possible inelastic interactions include phonon excitation, interband transitions, and
plasmon excitations. This technique is especially useful for detecting elemental
composition due to inner-shell ionization, but can determine a great deal more information
than the simple presence of an element, such as chemical composition, band structure and
properties, surface properties with plasmon peaks, and even ultra-low energy loss
vibrational spectroscopy.
301
The low-loss spectrum, below ~50 eV, contains zero-loss and plasmon peaks, which provide
information about the band structure and dielectric properties of the sample. Integration
of the zero-loss peak can be used to calculate sample thickness when compared to the
whole spectrum,
302
and low-energy peak shifts can provide pressure information about the
sample.
303
The high-loss region contains the ionization edges used for compositional
analysis, as they are characteristic to their elemental origins. An example is shown in
Figure 2.43.
In this work, collaborators used EELS in conjunction with STEM at Oak Ridge National
Laboratory. A Gatan Enfinium spectrometer, with a collection semi-angle of 33 mrad, an
energy dispersion of 1
eV per channel and pixel
dwell time of 0.4
seconds, was used to
acquire EELS datasets.
Principal component
analysis (PCA) was
performed for the EELS
spectrum data, to
improve the signal-to-
noise ratio. A power law
was used to model the
background signal prior
to the core-loss signal
for each element.

Figure 2.43 (a) Image showing an LaAlO3 capping layer, SrRuO3 layer and
SrTiO3 substrate. (b) Elemental maps for Ti L edge, O K edge and La M edge, for
the region highlighted as white box in (a). Each elemental map is normalized within
itself. (c) Extracted EEL spectra for Ti L edge, O K edge and La M edge, where the
color of each spectrum corresponds to the region of same color highlighted as boxes
in (a). Image: taken by collaborators.
74
2.7 Analysis

The methods employed for analyzing the data are as important to this body of work as
the data itself. The MATLAB scripts developed for compiling and analyzing the Auger
spectra, for instance, were used to process hundreds of thousands of data points. Density
functional theory was applied for calculating the surface energies of variable structures to
illuminate our experimental results. Monte Carlo simulations illustrated the principles
that we used to justify our decision making processes. Lastly, the parameter-free escape
depth model developed for analyzing expected Auger signal intensity served as a crucial
bridge between the observed data and our ability to understand it.

2.7.1 Density-Functional Theory
Density-functional theory calculations were conducted as part of a collaboration with the
group of Dr. Rohan Mishra at the University of Washington in St. Louis. A full
investigation of the methods used is, therefore, beyond the scope of this work. However,
their excellent and insightful analysis is presented in §5.5, and their reported methods
listed here, after a brief introduction.
Density-functional theory is a computational modeling method for quantum mechanical
systems, used to investigate the electronic structure of many-bodied systems. When
confronting these types of problems, a standard approach is to assume that the nuclei are
fixed with respect to the electrons because the difference in their oscillation frequencies is
about three orders of magnitude; this is the Born-Oppenheimer approximation. The task
then is to solve the many-electron time-independent Schrodinger equation, given by
Equation 2.21,
16
where, for an 𝑁 -electron system, ℋ
̂
is the Hamiltonian, 𝑇 ̂
is the kinetic
energy, 𝑉 ̂
is the potential energy from the external field due to the ‘stationary’ nuclei, 𝑈 ̂

is the electron-electron interaction energy, and 𝐸 is the total energy.
ℋ
̂
Ψ = [𝑇 ̂
+ 𝑉 ̂
+ 𝑈 ̂
]Ψ = [∑ ( −
ℏ
2
2𝑚 𝑖 ∇
𝑖 2
)+ ∑ 𝑉 ( 𝒓 𝑖 )
𝑁 𝑖 =1
+ ∑ 𝑈 ( 𝒓 𝑖 , 𝒓 𝑗 )
𝑁 𝑖 <𝑗 𝑁 𝑖 =1
] Ψ = 𝐸 Ψ (2.21)
What makes this a particularly challenging undertaking is the interaction term 𝑈 ̂
, for
which there are many very sophisticated approaches that require significant (and
sometimes vast) computational resources to calculate. In contrast, DFT elegantly works
around 𝑈 ̂
by mapping the many-bodied problem onto a single-bodied problem which does
not contain 𝑈 ̂
. This is accomplished by using a normalized wavefunction for the electron
density, 𝑛 ( 𝒓 ) , which can be reversed, making Ψ a functional of 𝑛 0
.
304
Expanding on this,
the ground-state energy, external potential contribution, kinetic energy, and electron-
75
electron interaction energy become functionals of the electron density, and minimizing the
energy functional will yield the ground state density 𝑛 0
and from there the other ground-
state observables.
Naturally DFT has become much more sophisticated since its inception, and modifications
are constantly being made in attempts to improve the functional and the approximations
for exchange and correlation interactions.
305,306
We used the Vienna Ab initio Simulation Package to carry out the DFT calculations.
307

The energy cutoff for the plane waves was set at 500 eV. The threshold for energy
convergence of the self-consistent loops was set to 10
-6
eV. During structure optimization,
the convergence criteria for forces on ions was set to 0.01 eV Å
-1
. We used projector
augmented-wave potentials and the generalized gradient approximation within the Perdew
– Burke - Ernzerhof parameterization to describe the electron-ion and the electronic
exchange-correlation interactions, respectively.
308,309

2.7.2 M odeling and Simulations
Modeling and simulations were used extensively in this work to, primarily, draw parallels
between the observed data and phenomena we believed to be taking place. This was
accomplished with a parameter-free escape depth model based on the principle of variable
inelastic mean free paths for Auger electrons escaping a solid, and Monte Carlo simulations
of plume dispersion in various pressure conditions for different species.
Parameter-Free Escape Depth Model
When considering the Auger signal expected from SrTiO
3
, without references the only
relevant information will be the relative signal intensities of the three elements present.
Assuming uniform volumetric illumination of the sample from the excitation source, each
element in an amorphous or polycrystalline sample would have a uniform volumetric
distribution. This would mean that the resulting ratio of signal intensity could be
attributed solely to elemental concentration, probability of the Auger transition, inelastic
mean free path (IMFP) of the Auger electrons, and element-specific absorbance and
fluorescence effects (ZAF). If we assume that the probability of Auger transition scales
uniformly, and element-specific absorbance and fluorescence effects are manifest in the
IMFP, which can be approximated with Equation 2.16,
274
the only remaining variable is
elemental concentration. This means that signal intensity as a function of IMFP should
be proportional to the amount of material present.
To construct the model, we will be viewing the net signal as the sum of contributions
from individual layers. The Auger electrons originating from the surface layer would have
76
a (reasonably) clear path to the Auger probe, but some of those originating in the layer
beneath it would interact with the top layer and never leave the material. As you look at
Auger electrons originating from deeper in the material the probability of escape continues
to decrease as a function of depth and inelastic mean free path of the species, with the
electron’s probable decay in intensity given by Equation 2.13, rewritten here as Equation
2.22, where 𝐼 0
is the original intensity, 𝑥 is the distance the electron has traveled from its
point of origin, and 𝜆 is its inelastic mean free path.
𝐼 = 𝐼 0
𝑒 ( −
𝑥 𝜆 )
(2.22)
Treating the surface as 𝑥 = 0, for a homogenous single-element material the composition
of each layer will be the same, and the bulk signal (𝐼 𝐵𝑢𝑙𝑘
) can be written as a function of
the signals from the individual monolayers (𝐼 𝑀𝑜𝑛𝑜𝑙𝑎𝑦𝑒𝑟 ) and their depths as Equation 2.23,
where 𝑎 is the atomic lattice spacing between each monolayer.
𝐼 𝐵𝑢𝑙𝑘
= 𝐼 𝑀𝑜𝑛𝑜𝑙𝑎𝑦𝑒𝑟 ∑ 𝑒 ( −
𝑛 𝑎 𝑏𝑢𝑙𝑘 𝜆 𝑏𝑢𝑙𝑘 )
∞
𝑛 =0
=
𝐼 𝑀𝑜𝑛𝑜𝑙𝑎𝑦𝑒𝑟 1−𝑒 ( −
𝑎 𝑏𝑢𝑙𝑘 𝜆 𝑏𝑢𝑙𝑘 )
(2.23)
If the material is not homogeneous, as would be the case with a nonstoichiometric film or
a heterostructure, the signal will behave somewhat differently because it will change as a
function of depth. If the model is applied to a heterostructure with alternating layers
several unit cells thick of two different elemental materials, the signal from the different
elements will be dependent on their stacking sequence. This is illustrated in Figure 2.44,
which shows how this case would be calculated for two different materials, and the
resulting divergence in their signal intensities due to stacking sequence even if they had
identical lattice parameters and mean free paths.
Figure 2.44 Illustration of the impact stacking sequence has on escape-depth signal, for two arbitrary
materials with equal arbitrary thickness, such that 𝜆 = 1, where Depth = 1 = 𝜆 . The deviation in signal is
immediately visible, even for only 2 layers of each material generating signal. Image: the author.
77
If the model is reexamined as an application to (001) SrTiO
3
, it might be expected that
the signal should be constant because it is a bulk material. However, because (001) SrTiO
3
can be viewed as alternating layers of SrO and TiO
2
, with a lattice constant of 3.905 Å,
the distance between layers (1.95 Å) is almost half the IMFP of Ti
MVV
(4.2 Å). Therefore,
if the surface is not a uniform distribution of both layer-types, there will be a signal
discrepancy as a result. Would that discrepancy be observable?
If we assume the IMFP for a given elemental Auger electron is constant within the
material, the net intensity for a given element would only be dependent on the position
of its layers in the termination hierarchy, the same as illustrated in Figure 2.44. Defining
the termination layer as the surface with a depth of zero, the net signal intensity of the
element with the terminating layer would be given by Equation 2.24, where 𝐼 𝑇𝑒𝑟𝑚𝑀𝐿 is the
unattenuated signal from a single monolayer with the specified elemental termination.
Likewise, the net signal intensity of the element found in the sublayer of the structure
would be Equation 2.25, where 𝐼 𝑆𝑢𝑏𝑀𝐿 is, again, the unattenuated signal from a single
monolayer with the specified elemental sublayer.
𝐼 𝑇𝑒𝑟𝑚𝐵𝑢𝑙𝑘 = 𝐼 𝑇𝑒𝑟𝑚𝑀𝐿 ∑ 𝑒 ( −
𝑛 𝑎 𝑏𝑢𝑙 𝑘 𝜆 𝑏𝑢𝑙𝑘 )
∞
𝑛 =0
=
𝐼 𝑇𝑒𝑟𝑚𝑀𝐿 1−𝑒 ( −
𝑎 𝑏𝑢𝑙𝑘 𝜆 𝑏𝑢𝑙𝑘 )
(2.24)
𝐼 𝑆 𝑢 𝑏𝐵𝑢𝑙𝑘 = 𝐼 𝑆𝑢𝑏𝑀𝐿 ∑ 𝑒 (−
( 𝑛 𝑎 𝑏𝑢𝑙𝑘 +
𝑎 𝑏𝑢𝑙𝑘 2
)
𝜆 𝑏𝑢𝑙𝑘 )
∞
𝑛 =0
=
𝐼 𝑆𝑢𝑏𝑀𝐿 𝑒 ( −
𝑎 𝑏𝑢𝑙𝑘 2𝜆 𝑏𝑢𝑙𝑘 )
1−𝑒 ( −
𝑎 𝑏𝑢𝑙𝑘 𝜆 𝑏𝑢𝑙𝑘 )
(2.25)
Using this model to examine how the net signal of an element would change during the
deposition of thin films simply requires selecting which layers contribute to the sum or
examining how a bulk signal decays beneath a different material. Applied to the case
where the signal is coming from a film being deposited would give Equation 2.26 for a
homogenous bulk material, where 𝑁 is the total number of atomic layers. As 𝑁 goes to
infinity 𝐼 𝐹𝑖𝑙𝑚 goes to 𝐼 𝐵𝑢𝑙𝑘
, as expected, if the lattice constant and IMFP for the film is
equivalent to those of the bulk material. For the same scenario with an intercalated
material, such as SrTiO
3
, the terminating signal would be calculated in the same way as
the homogenous film scenario above, while the sublayer signal intensity would be given
by Equation 2.27, which also means that if the lattice constant and IMFP for the film is
equivalent to those of the bulk material then 𝐼 𝑆𝑢𝑏 and 𝐼 𝑇𝑒𝑟𝑚 go to 𝐼 𝑆𝑢𝑏𝐵𝑢𝑙𝑘 and 𝐼 𝑇𝑒𝑟𝑚𝐵𝑢𝑙𝑘 ,
respectively, as 𝑁 goes to infinity.
𝐼 𝐹𝑖𝑙𝑚 = 𝐼 𝑀𝑜𝑛𝑜𝑙𝑎𝑦𝑒𝑟 ∑ 𝑒 (−
( 𝑛 −1) 𝑎 𝑓𝑖𝑙𝑚 𝜆 𝑓𝑖𝑙𝑚 )
𝑁 𝑛 =1
(2.26)
78
𝐼 𝑆𝑢𝑏𝐹𝑖𝑙𝑚 = 𝐼 𝑆𝑢𝑏𝑀𝐿 ∑ 𝑒 (−
( 𝑛 𝑎 𝑓𝑖𝑙𝑚 −
𝑎 𝑓𝑖𝑙𝑚 2
)
𝜆 𝑓𝑖𝑙𝑚 )
𝑁 𝑛 =1
(2.27)
For the reverse case, in which one material is being deposited on another, thereby
attenuating the signal, the signal intensity for the bulk case beneath 𝑁 layers is given by
Equation 2.28, while the signal for the terminating layer and sublayer in an intercalated
material would be given by the same expression for the terminating layer as for the bulk,
and Equation 2.29 for the sublayer.
𝐼 𝐵𝑢𝑙𝑘𝐵𝑢𝑟𝑖𝑒𝑑 = 𝐼 𝐵𝑢𝑙𝑘
𝑒 (−
𝑁 𝑎 𝑓𝑖𝑙𝑚 𝜆 𝑓𝑖𝑙𝑚 )
=
𝐼 𝑀𝑜𝑛𝑜𝑙𝑎𝑦𝑒𝑟 1−𝑒 ( −
𝑎 𝑏𝑢𝑙𝑘 𝜆 𝑏𝑢𝑙𝑘 )
𝑒 (−
𝑁 𝑎 𝑓𝑖𝑙𝑚 𝜆 𝑓𝑖𝑙𝑚 )
(2.28)
𝐼 𝑆𝑢𝑏𝐵𝑢𝑟𝑖𝑒𝑑 = 𝐼 𝑆𝑢𝑏𝐵𝑢𝑙𝑘 𝑒 (−
𝑁 𝑎 𝑓𝑖𝑙𝑚 +
𝑎 𝑏𝑢𝑙𝑘 2
𝜆 𝑓𝑖𝑙𝑚 )
=
𝐼 𝑆 𝑢𝑏𝑀𝐿 𝑒 ( −
𝑎 𝑏𝑢𝑙𝑘 2𝜆 𝑏𝑢𝑙𝑘 )
1−𝑒 ( −
𝑎 𝑏𝑢𝑙𝑘 𝜆 𝑏𝑢𝑙𝑘 )
𝑒 (−
𝑁 𝑎 𝑓𝑖𝑙𝑚 +
𝑎 𝑏𝑢𝑙𝑘 2
𝜆 𝑓𝑖𝑙𝑚 )
(2.29)
Using this modeling method to examine the relative signal intensity shifts during thin film
growth as a result of termination switching events simply requires selecting appropriate
lattice constants for the species being grown and then counting the layers present.
Therefore, when switching from TiO
2
-terminated SrTiO
3
to SrO-termination using
SrRuO
3
(SRO), the net signals would shift from those shown above to Equation 2.30 and
2.31 after depositing a single atomic layer of SrRuO
3
.
𝐼 𝑇𝑖
= ∑ 𝑒 (−
( 𝑎 𝑆𝑅𝑂 +
𝑎 𝑆𝑇𝑂 2
+𝑛 𝑎 𝑆𝑇𝑂 )
𝜆 𝑇𝑖
)
∞
𝑛 =0
=
𝑒 (−
( 𝑎 𝑆𝑅𝑂 +
𝑎 𝑆𝑇𝑂 2
)
𝜆 𝑇𝑖
)
1−𝑒 ( −
𝑎 𝑆𝑇𝑂 𝜆 𝑇𝑖
)
(2.30)
𝐼 𝑆𝑟
= 1 + ∑ 𝑒 ( −
( 𝑎 𝑆𝑅𝑂 +𝑛 𝑎 𝑆𝑇𝑂 )
𝜆 𝑆𝑟
)
∞
𝑛 =0
= 1 +
𝑒 ( −
𝑎 𝑆𝑅𝑂 𝜆 𝑆𝑟
)
1−𝑒 ( −
𝑎 𝑆𝑇𝑂 𝜆 𝑆𝑟
)
(2.31)
A second atomic layer of SrRuO
3
would shift the signals to Equation 2.32 and 2.33.
𝐼 𝑇𝑖
= ∑ 𝑒 (−
( 2𝑎 𝑆𝑅𝑂 +
𝑎 𝑆𝑇𝑂 2
+𝑛 𝑎 𝑆𝑇𝑂 )
𝜆 𝑇𝑖
)
∞
𝑛 =0
=
𝑒 (−
( 2𝑎 𝑆𝑅𝑂 +
𝑎 𝑆𝑇𝑂 2
)
𝜆 𝑇𝑖
)
1−𝑒 ( −
𝑎 𝑆𝑇𝑂 𝜆 𝑇𝑖
)
(2.32)
79
𝐼 𝑆𝑟
= 1 + 𝑒 ( −
𝑎 𝑆𝑅𝑂 𝜆 𝑆𝑟
)
+ ∑ 𝑒 ( −
( 2𝑎 𝑆𝑅𝑂 +𝑛 𝑎 𝑆𝑇𝑂 )
𝜆 𝑆𝑟
)
∞
𝑛 =0
= 1 + 𝑒 ( −
𝑎 𝑆𝑅𝑂 𝜆 𝑆𝑟
)
+
𝑒 ( −
2𝑎 𝑆𝑅𝑂 𝜆 𝑆𝑟
)
1−𝑒 ( −
𝑎 𝑆𝑇𝑂 𝜆 𝑆𝑟
)
(2.33)
If the two layers of SrRuO
3
are then capped with two more atomic layers of SrO-
terminated SrTiO
3
, you would get net signals given by Equation 2.34 and 2.35 for the
first layer.
𝐼 𝑇𝑖
= 𝑒 (−
𝑎 𝑆𝑇𝑂 2
𝜆 𝑇𝑖
)
+ ∑ 𝑒 (−
( 2𝑎 𝑆𝑅𝑂 +
𝑎 𝑆𝑇𝑂 2
+𝑛 𝑎 𝑆𝑇𝑂 )
𝜆 𝑇𝑖
)
∞
𝑛 =0
= 𝑒 (−
𝑎 𝑆𝑇𝑂 2
𝜆 𝑇𝑖
)
+
𝑒 (−
( 2𝑎 𝑆𝑅𝑂 +
𝑎 𝑆𝑇𝑂 2
)
𝜆 𝑇𝑖
)
1−𝑒 ( −
𝑎 𝑆𝑇𝑂 𝜆 𝑇𝑖
)
(2.34)
𝐼 𝑆𝑟
= 1 + 𝑒 ( −
𝑎 𝑆𝑇𝑂 𝜆 𝑆𝑟
)
+ 𝑒 ( −
( 𝑎 𝑆𝑇𝑂 +𝑎 𝑆𝑅𝑂 )
𝜆 𝑆𝑟
)
+ ∑ 𝑒 ( −
( 𝑎 𝑆𝑇𝑂 +2𝑎 𝑆𝑅𝑂 +𝑛 𝑎 𝑆𝑇𝑂 )
𝜆 𝑆𝑟
)
∞
𝑛 =0

= 1 + 𝑒 ( −
𝑎 𝑆𝑇𝑂 𝜆 𝑆𝑟
)
+ 𝑒 ( −
( 𝑎 𝑆𝑇𝑂 +𝑎 𝑆𝑅𝑂 )
𝜆 𝑆𝑟
)
+
𝑒 ( −
( 𝑎 𝑆𝑇𝑂 +2𝑎 𝑆𝑅𝑂 )
𝜆 𝑆𝑟
)
1−𝑒 ( −
𝑎 𝑆𝑇𝑂 𝜆 𝑆𝑟
)
(2.35)
For the second capping layer of SrTiO
3
, the signals would be given by Equation 2.36 and
2.37.
𝐼 𝑇𝑖
= 𝑒 (−
𝑎 𝑆𝑇𝑂 2
𝜆 𝑇𝑖
)
+ 𝑒 (−
3𝑎 𝑆𝑇𝑂 2
𝜆 𝑇𝑖
)
+ ∑ 𝑒 (−
( 2𝑎 𝑆𝑅𝑂 +
3𝑎 𝑆𝑇𝑂 2
+𝑛 𝑎 𝑆 𝑇 𝑂 )
𝜆 𝑇𝑖
)
∞
𝑛 =0

= 𝑒 (−
𝑎 𝑆𝑇𝑂 2
𝜆 𝑇𝑖
)
+ 𝑒 (−
3𝑎 𝑆𝑇𝑂 2
𝜆 𝑇𝑖
)
+
𝑒 (−
( 2𝑎 𝑆𝑅𝑂 +
3𝑎 𝑆𝑇𝑂 2
)
𝜆 𝑇𝑖
)
1−𝑒 ( −
𝑎 𝑆𝑇𝑂 𝜆 𝑇𝑖
)
(2.36)
𝐼 𝑆𝑟
= 1 + 𝑒 ( −
𝑎 𝑆𝑇𝑂 𝜆 𝑆𝑟
)
+ 𝑒 ( −
2𝑎 𝑆𝑇𝑂 𝜆 𝑆𝑟
)
+ 𝑒 ( −
( 2𝑎 𝑆𝑇𝑂 +𝑎 𝑆𝑅𝑂 )
𝜆 𝑆𝑟
)
+ ∑ 𝑒 ( −
( 2𝑎 𝑆𝑇𝑂 +2𝑎 𝑆𝑅𝑂 +𝑛 𝑎 𝑆𝑇𝑂 )
𝜆 𝑆𝑟
)
∞
𝑛 =0

= 1 + 𝑒 ( −
𝑎 𝑆𝑇𝑂 𝜆 𝑆𝑟
)
+ 𝑒 ( −
2𝑎 𝑆𝑇𝑂 𝜆 𝑆𝑟
)
+ 𝑒 ( −
( 2𝑎 𝑆𝑇𝑂 +𝑎 𝑆𝑅𝑂 )
𝜆 𝑆𝑟
)
+
𝑒 ( −
( 2𝑎 𝑆𝑇𝑂 +2𝑎 𝑆𝑅𝑂 )
𝜆 𝑆𝑟
)
1−𝑒 ( −
𝑎 𝑆𝑇𝑂 𝜆 𝑆𝑟
)
(2.37)
These models, their applications, and their implications for film growth and
characterization, will be thoroughly discussed in the text.

80
Monte Carlo Simulations
To examine (and illustrate) the potential for preferential scattering to alter film
composition, I performed very simple Monte Carlo simulations of scattering as a function
of pressure and atomic mass. Because of the discrepancy in atomic weight between
different species, they will behave differently within the plume. Lighter weight elements
will be more readily diverted due to collisions, and higher pressure will increase the
probability of collision. Therefore, while higher pressure will increase collision frequency
and thus diffusion for both species, their compositional profile within the plume will be
different, and the stoichiometry of the plume-front which reaches the substrate will be
more heavy-element-rich with high pressure. The simulation performed depicts the flight
paths and final destinations of co-ablated Sr and Ti under two different pressure regimes,
and is shown in §4.2.
The Monte Carlo scattering simulations begin by assuming the Sr and Ti species have a
considerable amount of energy after being ablated,
94
on the order of a few hundred eV.
Converting that energy to velocity is dependent on the mass of the species, and yields
around 5x10
5
m/s with Ti a little higher (less mass) and Sr a little lower (more mass).
The mean free path in the background pressure is then estimated with the proportional
constant of 5 x 10
-3
cm in 1 Torr. Equipped with an initial velocity (which we assume to
be normal to the target surface) and a mean free path, we randomly select a path length
from a Gaussian distribution centered on the mean free path length, with a standard
deviation of 20% of the magnitude of the mean free path. At that point the atom will
collide with a member of the background gas, which has a random initial velocity vector
and a temperature-derived velocity, which is about 400 m/s for O
2
at room temperature.
After resolving the collision, the particle will have a new velocity vector and starting
location. This process repeats until a given threshold of collisions is reached or the particle
travels beyond the target or substrate planes, located at 0 and 75 mm, respectively. The
path is traced and final position on the x-axis recorded if the particle arrived at the
substrate plane. After a predetermined number of particles are simulated the paths are
plotted and a histogram of final positions generated. As should be seen, this simulation is
quite rudimentary, and ignores a great number of factors that would influence the results
in reality, such as plume pressure and interparticle interactions. These effects are discussed
in greater detail in §2.2.3. However, for the sake of illustrating the principle, this
simulation will suffice.

2.7.3 Spectral Processing
With the tremendous amount of data generated by the AugerProbe, signal processing and
analysis was extremely important. This was especially true for semi-quantitative analysis,
81
as the calculated values must be both self-consistent and repeatable. This development
took time, however, as the majority of techniques I discovered that were last developed
for Auger analysis of data collected with an RFA are quite out of date. In fact, one of the
primary benefits of the RFA at the time of its initial popularity was that the derivative
signal could be acquired electronically.
138
It was only after electron energy analyzers, such
as cylindrical mirror analyzers, developed sufficient performance that it was realized the
wealth of information being lost by ignoring the N(E) or I(E) signal. Reintroducing the
technology for high pressure application with the AugerProbe has allowed the
reexamination of RFA analysis techniques with digital processing capabilities.
The spectral data generated with the AugerProbe software is saved as ‘.lvm’ files
containing six columns of data and a footer, identified by the scan ID and the scan
number. The columns contain a timestamp, the energy in eV, the lock-in signal N(E), the
photomultiplier signal, and two columns of instrumental settings information. Depending
on the type of scans being conducted, the final data set may be a large number of
individual spectra or a small number of spectra composed of many summed scans.
Regardless, the first requirement for analysis was generating usable data, so a MATLAB
script was developed to compile the large number of spectra, often with different energy
ranges, into a single data file readable by MATLAB. This was done with a graphical user
interface (GUI) designed to allow the selection of the desired scans for processing, and
options for the processing parameters used. The output of the GUI is a single ‘.mat’ file
containing a data-tree organized by the scans collected, which also generates a sum of all
the spectra for a given scan.
For instance, let’s say a growth were conducted during which 10 scans of 3 energy ranges
were collected after every monolayer of deposition. The scans for each layer would be
directed to a unique folder, based on the operator’s preference. In this folder there would
be subfolders labeled with the names of the scans, and each subfolder would contain the
individual files for each spectrum. Inputting the file directory of the unique folder into the
GUI and telling it which scans to look for, and how many spectra to collect, would create
a ‘.mat’ data-tree. This structure contains individual structures for each scan, which
contain the individual spectra as column vectors, and a ‘sum’ column vector with the
lock-in and photomultiplier signal columns replaced by the sum of their equivalents from
the other spectra within the same scan. If, on the other hand, the growth required many
individual scans which were temporally sensitive and thus not required to sum, the output
of the GUI would be the same. This still allows the user to access the individual spectra,
labeled monotonically in the order they were collected.
Further analysis of the spectra was not conducted with the GUI because the individualized
parameters for any given growth and its scans were changing too frequently to justify the
time investment required to encode them into the GUI. Instead, scripts were developed
82
for signal processing that allowed the individual changes, unique to each experiment, to
be made. The principles used in spectral processing are provided here, and generalized
scripts provided in Appendix A.
Peak-to-Peak
Peak-to-peak (P2P) quantification is simply subtracting the minimum of the first
derivative of the Auger intensity from the maximum, illustrated in Figure 2.45. This
was the dominant mode of AES analysis for many years because the derivative signal
could be generated electronically using analog circuits. However, as mentioned above, this
became less popular due to information loss, but the peak-to-peak method of semi-
quantification of Auger signal remained in frequent use despite that fact.
310

Generating P2P values for a given spectrum is quite simple from a scripting perspective,
as it simply requires generating the differential by subtracting each intensity from the
next, and plotting the difference as a function of energy, then subtracting the minimum
value from the maximum. However, though this method certainly works it lacks the finesse
required for situations where the generic maximum and minimum of a given spectrum are
Figure 2.45 Demonstration of the peak-to-peak measurement for an OKLL peak plotted in N(E): the
dN/dE signal has a clear trough which can be subtracted from the clearly defined peak. Image: the
author.

83
not the correct values due to multiple peaks or spectral artifacts, such as for the spectrum
shown in Figure 2.46. These situations require a more deliberate approach, wherein the
maximum and minimum locations should be known ahead of time, so as to prevent
misidentification. However, an individual energy location of a given max or min is not
enough either. Peak shifting and drifting is not uncommon, and without accounting for it
the incorporation of artificial shifts in P2P intensity will be superimposed on the data set.
Therefore, the simplest solution is to select a local maximum from a given value plus or
minus some small delta. This effectively reduces the noise caused by weak signals, but
suffers from a common problem in this type of quantification: defining a floor.
What is a reasonable P2P value for a spectrum without a peak? Answering this question
also defines the data floor, or background noise, of the data set. There are many ways to
approach this problem, as it is certainly not new. However, in the context of Auger
analysis presented here, the concern is not necessarily about how to define it, but rather
how to prevent it from being incorrectly identified. This is a particular problem with peak
overlap, where one diminishing signal is accompanied by the increase of another, and there
is sufficient overlap in the spectrum that the increase is seen by the P2P calculation. For
instance, the Sr
MNN
peak located around 85 eV is next to the La
NVV
peak located around
75 eV. When LaAlO
3
is deposited on SrTiO
3
, the loss of Sr
MNN
signal can rebound after
the LaAlO
3
is thick enough for a significant La
NVV
signal to appear, because the peak
Figure 2.46 An example of a situation where a clear peak in the dN/dE signal
is not observed, because the energy-width of the SrMNN scan is too narrow, even
though the peak can clearly be seen. Image: the author.
84
selection algorithm is searching in too wide a
range. The comparison of the two peaks is shown
in Figure 2.47 as LaAlO
3
is deposited on SrTiO
3
.
It should be noted that this can also occur without
a wide peak selection range, but it certainly
exacerbates the problem. The only real solution is
selecting peaks which do not have overlap, but
that comes with the potential for other negative
effects such as loss of surface sensitivity.
Generalized P2P calculation scripts for the Auger
data used in this work are shown in Appendix A.
Area Under the Curve
Area under the curve (AUC) calculations are, as
can likely be guessed, the area beneath the curve,
illustrated in Figure 2.48. Because the area would
have to be calculated by hand for Auger spectra
which were not digital, AUC never usurped P2P
as a method of analysis. However, because it is a
more direct assessment of the same thing being
measured by P2P calculations, it stands to reason
that the curve width and height are proportional
to signal generation and, thus, the amount of the
characteristic element present. Furthermore,
attempts were made to develop
techniques that would arrive at
the intensity area, such as the
dynamic background
subtraction method,
138,311-315

because it was noted that the
area will provide consistent
data regardless of signal
modulation. The idea was to
obtain a linear background in
terms of the differentiated
spectrum, and then recover the
N(E) areas by integration of the
differentiated data. However,
the approach was troubled by
Figure 2.47 Incremental observation of the
deposition of LaAlO3 on SrTiO3, 1 unit cell at a
time, as indicated, up to 7 unit cells of LaAlO3
on SrTiO3. Note the shift in signal, both to lower
energy from the SrMNN peak at around 85 eV to
the LaNVV peak around 76 eV, and the loss of the
SrMNN peak around 60 eV. Image: the author.
Figure 2.48 Illustration of the peak area selected for an AUC
measurement, with a given start and end point. Image: the author.
85
inconsistencies in the results due to low-energy tails in the N(E) signal. This problem is
an excellent example of something easily addressed with modern data processing
capabilities.
Calculating the AUC from an Auger spectrum is superficially easy with scripting, but,
just as it was with P2P calculations, more complex in actuality. To calculate the AUC a
simple numerical integration of the selected energy range using the trapezoidal method
(Equation 2.38) will suffice. However, the same problem with peak location which arose
during P2P calculation is also present in AUC calculations, because the width of the
region selected for the area should not be arbitrary.
∫ 𝑓 ( 𝑥 ) 𝑑𝑥 𝑏 𝑎 ≈
𝑏 −𝑎 2𝑁 ∑ ( 𝑓 ( 𝑥 𝑛 )+ 𝑓 ( 𝑥 𝑛 +1
) )
𝑁 𝑛 (2.38)
To ensure that the AUC generation has limited noise and maximum self-consistency
requires subtracting the background and leveling the curve before calculating the area.
This limits instrumental error from alignment and peak drift, but requires some degree of
initial analysis to determine the best approach for eliminating the background. In most
energy regions the background for a narrow range will be approximately linear, allowing
a simple fit. However, the specific area selected to fit, and the manner the fit is chosen,
can both alter the final result. The most consistent manner for fitting the background for
subtraction has been an initial assessment of the spectra which surveys the background
and selects two energies to use as an approximation of the net background slope, shown
in Figure 2.48. The advantage of this approach is that consistency of the calculations is
not sacrificed, as any shift of the fit from one spectrum to the next will potentially be a
source of artificial signal overlay, caused by the selective adjustment, rather than an actual
signal.
While the proximity of other peaks can alter the net AUC value, just as it did with the
P2P, the problem is not as pronounced. This is because the primary impact will be to the
background fit, and it will not artificially create a peak. However, care should still be
taken when selecting the fit region, because potential conflict can be avoided simply by
making a deliberate choice for the start and end values.
Generalized AUC calculation scripts for the Auger data used in this work are shown in
Appendix A.

86

Chapter 3
In Situ Auger Electron
Spectroscopy

3.1 Introduction

In this chapter I report the incorporation of an in situ Auger electron spectroscopy probe
into a pulsed laser deposition chamber. The method of operation is discussed, and the
capabilities of the probe are explored including limits of operation. Collected spectra are
shown from a wide variety of elements and materials in various conditions, and the method
of spectral analysis is discussed. Real time operation with the development of a pulse-
probe method is demonstrated. Finally, I illustrate the use of the in situ AES probe during
the pulsed laser deposition of a CaTiO
3
/LaMnO
3
thin film superlattice on a NdGaO
3

substrate at various thickness intervals to identify the surface composition. Layer-by-layer
growth is demonstrated by reflection high energy electron diffraction, and in situ AES is
used to observe the compositional changes during the growth. These results are used to
verify the inelastic mean free path of the observed Auger electrons and thereby estimate
the depth-sensitivity of the probe.

87
3.2 The AugerProbe

The AugerProbe was designed by Philippe Staib to serve, specifically, as a surface
characterization method for in situ growth analysis;
133
in this, he was certainly successful.
However, incorporating the probe into a PLD chamber was not trivial. In order to acquire
sufficient signal for analysis, the system geometry and dimensions must be precise, and
the signal’s viability must be verified. While these requirements are certainly
surmountable, the natural sensitivity of Auger electrons to local electric and magnetic
fields makes them somewhat more complicated than they initially might appear. Preparing
the chamber and probe for its first spectral acquisition of room temperature polycrystalline
silver took longer than anticipated due to unforeseen complications arising from the
scintillator and magnetic steering. The next step, of collecting spectra from an SrTiO
3

substrate in deposition conditions, required similar unforeseen adjustments as the steering
had to be modified and charging sources within the chamber mitigated. Lastly, achieving
the surface-sensitivity we desired from the AugerProbe required intensive tuning and
alignment for every sample, which had not been
foreseen. Despite these setbacks, however, the final
result is a flexible and reliable probe with greater
surface sensitivity than originally anticipated.
Understanding the capabilities, limits, and specific
requirements of the system allows the AugerProbe
to generate and collect Auger spectra of consistent
quality.

3.2.1 Incorporation
While incorporating the AugerProbe into an
existing PLD chamber, a significant challenge was
ensuring that the target-substrate distance could be
maintained while still allowing ample room for
alignment. As shown in Figure 3.1, the probe sits
30° off the normal axis and is equipped with a 100
mm z-translator. The ideal target-probe distance is
55 mm and the target position is 75 mm normal to
the substrate; when in these positions, the probe
must not interfere with the propagation of the
ablated plume. Furthermore, the probe must not be
in contact with any targets, including those not
Figure 3.1 The Auger probe in 55 mm
position for spectral acquisition, 30° off the
normal axis. Image: the author.
88
currently being ablated, which share a rotating carousel with the primary target. Meeting
these requirements can be taxing, but is achievable. The primary limitations imposed by
these conditions arise from the target carousel, because target selection rotates the entire
apparatus and it can collide with the probe if they are both in operating position. This
means that the carousel or probe must be lowered to change or align targets. Additionally,
the relative positions of the carousel and probe limit the available space, and the target
slot adjacent to the active target should remain empty to achieve the desired target-
substrate distance. This limits the number of potential active targets per growth to three
without removing the carousel during growth or retracting/inserting the probe between
depositions.
The AugerProbe is backed by a turbomolecular pump limited by a gate-valve, which is
itself backed by the PLD system’s dry scroll pump. When active, the probe’s pump
influences the system pressure, and must be accounted for when changing the gas flow
rates and valve positions. The impact is relatively small, however, and no discernable
difference has been observed in film quality as a result of its presence.
The growth chamber laboratory is located on the top (7
th
) floor of a steel- and concrete-
construction building, which has extensive ventilation on the roof for both laboratories
and workspaces. As a result of this equipment, and the construction of the building, there
is a constant background frequency of approximately 0.4—0.6 Hz that is visible in
vibration of the RHEED specular spot as well as the Auger lock-in signal. This effect can
be minimized with alignment, but is difficult to remove entirely.
Electromagnetic fields are a potential problem due to the low energy (and thus easy
deflection) of the Auger electrons. This is observed in the Auger signal when the target is
rotating, but has not been observed to impact
the collected spectra. Other opportunities for
incipient fields to cause problems are abundant,
and must be countered if possible. For instance,
when perfecting the system parameters during
installation it was decided that stainless steel
mesh cloth should be installed on the interior
surfaces of the viewports to minimize the effects
of charging. In an oxide deposition system, the
stray ablated material is typically insulating,
and its buildup may result in charging on
surfaces which would otherwise be grounded,
and where it might not have been expected. This
is a common occurrence on the RHEED
phosphor screen, resulting in the gradual
Figure 3.2 Electron diffraction pattern of a poorly
aligned SrTiO3 substrate viewed on a phosphor
screen damaged after excessive deposition-buildup.
Image: the author.
89
reduction in intensity and, eventually, the destruction of the screen from excess thermal
stress, shown in Figure 3.2.
While many of the issues encountered while incorporating the probe are chamber-specific,
their generalities apply to all systems and illustrate some of the complications which must
be considered during implementation.

3.2.2 Use
Using the AugerProbe is a straightforward process, but requires an understanding of its
operation to properly align the system for maximum signal, and careful experimental
design to generate usable, repeatable, data. A succinct overview of the steps followed for
use are presented here, in Table 7, with more detailed explanation following.
Table 7 Outline of steps required for operation of the AugerProbe.
1. Prepare chamber and sample for acquisition.
2. Turn on electronics.
3. Start the software, initialize the probe, and set up the RHEED gun.
4. Align the sample to the RHEED gun and insert the probe with the
backing gate valve open.
5. Optimize photomultiplier (PM) signal at a fixed energy (O
KLL
≈ 518 eV
for oxides) with the electron gun controller (EGC).
6. Optimize PM signal at same fixed energy with heater azimuth and tilt.
7. Verify satisfactory lock-in (LI) signal at same fixed energy.
8. Optimize LI signal at 20 eV with steering.
9. Set up software for acquisition.
10. Collect spectra!
11. View and save the data.
12. Shut it down.

To begin using the AugerProbe the sample and chamber must be prepared such that the
electron gun can be operated without damage. The probe requires a lower pressure than
the gun, and as long as the background pressure is approximately ≤ 5 x 10
-2
mbar the
probe should be operable. The probe backing gate valve must be open, and the sample
should be at the temperature and pressure desired for observation.
After turning on the electronics required for operating the probe and electron gun, the
probe must be initialized via the software interface. The electron gun should be set to the
desired parameters for observation: typically, in an equivalent setup as used for RHEED.
90
Aligning the diffraction pattern is necessary for simultaneous AES and RHEED, and also
serves as a frame of reference for repeatability even if RHEED is not being performed.
The accelerating energy of the electron gun must be higher than the energy of the peaks
observed, but higher energy can create a larger background signal. However, the lower
the accelerating voltage, the greater the pressure-dependence of the specular spot
intensity. Therefore, a decision must be made to accommodate AES and RHEED signal
intensities, depending on the required data and conditions. Comparison of diffraction
patterns from a single substrate with 10 keV and 35 keV beam energies can be seen in
Figure 3.3.
After aligning the substrate to the electron beam, the probe must be aligned. The
deposition system should be in the same configuration during alignment as it will be
during spectral acquisition. Aligning the probe is an iterative process used to find a
multivariable maximum. The PM voltage should be set to a reasonable value based on
the energy of the characteristic signal and background pressure. Using the same PM
voltage every alignment (as well as time constant and modulation) is recommended to
verify consistency. Beginning with the electron beam position on the substrate, the
photomultiplier signal is maximized for a characteristic peak such as O
KLL
. Changing the
heater tilt can greatly increase the PM signal due to an increase in forward-scattered
electrons, but watching the lock-in signal verifies that the result is inconsequential. The
most important things to consider for heater tilt are the position of the specular spot for
RHEED and measurement repeatability.
The next alignment is of the probe itself, conducted with the triaxial tilt system, shown
in Figure 3.4. While watching the PM signal, the axes are adjusted one at a time to find
the local maximum. When complete, the electron beam alignment is repeated. The
observed energy is changed to 20 eV, and the LI signal maximized with the magnetic
steering controller, then the high energy LI signal is confirmed.
Figure 3.3 Comparison of the diffraction pattern formed from a SrTiO3 substrate with a 10 keV beam (left) and a
35 keV beam (right). Note the higher order diffraction spots on the lower energy diffraction pattern. Image: the author.
91
The software is arranged for spectral acquisition and the specific scans defined with
start/end energies, step size in eV, dwell per step in ms, sensitivity in mV, time constant
in ms, modulation, and PM voltage. Typical scans depend entirely on what is being
observed, and why. For instance, a standard scan to observe the Sr
MNN
peak at around
82 eV could have start/end energies of 72 eV to 92 eV, a step size of 1 eV, dwell time of
500 ms, and PM voltage of 600, resulting in a scan duration of approximately 10 seconds
plus dead time. In contrast, a precision scan might decrease the step size, and a scan of
lower energies would need to decrease the PM voltage to avoid overloading the PM.
Regardless, the scans are defined and the number to be conducted chosen, followed by
data collection until the program is complete or manually stopped.

Figure 3.4 The portion of the AugerProbe outside of the growth
chamber, with arrows identifying the two (of three) visible axes used to
control probe axial tilt. Image: the author.
92
3.3 Capabilities

Determining the capabilities of the AugerProbe was extremely important, as they dictate
the experiments that can be conducted. After setting up the probe and ensuring spectra
were reliably observed, a variety of samples were examined to determine if there were any
significant limits. Growth parameters were then adjusted to determine their impact on
the quality of the spectra collected and the limits of its observational capabilities.

3.3.1 Observables
This effectiveness of this Auger probe design had been demonstrated for surface analysis
in MBE systems with reported in situ observation of N, O, Si, Fe, Zn, Ga, Tb, and
Dy.
133,137,318,319
However, to the best of my knowledge, in situ observation of complex
oxides thin films grown by PLD using this probe design had yet to be reported in the
literature, though other efforts on simple oxides had been reported.
133,318

I observed characteristic Auger spectra of more than 24 elements collected from samples
ranging from single crystal and thin film complex oxides at 850℃ to polycrystalline metals
at room temperature. The atomic number of characteristic peaks ranged from 6 to 64,
with peak energies from 40 to 1760 eV, and signal observable from samples less than a
single monolayer thick. For many of the elements multiple characteristic transitions can
be seen, including fine structure. The spectra are shown with their corresponding dN/dE
signals in Figure 3.5.
The samples used to generate the spectra shown in Figure 3.5 were all either single crystal
substrates purchased from CrysTec GmbH, thin films grown by PLD, or metal foils, as
described and summarized in Table 8. The PLD targets used to grow the thin films were
dense polycrystalline targets fabricated by solid state reaction except for the LaAlO
3

target, which is a single crystal target purchased from CrysTec GmbH. Auger studies on
metal foils were conducted at room temperature, while all other spectra were obtained at
750 to 850°C. Spectra 1, 22, and 23 were collected from a single crystal MgAl
2
O
4
(111)
substrate, 2 and 10 are from a GdScO
3
(110) substrate, 3 is from a CaY
2
Co
2
Ge
3
O
12
thin
film, 4 and 17 are from a Y
3
Fe
5
O
12
thin film, 5 is from a LaAlO
3
thin film, 6 and 19 are
from a NdNiO
3
thin film, 7 is from contamination on Ag foil, 8 is from a SrRuO
3
thin
film, 9, 12, and 14 are from a CaTiO
3
thin film, 11 is from Ag foil, 13 is from a VO
x
thin
film, 15 is from a BaTiO
3
thin film, 16 is from a LaMnO
3
thin film, 18 and 21 are from a
NdGaO
3
(110) substrate, 20 is from Cu foil, 24 is from a SrTiO
3
(100) substrate, and 25
is from a BaZrO
3
thin film. Many of the characteristic elemental peaks have been
93
confirmed with multiple samples and sample types, such as Ba and Ti from BaTiO
3
films,
Ca and Zr from CaZrO
3
films, and Sr from SrTiO
3
thin films.

Figure 3.5 Auger spectra of 24 elements, intensity adjusted for visibility and comparison, plotted as E*N(E) (top)
and N(E) (top, inset), as well as dN/dE (bottom). Numbers correspond to spectra and peak identifications (for peaks
marked with arrows) found in Table 8. Image: the author.
94
Table 8 Identification of peaks denoted by arrows for the spectra shown and numbered in Figure 3.5, sorted by reference
number (column 1). Column 2 is the energy at which the peak was observed, and column 3 is the element responsible
for the peak, with questionable identifications noted. Column 4 is the atomic number, and column 5 is the identity of
the transition responsible for the peak. The reference numbers marked with the ○ symbol are from single crystal
substrates, the ⸸ symbol indicates they are from PLD-grown thin films, and the § symbol indicates they are from metal
foil.

This verified the viability of this technique, and probe, for spectral acquisition from
complex oxide thin films before, during, and after the growth process. With the ultimate
goal of real time observation of the growth process, the next step was to determine the
limits of its characterization ability.

95
3.3.2 Limits
With the only apparent fundamental limitation to observing elements (aside from H and
He) arising from the possibility of peak overlap, determining the experimental limits of
observation was a logical next step. To explore these limits, I investigated the impact of
the laser pulse on the Auger signal, the pressure-dependence of the AugerProbe, and ways
to address peak overlap.
The ideal scenario for real time compositional observation during deposition would be the
ability to collect spectra regardless of the laser plume. However, several factors limited
this possibility, foremost among them being the effect of the laser plume on the signal
received by the probe. When observing the real time PM signal during deposition, there
was a direct correlation between spikes in the signal and laser pulses. This likely occurs
from increased electron scattering near the surface during the relatively high pressure
arrival and condensation of the ablated plume. Unfortunately, however, the spikes in
signal rule out the acquisition of spectra simultaneously with deposition due to the
duration of the scans, and alternatives
would have to be investigated.
To identify the acceptable range of
growth pressures which can be used with
the in situ probe, I conducted spectral
acquisitions as a function of background
pressure on a NdGaO
3
(110) substrate
prepared in the same manner as those
used for typical growths. As can be seen
in Figure 3.6, the signal only decreases
by approximately 10% between 5x10
-7

and 5x10
-4
mbar, and 50% at 5x10
-3

mbar, with effectively complete loss at
5x10
-2
mbar, caused by increased electron
scattering in the higher-pressure
environment.
136
The O
KLL
direct peak,
shown in the inset of Figure 3.6 for the
pressures measured, is still clearly present
even at 5x10
-3
mbar. Therefore, adequate
signal for analysis should be attainable
using many of the conditions for growing
complex oxides with PLD, which are
often below this threshold.
37,96,114
Figure 3.6 Normalized O, Ga, and Nd peak-to-peak
intensities of spectra collected with 10 second total dwell time
as a function of chamber pressure, and (inset) the OKLL N(E)
spectra with denoted pressure, intensity shifted linearly for
clarity. Approximately 50% of the oxygen Auger signal is lost
at 5x10
-3
mbar and effectively all of it is lost by 5x10
-2
mbar,
as illustrated by the decreasing oxygen peak intensity shown in
the inset. The shift in oxygen peak energy is very likely due to
systematic errors arising from the instrument conditions. The
dashed line is a guide to the eye. Image: the author.

96
The increase in oxygen peak energy with increasing pressure is observed in the Nd and
Ga spectra as well and is of an equivalent magnitude for all three elements (~2 ev), shown
in Figure 3.7. This indicates the shift is very likely systematic, rather than chemical
changes on the surface which has been observed in prior AES surface oxidation
studies.
310,320-322
While the ultimate cause of the energy shift is difficult to determine in
this instance due to the low energy resolution of these scans (1 eV), the most likely source
is instrumental as equivalent amounts of peak drift in short periods of time have been
observed previously with this probe.
As mentioned previously, an inherent limitation of this system will arise from peak overlap
between different elements, especially when observing complex stoichiometry materials
like perovskite oxides. An example of this phenomenon can be seen in the spectra of
LaMnO
3
, for which peak overlap is observed between the primary La
MNN
peak and a
Mn
LMM
satellite peak, shown for different film thicknesses in Figure 3.8. This unfortunate
occurrence can make peak analysis more challenging, as peak deconvolution increases the
relative error. Luckily, however, the La
NVV
peak is visible at around 74 eV. Although the
peak appears weak due to the very large low energy background arising from secondary
electrons,
138
when viewed as E*N(E) or dN/dE instead of N(E) it becomes quite clear, as
seen in Figure 3.9, and can be consistently measured. The presence of multiple
characteristic peaks proves quite useful for surface characterization with AES for reasons
beyond peak overlap, as discussed shortly.
Figure 3.7 Peak shape and position comparison for spectra of characteristic O, Nd, and Ga peaks at increasing
background partial pressures of oxygen. For all three elements, the top spectra were collected at 5x10
-7
mbar, with each
successive row below it increasing by an order of magnitude until the bottom spectra, which were taken at 5x10
-3
mbar.
The dashed line connects the maximum position for each peak, showing that the energy shift observed is consistent
across all three elements. Image: the author.
97

Figure 3.8 The MnLMM and LaMNN peaks collected from LaMnO3 thin films of increasing
thicknesses between depositions, identified in the legend. Note the overlap between the LaMNN
peak located at approximately 636 eV and the MnLMM satellite peak located at approximately
643 eV. Spectra shown are 5 summed scans. Image: the author.
Figure 3.9 (right) The LaNVV peak collected
from LaMnO3 thin films of increasing
thicknesses between depositions, identified in
the legend. Despite the appearance of low signal
intensity due to the large secondary electron
background, the peak can be consistently
measured. Image: the author.
98
3.4 Spectral Analysis

The observation of spectra is, at most, a binary signal indicating the presence of an
element without some form of analysis. Qualitative analysis of this form has its uses when
the question only requires a binary solution-set. However, this can still be problematic
when, for example, the presence of carbon is ubiquitous due to fairly consistent surface
contamination, thereby making the observation of a carbon peak have little value.
Moving from qualitative to quantitative Auger analysis has its own (quite considerable)
challenges.
135
Take, for example, the extensive literature capturing “standard spectra.”
Palmberg et al published a collection in 1972
plotted in derivative mode with a 3 kV primary-
electron beam,
323
which was followed in 1976 with
a revision by Davis et al including spectra excited
with a 5 kV primary-electron beam.
324
In 1979
McGuire published a comparable collection which
included some high resolution scans, as well as
some plotted as a function of modulation
voltage.
325
Then, in 1982, another compilation was
published by Sekine et al.
326
The primary
significance of these collections is that none of
them match, demonstrating the importance of
measurement conditions, sample preparation,
laboratory standards, and the simple fact that
instruments designed differently behave
differently. Ultimately, the only reliable data will
be that collected in the same system, and only then
if sufficient care is taken to make results
repeatable.
Quantifying the intensity of Auger peaks is
pseudo-quantitative, and not the same thing as
quantitative Auger,
327-329
but it can be used for
self-referential analysis which certainly has its
uses. The primary method for measuring the
intensity of Auger peaks has historically been
peak-to-peak,
138
illustrated in Figure 3.10, which
compares the maximum and minimum signal of
the dN/dE signal. The major advantages of this
method are that it is easy to execute numerically
Figure 3.10 The OKLL peak plotted as N(E).
The top panel shows the accompanying dN/dE
signal with circles indicating the data points from
which the peak-to-peak value is calculated by
subtracting the local minimum from the local
maximum. The bottom panel is a visual
representation of the trapezoidal numerical
integration method, with the spectra split into
trapezoids of equal width. Image: the author.
99
on an electronically-generated signal, and the peak-to-peak signal clearly shows subtle
features which are difficult to see in the direct signal. While the second advantage is still
significant today, the first has lost its significance as a result of the abundance of digital
equipment. The spectra can now be manipulated in any way imaginable with software
analysis, removing peak-to-peak measurements from center stage.
The adoption of new intensity measurement techniques is especially useful for data with
a low signal-to-noise ratio, as the dN/dE signal still has the noise from the N(E) signal.
Instead, a method which has proven effective for quantifying intensity has been
measurement of area-under-the-curve. Simple to execute on a well-defined data set, the
AUC is calculated by performing a numerical integration of a given peak using the
trapezoidal method to acquire an approximate area, as shown in Figure 3.10. The
trapezoidal method uses the approximation in Equation 3.1, where 𝑁 + 1 is the number
of evenly spaced data points.
∫ 𝑓 ( 𝑥 ) 𝑑𝑥 𝑏 𝑎 ≈
𝑏 −𝑎 2𝑁 ∑ ( 𝑓 ( 𝑥 𝑛 )+ 𝑓 ( 𝑥 𝑛 +1
) )
𝑁 𝑛 (3.1)
Regardless of which intensity measurement method is adopted, the spectra collected with
the AugerProbe in our PLD chamber under film growth conditions tends to be noisier
than those collected with cylindrical mirror analyzers (CMAs) in UHV. The analytical
methods, therefore, require slightly more effort than the traditional techniques applied to
AES to get comparable results. One simple method for improving signal-to-noise is
collecting multiple spectra and summing them. Higher quality scans typically use a sum
of between 5 and 20 scans, depending on the parameters I am working with. By summing
the spectra, the error is additive whereas if the average is taken the error is exponential.
The large number of spectra collected require a great deal of analysis, so scripts were
developed in MATLAB to automate the P2P and AUC calculations, discussed in §2.7.3.
This has the additional – very important – benefit of applying an identical process to each
spectrum. The downside of scripted analysis is that the error increases because the script
will indiscriminately try to calculate intensity of a peak, regardless of whether or not a
peak is actually present. Additionally, slight shifts in peak location due to noise can add
that noise to the calculated intensity signal because the ‘true’ peak was not identified. To
combat this, the scripts were designed to adjust the local peak location and try to reduce
background noise by selecting max- and minima from a locally defined range and
subtracting the secondary electron background where appropriate. The methods used are
illustrated in Figure 3.11.
100

Figure 3.11 Examples of spectra plotted as E*N(E) and dN/dE, with visual examples of the area under the curve
and peak-to-peak methods of quantifying intensity, La and Mn peak-overlap, and narrow scan summing. Image: the
author.

101
3.5 M onitoring Composition in a
CaTiO
3
/ LaM nO
3
Superlattice

My experimental objective was to measure contrasting Auger signals from two different
materials to determine the sensitivity of the probe. Additionally, this would reveal how,
and if, signal from contrasting species in different materials interfered with one another.
The deposition of a superlattice is an ideal basis for this experiment because if it can be
demonstrated that the growth can be controlled at the atomic level, the Auger spectra
would have the same resolution. The CaTiO
3
/LaMnO
3
superlattice was chosen because
our laboratory group had experience growing it with very high quality (special thanks to
Yang Liu).
330
We optimized PLD growth parameters for the CaTiO
3
/LaMnO
3
superlattice on a single-
crystal (110) NdGaO
3
substrate to achieve long-lived layer-by-layer growth with a smooth
resulting surface as observed with RHEED, shown in Figure 3.12 (a). Deposition of
CaTiO
3
causes a very sharp increase in specular spot intensity, while the deposition of
LaMnO
3
causes a rapid decay in intensity. This results in a step-like RHEED pattern,
with the tops of the steps corresponding to the deposition of CaTiO
3
, and the bottoms
corresponding to LaMnO
3
deposition. Intensity oscillations corresponding to the growth
of a single monolayer are clear for the deposition of both materials, following the larger
intensity trends mentioned above. This clearly indicates that the superlattice grown
consists of alternating layers of CaTiO
3
and LaMnO
3
, where each layer consists of four
atomic layers of the complex oxide material.
The quality of the film during the deposition process was observed by monitoring the
electron diffraction pattern. At the end of the growth the capping layers of the superlattice
were LaMnO
3
, and the post-growth diffraction pattern demonstrates this in Figure 3.12
(b). The post-growth film was smooth, as indicated by the vertical streaking of the
diffraction pattern due to Bragg columns arising from the two-dimensional surface. The
mid-growth diffraction patterns of the CaTiO
3
surface and LaMnO
3
surface (Figure 3.12
(c) and (d)) show equivalent smoothness and no observable three-dimensionality, as would
be indicated by a grid of diffraction spots. Therefore, this is a suitable system for AES
surface composition studies. The stability of the layer-by-layer growth allows us to use
the same growth parameters used for the superlattice, with RHEED, to perform surface
composition analysis with single-atomic-layer resolution.
102

Figure 3.12 Electron diffraction data from the deposition of a CaTiO3/LaMnO3 superlattice on a NdGaO3 single
crystal substrate. (a) RHEED oscillations during growth, with four oscillations per material, per layer, and distinction
between the two materials indicated by background color, with the lower intensity oscillations from the LaMnO3 and the
higher intensity oscillations from the CaTiO3. (b-d) Electron diffraction patterns demonstrating smoothness of the as-
grown film after growth (b), after deposition of four atomic layers of CaTiO3 (c), and after deposition of four atomic
layers of LaMnO3 (d). Image: the author.

3.5.1 Experimental Design
With the CaTiO
3
/LaMnO
3
superlattice as an established system within which to work,
the specifics of the experiment were decided based on the expected signal evolution.
Ideally, the Auger signal would increase with deposition while masking the signal from
the layers beneath it. To verify that this is occurring requires confidence in the definition
of a lack of peak in the conducted scans. Therefore, the thickness of deposition should be
such that there is no doubt that the signal below the current material is masked, so it was
decided that the thickness of the alternating layers in the superlattice would be 20 unit
cells. To collect spectra of sufficient quality, the growth was stopped after 1, 2, 3, 4, 5, 7,
10, 15, and 20 total unit cells had been deposited to collect scans for the first 20 unit cell
layer of each material, and 2, 5, 10, and 20 unit cells for the second 20 unit cell layer of
each material, as illustrated in Figure 3.13. At each interval during which the growth
was stopped, spectra of the energy range that contained the characteristic Auger peaks
for Ca
LMM
, Ti
LMM
, O
KLL
, La
NVV
, and Mn
LMM
were collected 20 times and summed for a
total dwell time of 10 seconds per data point. The AUC values for the peak intensities
were normalized to the oxygen AUC, on the assumption that even if the overall signal
intensity shifted from layer to layer (due to elemental brightness, surface roughness, or
local field fluctuations driven by heater current changes) the effect would scale with the
103
oxygen signal. The resulting values were then
normalized to demonstrate overall elemental intensity
shifts as a function of film thickness.
The superlattice was grown via PLD with the laser
operating at a repetition rate of 1 to 5 Hz using dense
polycrystalline targets prepared by solid state
reaction. The laser spot size was varied from 2.5 to 6
mm
2
, resulting in a net fluence of 0.8 to 2 J cm
-2
. The
chamber was evacuated to a base pressure below
5x10
-7
mbar before flowing high purity oxygen to
create a background growth pressure of 5x10
-2
mbar.
The single-crystal (110) NdGaO
3
substrate was
annealed in flowing O
2
for three hours at 1100°C prior
to cleaning with acetone and then isopropyl alcohol
in an ultrasonic bath. The substrate was heated to
750°C in growth pressure. Growth rate and thickness
were observed with in situ RHEED using a 35 kV
accelerating voltage. During the acquisition of Auger
spectra, the chamber pressure was reduced to < 10
-6

mbar and the accelerating voltage of the electron
source decreased from the 35 kV used for RHEED
during deposition to 5 kV, with an emission current
of 5 µA. The tilt of the film relative to the electron
beam was increased from ~1° to 15° to maximize the
Auger signal. Tilt angle was controlled with a digital
microcontroller, thereby allowing translation between
the same precise tilt angles for deposition and AES.
After acquisitions the state of the chamber was
returned to that used for deposition, and vice-versa.

3.5.2 Data and Analysis
Surface composition of the superlattice, acquired with in situ AES, demonstrates that, as
expected, the elemental signal from a deposited material increases with the deposition of
atomic layers and saturates after a few (approximately five) monolayers, as shown in
Figure 3.14. Likewise, the elemental signal from the underlayer decreases with the
deposition of an alternate material on top and vanishes after approximately five
monolayers. With a lattice constant of approximately 4 Å, this corresponds to a depth
limit of around 2 nm for the Auger probe in this geometry before the signal can no longer
Figure 3.13 Cartoon of the stacking
sequence of deposition and Auger spectra
acquisition for the CaTiO3/LaMnO3
superlattice. The alternating colors
correspond to the regions deposited between
stopping for spectroscopy, as do the numbers.
Image: the author.
104
be differentiated from the noise. The large increase in signal observed after the deposition
of a single monolayer is more than adequate for measurement, demonstrating that the
probe is capable of sub-monolayer sensitivity. The ratio of A-site to B-site (Ca/Ti or
La/Mn) signal is consistent throughout the depositions, indicating that the data
acquisition and analysis methods used are suitable for this degree of qualitative analysis,
and, with the addition of standards, may be suitable for quantitative analysis.

Figure 3.14 A plot of normalized (to the oxygen signal) elemental Auger spectra signals during the growth of a
CaTiO3/LaMnO3 superlattice. Note that the A-site intensity (Ca and La) data is offset from the B-site intensity (Ti
and Mn) data for clarity. Each data point is from an area under the curve calculated from scans of the characteristic
peak for that specific element, with 10 seconds total dwell time, normalized to the oxygen signal from that specific
collection. The Auger spectra were collected in situ between depositions of known thickness, as calibrated by RHEED
and previous growths. Lines are exponential fits to the data in approximation of escape depth as outlined in the text.
Image: the author.
In order to perform quantitative Auger with this probe would require, at the very
minimum, a significant investment in the development of a standardization system. To
do so, the observed spectra would need to be compared to a sample with a precisely known
composition, which can be a challenge in itself. Furthermore, a persistent complication
with Auger quantification is the extreme surface sensitivity. This property is clearly
valuable in an analytical setting, but problematic when trying to develop standards. This
is because the 2 nm of surface-depth the probe can ‘see’ are extremely sensitive to,
essentially, everything. On exposure to atmosphere the presence of dust and organics
effectively guarantee the immediate surface contamination of a sample, and therefore
performing reliable standardization is a challenging prospect and a bit of a Catch-22. If it
is possible to characterize the sample sufficiently for standardization in the same
105
environment as the Auger analysis, why is the Auger even necessary? Despite this minor
conundrum, a work-around was developed and is discussed at length in §4.3.
The sensitivity of the Auger probe to a given material is dependent on the ability to
generate and observe Auger electrons. Generation is predominantly a function of the
amount of the material present, though other factors such as ionization cross-section play
a role. Observation of the Auger electrons is then dependent on their ability to escape the
material and find their way to the detector. The escape process is largely responsible for
surface-sensitivity, which is why AES is a valuable surface characterization tool, because
the relatively low energy of Auger electrons limits their ability to leave the parent
material. The proportion of Auger electrons which escape can be approximated using
Equation 3.2, rewritten here but discussed at length in §2.3.2, where z is the depth from
which they are generated and λ is the inelastic mean free path (λ
MFP
). However, the λ
MFP

for any given element is not a constant and depends on the other elements present as well
as the structure of the material, as there are numerous energy-dependent processes which
can prevent an Auger electron from escaping. For this reason, experimental escape depth
data was fit by Seah and Dench and a relation found, displayed in Equation 3.3, as a
function of nm instead of monolayers, as it was written previously in Equation 2.16, where
E is the energy of the Auger electron in eV, and λ
MFP
is in nm.
274

𝑁 𝑁 0
= 𝑒 (
−𝑧 𝜆 )
(3.2)
𝜆 𝑀𝐹𝑃
=
143
𝐸 2
+ 0.054( 𝐸 )
1
2
⁄
(3.3)
The deposition data from the CaTiO
3
/LaMnO
3
superlattice growth were fit to the basic
escape model (Eq. 2.13), with the results shown in Figure 3.15. The fits were applied
assuming a single layer thickness of 0.4 nm using Auger energies shown in Table 9 for
the observed transitions. The escape depths calculated from the fit are quite close to the
theoretical values and follow the expected trend. There are many reasons the observed
values may vary from expected,
331-333
including the structure or composition of the parent
material or the material deposited on it. These results show that this probe has the acute
surface sensitivity expected for AES, making it a viable tool for delicate real time in situ
surface characterization experiments.

106

Table 9 Escape depths calculated from fits made to deposition data shown in Fig. 3.15 following Eq. (3.2), compared to
escape depths calculated for the same elements following Eq. (3.3).
Element Transition Energy (eV) Fit λMFP (nm) Calculated λMFP (nm)
Ca LMM 291 0.60 0.92
Ti LMM 387 1.20 1.06
La NVV 80 0.51 0.51
Mn LMM 590 1.23 1.31

Figure 3.15 Fits to the normalized intensity data of CaTiO3/LaMnO3 superlattice layers
during deposition. Comparing the fits (long dash) to the calculated escape depths (short dash)
for these elements reveals that the escape depth is proportional to the Auger electron energy,
as expected, and has agreement within one unit cell (0.4 nm) for all elements. Image: the author.
107
3.6 Real Time Growth Observation

With the sensitivity of the probe proven, its suitability for real time observation could be
addressed. Unfortunately, when collecting spectra during laser ablation the pulse can
clearly be observed in the AugerProbe photomultiplier signal, indicating that it will likely
influence the resulting spectra. The signal interference from the laser pulse can be seen as
a spike in photomultiplier signal, indicating that it is associated with an increased number
of electrons passing through the collimator lens to the scintillator within the probe. The
spike in signal was observed at various acquisition energies, suggesting it was the result
of one of several possibilities: electrons with sufficient energy to pass the collimator lens,
high energy photons, or electrons of the energy the lens was tuned to. High energy
electrons could be the result of electron scatter from the high density plume near the
substrate surface, though it is not certain if even elastically backscattered electrons with
the accelerating energy of 5 keV would be able to pass the collimator lens if it was not
tuned to their energy. However, the target-probe geometry makes it unlikely the source
of the signal could be from anywhere other than the vicinity of the substrate. Therefore,
if it is not from electrons it would have to be from photons, which would explain their
ability to ignore the electromagnetic filter. However, though there has been extensive
imaging of plumes with high-speed cameras and spectrometers,
94,211,223
to my knowledge
there has not been investigation of this phenomenon. The last possibility is that the
excitation of the plume simply creates a sufficiently increased signal to cause the observed
spike, because the high pressure plume has a sufficiently increased electron mean free path
compared to the substrate to observe excitations from a larger volume. Regardless, the
answer to this question is beyond the scope of the current work, yet still poses the
important question of how it can be overcome for real time AES.
The technique I developed to
accommodate the fundamental limitation
of prohibited simultaneous Auger spectra
acquisition and deposition was to simply
work around it. This is achieved by
implementation of a pulse-probe pattern
of spectral acquisition, with the only
fundamental temporal limit the duration
of a spectral scan. With the AugerProbe
the amount of time required for a 1 second
scan of a single energy is about 5 seconds.
Of that 5 seconds, only the middle 1
second is actually collecting data, so if the
laser is appropriately timed it can pulse
Figure 3.16 Illustration of the SrMNN peak when plotted as
E*N(E). Image: the author.
108
with an interval greater than 1 second every 5 seconds. Between scans any number of
pulses can be shot, and of the only limitations will be self-imposed by the desired
deposition parameters. This has the fundamental advantage of allowing either low
resolution scans with a fast continuous deposition, or high resolution scans with sporadic
deposition, or anything in between. Depending on the needs of the deposition parameters
and spectra, it should be possible to tune the number of pulses and scan durations for a
wide variety of applications.
A typical scan duration for a single peak, such as Sr
MNN
at ~85 eV, shown in Figure
3.16, may take approximately 20 seconds for 0.5 second collection per eV over about 20
eV, including the time required for the probe to adjust its parameters between scans.
Depending on the number of pulses between scans, 20 to 30 seconds can provide the
surface ample time for recovery, in a manner similar to interval deposition, but without
necessarily depositing a full layer of material.
228,334
This technique has been shown to
create very smooth films due to the generous surface diffusion and reconstruction
period.
100,227,229
With fewer than a full monolayer of pulses between scans the effect is not
as dramatic as it is for interval deposition, but it is still effective at providing a similar
benefit, and does not interfere with the observation of RHEED oscillations, as shown in
Figure 3.17 for a growth using a 30 second interval between 3 pulses at 1 Hz.
Alternately, the scan width can be extremely narrow, creating a much noisier signal in
exchange for a higher pulse frequency. When the spectrum is this narrow, the scan is
effectively just glancing at the height of a peak, and careful observation must be conducted
prior to the growth to ensure
that the peak location is actually
being seen. Simple numerical
techniques can be used to
subtract a known background,
but the potential for a shifting
background over the course of
deposition makes this practice
potentially error-prone.
To improve signal quality, and
thereby improve temporal
resolution, the most effective
techniques simply rely on the
appropriate calibration and
alignment of the sample and
probe. However, a further tactic
which can be employed is
Figure 3.17 Demonstration of the pulse-probe method’s ability to
observe RHEED oscillations despite the long recovery time. Shown here
with 3 pulses between each scan. Image: the author.
109
selecting peaks which provide sufficient data for analysis over a narrower energy range,
thereby allowing the same energy resolution in a shorter scan duration. Because most
elements have multiple visible transition peaks, though only some are intense enough for
measurement, it may be possible to choose the peak measured. For instance, shown in
Figure 3.18, there is a Ti
LMM
transition around 381 eV, and a Ti
MVV
transition around
36 eV. The Ti
LMM
peak is quite intense, but also quite broad, whereas the Ti
MVV
peak is
weaker, but narrower. By scanning the
low energy Ti
MVV
peak the spectral
width can be narrower to observe the
entire peak, and low energy scans
provide ample signal when the
secondary electron background is
subtracted.
136,149,335

Adding AES to the semi-interval
deposition yields excellent results,
showing the viability of the pulse-probe
technique in practice. Using this
method, I performed real time in situ
observation of surface composition in
homoepitaxial SrTiO
3
, using both slow
and fast Auger spectra acquisition
rates, shown as simultaneous Auger and RHEED intensity in Figure 3.19. The observed
composition has some noise, but the A:B ratio is reasonably constant, indicating
stoichiometric deposition, or at least no significant composition evolution, for both
growths. Furthermore, the lack of correlation between RHEED specular spot intensity
and AES intensity indicates that in this geometry the Auger probe is insensitive to surface
roughness on the scale of a nm or so, as would be expected in layer-by-layer homoepitaxial
growth. This pulse-probe technique successfully shows sufficient temporal sensitivity to
monitor subtleties of dynamic complex oxide thin film growth events never before
witnessed with previously demonstrated in situ characterization methods.
Figure 3.18 Comparison between the high energy TiLMM
peak located at 381 eV and the TiMVV peak located at 36 eV,
with energy scale and intensity adjusted for demonstration.
Image: the author.
110

Figure 3.19 Measured intensity of RHEED specular spot (noted as RHEED intensity) and AES data
simultaneously collected using the pulse-probe method outlined in the text for the growth of homoepitaxial SrTiO3.
The Auger lines used to monitor Sr and Ti are SrMNN, and TiLMM transitions in (a), and to monitor O, Sr and
Ti are OKLL, SrMNN, and TiLMM transitions in (b). Laser pulses result in a roughening of the surface (a rapid
decrease in RHEED intensity) followed by recovery during Auger acquisition (gradual increase in RHEED
intensity). Depending on the duration of the spectral collection and the number of pulses between scans, the shape
of the resulting oscillations will vary with either long scans and long recovery time as shown in (a), or short scans
and thus short recovery as shown in (b). Likewise, the sensitivity of the scans to compositional changes is dependent
on the number of data points and their integration time, resulting in a trade-off between the quality of RHEED
and AES data. Image: the author.

111
3.7 Summary

A Staib AugerProbe was successfully incorporated into a pulsed laser deposition system,
and I have demonstrated in situ Auger electron spectroscopy of complex oxide thin film
surfaces grown by PLD. Characteristic Auger spectra have been collected in situ from
numerous elements sourced from various complex oxide single crystal substrates, PLD-
grown thin films, and metal foils. Pressure-dependent studies show that ample signal is
obtainable up to 5x10
-3
mbar, a pressure suitable for the synthesis of many complex oxides.
Challenges limiting our ability to quantify our results have been identified and discussed,
and strategies have been proposed to overcome them moving forward. Sustained layer-by-
layer growth of CaTiO
3
/LaMnO
3
superlattices on NdGaO
3
substrates was achieved,
thereby allowing monolayer depth-resolved Auger surface composition studies. We have
found that our Auger probe provides spectra from, approximately, the top 2 nm of the
thin film, with the possibility of sub-monolayer depth sensitivity. The prospect of applying
Auger electron surface analysis in real-time during pulsed laser deposition of complex
oxides has been addressed. These studies demonstrate the viability of AES for a wide
variety of materials analysis applications in our PLD growth chamber.

112

Chapter 4
In Situ Compositional
Quantification

4.1 Introduction

In this chapter I review the influence of PLD growth parameters on the composition of
thin films, and explore this relationship by conducting a pressure-fluence series of
homoepitaxial SrTiO
3
films. The composition of the films is approximated with XRD, by
measuring the c-axis lattice expansion, and compared to the Sr
MVV
and Ti
MNN
peak
intensities acquired with AES. A direct observation of the influence of growth parameters
on film composition is conducted, demonstrating that this is a viable process-control
technique. Furthermore, the observed behavior of the growth parameter-dependent film
signal cannot be explained without considering the surface termination to be shifting as a
function of composition. To understand this phenomenon, a parameter-free escape depth
model is developed to explore the relationship between film structure, termination, and
expected Auger intensity. The expected intensity of the structure grown is modeled and
comparison of fits clearly shows the sensitivity of the AugerProbe to surface termination.
Using this as a baseline, the model is expanded to quantify the pressure-fluence series of
homoepitaxial SrTiO
3
films and shows agreement with the literature.

113
4.2 Film Sensitivity to Growth Parameters

The existence of a relationship between growth parameters and film composition is no
secret, but the full extent of that relationship is difficult to identify,
336-360
as illustrated
by the persistent debate over the origins of the conducting LaAlO
3
/SrTiO
3
interface. This
difficulty arises from several factors: the existence of interdependencies between
parameters can confuse the picture,
95,361-365
specific answers are easier to generate than
general ones,
119,366,367
the impact of composition on growth dynamics is easily
overlooked,
96,190,368
and compositional changes are harder to observe than structural
changes.
120,127,136
These factors will be addressed individually.
Interdependencies between film growth parameters exist because the axes of the solution
space are not invariate and the solution is not linear.
100,140,369,370
When changes are made
to one parameter the influence of the others invariably change as well. For instance, the
temperature of the substrate (and shape of the heater) will, to the first order, change the
surface energy and diffusion of the incoming adatoms. In the second order, it will influence
the energy of the adatoms before they arrive;
211,371
and in third order, changing the energy
of the species in the plume will actually change the shape of the plume,
185
thereby
potentially altering the stoichiometry of the composition which arrives at the substrate.
This illustrates how trying to determine the influence of growth parameters can be set
awry by these types of relationships: altering one parameter can make it appear as if it is
affecting far more than it actually is, thereby undermining the process of untangling the
true impact of individual parameters.
The subtle complexity of the growth system can make developing effective growth
parameters for any individual material challenging, and it is not uncommon for the
parameters that suit one material to be a poor match for another. This frustrating reality
tends to promote the development of single-material solutions, because the system is
simply too complex to generate general solutions without a great deal of effort. This is
not to overlook the importance of developing material-specific recipes, but with the
exception of certain generalities, they typically do not translate with any significant
effectiveness between experimental setups. While there has been considerable effort to
develop general solutions to simplify the complexities of PLD systems, none have yet to
offer a systematic perspective of universal value. A promising approach to this problem is
the application of machine learning,
140,372-375
which can find patterns in complex data sets
more effectively than humans. However, even this method will require high quality data
to achieve the desired outcome, and the addition of more complex data will generate a
more complex solution-space.
114
Composition has the potential to significantly alter growth dynamics without being
observed. This is largely because composition itself has been challenging to observe, but
the result is the same: the true impact of growth parameters on growth dynamics is
difficult to ascertain. Subtle differences in composition can drastically change the
energetics of a system by altering the surface reconstruction,
376,377
lattice constant, or even
termination.
119,146
All of these examples influence surface energy, and will therefore, by
definition, change the dynamics of film growth.
120,145
However, while the impact of this
change may be observed with in situ RHEED oscillations or post-growth XRD, the
connection will be very difficult to determine without deliberately seeking it out or the
use of composition characterization tools.
Lastly, because compositional changes are currently harder to observe than structural
changes, the manifestation of compositional changes in the structure will not adequately
correlate growth parameters with composition.
114,142
The source of this problem is, in
many ways, the same as that of the previous ones: the complexity of the system is greater
than our observational ability. The result is like trying to map the ocean floor by watching
the waves, and the solution is the implementation of more comprehensive characterization
tools.

4.2.1 Laser Fluence and Spot Size
The role of laser fluence in film composition and the resulting properties has been
documented numerous times.
96,190,378
However, the conclusions often provide only weak
generalizations when the specifics do not extend to other growth chambers. Laser fluence
is thought to change the stoichiometry of the plume largely in two ways: through
preferential ablation and energy of the ablated species.
94,189,379,380
Preferential ablation is
a problem in complex oxide perovskites, and means that either A or B of ABO
3
is more
likely to be ablated. The difference between A and B ablation is dependent on the species
as well as the fluence. As an example, SrTiO
3
has mixed ionic-covalent bonding states,
with SrO exhibiting ionic bonding and TiO
2
covalent bonding. This means the energy-
dependent ablation of these two species is not identical. Specifically, Ti is more difficult
to ablate, causing a lower laser fluence to result in a Sr-rich plume. However, this also
means that the fluence-dependence of composition is not symmetric, and a higher laser
fluence may not produce a significant excess of Ti, if at all. The energy of the ablated
species is dependent on the laser fluence, though not linearly due to the complexity of
plasma-formation. A more direct correlation arises between energy of the ablated species
and the resulting plume shape, though it is also influenced by spot size and background
pressure.
115
The size of the spot ablating the target will obviously impact the flux of the plume, but
also determines its shape.
181,191,208,381,382
Beyond plume propagation, the flux
predominantly impacts the growth process. A larger flux requires higher surface diffusion
to grow a smooth film, as otherwise the arriving material will not have sufficient time to
settle into an ideal lattice position before more piles on top of them. However, if the
surface energy encourages 3D growth then using a large flux with large surface diffusion
can promote the formation of a smooth film, as is the case with SrRuO
3
.
104
A low flux
provides ample opportunity for the arriving material to settle into a low energy
position,
228-231
which is perfectly fine for some films, such as homoepitaxial SrTiO
3
.
Materials with the inclination to grow 3-dimensionally due to their nucleation energy
(such as, again, SrRuO
3
)
104
will have far too much time to ball up if allowed, and a low
flux will produce a rough film.
It should be noted that the relationship between pulse frequency and spot size can change
the dynamics of film growth in unexpected ways despite net flux. For instance, a low pulse
rate with large spot size may produce a given flux that is not equivalent to twice the
frequency and half the spot size, even though the flux per second may be the same.
Returning to SRO, this system needs a high flux to grow smooth films, but more
importantly it needs a high pulse frequency on the order of 5 – 10 Hz. Determining the
intricacies of these requirements for different material systems is important for
conceptualizing the growth process.
The influence of spot size and shape on the
resulting plume is not immediately intuitive.
If the aspect ratio of the laser spot is not 1:1,
the resulting deposition shadow demonstrates
the flip-over effect,
191
in which the plume
propagation is smallest in the longest
dimension of the spot, and vice-versa. This
arises due purely to gas-dynamics
phenomena, and is the result of the density of
interatomic collisions in the plume. In other
words, the plume spreads preferentially in the
short-dimension because the local pressure is
lower in that direction. Unfortunately, it is
rare for spot dimensions to be 1:1, especially
because mask shape is distorted by the 45°
angle of incidence of the laser. However, with
sufficient adjustment of the chamber and
growth parameters this phenomenon should
not be a problem, as long as it is accounted
Figure 4.1 Laser spot sizes on thermal paper for the
‘R2’ mask at different lens positions. Relative to the spot
with best focus, labeled with an arrow, increasing sizes
correspond to 5 mm increments of lens transit away
from the chamber, and decreasing sizes are 5 mm
increments towards the chamber. Ruler present for scale:
1 mm increments. Image: the author.
116
for that the disproportionate plume dispersion geometry will also influence alternate
species disproportionately due to the difference in their atomic masses. Example laser spot
sizes are shown in Figure 4.1.

4.2.2 Background Pressure
The chamber background pressure plays a critical role in the formation of the plasma
plume and propagation of atomic species, discussed extensively in §2.2. The nature of the
influence pressure has on the resulting structure and composition of the deposited film is
complex, but ultimately comes down to kinetics. The primary ways that pressure
influences film growth and stoichiometry are directly through plume dispersion and
preferential scattering,
180,181,382
oxidation of plume species,
218,234,371
and kinetic energy of
the arriving species. As with all other growth parameters, these factors are inextricably
linked with one another as well as other growth conditions, but there is still benefit in
exploring the first-order connections.
The shape of the plume as it travels from the target surface to the substrate is determined
by a balance between the linear momentum of the plume front and the braking force of
the background gas.
189,193
Plume propagation models approximate the shape using the
initial front velocity and driving mass of the plume,
185,187
which are determined by the
laser fluence and spot size, against the background pressure. With a constant fluence, low
Figure 4.2 A simple Monte Carlo simulation of the pressure-dependent flight paths between target and
substrate (bottom) and final particle distribution on the substrate surface (top) illustrates a significant deviation
in deposition composition for Sr and Ti arising solely from the difference in atomic mass. Comparing high
(left) versus low (right) pressure shows the dramatic influence of growth parameters on the resulting
composition. Image: the author.
117
pressure will not brake the plume, allowing it to maintain a higher velocity and more
confined shape, whereas higher pressure will slow plume propagation and can effectively
stop it, forcing the adoption of diffusive behavior to reach the substrate.
219
In addition to
the impact on growth kinetics affected by reducing the plume energy into the diffusive
regime, this can increase preferential scattering of different species in the plume, due to
the disparity in atomic mass.
To illustrate the potential for preferential scattering to alter film composition, I performed
very simple Monte Carlo simulations of scattering as a function of pressure and atomic
mass. Again, because of the discrepancy in atomic weight between Sr and Ti, the two
species will behave differently within the plume. Lighter weight Ti will be more readily
diverted due to collisions, and higher pressure will increase the probability of collision.
Therefore, while higher pressure will increase collision frequency and thus diffusion for
both species, their compositional profile within the plume will be different, and the
stoichiometry of the plume-front which reaches the substrate will be more Sr-rich with
high pressure. The simulation depicts the flight paths and final destinations of co-ablated
Sr and Ti under two different pressure regimes, and is shown in Figure 4.2. The difference
in atomic mass between the two species results in diverging deposition patterns highly
sensitive to background pressure.
The oxidation of plume species changes the nature of the interaction between pressure
and plume.
178
This is obvious in some effects, such as by altering species’ kinetic energy
after oxidation. However, in other ways the result is more subtle and less intuitive, for
instance in the effect oxidation plays on surface diffusivity of different species, thereby
completely changing the kinetics of the growth process.
276
An example of the impact
oxidation can have on stoichiometry is shown in Figure 4.3, which applies the Monte
Carlo simulation of scattering as a function of pressure and atomic mass to the oxidized
Figure 4.3 Simple Monte Carlo simulation showing the difference in projected flight paths for Sr and Ti when the
species are elemental (left) versus oxidized (right). Note that the pressure in both simulations is the same, and therefore
the dispersion difference is purely the result of less disparity in the species’ mass. Image: the author.
118
Sr and Ti species. The result is a decrease in disparity under otherwise the same
conditions, demonstrating the complexity of these interactions.
The kinetic energy that the plume species bring to the substrate during adsorption must
also be accounted for somehow. On ablation, cations of perovskite oxides can have kinetic
energies on the order of hundreds of eV,
94
while the temperature of the substrate will be
somewhat less than 1 eV. Much of that energy will be lost through gas collisions, but any
that remains on reaching the substrate will increase the surface temperature. The impact
this can have on growth kinetics is clearly considerable, but it can also significantly impact
the composition and properties of films through, for example, the formation of oxygen
defects.
188

4.2.3 Summary
The relationship between growth parameters and the resulting thin film is clearly complex,
with interdependencies that can confuse the simplest attempts at analysis. The ultimate
vision is to understand the system well enough to make rational decisions about growth
parameters in order to achieve specific results. The first step towards this goal must be to
continue developing a more comprehensive understanding of the growth phenomena
themselves, while tying them conceptually to the underlying growth conditions.

119
4.3 Structural Correlation to Composition

The perovskite structure is quite malleable, and will readily adopt an orthorhombic or
tetragonal unit cell, as dictated by the relative sizes of the cations and anions. This
relationship is captured for oxides in Equation 4.1, the Goldschmidt tolerance factor (t),
which serves as an indicator of the distortion and stability of the perovskite structure as
a function of the constituents’ atomic radii.
𝑡 =
𝑟 𝐴 +𝑟 𝑂 √2( 𝑟 𝐵 +𝑟 𝑂 )
(4.1)
The origin of t is geometric, and simply indicates whether or not the size of the ions can
fit diagonally within a cube. When the A ions are too large (or B too small) t will be
greater than 1 and the structure adopted will be hexagonal or tetragonal, as is the case
for BaTiO
3
. When the A ions are too small (or B too large) t will be less than 0.9 and the
structure adopted will be orthorhombic or rhombohedral, as is the case for CaTiO
3
. When
the atomic radii of A and B generate a t value between 0.9 and 1 the structure will be
cubic.
When thin film perovskite oxides are not stoichiometric, the structure will be forced to
respond. The exact nature of that response will depend on the specific material and in
what way it is nonstoichiometric. However, regardless of whether it is a cation or anion,
deficient or in excess, it is almost always associated with c-axis lattice expansion
proportional to the degree of nonstoichiometry.
347,384,385
This arises as a result of the
formation of oxygen vacancies to compensate the
cation deficit, which means that an excess of
cation A is effectively just a deficit of cation B,
and vice-versa. Oxygen vacancies cause the
regional cations to move towards them, which
tilts the local octahedra and thus expands the net
lattice constants,
383
illustrated in Figure 4.4.
The relationship between nonstoichiometry and
degree of expansion has been both measured
experimentally and modeled extensively,
386-388

with the standard approach reporting the lattice
expansion as a function of oxygen vacancy
radius.
383
This logically extends to the observed
principle that the expansion influence of the B-
site is (typically) twice that of the A-site.

Figure 4.4 Illustration of octahedral tilt
accompanying oxygen vacancies in cubic
perovskites, thereby causing lattice expansion.
Image: (Marrocchelli et al, 2015).
120
4.4 Homoepitaxial SrTiO
3
Growth-Parameter
Series

The first step in developing quantitative Auger is standardization. As an initial metric it
was assumed that the SrTiO
3
substrates purchased from Crystec were stoichiometric.
However, there remained the requirement of additional references to calibrate the
quantification, which meant nonstoichiometric SrTiO
3
. While other materials which had
known stoichiometric quantities of Sr and Ti (which were not SrTiO
3
) could technically
work, complications could still arise due to inconsistencies between the samples,
structures, or surfaces. Furthermore, there are few stoichiometric compounds containing
Sr and Ti which are not SrTiO
3
, and finding one equitable to a thin film would indeed be
challenging. As such, the simplest solution would be to deliberately grow
nonstoichiometric films and characterize them. This satisfied key requirements of
quantitative Auger: the ability to use consistent temperature, pressure, and surface quality
for spectral analysis.
Growing nonstoichiometric films is easy. As explained in prior sections, film composition
can be highly sensitive to growth conditions, and the challenge would actually be growing
a stoichiometric film. I grew nine thin films with varying laser fluences and growth
pressures, to create a series that could be used for quantification. All growth parameters
aside from those deliberately changed were kept constant, and the pressure used during
Auger spectral acquisition was made consistent across all growths. The substrates were
(001) SrTiO
3
purchased from Crystec;
cleaned and then annealed at 1100℃ for
3 hours in 100 SCCM flowing O
2
to
create a smooth surface, shown in
Figure 4.5.
I monitored the composition of 40 nm
thick homoepitaxial SrTiO
3
thin films
with in situ AES grown at 75 mm target-
substrate distance and 850℃. The laser
used a static spot size of 2.7 mm
2
, and
fluence adjustment was made solely by
changing the energy of the laser output
with the addition or removal of fused
quartz attenuators in the optical path.
The resulting fluences were 0.8, 1.8, and
2.8 J cm
-2
, which are typical low,
medium, and high fluences for our
Figure 4.5 Excellent SrTiO3 surface showing atomic steps
after etch treatment. Vertical scale is 5 nm. Image: the author.
121
growth chamber, respectively. The growth pressures used were 10
-4
, 2 x 10
-3
, and 5 x 10
-2

mbar, and were controlled by adjustment of O
2
flow rate and outlet area using a mass
flow controller and butterfly valve, respectively. Film thicknesses were determined by
monitoring in situ RHEED oscillations during the growth using an electron source
operating at 5 kV, which was also the source used for AES.
With the film-series in hand, another technique was required to reliably characterize their
compositions, in order to use them as a standard reference for quantification. There are
many options available for compositional characterization, but few that were suited for
this purpose. This was because most compositional characterization techniques usable for
40 nm thin films require their own standards for quantification due to inconsistencies in
samples and instrumentation. However, X-ray diffraction is an effective method because
there is an abundance of literature discussing the correlation between c-axis lattice
expansion of homoepitaxial SrTiO
3
thin films and their compositions.
378,389

4.4.1 X-Ray Diffraction
Thin film XRD was conducted on the nine homoepitaxial SrTiO
3
films, focusing on the
out-of-plane (002) peak. The XRD scans are plotted for comparison in Figure 4.6, with
fluence and pressure indicated, and substrate and film diffraction peak positions marked.
The obvious trend in film (002) peak position compared to the substrate shows that the
films grown with a medium laser fluence of 1.8 J cm
-2
have the least c-axis lattice
expansion, indicating that their compositions are closest to that of the substrate. The high
laser fluence (2.8 J cm
-2
) films have moderate expansion, while the low laser fluence (0.8
J cm
-2
) films demonstrate the largest deviation in (002) peak position. The films grown
with low laser fluence showed the largest variation in lattice expansion as a function of
pressure. At high fluences, although the expansion was small, nominally the pressure
dependence had an opposing trend to the low fluence series. The lattice parameter showed
the least deviation from the bulk SrTiO
3
for films grown with medium fluence and low
pressure.
122

Figure 4.6 XRD and AES E*N(E) of (001) oriented homoepitaxial STO thin films. The primary substrate (002) peak
at 2θ = 46.51° (lightly dashed line) is used as a reference to determine c-axis expansion of the thin films by comparing
its position to the lower 2θ (002) peak position (heavily dashed line). A clear trend can be observed in which the medium
laser fluence films (1.8 J cm
-2
) have the least lattice expansion, while the low laser fluence films (0.8 J cm
-2
) have the
greatest, and higher pressure exacerbates the low laser fluence trend. Comparing the SrMNN and TiMVV Auger spectra
shows a similar visible trend, in which higher laser fluence increases the intensity of the TiMNN peak, while lower laser
fluence and higher pressure increase the intensity of the SrMNN peak. Image: the author.

4.4.2 Auger Spectra
The Auger spectra were collected from each film after deposition at a pressure of 10
-4

mbar, for consistency. The E*N(E) spectra of the Ti
MVV
and Sr
MNN
peaks are plotted as
insets with their corresponding XRD scans in Figure 4.6. These peaks were chosen
deliberately for their low energies (~82 eV for Sr
MNN
and ~34 eV for Ti
MVV
) which have
greater surface-sensitivity than their higher energy counterparts. The plots shown are each
10 summed spectra, with collection times of 20-30 seconds per scan. The intensity of the
Auger peaks correlates with composition, as peak signal is proportional to quantity of the
species present. A simple, qualitative, inspection of the Auger peak intensities describes a
relationship between growth parameters and composition in agreement with expectations.
The intensity of the Sr
MNN
peak increases with decreasing laser fluence or increasing
pressure. Likewise, the intensity of the Ti
MVV
peak increases with increasing laser fluence.
These trends indicate that the Sr/Ti ratio scales proportionately with pressure but
inversely with fluence.
To better understand the observed trends in composition as a function of laser fluence
and growth pressure, the collected AES results were quantified to qualitatively compare
the composition information. Peak-to-peak values, representative of the composition, were
calculated from the summed Auger spectra for each set of growth parameters and
compared with the c-axis lattice constant calculated from the XRD data, summarized in
Figure 4.7. Demonstration of P2P calculations are illustrated in Chapter 3. The
correlation between Sr/Ti ratio and c-axis expansion for the lowest laser fluence series is
clear, with all films showing a higher Sr/Ti ratio than the substrate and corresponding
123
increase in lattice constant. The middle laser fluence films show a tendency toward Sr-
richness with increasing pressure and an equivalently slight increase in lattice constant.
The high laser fluence films show Ti-richness with higher pressure moving the Sr/Ti ratio
towards stoichiometric, and the lattice constant likewise decreasing towards that of the
substrate. Overall, the in situ AES signal indicates that film composition varies from ideal
stoichiometry following the gross trends reported in the literature, corroborated by c-axis
lattice expansion. They show that one could find an optimal set of fluence and pressure
conditions to grow metal-stoichiometric SrTiO
3
.

Figure 4.7 Auger signal intensity quantified as the ratio of the summed SrMNN peak-to-peak
value to that of TiMVV, for nine homoepitaxial SrTiO3 thin films grown with the indicated
laser fluences and pressures (bottom), as well as their c-axis lattice constants as derived from
thin film XRD (top). Horizontal dashed lines indicate the Sr/Ti ratio and c-axis lattice
constant of the single crystal SrTiO3 substrate. Image: the author.
124
4.5 Direct Observation of Parameter
Dependence

To develop a more comprehensive understanding of the compositional changes occurring
during the deposition process, a homoepitaxial SrTiO
3
thin film was monitored with in
situ AES while deliberately changing growth parameters to control composition. From the
results of the pressure/fluence series presented in Figure 4.6 and 4.7, the initial growth
parameters were chosen to maintain the same composition as the substrate in order to
verify my ability to do so. SrTiO
3
was deposited two unit cells at a time, between which
Auger spectra were collected for compositional analysis without altering any parameters
of the growth. The deposition occurred with a substrate temperature of 850°C in 10
-4
mbar
flowing O
2
. The Auger spectra were summed, and P2P values calculated, with Sr/Ti ratio
as a function of overall film thickness and laser fluence presented in Figure 4.8. After
the deposition of 20 unit cells of stoichiometric homoepitaxial SrTiO
3
using a laser fluence
of 1.8 J cm
-2
, fused quartz attenuators were placed in the laser path to reduce the laser
fluence to 0.8 J cm
-2
, making the deposition composition Sr-rich. The Sr/Ti ratio increased
linearly until saturation after approximately 16 unit cells, and remained steady during the
Figure 4.8 Sr/Ti ratio of the P2P calculated from summed Auger spectra collected after the deposition of every two
unit cells of homoepitaxial SrTiO3, as a function of the number of deposited layers. The laser fluence was changed
periodically during the deposition, as indicated by the shape and color of the markers. A single deposition of two unit
cells of TiO2 (equivalent to two half unit cells of SrTiO3) was performed after depositing 70 unit cells of SrTiO3,
demarcated as a star. The original composition of the SrTiO3 substrate, as determined with AES, is indicated with a
thin dashed gray line. The results of a parameter-free escape depth model used to explain the compositional variance
as a function of deposition is shown as thick dashed lines. Image: the author.
125
deposition of an additional 18 unit cells. The laser fluence was then increased back to 1.8
J cm
-2
by removing the attenuators, and an additional 16 unit cells were deposited during
which the Sr/Ti ratio steadily decreased, but did not return to the original signal ratio of
the substrate. At this point, two unit cells of TiO
2
(equivalent to two half unit cells of
SrTiO
3
) were deposited, bringing the Sr/Ti ratio below that of the substrate, indicating
Ti-richness. An additional 28 unit cells of SrTiO
3
were then deposited with the same laser
fluence of 1.8 J cm
-2
, bringing the Sr/Ti ratio back to that of the substrate, demonstrating
a return to the original composition after a total of 100 unit cells of SrTiO
3
.
If a homoepitaxial deposition has an off-stoichiometry plume, the resulting film must
somehow account for the excess species. While some of the literature provides explanations
for the manner in which the film incorporates an excess species (or creates a deficit),
389
many do not.
96,190,378
Regardless, the specifics will depend on the material and kinetics of
the growth event in question. For SrTiO
3
, plausible scenarios to account for small amounts
of excess Sr or Ti are the formation of interstitials and/or vacancies, or other types of
point and line defects.
349,390
However, the trends observed in the Sr/Ti ratio of the growth
shown in Figure 4.8 cannot be accounted for by bulk phenomena alone.
Referring to the IMFP calculations in §3.4, the energy of the Auger peaks observed (~82
eV for Sr
MNN
and ~34 eV for Ti
MVV
) should have mean free paths of approximately 0.52
and 0.42 nm for Sr
MNN
and Ti
MVV
, respectively. Assuming attenuation with typical
exponential decay, the signal associated with a shift in stoichiometry should plateau after
approximately 5 unit cells of deposition, just as it did in §3.4 (Figure 3.14). However, this
is clearly not observed in this film. The initial explanation for the impact of growth
parameters on composition is then no longer valid, and an alternative explanation must
be developed.

126
4.6 Parameter Free Escape Depth M odel

To explain the behavior of the Auger intensity observed in the growth shown in Figure
4.8, a model was developed using the principle of IMFP set forth in §3.4, and explained
in detail in §2.7. The essential premise of the model is that we should be able to predict
the relative intensities of Auger signal as a function of depth within the heterostructure.
By applying the model to the structures grown, we should be able to determine the actual
structure based on observed signal.

4.6.1 Principle and Description
If a deposited material is not homogeneous, as is the case with the deliberately
nonstoichiometric film grown in Figure 4.8, the signal will behave somewhat differently
than it would otherwise because it will change as a function of depth. To illustrate this
principle with the parameter-free escape depth model, consider Table 10, where the
predicted Auger signals for Ti
MVV
and Sr
MNN
are based on their calculated 𝜆 using
Equation 3.3, and the model outlined in §2.7 for the ‘bulk’ case.
Table 10 Predicted Auger signal intensity for TiMVV and SrMNN peaks as a function of layer using the parameter-free
escape depth model with the ‘bulk’ case (left), and the same model applied with 5% excess strontium in the highlighted
cells for 2 unit cells (middle), and 4 unit cells (right). Note the decrease in signal increase as a result of the exponential
decay of the escape depth.

127
Beginning with stoichiometric SrTiO
3
10 unit cells thick, the contributions from the
individual layers can be calculated with the equations in §2.7, assuming that the initial
signal from each layer is the same (I
0
= 1). The calculated contribution from each layer
and their sum is shown in Table 10 (left).
If 2 unit cells of 5% Sr-rich SrTiO
3
are deposited on top of the stoichiometric SrTiO
3
, the
resulting signal contributions, Table 10 (center), will change accordingly and the resulting
net Sr/Ti signal will increase by ~3.8%. However, with an additional 2 unit cells of Sr-
rich SrTiO
3
, the resulting signal contributions, shown in Table 10 (right), do not increase
as dramatically because the electrons from the buried layers are attenuated by the layers
above them, resulting in a gain of only an additional 0.9%.
When this model is applied to the previous nonstoichiometric growth, shown in Figure
4.8, it is clear that this bulk stoichiometric shift is not a suitable fit for the observed
intensity shift. Furthermore, if the model is adjusted to alter how the nonstoichiometry is
incorporated into the bulk it does not make a significant difference. The alternate ways
the nonstoichiometry can be incorporated in these models is by decreasing Ti along with
the increase in Sr, either by half or the full amount of Sr (Sr:Ti = 1.1:0.95 and 1.1:0.9).
This would be assuming that either Ti
4+
deficiency is the primary compensation for the
increase in Sr
2+
, or the excess compensation is accounted for with oxygen vacancies,
respectively. It must be noted that the only difference between these scenarios is in the
relative scale of Sr/Ti. When compared to the observed data the models must be
normalized, as their numeric values only have relative significance, and they appear
identical as a result. This means that the only way to account for the observed Auger
signal is with a completely different scenario.

4.6.2 Sensitivity to Termination
If the model is reexamined as an application to (001) SrTiO
3
, the stacking sequence of
alternate layers will influence the net signal, as explained in §2.7. If we apply this
termination-sensitive model to the Sr-rich example shown in Table 10, it is clear that the
results are effectively unchanged because, as was the case for the alternate
nonstoichiometric scenarios described, this scenario does not actually change the growth
event. This means that the most reasonable explanation for a shift in signal which occurs
over a duration greater than 5 unit cells is a shift in composition which takes place over
the same period of time, and the simplest explanation to account for this sort of shift in
composition is surface termination.
Consider Sr-rich SrTiO
3
deposited on a TiO
2
terminated substrate. After a single
monolayer, what happens to the excess Sr? The lowest energy configuration would likely
128
be sitting on the surface, and as layers continue being deposited the termination would
change incrementally with the excess Sr. If this scenario is examined with the ‘bulk’ model,
the result does not change. However, if it is approached with the ‘termination’ model, this
gentle shift in termination will be observable in the Auger signal.
Using this modeling method to examine the relative signal intensity shifts during thin film
growth as a result of termination switching events simply requires selecting appropriate
lattice constants for the species being grown and then counting the layers present.
Therefore, when switching from TiO
2
terminated SrTiO
3
to SrO termination using Sr-rich
deposition, the calculated values are shown in Table 11.
Table 11 Predicted Auger signal intensity for TiMVV and SrMNN peaks as a function of layer using the parameter-free
escape depth model with the ‘termination case for TiO2 termination (left) and SrO termination (right).

129
4.7 Termination Analysis

When the ‘termination’ model from §2.7 is applied to the previous nonstoichiometric
growth with incremental termination switching, now shown in Figure 4.9, we can
compare the two models. In the model which does not account for surface termination,
the ‘bulk’ model, deposition of Sr-rich SrTiO
3
would increase the Sr/Ti ratio exponentially
and then plateau, reaching saturation (99% of maximum signal) after the deposition of
only five unit cells. Likewise, when stoichiometric SrTiO
3
was then deposited on top of
the Sr-rich layers, the Sr-rich signal would again be masked after only five unit cells. If,
at that point, an additional layer of TiO
2
was deposited the Sr/Ti signal would drop
immediately and then recover as the double-layer of TiO
2
was buried. However, if we
apply a model accounting for surface termination, the ‘termination’ model, deposition of
Sr-rich SrTiO
3
will slowly increase the Sr/Ti signal as a function of excess Sr. After
completing the SrO terminating layer, the signal would then increase further, dependent
on the manner in which the excess Sr is incorporated into the film. If the composition of
the film after switching to SrO termination is simply Sr-rich in the same proportion as
that indicated by the rate of termination-switching, the signal would increase slightly then
plateau. Returning to stoichiometric deposition, the Sr/Ti signal would then decrease as
the excess Sr layers are buried, while maintaining the SrO termination. The deposition of
TiO
2
is then used to force the termination back to that of the initial substrate. At this
point, deposition of stoichiometric SrTiO
3
will bury the double TiO
2
layer while
maintaining TiO
2
termination. During deposition on the original and final TiO
2

terminated surfaces the composition does not appear to change, which is in agreement
with a stoichiometric deposition.
Comparison of the two models against the data in Figure 4.9 clearly shows that the
termination model provides a better fit, with χ
2
= 0.011, versus χ
2
= 0.370 for the bulk
model. Additionally, the rate of termination conversion observed with Sr-rich deposition
indicates a Sr/Ti compositional ratio of ~1.06. These results demonstrate how effective
the escape depth model can be at predicting intensity and explaining observed phenomena.
130

Figure 4.9 The data from the growth originally shown in Figure 4.8, in comparison to the parameter-free escape
depth model that accounts for termination. The termination model is clearly a better fit to the data, with χ
2
= 0.011,
versus χ
2
= 0.370 for the bulk model. Image: the author.
131
4.8 Quantification

Applying the same escape depth model outlined in §2.7 to determine Sr/Ti signal ratios
for thick film compositions yields a method for quantification. The assumption for these
signal ratios is based on expected intensity for post-surface-change nonstoichiometric
deposition. The results find agreement with the experiment shown in Figure 4.9. When
this quantification is applied to the Sr/Ti signal ratios of the films shown in Figure 4.6
and 4.7, a clear trend in composition and lattice parameter can be observed, presented in
Figure 4.10. Furthermore, the compositions and lattice parameters reported here are
consistent with values found in the literature.
385,389
These results indicate that the Auger
probe used to collect in situ spectral data in these experiments is not only capable of
providing gross compositional information, but with appropriate calibrations also capable
of quantifying in situ film composition.

Figure 4.10 Normalized Sr/Ti Auger signal versus c-axis lattice constant for the homoepitaxial SrTiO3 thin films
presented in Figure 4.6 and 4.7, also plotted as excess Sr quantified from the parameter free escape depth model outlined
in the text. For comparison, composition and c-axis data have been plotted from the literature,
385,389
showing agreement
with our reported trends. Image: the author.

132
4.9 Summary

Film sensitivity to thin film growth parameters was explored in terms of laser fluence and
background pressure. The complexity of teasing out specific correlations from the
complicated interrelationships of these systems was discussed. The impact of
nonstoichiometry on crystal structure was identified as a means of compositional
characterization, and applied to a series of homoepitaxial SrTiO
3
thin films deposited with
varying growth parameters. Thin film XRD and Auger spectra were compared, verifying
the existence of expected trends in compositional-dependence on deposition parameters.
An additional film was grown to demonstrate the deliberate modification of composition
by adjusting growth parameters, however the observed Auger signal response was
unexpected and could not be explained with a simple bulk escape depth model. By altering
the escape depth model to account for termination, the phenomenon observed in the data
was explained as the incremental shift in termination with nonstoichiometric deposition.
Applying the same model to the prior growth-series, the Auger signal ratios are quantified
and compared to c-axis lattice expansion, demonstrating agreement with expected values
based on the literature. These results validate this combination of Auger analysis and
signal modeling for compositional quantification.

133

Chapter 5
Direct Observation and Control of
Surface Termination

5.1 Introduction

In this chapter I elaborate on the relationship between thin film composition/structure
and properties, using classic examples to demonstrate this principle. Dynamic thin film
growth phenomena, wherein the expected film structure changes due to system energetics,
are discussed in principle and within the context of their observation. Surface termination
switching is explored, using the p-type LaAlO
3
/SrTiO
3
heterostructure and SrRuO
3
as
examples. Surface energy calculations of SrTiO
3
, SrRuO
3
, and their heterostructures are
introduced, and the results discussed. The design of thin film growths exhibiting deliberate
control of surface termination is introduced, and then executed by first using Auger
spectra collected every ¼ monolayer, then in pseudo real time using the pulse probe
method introduced in Chapter 3. The Auger data is analyzed and compared to parameter-
free escape depth models of the proposed structure both with and without termination-
switching, demonstrating that termination is indeed observed and controlled with this
method. The premise of this experiment is then extended beyond homoepitaxial SrTiO
3
,
to an SrZrO
3
/SrTiO
3
heterostructure, illustrating its flexibility.

134
5.2 Compositional and Structural Sensitivity of
Properties

Understanding process/structure/property relationships is the heart of materials science.
The influence of growth parameters on composition (process/structure) was explored in
Chapter 4, and I will now further investigate the structure/property relationship. To
begin, it is important to clarify that the interdependence of composition, atomic structure,
and electronic structure is a false trichotomy: these three properties are, in fact, the same.
This is because you cannot change one without influencing the others, and the
manifestation of that change will always be consistent. Altering the composition of a
material does not change the atomic lattice and electronic structure; altering the
composition of a material is altering the atomic lattice and electronic structure. However,
their manifestation as material properties does not share this quality, because the
properties are emergent and can arise from various configurations of composition and
structures.
The dependence of a material property on composition and structure is as complex as the
property itself. This statement is a tautology, yet it still illustrates the foundational
principle of the structure/property relationship: the gradation inherent to a property is
the result of the physics of its formation, and can therefore be manipulated within the
bounds of that formation. This is exemplified in the difference between material properties
that are the product of quantized states and those that are not. For instance, let us
compare transparency and resistivity. Transparency is the proportion of photons of a
specific wavelength that will pass through a material (be transmitted) without scattering.
Resistivity determines the ease with which electrons are able to travel through a material.
Both transparency and resistivity have an infinite range of possible values, yet resistivity
is countably infinite while transparency is uncountably infinite. This is, in effect, because
transparency is a probability while resistivity is not. Something can always be less
transparent without being opaque (0.1, 0.01, 0.001, …), but resistance is grounded in the
quantization of Hall conductance,
391
and therefore has fundamental limits. The
significance of this comparison is in how we approach materials design and the
structure/property relationship.
Understanding the effect that structure has on properties is crucial to the design and
engineering of thin films and heterostructures because, in the act of depositing films, what
we are trying to control is the structure. The properties we observe emerge from the
structures we create. This means that we are either trying to engineer the emergence of
something that didn’t exist before (uncountable), or we are trying to manifest an existing
property (countable). The distinction is important because it influences how we think
about, and try to solve, these types of problems. An example that illustrates different
135
approaches to applying the structure-property relationship is the LaAlO
3
/SrTiO
3

heterostructure.
The LaAlO
3
/SrTiO
3
heterostructure is famous for showing metallic 2D conductivity at
the interface between two wide bandgap insulators.
37
Though the precise reason why this
occurs is still under debate,
336-360
there is a general consensus that it is a product of the
polar/non-polar interface. As discussed previously, (001) SrTiO
3
can be thought of as
alternating layers of TiO
2
and SrO, with each layer maintaining charge-balance, illustrated
in Figure 5.1 in comparison to LaAlO
3
, which will have alternating layers of unbalanced
LaO
+
and AlO
2
-
, forming a polar structure with orientation dependent on the terminating
layer. When LaAlO
3
is grown on TiO
2
terminated SrTiO
3
it creates a charge imbalance
which increases as the LaAlO
3
grows thicker. At the critical thickness of four unit cells,
the polar catastrophe is averted by balancing this charge imbalance with a two-
dimensional electron gas at the LaAlO
3
/SrTiO
3
interface.
This discovery created a great deal of excitement about the prospect of all-oxide
electronics, and has inspired a tremendous amount of research to understand the formation
of the interface in order to promote the development of new devices.
24,25,35,40
However,
despite these efforts there is still a great deal of debate as to how much influence different
factors play in the formation of the interface and its resulting properties. For instance,
the addition of oxygen vacancies can dramatically increase the carrier density of this
structure,
156,392
but are inherently unstable and thus impractical for some applications.
The role of the polar catastrophe has been explored with superlattices and
heterostructures,
346,365
and the degree of polarization has been attempted to be measured
Figure 5.1 An illustration of the polar/nonpolar LaAlO3/SrTiO3 interface, showing the resulting formation of a 2-
dimensional electron gas. Image: the author.
136
with scanning tunneling microscopy,
351
gating,
358
and other methods.
393
Cation
intermixing has been offered as an explanation,
336
as has nonstoichiometry.
96

CaZrO
3
/SrTiO
3
heterostructures have also demonstrated the formation of 2DEGs, but
the explanation for their formation was structural-induced polarization,
394
and is similar
to structural arguments made about LaAlO
3
/SrTiO
3
.
357
What this brief survey ultimately
demonstrates is that this system has enough inherent flexibility that trying to spin a single
origin story for 2DEG formation is foolhardy. The conducting LaAlO
3
/SrTiO
3
interface
has complex, multifaceted, origins, which is also why there is such disparity in published
properties of the films. The interface is a non-binary spectrum of conduction, rather than
an on-off switch, and this quality must be accommodated for in its application to designs.

137
5.3 Dynamic Thin Film Growth Phenomena

With the inherently complicated nature of thin film growth, it should come as no surprise
that unexpected phenomena are frequently observed. When the structural design of a thin
film is not energetically preferable, it results in either the formation of metastable phases
(like diamond), or it reorganizes to a lower energy state (like graphite). With perovskite
oxides, epitaxial thin films are typically considered metastable because they are strained
due to lattice mismatch, but the restoring force is not sufficient to overcome the energy
barrier to relaxation. In some cases, however, the metastability fails and spontaneous
reorganization occurs, referred to here as dynamic thin film growth phenomena. A classic
example is the spontaneous layer rearrangement of Ruddlesden-Popper SrTiO
3
.
120

Ruddlesden-Popper phases are perovskite-related structures
with the general formula A
n+1
B
n
X
3n+1
, illustrated in Figure
5.2. They form two-dimensional perovskite-like slabs
intercalated with cations, forming a structure like a
combination of perovskite and rocksalt phases, where n is the
number of octahedral layers between the perovskite-like slabs.
These materials are similar to perovskites, and can possess
similarly interesting and useful properties.
174,395
However, the
synthesis of these compounds, especially in thin film form, has
proven challenging because many of them are
thermodynamically unstable and often exhibit intergrowth
defects.
120
To better understand the relationship between the growth
process and the stability of the phase-structure, a study was
conducted growing Sr
2
TiO
4
(n = 1) with an MBE system at
the Advanced Photon Source. This allowed in situ synchrotron
X-ray scattering surface analysis, which revealed that the
layered oxide films underwent dynamic layer rearrangement
during growth, resulting in completely different structures than
were intended or deposited.
This demonstrates how reconstruction can be caused by the energetics of the larger
structure, causing a complete rearrangement of entire planes of atoms to minimize energy.
This is an extreme case, but still applicable to superlattices and heterostructures which
are relying on interactions between multiple layers.
119,145
The overall energetics of complex
structures are challenging to predict, and it is not unfathomable that this sort of event
could occur without realizing it during deposition.
Figure 5.2 The Ruddlesden-
Popper structure of Sr3Ru2O7.
Image: the author.
138

5.4 Surface Termination

The high-energy states of surfaces make them active and interesting, and is why they can
be representative of the larger energy picture through reconstruction or other
compensation mechanisms. It should not be surprising then that surfaces are the ideal
location for energy-driven spontaneous reorganization.
87
The increased energy state
provides a lower energy barrier for dynamic events than would be encountered in the bulk,
and this is why we are interested in these surface phenomena.

5.4.1 LaAlO
3
/SrTiO
3
To summarize the earlier discussion of LaAlO
3
/SrTiO
3
: the 2DEG formation at this
interface is believed to be caused, at least in part, by electronic reconstruction to counter
the polar discontinuity.
357
This is observed when the polar/non-polar interface is
TiO
2
/LaO
+
, and the LaAlO
3
is at least 4 unit cells thick. According to the reconstruction
theory, however, if the opposite interface is created (SrO/AlO
2
-
) it should be compensated
by the formation of an equivalent two-dimensional hole gas, illustrated in Figure 5.3.
However, with very limited reported exceptions,
365
this has not been observed. Theories
as to why are varied, and range from compensation by oxygen vacancies to surface
contamination.
A recent study investigating
this phenomenon was
attempting to form the 2DHG
without allowing surface
compensation, but noticed
that the LaAlO
3
surface
termination was not as
expected.
119
Using pseudo in
situ angle-resolved XPS and
time-of-flight ion scattering
and recoil spectroscopy, the
termination was determined
to be AlO
2
-
, when the stacking
sequence with a SrO/AlO
2
-

interface should result in
LaO
+
termination, as shown
Figure 5.3 Ionic model behavior of the LaAlO3/SrTiO3 interface for either
a TiO2/LaO
+
interface (top) or a SrO/AlO2
-
interface (bottom), with
compensating surface gas formation. Image: the author.
139
in Figure 5.3. This discrepancy was explained with DFT calculations by observing that
the AlO
2
-
termination served to compensate the polar discontinuity without the formation
of a 2DHG.
This demonstrates how atomic reconstruction can occur as the result of electronic
reconstruction, rather than the other way around. Beyond the obvious disruption of the
planned 2DHG, this type of event can cause significant havoc in the planned design and
growth of heterostructures. Furthermore, it is very unlikely that this phenomenon is
limited to this system, as polar/non-polar interfaces are not uncommon, and there have
been reports of strain-induced electronic reconstructions of the same type.
394

5.4.2 SrRuO
3
Approached with the same layered-perspective as SrTiO
3
, SrRuO
3
is alternating sheets of
SrO (just like SrTiO
3
) and RuO
2
. What’s interesting about SrRuO
3
, however, is that there
is a significant divergence in the surface energy of the two terminations: SrO is much
lower energy, and thus more stable, than RuO
2
. The result is that when thin films of
SrRuO
3
are deposited, they always come
up SrO terminated. For most materials,
if the deposition is stoichiometric,
epitaxial growth will yield a surface with
the same termination as the starting
surface, be it AO, BO
2
, or mixed.
This was first reported in 2004,
146
when
it was noticed that the first unit cell of
SrRuO
3
deposited on a TiO
2
terminated
substrate took an extra half-oscillation
of time, observed through RHEED,
shown in Figure 5.4. Every subsequent
oscillation was of normal duration, and
when SrRuO
3
was grown on an SrO
terminated SrTiO
3
substrate the first
unit cell of deposition showed a normal
oscillation duration, as did subsequent
oscillations. The interpretation of this
phenomenon was that the extra half-
oscillation of duration was required to
deposit the extra half unit cell, resulting
in an SrO terminated surface.
Figure 5.4 RHEED intensity oscillations for SrRuO3 films
grown by pulsed laser deposition, showing an extra half-
oscillation when growing on a TiO2 terminated surface versus
a SrO terminated surface. Image: (Rijnders et al, 2004).
140
This demonstrates two important aspects of surface and film energetics. The first is that
this sort of dynamic event is possible: where what should happen simply does not because
the system is more complex than initially understood. The second is that without a
comprehensive understanding of the state of the system during deposition, you probably
don’t actually know what’s going on in there. Consider, for instance, if SrRuO
3
was
deposited between layers of SrTiO
3
before LAO was grown on top. The expected TiO
2

terminated SrTiO
3
would actually be SrO terminated, and the interface would not form
the anticipated 2DEG. Without knowing that SrRuO
3
was undermining your termination-
control, you would likely have no idea why the 2DEG didn’t form. However, if the
phenomenon is known to occur it can be used to deliberately control the surface
termination, and thereby engineer the nature of thin film interfaces.

141
5.5 Surface Energy Calculations

To better understand the surface energetics of SrTiO
3
, SrRuO
3
, and their heterostructures,
we used DFT calculations to develop an understanding of the energetics of termination
switching observed experimentally. The results in Figure 5.5 show that in bulk SrTiO
3
,
SrO-termination is more stable than TiO
2
-termination for a broad range of chemical
potentials, which is consistent with the literature.
396
In
the case of bulk SrRuO
3
(Figure 5.6), however, SrO-
termination is found to be more stable over the entire
chemical potential range where SrRuO
3
is stable.
Moreover, when thin films of SrRuO
3
are deposited on
SrTiO
3
substrates, the difference in energy between SrO-
termination and RuO
2
-termination increases over that of
the bulk SrRuO
3
, indicating that thin films tip the
energetic preference even further in favor of SrO-
termination. The stability difference between SrO- and
RuO
2
-termination in SrRuO
3
films can be understood
from their electronic structures. The layer-resolved
density of states of SrO-terminated SrRuO
3
on SrTiO
3

substrate (Figure 5.7) shows the presence of a lower
density of states around the Fermi energy compared to
RuO
2
-termination
(Figure 5.8), signifying
the presence of a greater number of dangling surface bonds
with RuO
2
termination, and thus its higher surface energy.
Further deposition of additional layers of SrTiO
3
on a
sandwiched layer of SrRuO
3
results in the stabilization of
the SrO-termination layer in SrTiO
3
instead of the
common TiO
2
-termination. Our calculations in Figure
5.9 show that SrO-termination has a lower surface energy
with reference to TiO
2
-termination over the entire range
of allowed chemical potential, in a manner similar to that
of SrRuO
3
on SrTiO
3
substrates. It is consistent with our
experimental observation that the termination layer is
switched from TiO
2
to SrO for the SrTiO
3
epilayers
deposited on SrRuO
3
. The lower surface energy of SrO-
terminated SrTiO
3
can be understood from the chemical
potential constraints of SrTiO
3
/SrRuO
3
hybrid structures.
The growth window for the formation of SrRuO
3
is
narrower than that of SrTiO
3
(Figure 5.10), and it is
Figure 5.5 (a) The slab models of SrO-
terminated and TiO2-terminated SrTiO3.
(b) Variation of surface energy of SrTiO3
with SrO and TiO2 termination as a
function of the chemical potential of SrO
(µSrO). Image: collaboration.
Figure 5.6 SrO- and RuO2-
terminated slab models of SrRuO3. (b)
Variation in surface energy of SrRuO3
with SrO and RuO2 termination as a
function of the chemical potential of
SrO. Image: collaboration.
142
expected that the chemical potential
constraints of SrRuO
3
will determine the
growth condition of SrRuO
3
/SrTiO
3

heterostructures. Due to the reduced
growth window of SrTiO
3
on SrRuO
3
, the
SrO-terminated surface of SrTiO
3
has a
lower surface energy across the entire
allowed chemical potential range. It is
different from that of pure bulk SrTiO
3
,
where a TiO
2
-terminated surface could
have a lower surface energy for a narrow
chemical potential range (Figure 5.5).

Figure 5.7 Layer-resolved, spin-polarized density of states
(DOS) of SrRuO3 on SrTiO3 substrates. The SrRuO3 film
is terminated by SrO surface. The DOS corresponding to
SrRuO3 layer is shaded in grey color and that of SrTiO3 are
shaded blue. The SrTiO3 or SrRuO3 layer corresponding to
the DOS is shown in the atomic model on the right. Image:
collaboration.
Figure 5.8 Layer-resolved, spin-polarized density of
states (DOS) of SrRuO3 on SrTiO3 substrates. The
SrRuO3 film is terminated by RuO2 surface. Image:
collaboration.
143

Figure 5.9 Model configuration of SrO-terminated and RuO2-terminated SrRuO3 films (left side) on an SrTiO3
substrate, as well as TiO2-terminated and SrO-terminated SrTiO3 films sandwiching an SrRuO3 layer (right side).
Comparison of the chemical potential-dependent surface energy of the SrRuO3- (top-center) and SrTiO3 -capped
(bottom-center) structures. Image: collaboration.

Figure 5.10 Phase diagram describing stability of SrTiO3 (a) and SrRuO3 (b) as a function of the chemical potential
of Sr (µSr) and O (µO). The phase boundaries are indicated by Sr-rich or Ti-rich (Ru-rich) conditions during growth of
SrTiO3 (SrRuO3). Image: collaboration.

144
5.6 Controlling Surface Termination in
Homoepitaxial SrTiO
3

My initial experimental objective was to observe the switch in termination precipitated
by the deposition of SrRuO
3
. If successful, however, the complementary objective would
be to demonstrate the ability to not only observe surface termination, but actively control
it as well. Doing so would, for the first time, demonstrate the ability to observe and
control surface termination in complex oxides with PLD.

5.6.1 Experimental Design
I chose to study surface termination in model heterostructures of SrRuO
3
and SrTiO
3
to
evaluate the suitability of this Auger technique to observe subtle changes in surface
composition. To demonstrate the surface-sensitivity, I synthesized a film with the
following structure: two unit cells of homoepitaxial SrTiO
3
on a TiO
2
-terminated SrTiO
3

substrate, two unit cells of SrRuO
3
used to switch from TiO
2
-termination to SrO-
termination, and a cap of two unit cells of SrTiO
3
to verify the SrO-termination. Note
that the two unit cells of SrRuO
3
used to switch termination are actually two and a half
unit cells, layered as SrO/RuO
2
/SrO/RuO
2
/SrO, as shown in Figure 5.11. I chose to
collect ten Auger spectra of each element, to improve
the signal quality, nominally after every ¼
th
monolayer
of growth. However, the growth rate would not be
observable using RHEED with this method, so a mock
growth was conducted to determine growth rate, shown
in Figure 5.12. The exact same conditions were used
for each growth (the same as reported in §4.4), on a
substrate cut in two to be used for both.
The growth observed with AES was conducted
immediately after the mock growth, to limit the
introduction of variance in conditions. The growth was
stopped every ¼ unit cell to collect spectra of the
Ti
LMM
, Sr
MNN
, and Ru
MNN
peaks, as planned, based on
the observed growth rates of the previous film. The area
under the curve were calculated from the sum of the
spectra collected at each point. It should be noted that
¼ unit cell of deposition, based on the data collected,
Figure 5.11 Illustration of the
heterostructure grown with termination
switching. Image: the author.
145
corresponds to 1 unit cell coverage of approximately ¼
the surface, rather than ¼ unit cell coverage over the
entire surface.
For the next growth, I demonstrated the ability to probe
and control the surface termination switching in situ and
in real time using the pulse-probe technique described in
§3.5. For this growth, I used TiO
2
-terminated SrTiO
3

and grew, in chronological order, four unit cells of
SrTiO
3
, one unit cell of SrRuO
3
, four unit cells of SrTiO
3
, one monolayer of TiO
2
, and
eight unit cells of SrTiO
3
, shown in Figure 5.13. The one unit cell of SrRuO
3
results in
SrO-termination, and the termination is then switched back to TiO
2
-termination with the
deposition of a single monolayer of TiO
2
. The area under the curve were calculated for
each scan, using the Ti
MVV
and Sr
MNN
peaks, exactly as was done for the previous scan.
To verify the grown structure, collaborators used aberration-corrected STEM imaging and
EELS chemical mapping to ascertain the layer-by-layer growth and composition of the
heterostructure. Figure 5.14 shows a high-angle annular dark filed (HAADF) image of
the heterostructure. The heterostructure consists of homoepitaxial SrTiO
3
on a TiO
2
-
terminated SrTiO
3
substrate, two-unit cells of SrRuO
3
used to switch from TiO
2
-
termination to SrO termination, ~two-unit cells of SrTiO
3
, and a capping layer of LaAlO
3
that was added to protect the SrRuO
3
and SrTiO
3
overlayers from potential damage
during TEM specimen preparation. We have performed EEL spectroscopy for the region
highlighted in Figure 5.14a with a white box, to map the distribution of Ti, O and La.
The elemental maps for Ti L edge, O K edge and La M edge are shown in Figure 5.14b.
The Ti L edge map shows a dip in the intensity for the ~2 unit-cell SrRuO
3
layer followed
by an increase in the intensity for the ~2 unit-cell-thick overlayer of SrTiO
3
. The
demarcation between the LaAlO
3
capping layer and the SrTiO
3
overlayer can also be seen
from the elemental maps of Ti L edge and La M edge. Figure 5.14c shows the extracted
Figure 5.133 Illustration of the second
heterostructure grown with termination
switching. Image: the author.
Figure 5.12 RHEED oscillations for the ¼ unit cell growth conducted
as a ‘dummy’ deposition to determine growth rate. Image: the author.
146
EEL spectra for regions highlighted in Figure 1a with the same color. Ru L edge at 2838
eV is noisy, which makes accurate quantification of the SrRuO
3
layer thickness without
causing too much beam-damage challenging.

Figure 5.14 (a) Wide field-of-view HAADF image showing the LaAlO3 capping layer, SrRuO3 layer and SrTiO3
substrate. Scale bar corresponds to 2 nm. (b) Elemental maps for Ti L edge, O K edge and La M edge, for the region
highlighted as white box in (a). Each elemental map is normalized within itself. (d) Extracted EEL spectra for Ti L
edge, O K edge and La M edge, where the color of each spectrum corresponds to the region of same color highlighted
as boxes in (a). Image: collaborators.
The layer-by-layer structure of the heterostructure can be further inferred by comparing
the intensity of the atomic columns in the HAADF image. Figure 5.15 shows an atomic
resolution HAADF image of the perovskite oxide heterostructure. Since, the intensity in
a HAADF image is approximately proportional to the squared atomic number (~Z
2
) of
the atomic column,
1
each atomic column can be labelled based on the atomic number of
the elements, given that the atomic density remains constant. For the [100] orientation of
the perovskite structure, both the A-site and B-site cations have the same atomic density,
therefore, each atomic column can be directly labelled based on its atomic number. In this
image, La (Z = 57) atomic columns appear brightest followed by Ru (Z = 44), Sr (Z =
38), Ti (Z = 22) and Al (Z = 13). The oxygen atomic columns are not visible due to the
limited dynamic range of the detector. Based on the HAADF intensities, LaAlO
3
, SrRuO
3

and SrTiO
3
layers, along with the TiO
2
, SrO and RuO
2
layers are illustrated in Figure 2a.
Furthermore, for the part of the HAADF image shown in Figure 5.15b, we have shown
the intensities profiles of A and B-site cations in Figure 5.15c. La, Sr and Ti atomic
columns can be clearly distinguished in the HAADF intensity profiles. However, Ru
(Z=44) atomic columns have a lower HAADF intensity than Sr (Z=38) atomic columns.
This lowering of the HAADF intensity for Ru atomic columns is a result of intermixing
147
between Ru (Z=44) and Ti (Z=22) atoms. The atomic resolution HAADF image analysis
confirms the growth of two-unit cells of SrRuO
3
used to switch from TiO
2
-termination to
SrO termination on a TiO
2
-terminated SrTiO
3
substrate.

Figure 5.15 (a) Atomic resolution HAADF image showing the oxide heterostructure with LaAlO3 capping layer,
SrRuO3 layer and SrTiO3 substrate. Scale bar corresponds to 1 nm. (b) A part of the HAADF image chosen for
intensity analysis. (c) The A and B-site HAADF intensity profiles for the regions highlighted in (b). Image:
collaborators.
148
5.6.2 Data and Analysis
The intensity values for the first growth, with Auger spectra collected every ¼ unit cell
beginning with the pre-growth substrate, are plotted as a function of film thickness in
Figure 5.16. The Sr/Ti signal-ratio remains approximately constant, and consistent with
the substrate, during the first two monolayers of deposition. During the subsequent
deposition of two monolayers of SrRuO
3
, the Sr and Ru signals increase while the Ti signal
decreases exponentially, as would be expected with the corresponding increase/decrease
in Sr, Ru, and Ti content relative to the surface of the film. The last two monolayers of
deposited SrTiO
3
show a sustained higher level of Sr signal, decay of the Ru signal, and
a recovery of the Ti signal, with the resulting Sr/Ti signal-ratio opposing that observed
in TiO
2
-terminated SrTiO
3
. This indicates that the termination was switched successfully
from a TiO
2
surface to a SrO surface and maintained through the deposition of SrTiO
3
.
The intensity values for the
second growth, conducted in
real time, are plotted as a
function of thickness in
Figure 5.17. The relative
signals of Sr and Ti follow the
same trend as that observed
in the heterostructure shown
in Figure 5.16 during the
switch from the TiO
2
-
terminated substrate to SrO
termination, before reversing
to the original TiO
2

termination. Ru signal was
measured in this growth, but
the signal was too weak to
observe signal beyond the
noise with only one scan per
data point. The same
structure was grown again,
stopping to collect more scans
of the Ru
MNN
peak, showing
that it is possible to observe
the Ru signal from a single
monolayer in Figure 5.18.
Figure 5.16 Auger electron spectra signal intensity calculated from areas
beneath the curves for spectra collected at known thickness intervals during
the deposition of homoepitaxial SrTiO3 on a TiO2-terminated SrTiO3
substrate, with termination switching controlled by the deposition of SrRuO3.
The Auger lines used to monitor Sr, Ru, and Ti are SrMNN, RuMNN, and
TiLMM transitions and their intensities clearly track the surface termination
of the film both before and after the switching event. The dashed lines are
modeled signal intensity for the structures, described in the text, with thick
lines corresponding to the assumed switching events shown in the structure
beneath the plot, and thin lines corresponding to the same growth without
termination switching. Marker size is proportional to error. Image: the
author.
149
To quantitatively understand the termination changes in these heterostructures, we used
the parameter-free model based on Auger electron escape depths, explained in Chapter 4.
The model accounts for variation in Auger signal due to compositional changes at the
sub-monolayer scale by generating relative signal intensity shifts as a function of deposited
layer thickness. It predicts the relative Auger electron intensities by examining how the
signals from alternating layers in (100)-oriented SrTiO
3
, and epitaxial SrRuO
3
, are
attenuated by depth using calculated inelastic mean free paths for the transitions
observed.
I compared the calculated
relative AES intensities for
each element with and
without a termination
switching event using the
measured Auger data in
Figure 5.16 and 5.17. The
quality of the models’ fits to
the data is compared by
calculating χ
2
, with the
results normalized for
comparison such that the
non-switching χ
2
= 1. While
the switching versus non-
switching models for the Ti
and Ru signals appear
similar in both the scenarios
for the first growth, as
shown in Figure 5.16, the
rate of decay of Ti signal
and its recovery during and
after the deposition of
SrRuO
3
, and the inverse for
Ru, has a better fit for the
model with termination
switching (for Ru χ
2
= 0.45,
and for Ti χ
2
= 0.30). Likewise, the Sr signal follows the switching model exceedingly well,
and demonstrates a deviation which cannot be explained without a termination switching
event from TiO
2
to SrO (Sr χ
2
= 0.02).
Figure 5.17 Auger electron spectra signal intensity calculated from areas
beneath the curves for spectra collected in real time, during the deposition of
homoepitaxial SrTiO3 on a TiO2-terminated STO substrate, with termination
switching controlled by the deposition of SrRuO3 or TiO2. The Auger lines
used to monitor Ti and Sr are the TiMVV and SrMNN transitions, and their
intensities clearly track the surface termination of the film both before and
after all switching events. The dashed lines are modeled signal intensity for the
structures, described in the text, with thick lines corresponding to the assumed
switching events shown in the structure beneath the plot, and thin lines
corresponding to the same growth without termination switching. Marker size
is proportional to error. Image: the author.
150
The model with termination
switching accurately predicts the
evolution of the AES signal intensity
collected in real time during the
second growth, including subtleties of
the Ti signal recovery after the
deposition of SrRuO
3
, as shown in
Figure 5.17 (for Sr χ
2
= 0.02, and for
Ti χ
2
= 0.12). The model depicting
the signal from the same structure
without the assumption of a
switching event has an RuO
2
-
terminated SrRuO
3
monolayer, rather
than SrO-terminated, and the
deposition of TiO
2
results in a double-layer of TiO
2
. The model without the switching
event is clearly not a fit to the data for either Sr or Ti. The Ti signal without a switching
event would increase above the initial intensity and then decay, which is not observed
experimentally. The Sr signal would remain nearly constant after the deposition of
SrRuO
3
, with slight variation due to the shift in lattice constant, then decay sharply with
the addition of another layer of TiO
2
, neither of which were observed experimentally. This
clearly shows, without ambiguity, that the measured AES data demonstrates the first real
time and in situ observation and control of surface termination during complex oxide thin
film deposition. The ability of such a simple model to accurately predict relative shifts in
Auger signal using only IMFP and lattice dimensions is a testament to the sensitivity of
this technique to subtle compositional changes on surfaces.

Figure 5.18 A repetition of the growth shown in Figure 5.17,
with improved spectral-acquisition methods to observe Ru from the
single deposited monolayer. Image: the author.
151
5.7 Extension to SrZrO
3

To determine if the surface-sensitivity to termination, and the ability to control it,
extended beyond homoepitaxial SrTiO
3
, a growth was conducting with SrZrO
3
instead of
SrTiO
3
. The conditions for growth and Auger spectra collection remained the same. The
heterostructure was designed as follows, beginning with a TiO
2
terminated (001) SrTiO
3
substrate: 4 unit cells SrTiO
3
to verify TiO
2
termination, 8 unit cells SrZrO
3
to verify the
ability to observe ZrO
2
termination, two unit cells SRO to switch to SrO termination, 4
unit cells SrZrO
3
to observe SrO termination, 4 unit cells SrTiO
3
to verify the SrO
termination.
The growth was stopped after every 2 unit cells of deposition to collect 5 Auger spectra
per element. The Auger peaks observed in this growth were the Sr
MNN
, Ti
MVV
, and Zr
MNN
.
The area under the curve were calculated in the same manner as prior growths, from the
summed spectra of each element. The AUC intensity ratio is plotted as a function of film
thickness in Figure 5.19.
Figure 5.19 Auger electron spectra signal intensity calculated from areas beneath the curves for spectra collected
at known thickness intervals during the deposition of homoepitaxial STO on a TiO2-terminated STO substrate followed
by epitaxial SrZrO3 with termination switching controlled by the deposition of SrRuO3. The Auger lines used to monitor
Sr, Zr, and Ti are SrMNN, ZrMNN, and TiMVV transitions and their intensity-ratios track the surface termination of the
film both before and after the switching event. The dashed lines are modeled signal intensity for the structures, described
in the text. Image: the author.
152

The Sr/Ti ratio is constant in agreement with the substrate for the first 4 unit cells of
homoepitaxial SrTiO
3
. The ratio then increases as SrZrO
3
is deposited due to the loss of
Ti signal. When SrTiO
3
is again deposited, for the last 4 unit cells, the Sr/Ti ratio returns,
dropping as expected. The Sr/Ti model for the assumed structure is also plotted, and
follows the data well. The difference in Sr/Ti ratio for TiO
2
versus SrO termination is
clear.
The Sr/Zr ratio begins with the first Auger collection containing Zr, after a total of 6 unit
cells were deposited (4 SrTiO
3
, 2 SrZrO
3
). Over the deposition of SrZrO
3
the Sr/Zr ratio
declines slightly, as expected with an increase in Zr signal and constant Sr signal. With
the deposition of SrRuO
3
the Zr signal drops, causing the Sr/Zr signal to spike, then
decline again as SrZrO
3
is once again deposited. When SrTiO
3
is again deposited on top
of the SrZrO
3
, the Zr signal falls again, causing the Sr/Zr ratio to spike again. Of note is
the Sr/Zr ratio, which shows a clear difference between ZrO
2
termination and SrO
termination, though it is smaller than that showed for Sr and Ti. The modeled Auger
signal for this structure follows the Sr/Zr data extremely well, including the slight
deviation in signal ratio for the two terminations.
This experiment validates this technique for other oxide systems grown with PLD.
Furthermore, it demonstrates the flexibility of AES and the probe for different
compounds.

153
5.8 Summary

The sensitivity of thin film properties to composition and structure was examined,
demonstrating the importance of its understanding for materials science. The
LaAlO
3
/SrTiO
3
heterostructure was examined as an example of the complex role structure
and composition play in properties. Dynamic thin film growth phenomena were explored,
with the example of MBE-grown Ruddlesden-Popper phases of SrTiO
3
used to
demonstrate the possibility of dynamic layer rearrangement. Surface termination was
investigated as another avenue for dynamic events, and examples shown with the
LaAlO
3
/SrTiO
3
heterostructure and SrRuO
3
. Surface energy calculations with DFT were
conducted to illuminate the role of termination in thin films. The calculations predicted
that RuO
2
termination is intrinsically unstable as a result of Fermi level density of states,
explaining the observation of termination switching in this material. Surface termination
observation and control was shown using AES on homoepitaxial SrTiO
3
. Films grown
used SrRuO
3
and TiO
2
targets to control termination, as verified by a combination of AES
and parameter-free escape depth modeling. This technique was extended to termination
switching using SrRuO
3
in SrZrO
3
, as verified with modeling and AES.
The simplicity shown for both modeling and data collection will enable a direct
understanding of growth mechanisms and compositional evolution during thin film
growth, and will enable engineering complex oxide heterostructures for a variety of
applications.

154

Chapter 6
Future Work and Conclusions

6.1 Future Work

The three primary directions this project is currently heading are: application of
termination observation to the LaAlO
3
/SrTiO
3
heterostructure, machine learning for
improving spectral quality, and the development of chemical characterization techniques.
Although these projects are all currently underway in some form as of this writing, their
potential will be discussed here.
Termination observation of the p-type LaAlO
3
/SrTiO
3
heterostructure would be quite an
achievement. As discussed previously in this work, the p-type LaAlO
3
/SrTiO
3
heterostructure has been reported to switch termination in vacuo as a compensation
mechanism for polar catastrophe.
110
If this is the case, and termination observation is
possible in the LaAlO
3
/SrTiO
3
system, then this experiment should be trivial. However,
attempts have thus far failed to make this observation. The primary reason I believe it
has not yet been successful is a combination of deposition nonstoichiometry and Auger
peak selection. As observed with SrTiO
3
in Chapter 4, growing off-stoichiometry films is
not particularly difficult. If the LaAlO
3
system behaves somewhat similarly to SrTiO
3
, the
problem should be exacerbated due to the larger atomic mass difference between the two
species. If the deposition has consistently been La-rich, as is expected, the results would
be observed. The Auger peak selection is also a potential culprit, due to the close energy-
proximity of the Al
LVV
and La
NVV
peaks (~50 eV and ~72 eV, respectively). Overcoming
these problems simply requires the careful adoption of the techniques outlined in this
work. Namely, the deliberate design of LaAlO
3
depositions to observe the composition and
attempt to control it with growth parameters, and the adoption of mean free path
modeling with multiple peaks observed simultaneously from each element (adding Al
KLL

and La
MNN
) to observe compositional shifts as a function of depth.
The possibility of using machine learning to improve spectral quality is currently being
tested and verified for statistical integrity. Essentially, the idea is simply to train the
155
neural network to predict high-resolution spectra from low-resolution spectra. This would
improve temporal resolution and data quality by making spectral acquisition faster.
The development of chemical characterization techniques with the AugerProbe is a
natural next step, considering that Auger electrons are particularly sensitive to chemical
composition due to their outer-shell origins. Achieving this primarily requires an
investment of time currently underway to validate the observation of chemical states.
Potential avenues of investigation with this technique are plentiful, but limited by the
high pressure atmosphere of the PLD chamber. For instance, Ti
LMM
satellite peaks are
often used to determine the valence of Ti, but it would be a significant challenge to observe
it in a state other than Ti
4+
due to its tendency to oxidize even at low pressure.
296

Regardless, other methods are currently underway to demonstrate chemical
characterization, and with clever experimental design they will certainly succeed.

156
6.2 Conclusions

Thin films find abundant applications in modern technology as a direct result of our
ability to manipulate their properties through careful and deliberate construction. With
continuous innovations in this field, it is clear that thin films will continue being an
important part of our technological landscape. However, as we look to the future, the path
has begun looking less clear as a result of fundamental limits on the capabilities of Si
electronics. In response, there is currently a great deal of interest in alternative approaches
to device engineering. One such approach is the integration of complex oxide electronics
for their multifunctional properties. Perovskites are some of the most studied complex
oxide materials, with a long rich history of electronic applications. They are currently
being researched for applications that utilize the correlated nature of their electronic
structures, resulting in emergent properties that can be applied to devices in exciting new
ways. Examples include phase-transition transistors, negative capacitors, and correlation
photovoltaics. However, despite their rich history, complex oxides are still difficult to work
with compared to traditional semiconductors, and the correlated nature of their properties
makes them that much harder to understand and engineer. Yet in the face of these
challenges, thin film synthesis has time and again risen to the occasion to improve the
precision and quality of heterostructures and their properties. With new in situ
compositional characterization techniques for the challenging oxide PLD environment, a
next step in complex oxide thin film sophistication is about to be taken. Using these
techniques to better understand the subtleties of the process/structure/property
relationships of complex oxide thin films and heterostructures will enable the continued
development of this field and the next generation of electronic, photonic, and energy
applications and devices.
This work is centered on an Auger electron spectroscopy probe recently developed for high
pressure oxide growth systems. The AugerProbe is integrated into an existing PLD
chamber and its capabilities and limits explored. It is used to observe characteristic Auger
spectra from dozens of elements sourced from oxide single crystals and thin films, various
polycrystalline materials, and metal foils. Peaks were observed from 40 to 1740 eV, from
carbon (Z = 6) to gadolinium (Z = 64), in various pressures and temperatures used for
thin film deposition. The primary limitations of the probe are discovered to be its reliance
on a relatively low (but still high) pressure ≤ ~10
-2
mbar, and the lengthy spectral
acquisition time which is not compatible with laser pulses. To implement pseudo real time
capabilities, a pulse-probe technique was developed which still allows RHEED observation
despite the slow and disjointed growth rate. The probe is used to observe the deposition
157
of a superlattice, verifying its sensitivity and efficacy while also demonstrating the
correlation of signal intensity with inelastic mean free path of the Auger electrons.

Figure 6.1 Example Auger spectra from the full energy range available, collected in our PLD chamber with the
AugerProbe. Image: the author.

While the probe had been demonstrated as functional, it was not yet proven valuable. To
illustrate the AugerProbe’s effectiveness, the sensitivity of thin film growth to deposition
parameters is used to demonstrate its capability of quantification. A series of
homoepitaxial SrTiO
3
thin films were grown while modifying the laser fluence and growth
pressure, and characteristic Auger spectra taken for each film. Thin film XRD investigates
c-axis lattice expansion, with known relation to composition, and correlates the observed
values with the Auger intensities. A film is grown with continuous observation by AES
while deliberately modifying growth parameters, to demonstrate the capability of real
time process control. During the growth, Auger signal intensity behaves in a way not
explained by the current model used for these systems. A parameter-free escape depth
model is developed for layered (001) perovskite structures which accounts for surface
termination. When applied to the prior growth, it demonstrates that gradual termination
change as a function of nonstoichiometry was directly observed. This result is used to
quantify the SrTiO
3
series, which finds excellent agreement with the literature.
Now that the probe has demonstrated its ability to monitor novel surface phenomena, it
is time to apply it to the direct observation and control of surface termination. Exploring
the sensitivity of film properties to their composition and structure, as well as reported
thin film growth events involving dynamic layer rearrangement, the potential applications
for the AugerProbe are quite valuable. DFT surface energy calculations are used to explain
158
the termination-flipping seen in SrRuO
3
, with the high surface energy of RuO
2
to blame,
caused by a large density of states near the Fermi level. Applied to a homoepitaxial SrTiO
3

thin film growth using SrRuO
3
to flip surface termination from TiO
2
to SrO, the
AugerProbe clearly observes the change, with reported signal intensity verified by the
parameter-free escape depth model. A second growth is conducted, repeating the
observation of termination control by SrRuO
3
deposition. However, this growth uses real
time observation and, furthermore, uses TiO
2
deposition to return the film to its initial
termination. These methods are repeated a third time to demonstrate the same technique
in SrZrO
3
grown on SrTiO
3
, using SrRuO
3
to control termination.
With the conclusion of these experiments, the groundwork has been completed for this
probe’s application for compositional characterization of complex oxide perovskite thin
films grown by PLD. This development brings us one step closer to a more comprehensive
understanding of the complete thin film growth process, including the interrelationships
between growth parameters, their influence on the composition and structure of thin films,
and the resulting emergent properties. Developing the holistic perspective required to
apply these disparate concepts to any materials system is a tremendous challenge, but one
that will allow the integration of multifunctional complex oxides with the next generation
of modern electronic devices, promising a future of everyday correlated quantum
phenomena.

159

References

1. Brooks, A.S., Yellen, J.E., Potts, R., Behrensmeyer, A.K., Deino, A.L., Leslie, D.E.,
Ambrose, S.H., Ferguson, J.R., d’Errico, F., Zipkin, A.M. and Whittaker, S., 2018.
Long-distance stone transport and pigment use in the earliest Middle Stone
Age. Science, 360(6384), pp.90-94.
2. Ohring, M., 2001. Materials science of thin films. Elsevier.
3. Fester, G.A., 1962. Copper and copper alloys in ancient Argentina. Chymia, 8,
pp.21-31.
4. Murmann, B. and Höfflinger, B. eds., 2020. NANO-CHIPS 2030: On-chip AI for
an Efficient Data-driven World. Springer Nature.
5. Hisamoto, D., Lee, W.C., Kedzierski, J., Takeuchi, H., Asano, K., Kuo, C.,
Anderson, E., King, T.J., Bokor, J. and Hu, C., 2000. FinFET-a self-aligned double-
gate MOSFET scalable to 20 nm. IEEE transactions on electron devices, 47(12),
pp.2320-2325.
6. Lee, H., Yu, L.E., Ryu, S.W., Han, J.W., Jeon, K., Jang, D.Y., Kim, K.H., Lee, J.,
Kim, J.H., Jeon, S.C.J.S.C. and Lee, G.S.L.G.S., 2006, June. Sub-5nm all-around
gate FinFET for ultimate scaling. In 2006 Symposium on VLSI Technology, 2006.
Digest of Technical Papers. (pp. 58-59). IEEE.
7. Rostami, M. and Mohanram, K., 2011. Dual-$ V_ {th} $ independent-gate
FinFETs for low power logic circuits. IEEE Transactions on Computer-Aided
Design of Integrated Circuits and Systems, 30(3), pp.337-349.
8. Ionescu, A.M., 2010. Nanowire transistors made easy. Nature nanotechnology, 5(3),
pp.178-179.
9. Colinge, J.P., Kranti, A., Yan, R., Lee, C.W., Ferain, I., Yu, R., Akhavan, N.D.
and Razavi, P., 2011. Junctionless nanowire transistor (JNT): Properties and
design guidelines. Solid-State Electronics, 65, pp.33-37.
160
10. Asenov, A., Wang, Y., Cheng, B., Wang, X., Asenov, P., Al-Ameri, T. and
Georgiev, V.P., 2016, March. Nanowire transistor solutions for 5nm and beyond.
In 2016 17th International Symposium on Quality Electronic Design (ISQED) (pp.
269-274). IEEE.
11. Yu, R., Georgiev, Y.M., Das, S., Hobbs, R.G., Povey, I.M., Petkov, N., Shayesteh,
M., O'Connell, D., Holmes, J.D. and Duffy, R., 2014. Junctionless nanowire
transistor fabricated with high mobility Ge channel. physica status solidi (RRL)–
Rapid Research Letters, 8(1), pp.65-68.
12. Yeh, M.S., Luo, G.L., Hou, F.J., Sung, P.J., Wang, C.J., Su, C.J., Wu, C.T.,
Huang, Y.C., Hong, T.C., Chen, B.Y. and Chen, K.M., 2018. Ge FinFET CMOS
inverters with improved channel surface roughness by using In-situ ALD digital O
3 treatment. IEEE Journal of the Electron Devices Society, 6, pp.1227-1232.
13. Mimura, T., 2002. The early history of the high electron mobility transistor
(HEMT). IEEE Transactions on microwave theory and techniques, 50(3), pp.780-
782.
14. Ye, P.D., Yang, B., Ng, K.K., Bude, J., Wilk, G.D., Halder, S. and Hwang, J.C.M.,
2005. GaN MOS-HEMT using atomic layer deposition Al
2
O
3
as gate dielectric and
surface passivation. High Performance Devices (pp. 167-172).
15. Juneja, S., Pratap, R. and Sharma, R., 2020. Semiconductor technologies for 5G
implementation at millimeter wave frequencies–Design challenges and current state
of work. Engineering Science and Technology, an International Journal.
16. Ziman, J.M., 1972. Principles of the Theory of Solids. Cambridge university press.
17. Biring, S., 2014, November. Auto-metrology on TEM images of FinFET. In ISTFA 2014:
Conference Proceedings from the 40th International Symposium for Testing and Failure
Analysis (p. 327). ASM International.
18. Courtland, R., 2016. The next high-performance transistor [news]. IEEE Spectrum, 53(10),
pp.11-12.
19. Markov, I.L., 2014. Limits on fundamental limits to
computation. Nature, 512(7513), pp.147-154.
20. Mack, C., 2015. The multiple lives of Moore's law. IEEE Spectrum, 52(4), pp.31-
31.
21. Shalf, J.M. and Leland, R., 2015. Computing beyond moore's
law. Computer, 48(12), pp.14-23.
22. International Roadmap for Devices and Systems 2018 Update: Beyond CMOS.
2018. IEEE.
161
23. Manipatruni, S., Nikonov, D.E., Lin, C.C., Gosavi, T.A., Liu, H., Prasad, B.,
Huang, Y.L., Bonturim, E., Ramesh, R. and Young, I.A., 2019. Scalable energy-
efficient magnetoelectric spin–orbit logic. Nature, 565(7737), pp.35-42.
24. Coll, M., Fontcuberta, J., Althammer, M., Bibes, M., Boschker, H., Calleja, A.,
Cheng, G., Cuoco, M., Dittmann, R., Dkhil, B. and El Baggari, I., 2019. Towards
oxide electronics: a roadmap. Applied surface science, 482, pp.1-93.
25. Sønsteby, H.H., Skaar, E., Fjellvåg, Ø.S., Bratvold, J.E., Fjellvåg, H. and Nilsen,
O., 2020. A foundation for complex oxide electronics-low temperature perovskite
epitaxy. Nature communications, 11(1), pp.1-7.
26. Tsukazaki, A., Ohtomo, A., Kita, T., Ohno, Y., Ohno, H. and Kawasaki, M., 2007.
Quantum Hall effect in polar oxide heterostructures. Science, 315(5817), pp.1388-
1391.
27. Ramirez, A.P., 2007. Oxide electronics emerge. Science, 315(5817), pp.1377-1378.
28. Jonker, G.H. and Van Santen, J.H., 1950. Ferromagnetic compounds of manganese
with perovskite structure. physica, 16(3), pp.337-349.
29. von Helmolt, R., Wecker, J., Holzapfel, B., Schultz, L. and Samwer, K., 1993. Giant
negative magnetoresistance in perovskitelike La 2/3 Ba 1/3 MnO x ferromagnetic
films. Physical Review Letters, 71(14), p.2331.
30. Ramirez, A.P., 1997. Colossal magnetoresistance. Journal of Physics: Condensed
Matter, 9(39), p.8171.
31. Bednorz, J.G. and Müller, K.A., 1986. Possible high T c superconductivity in the
Ba − La − Cu − O system. Zeitschrift für Physik B Condensed Matter, 64(2), pp.189-
193.
32. Wu, M.K., Ashburn, J.R., Torng, C., Hor, P.H., Meng, R.L., Gao, L., Huang, Z.J.,
Wang, Y.Q. and Chu, A., 1987. Superconductivity at 93 K in a new mixed-phase
Y-Ba-Cu-O compound system at ambient pressure. Physical review letters, 58(9),
p.908.
33. Dijkkamp, D., Venkatesan, T., Wu, X.D., Shaheen, S.A., Jisrawi, N., Min‐Lee,
Y.H., McLean, W.L. and Croft, M., 1987. Preparation of Y‐Ba‐Cu oxide
superconductor thin films using pulsed laser evaporation from high T c bulk
material. Applied Physics Letters, 51(8), pp.619-621.
34. Tilley, R.J., 2016. Perovskites: structure-property relationships. John Wiley &
Sons.
35. Schlom, D.G., Chen, L.Q., Pan, X., Schmehl, A. and Zurbuchen, M.A., 2008. A
thin film approach to engineering functionality into oxides. Journal of the
American Ceramic Society, 91(8), pp.2429-2454.
162
36. Orvis, T., Surendran, M., Liu, Y., Niu, S., Muramoto, S., Grutter, A.J. and
Ravichandran, J., 2019. Electron doping BaZrO3 via topochemical reduction. ACS
applied materials & interfaces, 11(24), pp.21720-21726.
37. Ohtomo, A. and Hwang, H.Y., 2004. A high-mobility electron gas at the LaAlO
3/SrTiO 3 heterointerface. Nature, 427(6973), pp.423-426.
38. Moore, G.E., 1965. Cramming more components onto integrated circuits.
39. Grundman, M., 2010. The physics of semiconductors: an introduction including
nanophysics and applications. XXXVII, 864, pp.883-6.
40. Ha, S.D. and Ramanathan, S., 2011. Adaptive oxide electronics: A review. Journal
of applied physics, 110(7), p.14.
41. Lorenz, M., Rao, M.R., Venkatesan, T., Fortunato, E., Barquinha, P., Branquinho,
R., Salgueiro, D., Martins, R., Carlos, E., Liu, A. and Shan, F.K., 2016. The 2016
oxide electronic materials and oxide interfaces roadmap. Journal of Physics D:
Applied Physics, 49(43), p.433001.
42. Hassan, N., Yau, K.L.A. and Wu, C., 2019. Edge computing in 5G: A review. IEEE
Access, 7, pp.127276-127289.
43. Sittón-Candanedo, I., Alonso, R.S., Corchado, J.M., Rodríguez-González, S. and
Casado-Vara, R., 2019. A review of edge computing reference architectures and a
new global edge proposal. Future Generation Computer Systems, 99, pp.278-294.
44. McKee, R.A., Walker, F.J. and Chisholm, M.F., 1998. Crystalline oxides on silicon:
the first five monolayers. Physical Review Letters, 81(14), p.3014.
45. Baek, S.H. and Eom, C.B., 2013. Epitaxial integration of perovskite-based
multifunctional oxides on silicon. Acta Materialia, 61(8), pp.2734-2750.
46. Wang, J., Zheng, H., Ma, Z., Prasertchoung, S., Wuttig, M., Droopad, R., Yu, J.,
Eisenbeiser, K. and Ramesh, R., 2004. Epitaxial BiFeO 3 thin films on Si. Applied
Physics Letters, 85(13), pp.2574-2576.
47. Kahn, A.H. and Leyendecker, A.J., 1964. Electronic energy bands in strontium
titanate. Physical Review, 135(5A), p.A1321.
48. Kowalczyk, S.P., McFeely, F.R., Ley, L., Gritsyna, V.T. and Shirley, D.A., 1977.
The electronic structure of SrTiO3 and some simple related oxides (MgO, Al2O3,
SrO, TiO2). Solid State Communications, 23(3), pp.161-169.
49. Uwe, H., Yoshizaki, R., Sakudo, T., Izumi, A. and Uzumaki, T., 1985. Conduction
band structure of SrTiO3. Japanese Journal of Applied Physics, 24(S2), p.335.
163
50. Heifets, E., Eglitis, R.I., Kotomin, E.A., Maier, J. and Borstel, G., 2002. First-
principles calculations for SrTiO3 (100) surface structure. Surface science, 513(1),
pp.211-220.
51. Pai, Y.Y., Tylan-Tyler, A., Irvin, P. and Levy, J., 2018. Physics of SrTiO3-based
heterostructures and nanostructures: a review. Reports on Progress in
Physics, 81(3), p.036503.
52. Dylla, M.T., Kang, S.D. and Snyder, G.J., 2019. Effect of Two‐Dimensional Crystal
Orbitals on Fermi Surfaces and Electron Transport in Three‐Dimensional
Perovskite Oxides. Angewandte Chemie, 131(17), pp.5557-5566.
53. Lybye, D., Poulsen, F.W. and Mogensen, M., 2000. Conductivity of A-and B-site
doped LaAlO3, LaGaO3, LaScO3 and LaInO3 perovskites. Solid State
Ionics, 128(1-4), pp.91-103.
54. Bohn, H.G. and Schober, T., 2000. Electrical conductivity of the high‐temperature
proton conductor BaZr0. 9Y0. 1O2. 95. Journal of the American Ceramic
Society, 83(4), pp.768-772.
55. Rashid, N.L.R.M., Samat, A.A., Jais, A.A., Somalu, M.R., Muchtar, A.,
Baharuddin, N.A. and Isahak, W.N.R.W., 2019. Review on zirconate-cerate-based
electrolytes for proton-conducting solid oxide fuel cell. Ceramics
International, 45(6), pp.6605-6615.
56. Adams, T.B., Sinclair, D.C. and West, A.R., 2002. Giant barrier layer capacitance
effects in CaCu3Ti4O12 ceramics. Advanced Materials, 14(18), pp.1321-1323.
57. Sawaguchi, E., 1953. Ferroelectricity versus antiferroelectricity in the solid
solutions of PbZrO3 and PbTiO3. Journal of the physical society of Japan, 8(5),
pp.615-629.
58. Orlandi, F., Righi, L., Cabassi, R., Delmonte, D., Pernechele, C., Bolzoni, F.,
Mezzadri, F., Solzi, M., Merlini, M. and Calestani, G., 2014. Structural and electric
evidence of ferrielectric state in Pb2MnWO6 double perovskite system. Inorganic
chemistry, 53(19), pp.10283-10290.
59. Chen, J., Chan, H.M. and Harmer, M.P., 1989. Ordering structure and dielectric
properties of undoped and La/Na‐doped Pb (Mg1/3Nb2/3) O3. Journal of the
American Ceramic Society, 72(4), pp.593-598.
60. Martin, M.C., Shirane, G., Endoh, Y., Hirota, K., Moritomo, Y. and Tokura, Y.,
1996. Magnetism and structural distortion in the La 0.7 Sr 0.3 Mn O 3 metallic
ferromagnet. Physical Review B, 53(21), p.14285.
61. Müller, K.A., Von Waldkirch, T., Berlinger, W. and Faughnan, B.W., 1971.
Photochromic Fe5+ (3d3) in SrTiO3 evidence from paramagnetic resonance. Solid
State Communications, 9(13), pp.1097-1101.
164
62. Meiklejohn, W.H. and Bean, C.P., 1956. New magnetic anisotropy. Physical
review, 102(5), p.1413.
63. Mizumaki, M., Chen, W.T., Saito, T., Yamada, I., Attfield, J.P. and Shimakawa,
Y., 2011. Direct observation of the ferrimagnetic coupling of A-site Cu and B-site
Fe spins in charge-disproportionated CaCu 3 Fe 4 O 12. Physical Review B, 84(9),
p.094418.
64. Yuan, Y., Feng, H.L., Ghimire, M.P., Matsushita, Y., Tsujimoto, Y., He, J.,
Tanaka, M., Katsuya, Y. and Yamaura, K., 2015. High-pressure synthesis, crystal
structures, and magnetic properties of 5d double-perovskite oxides Ca2MgOsO6
and Sr2MgOsO6. Inorganic chemistry, 54(7), pp.3422-3431.
65. Wang, J.B.N.J., Neaton, J.B., Zheng, H., Nagarajan, V., Ogale, S.B., Liu, B.,
Viehland, D., Vaithyanathan, V., Schlom, D.G., Waghmare, U.V. and Spaldin,
N.A., 2003. Epitaxial BiFeO3 multiferroic thin film
heterostructures. science, 299(5613), pp.1719-1722.
66. Longo, J.M., Raccah, P.M. and Goodenough, J.B., 1968. Magnetic properties of
SrRuO3 and CaRuO3. Journal of Applied Physics, 39(2), pp.1327-1328.
67. Shi, Y.G., Guo, Y.F., Yu, S., Arai, M., Belik, A.A., Sato, A., Yamaura, K.,
Takayama-Muromachi, E., Tian, H.F., Yang, H.X. and Li, J.Q., 2009. Continuous
metal-insulator transition of the antiferromagnetic perovskite NaOsO 3. Physical
Review B, 80(16), p.161104.
68. Sleight, A.W., Gillson, J.L. and Bierstedt, P.E., 1993. High-temperature
superconductivity in the BaPb1 − xBixO3 system. Solid State
Communications, 88(11-12), pp.841-842.
69. Zeng, Z., Greenblatt, M. and Croft, M., 1999. Large magnetoresistance in
antiferromagnetic CaMnO 3 − δ. Physical Review B, 59(13), p.8784.
70. Hundley, M.F., Hawley, M., Heffner, R.H., Jia, Q.X., Neumeier, J.J., Tesmer, J.,
Thompson, J.D. and Wu, X.D., 1995. Transport‐magnetism correlations in the
ferromagnetic oxide La0. 7Ca0. 3MnO3. Applied physics letters, 67(6), pp.860-862.
71. Chen, J., Xing, X., Sun, C., Hu, P., Yu, R., Wang, X. and Li, L., 2008. Zero
thermal expansion in PbTiO3-based perovskites. Journal of the American
Chemical Society, 130(4), pp.1144-1145.
72. Laguta, V.V., Glinchuk, M.D., Slipenyuk, A.M. and Bykov, I.P., 2000. Light-
induced intrinsic defects in PLZT ceramics. Physics of the Solid State, 42(12),
pp.2258-2264.
73. Weis, R.S. and Gaylord, T.K., 1985. Lithium niobate: summary of physical
properties and crystal structure. Applied Physics A, 37(4), pp.191-203.
165
74. Yi, D., Wang, Y., van ʼt Erve, O.M., Xu, L., Yuan, H., Veit, M.J., Balakrishnan,
P.P., Choi, Y., N’Diaye, A.T., Shafer, P. and Arenholz, E., 2020. Emergent electric
field control of phase transformation in oxide superlattices. Nature
communications, 11(1), pp.1-8.
75. Yadav, A.K., Nguyen, K.X., Hong, Z., García-Fernández, P., Aguado-Puente, P.,
Nelson, C.T., Das, S., Prasad, B., Kwon, D., Cheema, S. and Khan, A.I., 2019.
Spatially resolved steady-state negative capacitance. Nature, 565(7740), pp.468-
471.
76. Pérez-Tomás, A., Mingorance, A., Tanenbaum, D. and Lira-Cantú, M., 2018. Metal
oxides in photovoltaics: all-oxide, ferroic, and perovskite solar cells. In The future
of semiconductor oxides in next-generation solar cells (pp. 267-356). Elsevier.
77. Kumar, D., Aluguri, R., Chand, U. and Tseng, T.Y., 2017. Metal oxide resistive
switching memory: materials, properties and switching mechanisms. Ceramics
International, 43, pp.S547-S556.
78. Gao, W., Zhu, Y., Wang, Y., Yuan, G. and Liu, J.M., 2020. A review of flexible
perovskite oxide ferroelectric films and their application. Journal of
Materiomics, 6(1), pp.1-16.
79. Hambe, M., Petraru, A., Pertsev, N.A., Munroe, P., Nagarajan, V. and Kohlstedt,
H., 2010. Crossing an interface: Ferroelectric control of tunnel currents in magnetic
complex oxide heterostructures. Advanced Functional Materials, 20(15), pp.2436-
2441.
80. Chang, S.J., Chen, S.Y., Chen, P.W., Huang, S.J. and Tseng, Y.C., 2019. Pulse-
Driven Nonvolatile Perovskite Memory with Photovoltaic Read-Out
Characteristics. ACS applied materials & interfaces, 11(37), pp.33803-33810.
81. Li, W., Shi, J., Zhang, K.H. and MacManus-Driscoll, J.L., 2020. Defects in complex
oxide thin films for electronics and energy applications: challenges and
opportunities. Materials Horizons, 7(11), pp.2832-2859.
82. Ortmann, J.E., Borisevich, A.Y., Kwon, S., Posadas, A., Kim, M.J. and Demkov,
A.A., 2021. Three-Dimensional Integration of Functional Oxides and Crystalline
Silicon for Optical Neuromorphic Computing Using Nanometer-Scale Oxygen
Scavenging Barriers. ACS Applied Nano Materials.
83. Lapano, J., Brahlek, M., Zhang, L., Roth, J., Pogrebnyakov, A. and Engel-Herbert,
R., 2019. Scaling growth rates for perovskite oxide virtual substrates on
silicon. Nature communications, 10(1), pp.1-7.
84. Yin, Y. and Li, Q., 2017. A review on all-perovskite multiferroic tunnel
junctions. Journal of Materiomics, 3(4), pp.245-254.
166
85. Varignon, J., Bibes, M. and Zunger, A., 2019. Origin of band gaps in 3 d perovskite
oxides. Nature communications, 10(1), pp.1-11.
86. Kim, J.H., Aghaeimeibodi, S., Carolan, J., Englund, D. and Waks, E., 2020. Hybrid
integration methods for on-chip quantum photonics. Optica, 7(4), pp.291-308.
87. Hwang, H.Y., Iwasa, Y., Kawasaki, M., Keimer, B., Nagaosa, N. and Tokura, Y.,
2012. Emergent phenomena at oxide interfaces. Nature materials, 11(2), pp.103-
113.
88. Seo, J.W., Fullerton, E.E., Nolting, F., Scholl, A., Fompeyrine, J. and Locquet,
J.P., 2008. Antiferromagnetic LaFeO3 thin films and their effect on exchange
bias. Journal of Physics: Condensed Matter, 20(26), p.264014.
89. Schwabl, F., 2006. Statistical Mechanics. Springer.
90. Beekman, A.J., Rademaker, L. and van Wezel, J., 2019. An introduction to
spontaneous symmetry breaking. SciPost Phys. Lect. Notes, 11, p.21468.
91. Bennemann, K.H. and Ketterson, J.B. eds., 2008. Superconductivity: Volume 1:
Conventional and Unconventional Superconductors Volume 2: Novel
Superconductors. Springer Science & Business Media.
92. Aziz, M.J., 2008. Film growth mechanisms in pulsed laser deposition. Applied
Physics A, 93(3), pp.579-587.
93. Warrender, J.M. and Aziz, M.J., 2007. Kinetic energy effects on morphology
evolution during pulsed laser deposition of metal-on-insulator films. Physical
Review B, 75(8), p.085433.
94. Ojeda‐G‐P, A., Döbeli, M. and Lippert, T., 2018. Influence of plume properties on
thin film composition in pulsed laser deposition. Advanced Materials
Interfaces, 5(18), p.1701062.
95. Chen, J., Döbeli, M., Stender, D., Conder, K., Wokaun, A., Schneider, C.W. and
Lippert, T., 2014. Plasma interactions determine the composition in pulsed laser
deposited thin films. Applied Physics Letters, 105(11), p.114104.
96. Breckenfeld, E., Bronn, N., Karthik, J., Damodaran, A.R., Lee, S., Mason, N. and
Martin, L.W., 2013. Effect of growth induced (non) stoichiometry on interfacial
conductance in LaAlO 3/SrTiO 3. Physical review letters, 110(19), p.196804.
97. Lowndes, D.H., Geohegan, D.B., Puretzky, A.A., Norton, D.P. and Rouleau, C.M.,
1996. Synthesis of novel thin-film materials by pulsed laser
deposition. Science, 273(5277), pp.898-903.
98. Willmott, P.R. and Huber, J.R., 2000. Pulsed laser vaporization and
deposition. Reviews of Modern Physics, 72(1), p.315.
167
99. Christen, H.M. and Eres, G., 2008. Recent advances in pulsed-laser deposition of
complex oxides. Journal of Physics: Condensed Matter, 20(26), p.264005.
100. Blank, D.H., Dekkers, M. and Rijnders, G., 2013. Pulsed laser deposition
in Twente: from research tool towards industrial deposition. Journal of physics D:
applied physics, 47(3), p.034006.
101. Masood, K.B., Kumar, P., Malik, M.A. and Singh, J., 2021. A
comprehensive tutorial on the pulsed laser deposition technique and developments
in the fabrication of low dimensional systems and nanostructures. Emergent
Materials, pp.1-18.
102. Rodríguez-Hernández, P.E., Quiñones-Galván, J.G., Meléndez-Lira, M.,
Santos-Cruz, J., Contreras-Puente, G. and de Moure-Flores, F., 2020. Effect of
laser fluence on structural and optical properties of CuxS films grown by pulsed
laser deposition at different wavelengths. Materials Research Express, 7(1),
p.015908.
103. Jaber, N., Wolfman, J., Daumont, C., Negulescu, B., Ruyter, A.,
Sauvage, T., Courtois, B., Bouyanfif, H., Longuet, J.L., Autret-Lambert, C. and
Gervais, F., 2017. Laser fluence and spot size effect on compositional and structural
properties of BiFeO3 thin films grown by Pulsed Laser Deposition. Thin Solid
Films, 634, pp.107-111.
104. Schraknepper, H., Bäumer, C., Gunkel, F., Dittmann, R. and De Souza,
R.A., 2016. Pulsed laser deposition of SrRuO3 thin-films: The role of the pulse
repetition rate. APL materials, 4(12), p.126109.
105. Panchal, G., Rawat, R., Bagri, A., Mandal, A.K., Choudhary, R.J. and
Phase, D.M., 2019. Effect of oxygen partial pressure on the electronic and magnetic
properties of epitaxial SrRuO3 thin films. Physica B: Condensed Matter, 572,
pp.190-194.
106. Ohnishi, T. and Takada, K., 2011. Epitaxial thin-film growth of SrRuO3,
Sr3Ru2O7, and Sr2RuO4 from a SrRuO3 target by pulsed laser deposition. Applied
physics express, 4(2), p.025501.
107. Bhattacharya, D., Singh, R.K. and Holloway, P.H., 1991. Laser‐target
interactions during pulsed laser deposition of superconducting thin films. Journal
of applied physics, 70(10), pp.5433-5439.
108. Vakulov, Z., Zamburg, E., Golosov, D.A., Zavadskiy, S.M., Dostanko,
A.P., Miakonkikh, A.V., Klemente, I.E., Rudenko, K.V. and Ageev, O.A., 2017,
November. Influence of target-substrate distance during pulsed laser deposition on
properties of LiNbO3 thin films. In Journal of Physics: Conference Series (Vol.
917, No. 3, p. 032024). IOP Publishing.
168
109. Plonczak, P., Bieberle‐Hütter, A., Søgaard, M., Ryll, T., Martynczuk, J.,
Hendriksen, P.V. and Gauckler, L.J., 2011. Tailoring of LaxSr1‐xCoyFe1‐yO3‐δ
Nanostructure by Pulsed Laser Deposition. Advanced Functional
Materials, 21(14), pp.2764-2775.
110. Shin, J., Kalinin, S.V., Lee, H.N., Christen, H.M., Moore, R.G., Plummer,
E.W. and Baddorf, A.P., 2005. Surface stability of epitaxial SrRuO3 films. Surface
science, 581(2-3), pp.118-132.
111. Meng, J., Chen, Z. and Jiang, A., 2015. Investigation of the growth
mechanism of SrRuO3 thin films fabricated by pulsed laser deposition. Japanese
Journal of Applied Physics, 54(5), p.055502.
112. Ravichandran, J., Yadav, A.K., Cheaito, R., Rossen, P.B., Soukiassian,
A., Suresha, S.J., Duda, J.C., Foley, B.M., Lee, C.H., Zhu, Y. and Lichtenberger,
A.W., 2014. Crossover from incoherent to coherent phonon scattering in epitaxial
oxide superlattices. Nature materials, 13(2), pp.168-172.
113. Yadav, A.K., Nelson, C.T., Hsu, S.L., Hong, Z., Clarkson, J.D.,
Schlepütz, C.M., Damodaran, A.R., Shafer, P., Arenholz, E., Dedon, L.R. and
Chen, D., 2016. Observation of polar vortices in oxide
superlattices. Nature, 530(7589), pp.198-201.
114. Liu, Y., Wang, Z., Thind, A.S., Orvis, T., Sarkar, D., Kapadia, R.,
Borisevich, A.Y., Mishra, R., Khan, A.I. and Ravichandran, J., 2019. Epitaxial
growth and dielectric characterization of atomically smooth 0.5 Ba (Zr0. 2Ti0. 8)
O3–0.5 (Ba0. 7Ca0. 3) TiO3 thin films. Journal of Vacuum Science & Technology
A: Vacuum, Surfaces, and Films, 37(1), p.011502.
115. Koster, G. and Rijnders, G. eds., 2011. In situ characterization of thin
film growth. Elsevier.
116. Soto, G., De La Cruz, W., Dıaz, J.A., Machorro, R., Castillón, F.F. and
Farıas, M.H., 2003. Characterization of tungsten oxide films produced by reactive
pulsed laser deposition. Applied surface science, 218(1-4), pp.282-290.
117. Soto, G., Dıaz, J.A. and De La Cruz, W., 2003. Copper nitride films
produced by reactive pulsed laser deposition. Materials Letters, 57(26-27), pp.4130-
4133.
118. Soto, G., De la Cruz, W., Castillon, F.F., Dıaz, J.A., Machorro, R. and
Farıas, M.H., 2003. Tungsten nitride films grown via pulsed laser deposition studied
in situ by electron spectroscopies. Applied surface science, 214(1-4), pp.58-67.
119. Singh-Bhalla, G., Rossen, P.B., Pálsson, G.K., Mecklenburg, M., Orvis,
T., Das, S., Tang, Y.L., Suresha, J.S., Yi, D., Dasgupta, A. and Doenning, D.,
169
2018. Unexpected termination switching and polarity compensation in LaAlO
3/SrTiO 3 heterostructures. Physical Review Materials, 2(11), p.112001.
120. Lee, J.H., Luo, G., Tung, I.C., Chang, S.H., Luo, Z., Malshe, M., Gadre,
M., Bhattacharya, A., Nakhmanson, S.M., Eastman, J.A. and Hong, H., 2014.
Dynamic layer rearrangement during growth of layered oxide films by molecular
beam epitaxy. Nature materials, 13(9), pp.879-883.
121. Arthur, J.R. and LePore, J.J., 1969. GaAs, GaP, and GaAs x P 1 − x
Epitaxial Films Grown by Molecular Beam Deposition. Journal of Vacuum Science
and Technology, 6(4), pp.545-548.
122. Pendry, J.B., 1980. Reliability factors for LEED calculations. Journal of
Physics C: Solid State Physics, 13(5), p.937.
123. Ino, S., 1977. Some new techniques in reflection high energy electron
diffraction (RHEED) application to surface structure studies. Japanese Journal of
Applied Physics, 16(6), p.891.
124. Harris, J.J., Joyce, B.A. and Dobson, P.J., 1981. Oscillations in the
surface structure of Sn-doped GaAs during growth by MBE. Surface Science
Letters, 103(1), pp.L90-L96.
125. Rijnders, G.J., Koster, G., Blank, D.H. and Rogalla, H., 1997. In situ
monitoring during pulsed laser deposition of complex oxides using reflection high
energy electron diffraction under high oxygen pressure. Applied physics
letters, 70(14), pp.1888-1890.
126. Wrobel, F., Mark, A.F., Christiani, G., Sigle, W., Habermeier, H.U., van
Aken, P.A., Logvenov, G., Keimer, B. and Benckiser, E., 2017. Comparative study
of LaNiO3/LaAlO3 heterostructures grown by pulsed laser deposition and oxide
molecular beam epitaxy. Applied Physics Letters, 110(4), p.041606.
127. Sun, H.Y., Mao, Z.W., Zhang, T.W., Han, L., Zhang, T.T., Cai, X.B.,
Guo, X., Li, Y.F., Zang, Y.P., Guo, W. and Song, J.H., 2018. Chemically specific
termination control of oxide interfaces via layer-by-layer mean inner potential
engineering. Nature communications, 9(1), pp.1-8.
128. Hsu, M.H.M., Merckling, C., El Kazzi, S., Pantouvaki, M., Richard, O.,
Bender, H., Meersschaut, J., Van Campenhout, J., Absil, P. and Van Thourhout,
D., 2016. Diffraction studies for stoichiometry effects in BaTiO3 grown by
molecular beam epitaxy on Ge (001). Journal of Applied Physics, 120(22),
p.225114.
129. Fujii, T., De Groot, F.M.F., Sawatzky, G.A., Voogt, F.C., Hibma, T. and
Okada, K., 1999. In situ XPS analysis of various iron oxide films grown by NO 2-
assisted molecular-beam epitaxy. Physical review B, 59(4), p.3195.
170
130. Thapa, S., Paudel, R., Blanchet, M.D., Gemperline, P.T. and Comes,
R.B., 2021. Probing surfaces and interfaces in complex oxide films via in situ X-
ray photoelectron spectroscopy. Journal of Materials Research, pp.1-26.
131. Eres, G., Rouleau, C.M., Lu, Q., Zhang, Z., Benda, E., Lee, H.N.,
Tischler, J.Z. and Fong, D.D., 2019. Experimental setup combining in situ hard X-
ray photoelectron spectroscopy and real-time surface X-ray diffraction for
characterizing atomic and electronic structure evolution during complex oxide
heterostructure growth. Review of Scientific Instruments, 90(9), p.093902.
132. Sasaki, T., Ishikawa, F., Yamaguchi, T. and Takahasi, M., 2016. Nitride-
MBE system for in situ synchrotron X-ray measurements. Japanese Journal of
Applied Physics, 55(5S), p.05FB05.
133. Staib, P.G., 2011. In situ real time Auger analyses during oxides and alloy
growth using a new spectrometer design. Journal of Vacuum Science & Technology
B, Nanotechnology and Microelectronics: Materials, Processing, Measurement,
and Phenomena, 29(3), p.03C125.
134. Chang, C.C., 1971. Auger electron spectroscopy. Surface Science, 25(1),
pp.53-79.
135. Chambers, S.A., Tran, T.T. and Hileman, T.A., 1995. Auger electron
spectroscopy as a real‐time compositional probe in molecular beam
epitaxy. Journal of Vacuum Science & Technology A: Vacuum, Surfaces, and
Films, 13(1), pp.83-91.
136. Orvis, T., Surendran, M., Liu, Y., Cunniff, A. and Ravichandran, J.,
2019. In situ Auger electron spectroscopy of complex oxide surfaces grown by
pulsed laser deposition. Journal of Vacuum Science & Technology A: Vacuum,
Surfaces, and Films, 37(6), p.061401.
137. Laws Calley, W., Greenlee, J.D., Henderson, W.E., Lowder, J., Moseley,
M.W., Alan Doolittle, W. and Staib, P.G., 2013. In situ Auger probe enabling
epitaxy composition control of alloys by elemental surface analysis. Journal of
Vacuum Science & Technology B, Nanotechnology and Microelectronics:
Materials, Processing, Measurement, and Phenomena, 31(3), p.03C126.
138. Ferguson, I.F., 1989. Auger microprobe analysis. CRC Press.
139. Cazaux, J., 2010. Secondary electron emission and charging mechanisms
in Auger electron spectroscopy and related e-beam techniques. Journal of Electron
Spectroscopy and Related Phenomena, 176(1-3), pp.58-79.
140. Young, S.R., Maksov, A., Ziatdinov, M., Cao, Y., Burch, M.,
Balachandran, J., Li, L., Somnath, S., Patton, R.M., Kalinin, S.V. and Vasudevan,
171
R.K., 2018. Data mining for better material synthesis: The case of pulsed laser
deposition of complex oxides. Journal of Applied Physics, 123(11), p.115303.
141. Sun, Z., Liao, T., Dou, Y., Hwang, S.M., Park, M.S., Jiang, L., Kim, J.H.
and Dou, S.X., 2014. Generalized self-assembly of scalable two-dimensional
transition metal oxide nanosheets. Nature communications, 5(1), pp.1-9.
142. Liu, Y.T., Zhang, P., Sun, N., Anasori, B., Zhu, Q.Z., Liu, H., Gogotsi,
Y. and Xu, B., 2018. Self‐assembly of transition metal oxide nanostructures on
MXene nanosheets for fast and stable lithium storage. Advanced Materials, 30(23),
p.1707334.
143. Aimon, N.M., Choi, H.K., Sun, X.Y., Kim, D.H. and Ross, C.A., 2014.
Templated Self‐Assembly of Functional Oxide Nanocomposites. Advanced
Materials, 26(19), pp.3063-3067.
144. Schuster, F., Furtmayr, F., Zamani, R., Magén, C., Morante, J.R.,
Arbiol, J., Garrido, J.A. and Stutzmann, M., 2012. Self-assembled GaN nanowires
on diamond. Nano letters, 12(5), pp.2199-2204.
145. Spurgeon, S.R., Sushko, P.V., Chambers, S.A. and Comes, R.B., 2017.
Dynamic interface rearrangement in LaFeO 3/n − SrTiO 3
heterojunctions. Physical Review Materials, 1(6), p.063401.
146. Rijnders, G., Blank, D.H., Choi, J. and Eom, C.B., 2004. Enhanced
surface diffusion through termination conversion during epitaxial SrRuO 3
growth. Applied physics letters, 84(4), pp.505-507.
147. Chapter 2:
148. De Graef, M. and McHenry, M.E., 2012. Structure of materials: an
introduction to crystallography, diffraction and symmetry. Cambridge University
Press.
149. Koster, G., Kropman, B.L., Rijnders, G.J., Blank, D.H. and Rogalla, H.,
1998. Quasi-ideal strontium titanate crystal surfaces through formation of
strontium hydroxide. Applied Physics Letters, 73(20), pp.2920-2922.
150. Orvis, T., Cao, T. Surendran, M., Kumarasubramanian, H., Cunniff, A., Mishra,
R. and Ravichandran, J., Nano Letters (under review) (2021). arXiv:2102.05022
151. Kern, W., 1990. The evolution of silicon wafer cleaning
technology. Journal of the Electrochemical Society, 137(6), p.1887.
152. f450c.org/infographic/ - accessed February, 2021
153. Tateno, Y., Endo, K., Arisawa, S., Vlaicu, A.M., Nedelcu, L., Preda, N.,
Secu, M., Iordanescu, R., Kuncser, A.C. and Badica, P., 2018. Growth of SrTiO3
172
Single Crystals with a Diameter of about 30 mm by the Verneuil Method. Crystal
Growth & Design, 19(2), pp.604-612.
154. Horn, F.H., 1962. Barium Titanate Crystals Grown from the
Melt. Journal of Applied Physics, 33(4), pp.1615-1616.
155. Belruss, V., Kalnajs, J., Linz, A. and Folweiler, R.C., 1971. Top-seeded
solution growth of oxide crystals from non-stoichiometric melts. Materials
Research Bulletin, 6(10), pp.899-905.
156. Shimamura, K., Tabata, H., Takeda, H., Kochurikhin, V.V. and Fukuda,
T., 1998. Growth and characterization of (La, Sr)(Al, Ta) O3 single crystals as
substrates for GaN epitaxial growth. Journal of crystal growth, 194(2), pp.209-213.
157. Kalabukhov, A. and Gunnarsson, B., 2007. RJ; Olsson, E.; Claeson, T.;
Winkler, D. Phys. Rev. B, 75, p.121404.
158. Becerra-Toledo, A.E. and Marks, L.D., 2010. Strontium oxide segregation
at SrLaAlO4 surfaces. Surface science, 604(17-18), pp.1476-1480.
159. Koo, B., Kim, K., Kim, J.K., Kwon, H., Han, J.W. and Jung, W., 2018.
Sr segregation in perovskite oxides: why it happens and how it exists. Joule, 2(8),
pp.1476-1499.
160. Zhu, Y., Salvador, P.A. and Rohrer, G.S., 2017. Controlling the
termination and photochemical reactivity of the SrTiO 3 (110) surface. Physical
Chemistry Chemical Physics, 19(11), pp.7910-7918.
161. Tomio, T., Miki, H., Tabata, H., Kawai, T. and Kawai, S., 1994. Control
of electrical conductivity in laser deposited SrTiO3 thin films with Nb
doping. Journal of Applied Physics, 76(10), pp.5886-5890.
162. Zhao, T., Lu, H., Chen, F., Dai, S., Yang, G. and Chen, Z., 2000. Highly
conductive Nb doped SrTiO3 epitaxial thin films grown by laser molecular beam
epitaxy. Journal of crystal growth, 212(3-4), pp.451-455.
163. Chen, H.C., Huang, C.W., Wu, J.C. and Lin, S.T., 2012. Theoretical
investigation of the metal-doped SrTiO3 photocatalysts for water splitting. The
Journal of Physical Chemistry C, 116(14), pp.7897-7903.
164. Savinov, M., Trepakov, V.A., Syrnikov, P.P., Železný, V., Pokorný, J.,
Dejneka, A., Jastrabik, L. and Galinetto, P., 2008. Dielectric properties of Mn
doped SrTiO3. Journal of Physics: Condensed Matter, 20(9), p.095221.
165. Escudero, M.J., Irvine, J.T.S. and Daza, L., 2009. Development of anode
material based on La-substituted SrTiO3 perovskites doped with manganese
and/or gallium for SOFC. Journal of Power Sources, 192(1), pp.43-50.
173
166. Leca, V., Rijnders, G., Koster, G., Blank, D.H. and Rogalla, H., 1999.
Wet etching methods for perovskite substrates. MRS Online Proceedings
Library, 587(1), pp.O3-6.
167. Gellé, F., Chirita, R., Mertz, D., Rastei, M.V., Dinia, A. and Colis, S.,
2018. Guideline to atomically flat TiO2-terminated SrTiO3 (001) surfaces. Surface
Science, 677, pp.39-45.
168. Connell, J.G., Isaac, B.J., Ekanayake, G.B., Strachan, D.R. and Seo,
S.S.A., 2012. Preparation of atomically flat SrTiO3 surfaces using a deionized-water
leaching and thermal annealing procedure. Applied Physics Letters, 101(25),
p.251607.
169. Caravati, E.M., 1988. Acute hydrofluoric acid exposure. The American
journal of emergency medicine, 6(2), pp.143-150.
170. Wang, X., Zhang, Y., Ni, L., You, C., Ye, C., Jiang, R., Liu, L., Liu, J.
and Han, C., 2014. A review of treatment strategies for hydrofluoric acid burns:
current status and future prospects. Burns, 40(8), pp.1447-1457.
171. Horton, D.K., Berkowitz, Z. and Kaye, W.E., 2004. Hydrofluoric acid
releases in 17 states and the acute health effects associated, 1993–2001. Journal of
occupational and environmental medicine, 46(5), pp.501-508.
172. Nie, Y.F., Zhu, Y., Lee, C.H., Kourkoutis, L.F., Mundy, J.A., Junquera,
J., Ghosez, P., Baek, D.J., Sung, S., Xi, X.X. and Shen, K.M., 2014. Atomically
precise interfaces from non-stoichiometric deposition. Nature
communications, 5(1), pp.1-8.
173. Basting, D. and Marowsky, G. eds., 2005. Excimer laser technology.
Springer Science & Business Media.
174. Liu, X., Cao, Y., Pal, B., Middey, S., Kareev, M., Choi, Y., Shafer, P.,
Haskel, D., Arenholz, E. and Chakhalian, J., 2017. Synthesis and electronic
properties of Ruddlesden-Popper strontium iridate epitaxial thin films stabilized
by control of growth kinetics. Physical Review Materials, 1(7), p.075004.
175. Nishio, K., Hwang, H.Y. and Hikita, Y., 2016. Thermodynamic guiding
principles in selective synthesis of strontium iridate Ruddlesden-Popper epitaxial
films. APL Materials, 4(3), p.036102.
176. von Wenckstern, H., Kneiß, M., Hassa, A., Storm, P., Splith, D. and
Grundmann, M., 2020. A Review of the Segmented‐Target Approach to
Combinatorial Material Synthesis by Pulsed‐Laser Deposition. physica status solidi
(b), 257(7), p.1900626.
177. Cheung, J. and Horwitz, J., 1992. Pulsed laser deposition history and
laser-target interactions. MRS bulletin, 17(2), pp.30-36.
174
178. Sellappan, P., Tang, C., Shi, J. and Garay, J.E., 2017. An integrated
approach to doped thin films with strain-tunable magnetic anisotropy: powder
synthesis, target preparation and pulsed laser deposition of Bi: YIG. Materials
Research Letters, 5(1), pp.41-47.
179. Guillon, O., Gonzalez‐Julian, J., Dargatz, B., Kessel, T., Schierning, G.,
Räthel, J. and Herrmann, M., 2014. Field‐assisted sintering technology/spark
plasma sintering: mechanisms, materials, and technology developments. Advanced
Engineering Materials, 16(7), pp.830-849.
180. Smith, D.L. and Hoffman, D.W., 1996. Thin-film deposition: principles
and practice. Physics Today, 49(4), p.60.
181. Mele, A., Guidoni, A.G., Kelly, R., Miotello, A., Orlando, S. and Teghil,
R., 1996. Spatial distribution of laser-ablated material by probing a plasma plume
in three dimensions. Applied surface science, 96, pp.102-111.
182. Tyunina, M. and Leppävuori, S., 2000. Effects of laser fluence, size, and
shape of the laser focal spot in pulsed laser deposition using a multielemental
target. Journal of Applied Physics, 87(11), pp.8132-8142.
183. Kang, D.W., Jeon, I.J., Song, J.S. and Kim, D., 2003. Spatial and energy
distribution of Co, Ag and Pt particles in pulsed laser deposition: in view of the
fabrication of nanometer multilayer film. Applied Physics A, 77(3), pp.449-453.
184. Zhigilei, L.V., 2003. Dynamics of the plume formation and parameters of
the ejected clusters in short-pulse laser ablation. Applied Physics A, 76(3), pp.339-
350.
185. Willmott, P.R., Herger, R., Schlepütz, C.M., Martoccia, D. and
Patterson, B.D., 2006. Energetic surface smoothing of complex metal-oxide thin
films. Physical review letters, 96(17), p.176102.
186. Sambri, A., Amoruso, S., Wang, X., Radovic’, M., Miletto Granozio, F.
and Bruzzese, R., 2007. Substrate heating influence on plume propagation during
pulsed laser deposition of complex oxides. Applied Physics Letters, 91(15),
p.151501.
187. Phipps, C. ed., 2007. Laser ablation and its applications (Vol. 129).
Springer.
188. Sambri, A., Amoruso, S., Wang, X., Granozio, F.M. and Bruzzese, R.,
2008. Plume propagation dynamics of complex oxides in oxygen. Journal of Applied
Physics, 104(5), p.053304.
189. Aruta, C., Amoruso, S., Ausanio, G., Bruzzese, R., Di Gennaro, E.,
Lanzano, M., Miletto Granozio, F., Riaz, M., Sambri, A., Scotti di Uccio, U. and
Wang, X., 2012. Critical influence of target-to-substrate distance on conductive
175
properties of LaGaO3/SrTiO3 interfaces deposited at 10 − 1 mbar oxygen
pressure. Applied Physics Letters, 101(3), p.031602.
190. Wicklein, S., Sambri, A., Amoruso, S., Wang, X., Bruzzese, R., Koehl, A.
and Dittmann, R., 2012. Pulsed laser ablation of complex oxides: The role of
congruent ablation and preferential scattering for the film stoichiometry. Applied
physics letters, 101(13), p.131601.
191. Breckenfeld, E., Bronn, N., Mason, N. and Martin, L.W., 2014.
Tunability of conduction at the LaAlO3/SrTiO3 heterointerface: thickness and
compositional studies. Applied Physics Letters, 105(12), p.121610.
192. Ojeda-GP, A., Schneider, C.W., Döbeli, M., Lippert, T. and Wokaun, A.,
2015. The flip-over effect in pulsed laser deposition: Is it relevant at high
background gas pressures?. Applied Surface Science, 357, pp.2055-2062.
193. Wang, J., Rijnders, G. and Koster, G., 2018. Complex plume
stoichiometry during pulsed laser deposition of SrVO3 at low oxygen
pressures. Applied physics letters, 113(22), p.223103.
194. Farid, N., Harilal, S.S., Ding, H. and Hassanein, A., 2013. Dynamics of
ultrafast laser plasma expansion in the presence of an ambient. Applied Physics
Letters, 103(19), p.191112.
195. Herman, M.A., Richter, W. and Sitter, H., 2013. Epitaxy: physical
principles and technical implementation (Vol. 62). Springer Science & Business
Media.
196. Phipps, C. ed., 2007. Laser ablation and its applications (Vol. 129).
Springer.
197. Barrett, H.H., 1964. Dielectric Breakdown of Single‐Crystal Strontium
Titanate. Journal of applied physics, 35(5), pp.1420-1425.
198. Shende, R.V., Krueger, D.S., Rossetti, G.A. and Lombardo, S.J., 2001.
Strontium zirconate and strontium titanate ceramics for high‐voltage applications:
synthesis, processing, and dielectric properties. Journal of the American Ceramic
Society, 84(7), pp.1648-1650.
199. Heiroth, S., Koch, J., Lippert, T., Wokaun, A., Günther, D., Garrelie, F.
and Guillermin, M., 2010. Laser ablation characteristics of yttria-doped zirconia in
the nanosecond and femtosecond regimes. Journal of applied physics, 107(1),
p.014908.
200. Eason, R. ed., 2007. Pulsed laser deposition of thin films: applications-
led growth of functional materials. John Wiley & Sons.
176
201. Willmott, P.R., Manoravi, P. and Holliday, K., 2000. Production and
characterization of Nd, Cr: GSGG thin films on Si (001) grown by pulsed laser
ablation. Applied Physics A, 70(4), pp.425-429.
202. D. B. Chrisey, G. K. Hubler, Pulsed Laser Deposition of Thin Films, Vol.
1, Wiley, New York 1994.
203. Liu, H.C., Mao, X.L., Yoo, J.H. and Russo, R.E., 1999. Early phase laser
induced plasma diagnostics and mass removal during single-pulse laser ablation of
silicon. Spectrochimica Acta Part B: Atomic Spectroscopy, 54(11), pp.1607-1624.
204. Chang, J.J. and Warner, B.E., 1996. Laser‐plasma interaction during
visible‐laser ablation of methods. Applied physics letters, 69(4), pp.473-475.
205. Timm, R., Willmott, P.R. and Huber, J.R., 1996. Ablation and blow‐off
characteristics at 248 nm of Al, Sn and Ti targets used for thin film pulsed laser
deposition. Journal of applied physics, 80(3), pp.1794-1802.
206. Singh, A.V., Mehra, R.M., Buthrath, N., Wakahara, A. and Yoshida, A.,
2001. Highly conductive and transparent aluminum-doped zinc oxide thin films
prepared by pulsed laser deposition in oxygen ambient. Journal of Applied
Physics, 90(11), pp.5661-5665.
207. Kelly, R., 1992. Gas dynamics of the pulsed emission of a perfect gas with
applications to laser sputtering and to nozzle expansion. Physical Review A, 46(2),
p.860.
208. Sibold, D. and Urbassek, H.M., 1991. Kinetic study of pulsed desorption
flows into vacuum. Physical Review A, 43(12), p.6722.
209. Kelly, R. and Miotello, A., 1994. Laser-pulse sputtering of atoms and
molecules Part II. Recondensation effects. Nuclear Instruments and Methods in
Physics Research Section B: Beam Interactions with Materials and Atoms, 91(1-
4), pp.682-691.
210. Kelly, R. and Miotello, A., 1993. Pulsed-laser sputtering of atoms and
molecules. Part I: Basic solutions for gas-dynamic effects. Applied Physics B, 57(2),
pp.145-158.
211. Ojeda-GP, A., Schneider, C.W., Döbeli, M., Lippert, T. and Wokaun, A.,
2017. Plasma plume dynamics, rebound, and recoating of the ablation target in
pulsed laser deposition. Journal of Applied Physics, 121(13), p.135306.
212. Ojeda-GP, A., Schneider, C.W., Lippert, T. and Wokaun, A., 2016.
Pressure and temperature dependence of the laser-induced plasma plume
dynamics. Journal of Applied Physics, 120(22), p.225301.
177
213. Harilal, S.S., Bindhu, C.V., Tillack, M.S., Najmabadi, F. and Gaeris,
A.C., 2003. Internal structure and expansion dynamics of laser ablation plumes
into ambient gases. Journal of applied physics, 93(5), pp.2380-2388.
214. Dawood, M.S., Hamdan, A. and Margot, J., 2015. Influence of
surrounding gas, composition and pressure on plasma plume dynamics of
nanosecond pulsed laser-induced aluminum plasmas. AIP Advances, 5(10),
p.107143.
215. Canulescu, S., Papadopoulou, E.L., Anglos, D., Lippert, T., Schneider,
C.W. and Wokaun, A., 2009. Mechanisms of the laser plume expansion during the
ablation of LiMn 2 O 4. Journal of Applied Physics, 105(6), p.063107.
216. Gurlui, S., Agop, M., Nica, P., Ziskind, M. and Focsa, C., 2008.
Experimental and theoretical investigations of a laser-produced aluminum
plasma. Physical Review E, 78(2), p.026405.
217. Sambri, A., Aruta, C., Di Gennaro, E., Wang, X., Scotti di Uccio, U.,
Miletto Granozio, F. and Amoruso, S., 2016. Effects of oxygen background pressure
on the stoichiometry of a LaGaO3 laser ablation plume investigated by time and
spectrally resolved two-dimensional imaging. Journal of Applied Physics, 119(12),
p.125301.
218. Geohegan, D.B., 1993. Imaging and blackbody emission spectra of
particulates generated in the KrF‐laser ablation of BN and YBa2Cu3O7 −
x. Applied physics letters, 62(13), pp.1463-1465.
219. Orsel, K., Groenen, R., Bastiaens, B., Koster, G., Rijnders, G. and Boller,
K.J., 2015. Influence of the oxidation state of SrTiO3 plasmas for stoichiometric
growth of pulsed laser deposition films identified by laser induced
fluorescence. APL materials, 3(10), p.106103.
220. Amoruso, S., Toftmann, B. and Schou, J., 2004. Thermalization of a UV
laser ablation plume in a background gas: From a directed to a diffusionlike
flow. Physical Review E, 69(5), p.056403.
221. Geohegan, D.B. and Puretzky, A.A., 1995. Species-resolved imaging and
gated photon counting spectroscopy of laser ablation plume dynamics during KrF-
and ArF-laser PLD of amorphous diamond films. MRS Online Proceedings
Library, 397(1), pp.55-68.
222. O’Mahony, D., Lunney, J., Dumont, T., Canulescu, S., Lippert, T. and
Wokaun, A., 2007. Laser-produced plasma ion characteristics in laser ablation of
lithium manganate. Applied Surface Science, 254(4), pp.811-815.
223. D. M. Packwood, S. Shiraki, T. Hitosugi, Phys. Rev. Lett. 2013, 111,
036101.
178
224. Ojeda-GP, A., Schneider, C.W., Döbeli, M., Lippert, T. and Wokaun, A.,
2016. The importance of pressure and mass ratios when depositing multi-element
oxide thin films by pulsed laser deposition. Applied Surface Science, 389, pp.126-
134.
225. Donnelly, T., Lunney, J.G., Amoruso, S., Bruzzese, R., Wang, X. and Ni,
X., 2010. Angular distributions of plume components in ultrafast laser ablation of
metal targets. Applied Physics A, 100(2), pp.569-574.
226. Rosenfeld, G., Poelsema, B. and Comsa, G., 1995. The concept of two
mobilities in homoepitaxial growth. Journal of crystal growth, 151(1-2), pp.230-
233.
227. King, D.A. and Woodruff, D.P., 1997. Growth and properties of ultrathin
epitaxial layers. Elsevier.
228. Koster, G., Rijnders, G.J., Blank, D.H. and Rogalla, H., 1999. Imposed
layer-by-layer growth by pulsed laser interval deposition. Applied physics
letters, 74(24), pp.3729-3731.
229. Blank, D.H., Rijnders, G.J., Koster, G. and Rogalla, H., 2000. A new
approach in layer-by-layer growth of oxide materials by pulsed laser
deposition. Journal of electroceramics, 4(2), pp.311-318.
230. Blank, D.H., Koster, G., Rijnders, G.A., van Setten, E., Slycke, P. and
Rogalla, H., 2000. Epitaxial growth of oxides with pulsed laser interval
deposition. Journal of crystal growth, 211(1-4), pp.98-105.
231. Rijnders, G., Koster, G., Leca, V., Blank, D.H. and Rogalla, H., 2000.
Imposed layer-by-layer growth with pulsed laser interval deposition. Applied
surface science, 168(1-4), pp.223-226.
232. Koster, G., Verbist, K., Rijnders, G., Rogalla, H., van Tendeloo, G. and
Blank, D.H., 2001. Structure and properties of (Sr, Ca) CuO2–BaCuO2
superlattices grown by pulsed laser interval deposition. Physica C:
Superconductivity, 353(3-4), pp.167-183.
233. Lippmaa, M., Nakagawa, N., Kawasaki, M., Ohashi, S. and Koinuma, H.,
2000. Growth mode mapping of SrTiO 3 epitaxy. Applied Physics Letters, 76(17),
pp.2439-2441.
234. Koster, G., Huijben, M. and Rijnders, G. eds., 2015. Epitaxial growth of
complex metal oxides. Elsevier.
235. Groenen, R., Smit, J., Orsel, K., Vailionis, A., Bastiaens, B., Huijben,
M., Boller, K., Rijnders, G. and Koster, G., 2015. Research Update: Stoichiometry
controlled oxide thin film growth by pulsed laser deposition. APL materials, 3(7),
p.070701.
179
236. Lucas, L. and Zhang, J., 2012. Femtosecond laser micromachining: A
back-to-basics primer. Applied Energetics, 27(4), p.29.
237. Ichimiya, A., Cohen, P.I. and Cohen, P.I., 2004. Reflection high-energy
electron diffraction. Cambridge University Press.
238. Neave, J.H., Joyce, B.A., Dobson, P.J. and Norton, N., 1983. Dynamics
of film growth of GaAs by MBE from RHEED observations. Applied Physics
A, 31(1), pp.1-8.
239. Lent, C.S. and Cohen, P.I., 1984. Diffraction from stepped surfaces: I.
Reversible surfaces. Surface science, 139(1), pp.121-154.
240. Pukite, P.R., Lent, C.S. and Cohen, P.I., 1985. Diffraction from stepped
surfaces: II. Arbitrary terrace distributions. Surface science, 161(1), pp.39-68.
241. Kawamura, T., Sakamoto, T. and Ohta, K., 1986. Origin of azimuthal
effect of RHEED intensity oscillations observed during MBE. Surface
science, 171(1), pp.L409-L414.
242. Cohen, P.I., Petrich, G.S., Pukite, P.R., Whaley, G.J. and Arrott, A.S.,
1989. Birth-death models of epitaxy: I. Diffraction oscillations from low index
surfaces. Surface science, 216(1-2), pp.222-248.
243. Jiang, Q.D. and Zegenhagen, J., 1999. c (6× 2) and c (4× 2) reconstruction
of SrTiO3 (001). Surface Science, 425(2-3), pp.343-354.
244. Najjar, R., André, R., Boukari, H., Mariette, H. and Tatarenko, S., 2008.
Intensity beats on RHEED oscillations during MBE growth of ZnTe. Surface
science, 602(3), pp.744-746.
245. Kajdos, A.P. and Stemmer, S., 2014. Surface reconstructions in molecular
beam epitaxy of SrTiO3. Applied Physics Letters, 105(19), p.191901.
246. Song, J.H. and Jeong, Y.H., 2003. SrTiO3 homoepitaxy by the pulsed
laser deposition method: island, layer-by-layer, and step-flow growth. Solid state
communications, 125(10), pp.563-566.
247. Hedin, L. and Johansson, A., 1969. Polarization corrections to core
levels. Journal of Physics B: Atomic and Molecular Physics, 2(12), p.1336.
248. Haas, T.W., Grant, J.T. and Dooley, G.J., 1970. Auger-electron
spectroscopy of transition metals. Physical Review B, 1(4), p.1449.
249. Packer, M.E. and Wilson, J.M., 1971. Auger transitions.
250. Coad, J.P. and Riviere, J.C., 1971. Auger spectroscopy of carbon on
nickel. Surface Science, 25(3), pp.609-624.
180
251. Chung, M.F. and Jenkins, L.H., 1970. Auger electron energies of the outer
shell electrons. Surface Science, 22(2), pp.479-485.
252. WA Coghlan and RE Clausing, ORNL-TM-3576 (1971)
253. Coghlan, W.A. and Clausing, R.E., 1972. A description of a catalog of
calculated Auger transitions for the elements. Surface Science, 33(2), pp.411-413.
254. Coghlan, W.A. and Clausing, R.E., 1973. Auger catalog calculated
transition energies listed by energy and element. Atomic Data and Nuclear Data
Tables, 5(4), pp.317-469.
255. Shirley, D.A., 1972. Relaxation effects on auger energies. Chemical
Physics Letters, 17(3), pp.312-315.
256. Shirley, D.A., 1973. Theory of KLL auger energies including static
relaxation. Physical Review A, 7(5), p.1520.
257. Larkins, F.P., 1975. Atomic Inner-Shell Processes. en. B. Crasemann
Academic, New York, 1, p.377.
258. Larkins, F.P., 1977. Semiempirical Auger-electron energies for elements
10≤ Z ≤ 100. Atomic Data and Nuclear Data Tables, 20(4), pp.311-387.
259. Fiermans, L. and Vennik, J., 1977. Electron Beams as Analytical Tools
in Surface Research: LEED and AES. In Advances in Electronics and Electron
Physics (Vol. 43, pp. 139-203). Academic Press.
260. Hörnfeldt, O., Fahlman, A., and Nordling C, 1965. Arkiv Fysik 23, 155.
261. Coster, D. and Kronig, R.D.L., 1935. New type of auger effect and its
influence on the x-ray spectrum. Physica, 2(1-12), pp.13-24.
262. Burhop, E.H.S. and Asaad, W.N., 1972. The auger effect. In Advances in
Atomic and Molecular Physics (Vol. 8, pp. 163-284). Academic Press.
263. Yin, L., Adler, I., Tsang, T., Chen, M.H. and Crasemann, B., 1973. Quasi-
atomic L-MM Auger spectra of solid Cu and Zn. Physics Letters A, 46(2), pp.113-
114.
264. Cini, M., 1976. Density of states of two interacting holes in a solid. Solid
State Communications, 20(6), pp.605-607.
265. Cini, M., 1978. Comment on quasiatomic Auger spectra in narrow-band
metals. Physical Review B, 17(6), p.2788.
266. Sawatzky, G.A., 1977. Quasiatomic Auger spectra in narrow-band
metals. Physical Review Letters, 39(8), p.504.
181
267. Antonides, E., Janse, E.C. and Sawatzky, G.A., 1977. LMM Auger
spectra of Cu, Zn, Ga, and Ge. I. Transition probabilities, term splittings, and
effective Coulomb interaction. Physical Review B, 15(4), p.1669.
268. Antonides, E., Janse, E.C. and Sawatzky, G.A., 1977. LMM auger spectra
of Cu, Zn, Ga, and Ge, II. Relationship with the L 23 photoelectron spectra via
the L 2 L 3 M 45 Coster-Kronig process. Physical Review B, 15(10), p.4596.
269. Krolikowski, W.F. and Spicer, W.E., 1970. Photoemission studies of the
noble metals. II. Gold. Physical Review B, 1(2), p.478.
270. Quinn, J.J., 1962. Range of excited electrons in metals. Physical
Review, 126(4), p.1453.
271. Kanter, H., 1970. Slow-electron mean free paths in aluminum, silver, and
gold. Physical Review B, 1(2), p.522.
272. Powell, C.J., 1974. Attenuation lengths of low-energy electrons in
solids. Surface Science, 44(1), pp.29-46.
273. Powell, C.J., 1985. Energy and material dependence of the inelastic mean
free path of low‐energy electrons in solids. Journal of Vacuum Science &
Technology A: Vacuum, Surfaces, and Films, 3(3), pp.1338-1342.
274. Powell, C.J., 1985. The energy dependence of electron attenuation
lengths. Surface and interface analysis, 7(6), pp.256-262.
275. Seah, M.P. and Dench, W.A., 1979. Quantitative electron spectroscopy
of surfaces: A standard data base for electron inelastic mean free paths in
solids. Surface and interface analysis, 1(1), pp.2-11.
276. Koch, J., 1974. A small retarding field analyzer for Auger electron
spectroscopy. Review of Scientific Instruments, 45(10), pp.1212-1213.
277. NARUMAND, D.H. and CHILDS, K.D., 2004. Auger spectrometers: A
tutorial review. Applied Spectroscopy Reviews, 34(3), pp.139-158.
278. Staib, P., 1972. An improved retarding field analyser. Journal of Physics
E: Scientific Instruments, 5(5), p.484.
279. Leder, L.B. and Simpson, J.A., 1958. Improved electrical differentiation
of retarding potential measurements. Review of Scientific Instruments, 29(7),
pp.571-574.
280. Spicer, W.E. and Berglund, C.N., 1964. Measurement of photoemitted
electron energy distributions by an ac method. Review of Scientific
Instruments, 35(12), pp.1665-1667.
182
281. Harris, L.A., 1968. Some observations of surface segregation by Auger
electron emission. Journal of Applied Physics, 39(3), pp.1428-1431.
282. Tharp, L.N. and Scheibner, E.J., 1967. Energy spectra of inelastically
scattered electrons and LEED studies of tungsten. Journal of Applied
Physics, 38(8), pp.3320-3330.
283. Palmberg, P.W. and Rhodin, T.N., 1968. Auger electron spectroscopy of
fcc metal surfaces. Journal of Applied Physics, 39(5), pp.2425-2432.
284. Bishop, H.E. and Riviere, J.C., 1969. Estimates of the Efficiencies of
Production and Detection of Electron‐Excited Auger Emission. Journal of Applied
Physics, 40(4), pp.1740-1744.
285. Somorjai, G.A. and Szalkowski, F.J., 1971. Simple rules to predict the
structure of adsorbed gases on crystal surfaces. The Journal of Chemical
Physics, 54(1), pp.389-399.
286. Prutton, M., 1977. A theoretical comparison of electron energy analysers
degraded so as to obtain high sensitivity. Journal of Electron Spectroscopy and
Related Phenomena, 11(2), pp.197-204.
287. Weber, R.E. and Peria, W.T., 1967. Use of LEED apparatus for the
detection and identification of surface contaminants. Journal of Applied
Physics, 38(11), pp.4355-4358.
288. Tersoff, J. and Hamann, D.R., 1983. Theory and application for the
scanning tunneling microscope. Physical review letters, 50(25), p.1998.
289. Tersoff, J. and Hamann, D.R., 1985. Theory of the scanning tunneling
microscope. Physical Review B, 31(2), p.805.
290. Meyer, E.R.N.S.T., 1992. Atomic force microscopy. Progress in surface
science, 41(1), pp.3-49.
291. Rugar, D. and Hansma, P., 1990. Atomic force microscopy. Physics
today, 43(10), pp.23-30.
292. Reimer, L., 2000. Scanning electron microscopy: physics of image
formation and microanalysis.
293. Carter, C.B. and Williams, D.B. eds., 2016. Transmission electron
microscopy: Diffraction, imaging, and spectrometry. Springer.
294. Erni, R., Rossell, M.D., Kisielowski, C. and Dahmen, U., 2009. Atomic-
resolution imaging with a sub-50-pm electron probe. Physical review letters, 102(9),
p.096101.
183
295. Girit, Ç.Ö., Meyer, J.C., Erni, R., Rossell, M.D., Kisielowski, C., Yang,
L., Park, C.H., Crommie, M.F., Cohen, M.L., Louie, S.G. and Zettl, A., 2009.
Graphene at the edge: stability and dynamics. science, 323(5922), pp.1705-1708.
296. Meyer, J.C., Kisielowski, C., Erni, R., Rossell, M.D., Crommie, M.F. and
Zettl, A., 2008. Direct imaging of lattice atoms and topological defects in graphene
membranes. Nano letters, 8(11), pp.3582-3586.
297. Thompson, J., Braun, G., Tierney, D., Wessels, L., Schmitzer, H., Rossa,
B., Wagner, H.P. and Dultz, W., 2018. Rosalind Franklin's X-ray photo of DNA
as an undergraduate optical diffraction experiment. American Journal of
Physics, 86(2), pp.95-104.
298. Gerstner, E., 2011. X-ray crystallography goes viral. Nature
Physics, 7(3), pp.194-194.
299. Baron, A.Q., 2015. Introduction to high-resolution inelastic x-ray
scattering. arXiv preprint arXiv:1504.01098.
300. Shindo, D. and Oikawa, T., 2013. Analytical electron microscopy for
materials science. Springer Science & Business Media.
301. Samuelson, D.A., 1998. Energy dispersive X-ray microanalysis. In Free
radical and antioxidant protocols (pp. 413-424). Humana Press.
302. 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. and Egerton, R.F., 2014.
Vibrational spectroscopy in the electron microscope. Nature, 514(7521), pp.209-
212.
303. Iakoubovskii, K., Mitsuishi, K., Nakayama, Y. and Furuya, K., 2008.
Thickness measurements with electron energy loss spectroscopy. Microscopy
research and technique, 71(8), pp.626-631.
304. Iakoubovskii, K., Mitsuishi, K. and Furuya, K., 2008. Structure and
pressure inside Xe nanoparticles embedded in Al. Physical Review B, 78(6),
p.064105.
305. Hohenberg, P. and Kohn, W., 1964. Inhomogeneous electron
gas. Physical review, 136(3B), p.B864.
306. Grimme, S., 2006. Semiempirical hybrid density functional with
perturbative second-order correlation. The Journal of chemical physics, 124(3),
p.034108.
307. Tkatchenko, A. and Scheffler, M., 2009. Accurate molecular van der
Waals interactions from ground-state electron density and free-atom reference
data. Physical review letters, 102(7), p.073005.
184
308. Kresse, G. and Furthmüller, J., 1996. Efficient iterative schemes for ab
initio total-energy calculations using a plane-wave basis set. Physical review
B, 54(16), p.11169.
309. Blöchl, P.E., 1994. Projector augmented-wave method. Physical review
B, 50(24), p.17953.
310. Perdew, J.P., Burke, K. and Ernzerhof, M., 1996. Generalized gradient
approximation made simple. Physical review letters, 77(18), p.3865.
311. Tass, Z., Horvath, G. and Josepovits, V.K., 1995. Investigation of the
titanium oxidation states by Auger electron spectroscopy. Surface science, 331,
pp.272-276.
312. Houston, J.E., 1973. Exact corrections for potential modulation distortion
in Auger yield measurements. Surface Science, 38(2), pp.283-291.
313. Houston, J.E., 1974. Dynamic background subtraction and the retrieval
of threshold signals. Review of Scientific Instruments, 45(7), pp.897-903.
314. Grant, J.T. and Haas, T.W., 1974. Corrections of auger electron signal
strengths for modulation amplitude distortion in a 4-grid retarding potential energy
analyzer. Surface Science, 44(2), pp.617-623.
315. Grant, J.T., Hooker, M.P. and Haas, T.W., 1975. Appearance potential
spectroscopy: Relative signal strengths from 3d transition metals. Surface
Science, 51(2), pp.433-440.
316. Grant, J.T., Hooker, M.P. and Haas, T.W., 1976. Auger current
measurements for quantitative Auger electron spectroscopy of solids. Journal of
Colloid and Interface Science, 55(2), pp.370-376.
317. Winklehner, D., Leitner, D., Cole, D., Machicoane, G. and Tobos, L.,
2014. Space-charge compensation measurements in electron cyclotron resonance ion
source low energy beam transport lines with a retarding field analyzer. Review of
Scientific Instruments, 85(2), p.02A739.
318. Marturi, N., 2013. Vision and visual servoing for nanomanipulation and
nanocharacterization in scanning electron microscope (Doctoral dissertation,
Université de Franche-Comté).
319. Tellekamp, M.B., Greenlee, J.D., Shank, J.C. and Doolittle, W.A., 2015.
Molecular beam epitaxy growth of niobium oxides by solid/liquid state oxygen
source and lithium assisted metal-halide chemistry. Journal of Crystal
Growth, 425, pp.225-229.
320. Svensson, S.P., Sarney, W.L., Yu, K.M., Ting, M., Calley, W.L., Novikov,
S.V., Foxon, C.T. and Walukiewicz, W., 2015. Determination of N-/Ga-rich growth
185
conditions, using in-situ auger electron spectroscopy. Journal of Crystal
Growth, 425, pp.2-4.
321. Matsushima, T., Almy, D.B. and White, J.M., 1977. The reactivity and
auger chemical shift of oxygen adsorbed on platinum. Surface Science, 67(1),
pp.89-108.
322. Axelsson, K.O., Keck, K.E. and Kasemo, B., 1985. AES and EEL spectra
from Zr, Zr+ O2 and ZrO2; peak energy shifts and intensity variations at various
stages of oxidation in the pressure range 10 − 6− 103 Torr. Surface science, 164(1),
pp.109-126.
323. Zvanut, M.E., Jeddy, S., Towett, E., Janowski, G.M., Brooks, C. and
Schlom, D., 2008. An annealing study of an oxygen vacancy related defect in SrTiO
3 substrates. Journal of Applied Physics, 104(6), p.064122.
324. Palmberg, P. W. (1972). Handbook of Auger electron spectroscopy.
325. Davis, L. E. (1976). Handbook of Auger electron spectroscopy.
326. McGuire (1979). Handbook of Auger electron spectroscopy.
327. Sekine, (1982). Handbook of Auger electron spectroscopy.
328. Shimizu, R., 1983. Quantitative analysis by Auger electron
spectroscopy. Japanese journal of applied physics, 22(11R), p.1631.
329. Grant, J.T., Haas, T.W. and Houston, J.E., 1973. Quantitative Auger
analysis using integration techniques. Physics Letters A, 45(4), pp.309-310.
330. Chang, C.C., 1975. General formalism for quantitative Auger
analysis. Surface Science, 48(1), pp.9-21.
331. Ravichandran, J., 2017. Thermoelectric and thermal transport properties
of complex oxide thin films, heterostructures and superlattices. Journal of
Materials Research, 32(1), pp.183-203.
332. Lindau, I. and Spicer, W.E., 1974. The probing depth in photoemission
and Auger-electron spectroscopy. Journal of Electron Spectroscopy and Related
Phenomena, 3(5), pp.409-413.
333. Jablonski, A. and Powell, C.J., 2003. Information depth and the mean
escape depth in Auger electron spectroscopy and X-ray photoelectron
spectroscopy. Journal of Vacuum Science & Technology A: Vacuum, Surfaces, and
Films, 21(1), pp.274-283.
334. Jablonski, A. and Powell, C.J., 2009. Practical expressions for the mean
escape depth, the information depth, and the effective attenuation length in Auger-
electron spectroscopy and x-ray photoelectron spectroscopy. Journal of Vacuum
Science & Technology A: Vacuum, Surfaces, and Films, 27(2), pp.253-261.
186
335. Bachelet, R., Sánchez, F., Santiso, J. and Fontcuberta, J., 2008.
Reversible growth-mode transition in SrRuO 3 epitaxy. Applied Physics
Letters, 93(15), p.151916.
336. Orvis, T., Kumarasubramanian, H., Surendran, M., Kutagulla, S.,
Cunniff, A. and Ravichandran, J., 2021. In Situ Monitoring of Composition and
Sensitivity to Growth Parameters of Pulsed Laser Deposition. ACS Applied
Electronic Materials.
337. Nakagawa, N., Hwang, H.Y. and Muller, D.A., 2006. Why some interfaces
cannot be sharp. Nature materials, 5(3), pp.204-209.
338. Thiel, S., Hammerl, G., Schmehl, A., Schneider, C.W. and Mannhart, J.,
2006. Tunable quasi-two-dimensional electron gases in oxide
heterostructures. Science, 313(5795), pp.1942-1945.
339. Huijben, M., Rijnders, G., Blank, D.H., Bals, S., Van Aert, S., Verbeeck,
J., Van Tendeloo, G., Brinkman, A. and Hilgenkamp, H., 2006. Electronically
coupled complementary interfaces between perovskite band insulators. Nature
materials, 5(7), pp.556-560.
340. Brinkman, A., Huijben, M., Van Zalk, M., Huijben, J., Zeitler, U., Maan,
J.C., van der Wiel, W.G., Rijnders, G.J.H.M., Blank, D.H. and Hilgenkamp, H.,
2007. Magnetic effects at the interface between non-magnetic oxides. Nature
materials, 6(7), pp.493-496.
341. Siemons, W., Koster, G., Yamamoto, H., Harrison, W.A., Lucovsky, G.,
Geballe, T.H., Blank, D.H. and Beasley, M.R., 2007. Origin of charge density at
LaAlO 3 on SrTiO 3 heterointerfaces: possibility of intrinsic doping. Physical
review letters, 98(19), p.196802.
342. Willmott, P.R., Pauli, S.A., Herger, R., Schlepütz, C.M., Martoccia, D.,
Patterson, B.D., Delley, B., Clarke, R., Kumah, D., Cionca, C. and Yacoby, Y.,
2007. Structural basis for the conducting interface between LaAlO 3 and SrTiO
3. Physical Review Letters, 99(15), p.155502.
343. Huijben, M., Brinkman, A., Koster, G., Rijnders, G., Hilgenkamp, H. and
Blank, D.H., 2009. Structure–property relation of SrTiO3/LaAlO3
interfaces. Advanced Materials, 21(17), pp.1665-1677.
344. Bell, C., Harashima, S., Hikita, Y. and Hwang, H.Y., 2009. Thickness
dependence of the mobility at the LaAlO 3/SrTiO 3 interface. Applied Physics
Letters, 94(22), p.222111.
345. Chen, Y.Z., Zhao, J.L., Sun, J.R., Pryds, N. and Shen, B.G., 2010.
Resistance switching at the interface of LaAlO 3/SrTiO 3. Applied Physics
Letters, 97(12), p.123102.
187
346. Singh-Bhalla, G., Bell, C., Ravichandran, J., Siemons, W., Hikita, Y.,
Salahuddin, S., Hebard, A.F., Hwang, H.Y. and Ramesh, R., 2011. Built-in and
induced polarization across LaAlO 3/SrTiO 3 heterojunctions. Nature
Physics, 7(1), pp.80-86.
347. Pentcheva, R., Huijben, M., Otte, K., Pickett, W.E., Kleibeuker, J.E.,
Huijben, J., Boschker, H., Kockmann, D., Siemons, W., Koster, G. and Zandvliet,
H.J., 2010. Parallel electron-hole bilayer conductivity from electronic interface
reconstruction. Physical review letters, 104(16), p.166804.
348. Chambers, S.A., Engelhard, M.H., Shutthanandan, V., Zhu, Z., Droubay,
T.C., Qiao, L., Sushko, P.V., Feng, T., Lee, H.D., Gustafsson, T. and Garfunkel,
E., 2010. Instability, intermixing and electronic structure at the epitaxial
LaAlO3/SrTiO3 (001) heterojunction. Surface Science Reports, 65(10-12), pp.317-
352.
349. Bark, C.W., Felker, D.A., Wang, Y., Zhang, Y., Jang, H.W., Folkman,
C.M., Park, J.W., Baek, S.H., Zhou, H., Fong, D.D. and Pan, X.Q., 2011. Tailoring
a two-dimensional electron gas at the LaAlO3/SrTiO3 (001) interface by epitaxial
strain. Proceedings of the National Academy of Sciences, 108(12), pp.4720-4724.
350. Qiao, L., Droubay, T.C., Kaspar, T.C., Sushko, P.V. and Chambers, S.A.,
2011. Cation mixing, band offsets and electric fields at LaAlO3/SrTiO3 (001)
heterojunctions with variable La: Al atom ratio. Surface science, 605(15-16),
pp.1381-1387.
351. Chen, Y.Z., Christensen, D.V., Trier, F., Pryds, N., Smith, A. and
Linderoth, S., 2012. On the origin of metallic conductivity at the interface of
LaAlO3/SrTiO3. Applied surface science, 258(23), pp.9242-9245.
352. Huang, B.C., Chiu, Y.P., Huang, P.C., Wang, W.C., Tra, V.T., Yang,
J.C., He, Q., Lin, J.Y., Chang, C.S. and Chu, Y.H., 2012. Mapping band alignment
across complex oxide heterointerfaces. Physical review letters, 109(24), p.246807.
353. Warusawithana, M.P., Richter, C., Mundy, J.A., Roy, P., Ludwig, J.,
Paetel, S., Heeg, T., Pawlicki, A.A., Kourkoutis, L.F., Zheng, M. and Lee, M.,
2013. LaAlO 3 stoichiometry is key to electron liquid formation at LaAlO 3/SrTiO
3 interfaces. Nature communications, 4(1), pp.1-9.
354. Park, J., Soh, Y.A., Aeppli, G., David, A., Lin, W. and Wu, T., 2014.
Influence of oxygen pressure and aging on LaAlO3 films grown by pulsed laser
deposition on SrTiO3 substrates. Applied Physics Letters, 104(8), p.081604.
355. Hong, S., Nakhmanson, S.M. and Fong, D.D., 2016. Screening
mechanisms at polar oxide heterointerfaces. Reports on Progress in Physics, 79(7),
p.076501.
188
356. Liu, G., Lei, Q., Wolak, M.A., Li, Q., Chen, L.Q., Winkler, C., Sloppy,
J., Taheri, M.L. and Xi, X., 2016. Epitaxial strain and its relaxation at the
LaAlO3/SrTiO3 interface. Journal of Applied Physics, 120(8), p.085302.
357. Lee, P.W., Singh, V.N., Guo, G.Y., Liu, H.J., Lin, J.C., Chu, Y.H., Chen,
C.H. and Chu, M.W., 2016. Hidden lattice instabilities as origin of the conductive
interface between insulating LaAlO 3 and SrTiO 3. Nature communications, 7(1),
pp.1-8.
358. Guo, H., Saidi, W.A. and Zhao, J., 2016. Tunability of the two-
dimensional electron gas at the LaAlO 3/SrTiO 3 interface by strain-induced
ferroelectricity. Physical Chemistry Chemical Physics, 18(41), pp.28474-28484.
359. Vaz, D.C., Lesne, E., Sander, A., Naganuma, H., Jacquet, E., Santamaria,
J., Barthélémy, A. and Bibes, M., 2017. Tuning up or down the critical thickness
in LaAlO3/SrTiO3 through in situ deposition of metal overlayers. Advanced
Materials, 29(28), p.1700486.
360. Ohsawa, T., Saito, M., Shimizu, R., Iwaya, K., Shiraki, S., Ikuhara, Y.
and Hitosugi, T., 2018. Impact of a surface TiO2 atomic sheet on the electronic
transport properties of LaAlO3/SrTiO3 heterointerfaces. Applied Physics
Letters, 113(14), p.141602.
361. Piyanzina, I.I., Eyert, V., Lysogorskiy, Y.V., Tayurskii, D.A. and Kopp,
T., 2019. Oxygen vacancies and hydrogen doping in LaAlO3/SrTiO3
heterostructures: electronic properties and impact on surface and interface
reconstruction. Journal of Physics: Condensed Matter, 31(29), p.295601.
362. Lee, H.N., Seo, S.S.A., Choi, W.S. and Rouleau, C.M., 2016. Growth
control of oxygen stoichiometry in homoepitaxial SrTiO 3 films by pulsed laser
epitaxy in high vacuum. Scientific reports, 6(1), pp.1-7.
363. Preziosi, D., Sander, A., Barthélémy, A. and Bibes, M., 2017.
Reproducibility and off-stoichiometry issues in nickelate thin films grown by pulsed
laser deposition. AIP Advances, 7(1), p.015210.
364. Posadas, A.B., Kormondy, K.J., Guo, W., Ponath, P., Geler-Kremer, J.,
Hadamek, T. and Demkov, A.A., 2017. Scavenging of oxygen from SrTiO3 during
oxide thin film deposition and the formation of interfacial 2DEGs. Journal of
Applied Physics, 121(10), p.105302.
365. Gabel, J., Zapf, M., Scheiderer, P., Schütz, P., Dudy, L., Stübinger, M.,
Schlueter, C., Lee, T.L., Sing, M. and Claessen, R., 2017. Disentangling specific
versus generic doping mechanisms in oxide heterointerfaces. Physical Review
B, 95(19), p.195109.
189
366. Lee, H., Campbell, N., Lee, J., Asel, T.J., Paudel, T.R., Zhou, H., Lee,
J.W., Noesges, B., Seo, J., Park, B. and Brillson, L.J., 2018. Direct observation of
a two-dimensional hole gas at oxide interfaces. Nature materials, 17(3), pp.231-
236.
367. Ohshima, R., Ando, Y., Matsuzaki, K., Susaki, T., Weiler, M., Klingler,
S., Huebl, H., Shikoh, E., Shinjo, T., Goennenwein, S.T. and Shiraishi, M., 2017.
Strong evidence for d-electron spin transport at room temperature at a LaAlO
3/SrTiO 3 interface. Nature materials, 16(6), pp.609-614.
368. Moon, S.Y., Moon, C.W., Chang, H.J., Kim, T., Kang, C.Y., Choi, H.J.,
Kim, J.S., Baek, S.H. and Jang, H.W., 2016. Thermal stability of 2DEG at
amorphous LaAlO 3/crystalline SrTiO 3 heterointerfaces. Nano convergence, 3(1),
pp.1-6.
369. Singh, V. and Pulikkotil, J.J., 2017. Mixed valence as a necessary criteria
for quasi-two dimensional electron gas in oxide hetero-interfaces. Solid State
Communications, 251, pp.28-31.
370. Afonso, C.N., Gonzalo, J., Vega, F., Dieguez, E., Cheang Wong, J.C.,
Ortega, C., Siejka, J. and Amsel, G., 1995. Correlation between optical properties,
composition, and deposition parameters in pulsed laser deposited LiNbO3
films. Applied physics letters, 66(12), pp.1452-1454.
371. Lee, J.Y., Wang, T.C., Chen, S.F., Juang, J.Y., Lin, J.Y., Wu, K.H.,
Uen, T.M. and Gou, Y.S., 2001. Growth kinetics of homoepitaxial strontium
titanate films by interrupted pulsed laser deposition. Chinese Journal of
Physics, 39(4), pp.299-304.
372. Sambri, A., Cristensen, D.V., Trier, F., Chen, Y.Z., Amoruso, S., Pryds,
N., Bruzzese, R. and Wang, X., 2012. Plasma plume effects on the conductivity of
amorphous-LaAlO3/SrTiO3 interfaces grown by pulsed laser deposition in O2 and
Ar. Applied Physics Letters, 100(23), p.231605.
373. Huo, H., Rong, Z., Kononova, O., Sun, W., Botari, T., He, T., Tshitoyan,
V. and Ceder, G., 2019. Semi-supervised machine-learning classification of
materials synthesis procedures. npj Computational Materials, 5(1), pp.1-7.
374. Suh, C., Fare, C., Warren, J.A. and Pyzer-Knapp, E.O., 2020. Evolving
the materials genome: how machine learning is fueling the next generation of
materials discovery. Annual Review of Materials Research, 50, pp.1-25.
375. Hiszpanski, A.M., Gallagher, B., Chellappan, K., Li, P., Liu, S., Kim, H.,
Han, J., Kailkhura, B., Buttler, D.J. and Han, T.Y.J., 2020. Nanomaterial synthesis
insights from machine learning of scientific articles by extracting, structuring, and
190
visualizing knowledge. Journal of chemical information and modeling, 60(6),
pp.2876-2887.
376. Kim, E., Jensen, Z., van Grootel, A., Huang, K., Staib, M., Mysore, S.,
Chang, H.S., Strubell, E., McCallum, A., Jegelka, S. and Olivetti, E., 2020.
Inorganic materials synthesis planning with literature-trained neural
networks. Journal of chemical information and modeling, 60(3), pp.1194-1201.
377. Hu, M., Zhang, Q., Gu, L., Guo, Q., Cao, Y., Kareev, M., Chakhalian, J.
and Guo, J., 2018. Reconstruction-stabilized epitaxy of LaCoO3/SrTiO3 (111)
heterostructures by pulsed laser deposition. Applied Physics Letters, 112(3),
p.031603.
378. Gerhold, S., Riva, M., Yildiz, B., Schmid, M. and Diebold, U., 2016.
Adjusting island density and morphology of the SrTiO3 (110)-(4× 1) surface:
Pulsed laser deposition combined with scanning tunneling microscopy. Surface
Science, 651, pp.76-83.
379. Lei, Q.Y., Liu, G.Z. and Xi, X.X., 2013. Structural Characterization of
Homoepitaxial SrTiO3 Films Grown by Pulsed Laser Deposition. Integrated
Ferroelectrics, 141(1), pp.128-133.
380. Schou, J., 2009. Physical aspects of the pulsed laser deposition technique:
The stoichiometric transfer of material from target to film. Applied Surface
Science, 255(10), pp.5191-5198.
381. Arnold, C.B. and Aziz, M.J., 1999. Stoichiometry issues in pulsed-laser
deposition of alloys grown from multicomponent targets. Applied Physics A, 69(1),
pp.S23-S27.
382. Schou, J., Toftmann, B. and Amoruso, S., 2004, September. Dynamics of
a laser-produced silver plume in an oxygen background gas. In High-Power Laser
Ablation V (Vol. 5448, pp. 110-120). International Society for Optics and
Photonics.
383. Pryds, N., Schou, J. and Linderoth, S., 2007. The spatial thickness
distribution of metal films produced by large area pulsed laser deposition. Applied
surface science, 253(19), pp.8231-8234.
384. Marrocchelli, D., Perry, N.H. and Bishop, S.R., 2015. Understanding
chemical expansion in perovskite-structured oxides. Physical Chemistry Chemical
Physics, 17(15), pp.10028-10039.
385. Jalan, B., Engel-Herbert, R., Wright, N.J. and Stemmer, S., 2009.
Growth of high-quality SrTiO 3 films using a hybrid molecular beam epitaxy
approach. Journal of Vacuum Science & Technology A: Vacuum, Surfaces, and
Films, 27(3), pp.461-464.
191
386. Ohnishi, T., Shibuya, K., Yamamoto, T. and Lippmaa, M., 2008. Defects
and transport in complex oxide thin films. Journal of Applied Physics, 103(10),
p.103703.
387. Atkinson, A. and Ramos, T.M.G.M., 2000. Chemically-induced stresses
in ceramic oxygen ion-conducting membranes. Solid State Ionics, 129(1-4), pp.259-
269.
388. Nguyen, T.L., Dokiya, M., Wang, S., Tagawa, H. and Hashimoto, T.,
2000. The effect of oxygen vacancy on the oxide ion mobility in LaAlO3-based
oxides. Solid State Ionics, 130(3-4), pp.229-241.
389. Jiang, L.Q., Guo, J.K., Liu, H.B., Zhu, M., Zhou, X., Wu, P. and Li,
C.H., 2006. Prediction of lattice constant in cubic perovskites. Journal of Physics
and Chemistry of Solids, 67(7), pp.1531-1536.
390. Brooks, C.M., Kourkoutis, L.F., Heeg, T., Schubert, J., Muller, D.A. and
Schlom, D.G., 2009. Growth of homoepitaxial SrTiO 3 thin films by molecular-
beam epitaxy. Applied physics letters, 94(16), p.162905.
391. Robertson, J. and Clark, S.J., 2011. Limits to doping in oxides. Physical
Review B, 83(7), p.075205.
392. Klitzing, K.V., Dorda, G. and Pepper, M., 1980. New method for high-
accuracy determination of the fine-structure constant based on quantized Hall
resistance. Physical review letters, 45(6), p.494.
393. Zhong, Z., Xu, P.X. and Kelly, P.J., 2010. Polarity-induced oxygen
vacancies at LaAlO 3 ∕ SrTiO 3 interfaces. Physical Review B, 82(16), p.165127.
394. Yan, Y., Wei, L., Guo, L., Zhang, F., Dai, J. and Zeng, C., 2019. Gate-
tunable anomalous transverse voltage at the superconducting LaAlO3/SrTiO3
interface. Applied Physics Letters, 115(6), p.061603.
395. Chen, Y., Trier, F., Kasama, T., Christensen, D.V., Bovet, N., Balogh,
Z.I., Li, H., Thydén, K.T.S., Zhang, W., Yazdi, S. and Norby, P., 2015. Creation
of high mobility two-dimensional electron gases via strain induced polarization at
an otherwise nonpolar complex oxide interface. Nano letters, 15(3), pp.1849-1854.
396. Seo, S.S.A., Nichols, J., Hwang, J., Terzic, J., Gruenewald, J.H., Souri,
M., Thompson, J., Connell, J.G. and Cao, G., 2016. Selective growth of epitaxial
Sr2IrO4 by controlling plume dimensions in pulsed laser deposition. Applied
Physics Letters, 109(20), p.201901.
397. Zhang, X. and Demkov, A.A., 2002. Steps on the (001) SrTiO 3
surface. Journal of Vacuum Science & Technology B: Microelectronics and
Nanometer Structures Processing, Measurement, and Phenomena, 20(4), pp.1664-
1670.
192

Appendix A
M ATLAB Scripts

193
% Creates Data Trees with the following column structure:
% (energy)(E*N(E))(FitLine)(ENE - FitLine, norm to 0)

% As well as AUC with:
% (Trapz from (4) with limited width)(first normalized)

% Fit Line Start/End Points for Sr and Ti:

SrFLS = 10;
SrFLE = 26;

TiFLS = 1;
TiFLE = 16;

% Peak Position (index):

SrP = 19;
TiP = 7;

% Diff (dN/dE) Peak/Trough Positions (index):

SrDP = 16;
TiDP = 2;

SrDT = 23;
TiDT = 9;

% From +/- 0 to o

for o = 0:3

% Trapz (numeric integration) Start/End Points:

SrTS = SrP-o;
SrTE = SrP+o;
TiTS = TiP-o;
TiTE = TiP+o;

clear SrLaDat;
clear TiAlDat;
clear TiPDat;
clear TFxSr;
clear TFxTi;
clear TFxTP;

% scan numbers, from na to nb:

na = 1;
nb = 21;

for n = na:nb
194

curdat = ['x' num2str(n)];
SrLaDat = TF271.SrLa.(curdat);
TiAlDat = TF271.TiAl.(curdat);

TFxSr.(curdat) = SrLaDat(:,2:3);
TFxTi.(curdat) = TiAlDat(:,2:3);

TFxSr.(curdat)(:,2) = TFxSr.(curdat)(:,1) .* TFxSr.(curdat)(:,2);
TFxTi.(curdat)(:,2) = TFxTi.(curdat)(:,1) .* TFxTi.(curdat)(:,2);

TFxSr.(curdat)(2:end,3) = diff(TFxSr.(curdat)(:,2));
TFxTi.(curdat)(2:end,3) = diff(TFxTi.(curdat)(:,2));

% Fit Lines from two selected points:

SrLine = TFxSr.(curdat)(SrFLS,1:2);
SrLine(2,:) = TFxSr.(curdat)(SrFLE,1:2);

TiLine = TFxTi.(curdat)(TiFLS,1:2);
TiLine(2,:) = TFxTi.(curdat)(TiFLE,1:2);

% Creating Background Fits:

SrFit = fit(SrLine(:,1),SrLine(:,2),'poly1');
TiFit = fit(TiLine(:,1),TiLine(:,2),'poly1');

TFxSr.(curdat)(SrFLS:SrFLE,4) = SrFit.p1 .* SrLaDat(SrFLS:SrFLE,2) +
SrFit.p2;
TFxTi.(curdat)(TiFLS:TiFLE,4) = TiFit.p1 .* TiAlDat(TiFLS:TiFLE,2) +
TiFit.p2;

TFxSr.(curdat)(SrFLS:SrFLE,5) = TFxSr.(curdat)(SrFLS:SrFLE,2) -
TFxSr.(curdat)(SrFLS:SrFLE,4) - min(TFxSr.(curdat)(SrFLS:SrFLE,2) -
TFxSr.(curdat)(SrFLS:SrFLE,4));
TFxTi.(curdat)(TiFLS:TiFLE,5) = TFxTi.(curdat)(TiFLS:TiFLE,2) -
TFxTi.(curdat)(TiFLS:TiFLE,4) - min(TFxTi.(curdat)(TiFLS:TiFLE,2) -
TFxTi.(curdat)(TiFLS:TiFLE,4));

% in case it’s trying to look past the end of the data:

TiSTART = TiDP-o;
if TiSTART < 1
TiSTART = 1;
end

if o > 0
TFxSr.AUCene(n,1) =
trapz(TFxSr.(curdat)(SrTS:SrTE,1),TFxSr.(curdat)(SrTS:SrTE,5));
TFxTi.AUCene(n,1) =
trapz(TFxTi.(curdat)(TiTS:TiTE,1),TFxTi.(curdat)(TiTS:TiTE,5));

195

TFxSr.P2P(n,1) = max(TFxSr.(curdat)(SrDP-o:SrDP+o,3)) -
min(TFxSr.(curdat)(SrDT-o:SrDT+o,3));
TFxTi.P2P(n,1) = max(TFxTi.(curdat)(TiSTART:TiDP+o,3)) -
min(TFxTi.(curdat)(TiDT-o:TiDT+o,3));

end

if o == 0
TFxSr.AUCene(n,1) = TFxSr.(curdat)(SrP,5);
TFxTi.AUCene(n,1) = TFxTi.(curdat)(TiP,5);

TFxSr.P2P(n,1) = max(TFxSr.(curdat)(SrDP-o:SrDP+o,3)) -
min(TFxSr.(curdat)(SrDT-o:SrDT+o,3));
TFxTi.P2P(n,1) = max(TFxTi.(curdat)(TiSTART:TiDP+o,3)) -
min(TFxTi.(curdat)(TiDT-o:TiDT+o,3));

end

if o == 3
TFxSr.AUCene(n,1) =
trapz(TFxSr.(curdat)(SrFLS:SrFLE,1),TFxSr.(curdat)(SrFLS:SrFLE,5));
TFxTi.AUCene(n,1) =
trapz(TFxTi.(curdat)(TiFLS:TiFLE,1),TFxTi.(curdat)(TiFLS:TiFLE,5));

TFxSr.P2P(n,1) = max(TFxSr.(curdat)(SrDP-o:SrDP+o,3)) -
min(TFxSr.(curdat)(SrDT-o:SrDT+o,3));
TFxTi.P2P(n,1) = max(TFxTi.(curdat)(TiSTART:TiDP+o,3)) -
min(TFxTi.(curdat)(TiDT-o:TiDT+o,3));

end

end

widthrun = ['T' num2str(o)];
SrAUC.(widthrun) = TFxSr.AUCene;
TiAUC.(widthrun) = TFxTi.AUCene;

SrP2P.(widthrun) = TFxSr.P2P;
TiP2P.(widthrun) = TFxTi.P2P;

end

Abstract (if available)

Abstract Complex oxides play host to myriad unconventional physical properties ranging from colossal magneto-resistance to high-temperature superconductivity and serve a foundational role in numerous emerging technologies. When these materials are dimensionally confined as thin films and heterostructures, broken symmetry leads to the emergence of new metastable phases bearing functional properties and phenomena such as two dimensional electron gases and quantum phase transitions. The ability to engineer thin films and interfaces with sufficient precision to observe and study these phenomena requires a comprehensive understanding of the deposition and growth process, and the limitations thereof—this is the primary obstacle to the continued development of this field and the next-generation devices it promises. Structural characterization of films during the growth process has advanced considerably over the past several decades, with in situ monitoring of structure and thickness with atomic layer resolution during deposition now commonplace. This progress has been foundational to current achievements in the quality and control of thin film deposition techniques, but is likewise limited in scope. The compositional and chemical characterization of thin films is still largely limited to pseudo in situ and ex situ techniques which fail to fully unravel the complexities of growth dynamics during the deposition process. Although surface composition characterization techniques are prolific, they are all too often cumbersome in design or sensitive during application. Their incorporation into the harsh oxygen-rich environment required for complex oxide deposition, with techniques such as pulsed laser deposition, is therefore a difficult task. Recent developments in Auger electron spectroscopy probe-design, however, have allowed its incorporation into a pulsed laser deposition system, making the real time in situ observation of atomic and chemical composition possible for the first time. We have performed a series of experiments to elucidate its capabilities and sensitivity down to the sub-monolayer scale. Through this process we have demonstrated the ability to qualitatively and quantitively observe real time changes in composition and surface termination of the prototypical complex oxide perovskite SrTiO₃ during the deposition process. These achievements open the door to a more robust understanding of the complexities inherent to the thin film deposition process, including temporal phenomena such as termination-switching and dynamic layer rearrangement. A more complete view of the deposition process will serve to deepen the community’s understanding of these material systems, and play a significant role in enabling research—comparable to that of in situ structural characterization decades ago. The next generation of fundamental science and thin film technology will be built on our ability to observe and understand these processes in action, and throughout this process in situ and real time compositional characterization will play a pivotal role.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Phase change heterostructures for electronic and photonic applications

PDF

Controlling polymer film patterning, morphologies, and chemistry using vapor deposition

PDF

Fabrication, deposition, and characterization of size-selected metal nanoclusters with a magnetron sputtering gas aggregation source

PDF

Printed electronics based on carbon nanotubes and two-dimensional transition metal dichalcogenides

PDF

Development of organic-inorganic optical microcavities for studying polymer thin films

PDF

Scale-up of vapor-phase deposition of polymers: towards large-scale processing

PDF

In situ process analysis for defect control during composites manufacturing

PDF

Molecular dynamics study of energetic materials

PDF

Water desalination: real-time membrane characterization for performance prediction and system analysis for energetic enhancement

PDF

Sustainable manufacturing of out-of-autoclave (OoA) prepreg: processing challenges

PDF

Plasmonic enhancement of catalysis and solar energy conversion

PDF

Synthesis, characterization, and device application of two-dimensional materials beyond graphene

PDF

Electronic structure analysis of challenging open-shell systems: from excited states of copper oxide anions to exchange-coupling in binuclear copper complexes

PDF

Void evolution in vacuum bag-only prepregs

PDF

Machine-learning approaches for modeling of complex materials and media

PDF

In2O3 COVID-19 biosensors and two-dimensional materials: synthesis and applications

PDF

Defect control in vacuum bag only processing of composite prepregs

PDF

Modeling and simulation of complex recovery processes

PDF

Feature and model based biomedical system characterization of cancer

PDF

Artificial photosynthesis on titanium oxide passivated III-V semiconductors

Asset Metadata

Creator Orvis, Thomas Ira (author)

Core Title Real time surface analysis of complex oxide thin films during pulsed laser deposition

School Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Materials Science

Publication Date 05/03/2021

Defense Date 03/09/2021

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag Auger electron spectroscopy,complex oxides,epitaxy,OAI-PMH Harvest,pulsed laser deposition,surface analysis,thin films

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Ravichandran, Jayakanth (committee chair), Armani, Andrea (committee member), Nakano, Aiichiro (committee member)

Creator Email orvis@usc.edu,Thomas.orvis@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-457513

Unique identifier UC11669359

Identifier etd-OrvisThoma-9579.pdf (filename),usctheses-c89-457513 (legacy record id)

Legacy Identifier etd-OrvisThoma-9579.pdf

Dmrecord 457513

Document Type Dissertation

Rights Orvis, Thomas Ira

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA