SUB-MICRON AUGER ELECTRON SPECTROSCOPY CHARACTERIZATION OF
LITHIUM NIOBATE FERROELECTRIC DOMAINS AND THEIR FABRICATION
by
Torrey John McLoughlin
A dissertation submitted in partial fulfillment
of the requirements for the degree
of
Doctor of Philosophy
in
Physics
MONTANA STATE UNIVERSITY
Bozeman, Montana
May 2022
©COPYRIGHT
by
Torrey John McLoughlin
2022
All Rights Reserved
ii
DEDICATION
I dedicate this to my Mom and Dad for their endless love and support.
iii
ACKNOWLEDGEMENTS
I would like to first and foremost thank my advisors Dr. William R. Babbitt and
Dr. Wataru Nakagawa. Dr. Babbitt’s hard work ethic and availability, even when to his
detriment, was fundamental to me finishing this work. Dr. Nakagawa’s guidance through
this project has been invaluable and I’ve learned a great deal from his scientific insight.
Thank you also to my other committee members Dr. Nick Borys, Dr. Kevin Repasky,
and Dr. Rufus Cone. Each of them has played an influential role either in helping me
when research was stuck or developing my interest in physics. I would like to specifically
acknowledge Margaret Jarrett for her encouragement and advocacy for me since the very
beginning. There are numerous other people, and some I will surely forget to mention, to
whom I am indebted for their support: Andrew Hohne, Tianbo Liu, Carol Baumbauer, Jed
Pai, Lillian Zimmerman, Ana Flores, Andy Lingley, Chris Ebbers, Dieter Jundt, Charles
Thiel, Tino Woodburn, Lewis Cox, and Nate Rieders; sincerely thank you all.
Most importantly, I would like to thank my parents for being a constant source of
strength, and especially my fiancée Natalie, for her unconditional support. A.M.D.G.
Funding Acknowledgements
This work was kindly supported by the National Science Foundation (grant number 1710128),
by the Montana Space Grant Consortium (MSGC), and by the Montana Research &
Economic Development Initiative (MREDI).
iv
TABLE OF CONTENTS
1. INTRODUCTION .................................................................................................. 1
1.1 Motivation ........................................................................................................ 1
1.2 QPM for wavelength conversion.......................................................................... 2
1.3 Requirement of sub-micron periodic poling.......................................................... 4
1.4 Overview of Ferroelectric Surface Domain Characterization Methods .................... 6
1.4.1 Need for characterization methods .............................................................. 6
1.4.2 PFM......................................................................................................... 7
1.4.2.1 Ferroelectric and Piezoelectric Coupling .............................................. 7
1.4.2.2 PFM Overview .................................................................................. 8
1.4.2.3 PFM Resolution ................................................................................ 9
1.4.2.4Other SPM Techniques for Ferroelectrics........................................... 11
1.4.3 SEM ....................................................................................................... 12
1.4.4 UV-PEEM .............................................................................................. 14
1.4.5 Optical Methods...................................................................................... 15
1.4.6 Domain Selective Chemical Etching .......................................................... 16
1.4.6.1 LN.................................................................................................. 16
1.4.6.2KTP ............................................................................................... 16
1.4.6.3Characterization of Domains after Etching ........................................ 17
1.5 Summary ........................................................................................................ 17
1.6 Overview of dissertation................................................................................... 18
2. ELECTRODE FABRICATION.............................................................................. 19
2.1 Photolithography............................................................................................. 19
2.1.1 Contact electrodes on insulating substrate................................................. 20
2.1.2 Electrodes on MgO:LiNbO3 substrate ....................................................... 25
2.2 Electron-beam lithography ............................................................................... 30
2.2.1 Wet etch method for nano-scale contact electrodes..................................... 32
2.2.2 Liftoff method for nano-scale contact electrodes......................................... 38
2.3 PMMA removal ............................................................................................... 44
2.4 Substrate Cleaning .......................................................................................... 46
2.5 Summary ........................................................................................................ 48
3. POLING .............................................................................................................. 50
3.1 Overview of poling process ............................................................................... 50
3.1.1 Determining Crystal Orientation............................................................... 50
3.2 Creation of Poling System ................................................................................ 52
3.3 Contact poling................................................................................................. 54
v
TABLE OF CONTENTS – CONTINUED
3.4 Poling with electrode fabricated on crystal surface ............................................. 60
3.5 Probe-tip poling .............................................................................................. 65
3.6 Summary ........................................................................................................ 70
4. AUGER ELECTRON SPECTROSCOPY FOR SURFACE
FERROELECTRIC DOMAIN DIFFERENTIATION IN
SELECTIVELY POLED MgO:LiNbO3................................................................... 73
4.1 Contribution of Authors and Co-Authors .......................................................... 73
4.2 Manuscript Information Page ........................................................................... 74
4.3 Abstract.......................................................................................................... 75
4.4 Introduction .................................................................................................... 75
4.5 Experiment ..................................................................................................... 79
4.6 Results and discussion...................................................................................... 83
4.7 Summary ........................................................................................................ 95
5. NANO-SCALE FERROELECTRIC DOMAIN DIFFERENTIATION IN
PERIODICALLY POLED LITHIUM NIOBATE WITH AUGER ELEC-
TRON SPECTROSCOPY..................................................................................... 97
5.1 Contribution of Authors and Co-Authors .......................................................... 97
5.2 Manuscript Information Page ........................................................................... 98
5.3 Abstract.......................................................................................................... 99
5.4 Introduction .................................................................................................... 99
5.5 Experiment ....................................................................................................101
5.6 Results...........................................................................................................101
5.7 Discussion ......................................................................................................108
5.8 Summary .......................................................................................................111
6. LITHIUM NIOBATE DOMAIN MAPPING WITH AUGER ELECTRON
SPECTROSCOPY...............................................................................................113
6.1 Overview........................................................................................................113
6.2 Background....................................................................................................113
6.3 Experiment ....................................................................................................114
6.4 Results and image processing ..........................................................................115
6.5 Discussion ......................................................................................................129
6.6 Conclusion .....................................................................................................131
vi
TABLE OF CONTENTS – CONTINUED
7. ADDITIONAL AUGER ELECTRON SPECTROSCOPY RESEARCH...................133
7.1 Introduction ...................................................................................................133
7.2 Energy inversion .............................................................................................134
7.3 AES on probe-tip poled chips ..........................................................................141
7.4 AES on MgLN with thin gold film ...................................................................147
7.5 Summary .......................................................................................................153
8. CONCLUSION....................................................................................................154
REFERENCES CITED.............................................................................................159
vii
LIST OF TABLES
Table Page
2.1 Photolithography recipe............................................................................. 26
2.2 Photolithography recipe for MgLN ............................................................. 29
2.3 PMMA removal solvent tests ..................................................................... 46
3.1 Coercive fields of common periodically poled crystals .................................. 51
viii
LIST OF FIGURES
Figure Page
1.1 QPM Conversion......................................................................................... 3
1.2 Poled grating period for input wave curves ................................................... 5
1.3 PFM response to +/-Z domains ................................................................... 9
1.4 SEM contrast switching ............................................................................. 14
2.1 Photolithography process for electrode patterning ....................................... 20
2.2 Shape of period grating electrode from first mask........................................ 21
2.3 Grating electrode with 8 µm period............................................................ 22
2.4 Micron-scale grating electrode SEM 1......................................................... 23
2.5 Micron-scale grating electrode SEM 2......................................................... 24
2.6 Film photomask on chips in contact aligner ................................................ 27
2.7 Sub-mm scale electrode on MgLN .............................................................. 28
2.8 Wet etch process for nano-scale electrode patterning ................................... 33
2.9 Alignment marks for electron beam lithography .......................................... 34
2.10 1 µm period wet etch method grating electrode........................................... 35
2.11 Nano-scale grating fingers from wet etch method......................................... 36
2.12 Incomplete PMMA removal ....................................................................... 37
2.13 Liftoff fabrication process for nano-scale electrode patterning....................... 39
2.14 Liftoff method electrode ............................................................................ 40
2.15 600 nm period liftoff method grating electrode ............................................ 41
2.16 Profile of bilayer and 240 nm metal lines .................................................... 43
2.17 Failed liftoff grating................................................................................... 44
2.18 PMMA stripping failure ............................................................................ 47
2.19 Chips cleaned with HCl ............................................................................. 49
3.1 Bulk polarity testing ................................................................................. 52
ix
LIST OF FIGURES – CONTINUED
Figure Page
3.2 Poling system ........................................................................................... 54
3.3 Periodic grating electrode .......................................................................... 57
3.4 Collared grating electrode.......................................................................... 58
3.5 Bulk through-poled hexagon in MgLN........................................................ 59
3.6 Un-etched micro-domains .......................................................................... 60
3.7 Electrode on MgLN................................................................................... 62
3.8 Over-poled MgLN ..................................................................................... 63
3.9 Striated edge micro-domains ...................................................................... 64
3.10 Probe tip for poling................................................................................... 66
3.11 Probe-tip poled MgLN .............................................................................. 67
3.12 Probe tip poled domain in MgLN............................................................... 69
3.13 Probe tip damage to MgLN ....................................................................... 70
3.14 Si oil limited over-poling............................................................................ 71
4.1 Contact poling system ............................................................................... 80
4.2 SEM image of MgLN sample...................................................................... 81
4.3 Survey areas on SEM image and Auger spectra........................................... 82
4.4 500 µm FOV SEM image with 10 survey areas and corre-
sponding spectra ....................................................................................... 84
4.5 100 µm FOV SEM image with 10 survey areas and corre-
sponding spectra ....................................................................................... 86
4.6 Peak energy drift across surveys................................................................. 87
4.7 Peak energy deviation from the mean across surveys ................................... 89
4.8 Peak energy drift across surveys, second data set ........................................ 91
4.9 Peak energy deviation from the mean across surveys, second
data set .................................................................................................... 92
x
LIST OF FIGURES – CONTINUED
Figure Page
5.1 PPLN SEM image and AES spectra ..........................................................102
5.2 78 µm FOV peak energy and peak amplitude deviation from mean .............103
5.3 9 µm, 3 µm, and 1 µm FOV SEM images and peak
amplitude deviation from mean.................................................................104
5.4 9 µm, 3 µm, and 1 µm FOV SEM images and peak
amplitude deviation from mean, second data set ........................................105
5.5 AES on aluminum with external applied voltage ........................................109
6.1 AES mapping of PPLN with 20 µm FOV ..................................................116
6.2 20 µm FOV map normalization and smoothing ..........................................117
6.3 Lineplots of a single map rows with range of sigma ....................................118
6.4 Lineplots of all map rows for range of sigma ..............................................119
6.5 Lineplots of all map rows and method 1 thresholding .................................120
6.6 Binary and three-level thresholding maps from method 1............................121
6.7 Lineplots of all map rows and method 2 thresholding .................................122
6.8 Binary and three-level thresholding maps from method 2............................123
6.9 Lineplots of all SEM image rows ...............................................................124
6.10 SEM image processing and mapping domain walls overlaid .........................125
6.11 AES mapping of PPLN with 7 µm FOV ....................................................127
6.12 AES mapping of PPLN with 41.4 µm FOV................................................128
6.13 200 µm FOV AES map of PPLN...............................................................130
7.1 AES energy inversion SEM image..............................................................135
7.2 AES energy inversion instability................................................................137
7.3 AES amplitude stable from inversion.........................................................139
7.4 Selectively removed photoresist on PPLN ..................................................142
7.5 SEM images of probe-tip poled chip and East AES survey areas .................144
xi
LIST OF FIGURES – CONTINUED
Figure Page
7.6 SEM images of probe-tip poled chip and West AES survey areas.................145
7.7 AES peak amplitudes from probe-tip poled domain ....................................146
7.8 AES peak energies from probe-tip poled domain ........................................146
7.9 AES on MgLN with thin gold film.............................................................149
7.10 AES on gold thin film on MgLN SEM image..............................................151
7.11 2 µm FOV AES SEM image drift with gold thin film .................................152
xii
NOMENCLATURE
AES - Auger electron spectroscopy
AFM - Atomic force microscopy
CSHG - Cherenkov second harmonic generation
DW - Domain wall
EBL - Electron-beam lithography
EBPVD - Electron-beam physical vapor deposition
FE-SEM - Field emission scanning electron microscope
HCl - Hydrochloric acid
HF - Hydrofluoric acid
HV - High voltage
KTP - Potassium titanyl phosphate
LN - Lithium niobate
MgLN - Magnesium-doped lithium niobate, (5 %mol unless specified otherwise)
O-KLL - Oxygen KLL Auger transition electrons
PFM - Piezoresponse force microscopy
PPLN - Periodically poled lithium niobate
PTP - Probe-tip poled
QPM - Quasi-phase matching
RS - Raman spectroscopy
SEM - Scanning electron microscopy
SFG - Sum frequency generation
SHG - Second harmonic generation
SPM - Scanning probe microscopy
UHV - Ultra-high vacuum
xiii
ABSTRACT
Ferroelectrics are a novel class of materials with a built-in electric polarization
state. Like ferromagnetic materials, by applying a strong external field, the direction of
ferroelectric polarization can be switched, or ‘poled’. Poling also switches the sign of
the nonlinear coefficient, which determines the strength of a material’s nonlinear optical
interactions. By controlling the ferroelectric poled domain structures, the switching sign of
the nonlinear coefficient can keep interacting optical waves in-phase, limiting deleterious
material dispersion during nonlinear optical interactions. Lithium niobate (LiNbO3) is
one such ferroelectric crystal, prominently used in nonlinear optics. Periodically poled
lithium niobate (PPLN) domain structures can produce the phase-matching conditions
described above in a process called quasi-phase matching, creating powerful nonlinear optical
devices. The applications of these devices are numerous, yet they have not reached their
full potential due to the limitations of fabricating and characterizing nano-scale patterned
domain structures.
We first explored nano-fabrication of electrodes as a precursor to nano-scale poling.
Periodic grating electrodes with 600 nm periods were fabricated using an innovative combined
photolithography and electron beam lithography (EBL) liftoff method to create HV poling
contact electrodes. A 10 kV bulk poling system was built and preliminary poling tests in
three distinct poling configurations were performed on magnesium-doped lithium niobate
(MgLN).
We then adapted Auger electron spectroscopy (AES) as a new method to address the
unique challenge of characterizing ferroelectric domains. In our initial AES characterization
method, polar ferroelectric domains (+/-Z directions) in MgLN were differentiated from
one another by the Auger O-KLL peak energy, with the -Z domains having higher peak
energy due to the lower surface potential. We then discovered that +/-Z domains in PPLN
can be differentiated with nano-scale resolution by the O-KLL peak amplitude, which is
larger for -Z domains. The principle of this AES peak amplitude separation method was
applied to mapping to achieve full imaging of PPLN’s +/-Z domains with fields of view
spanning from 7.5–200 µm. We ultimately demonstrate AES mapping as a new lithium
niobate domain imaging method that is non-destructive, non-contact, unambiguous, with
nano-scale resolution down to 67 nm.
1
CHAPTER ONE
INTRODUCTION
1.1 Motivation
The research for this thesis began with the goal of fabricating periodically poled
nonlinear optical crystals with sub-micron poling periods in order to achieve first-order
quasi-phase matching (QPM) [1] for efficient wavelength conversion. Of particular interest
was lithium niobate (LN), well known for the numerous applications of periodically poled
lithium niobate (PPLN). The applications for such devices are numerous but of particular
interest was the ability to convert low intensity infrared LIDAR signals, of perhaps only a
few photons, into the visible. Due to changing circumstances as well as the realization that
reliable small-scale characterization of poled domains would be necessary to pursue this goal,
the focus of the project evolved into characterization of ferroelectric domains in nonlinear
materials. New or improved characterization methods have the potential to greatly improve
understanding of PPLN and similar materials, and consequently aid ongoing fabrication
efforts, especially on the nano-scale. This shift in focus also allowed most of the work to
occur in-house in Montana State University’s Imaging and Chemical Analysis Laboratory
(ICAL) and Montana Microfabrication Facility (MMF).
In this chapter, after a brief description of some of the applications of nonlinear optical
wavelength conversion, the requirements of nano-scale periodic poling to achieve first-order
QPM are discussed. With interest in nano-scale periodically poled crystals in mind, an
overview of existing ferroelectric domain characterization methods is then given, showing
the need for new or improved methods to complement existing techniques. Finally, an
2
overview of this dissertation is given.
1.2 QPM for wavelength conversion
A number of applications require efficient wavelength conversion and others require
coherent light sources at wavelengths not currently readily accessible [2–4]. Of particular
interest are devices that can efficiently convert weak signals between infrared and visible
wavelengths. One example is LIDAR systems that detect weak infrared signals (such as
methane absorption at 1700 nm) [5], where efficient detectors are not readily available.
Another example is quantum optical systems that take advantage of low loss optical fibers
and circuits at telecom wavelengths but could also benefit from efficient quantum detectors
and memory devices in the visible and near infrared [6]. Nonlinear optical (NLO) materials
have the potential to solve this issue.
Normal NLO processes suffer from lowered efficiency due to the accumulating phase
mismatch as the input and converted beams propagate through the crystal. The source
of the phase mismatch is material dispersion, meaning the material’s index of refraction
is wavelength dependent and thus different for the input and converted beams [7]. The
basic concept of QPM is to periodically switch the sign of the nonlinear susceptibility
to reverse the direction of de-phasing [1, 8], thus overcoming this problem. Physically,
inverting the nonlinear susceptibility is accomplished by switching the direction of the
crystal’s ferroelectric polarization state; a process named ‘poling’. Repeated poling (periodic)
at the right interval throughout a region can ensure that the beams are never more than
90◦ out of phase, resulting in high efficiency conversion. This is illustrated in figure 1.1,
where the crystal’s regional nonlinear susceptibility is denoted by the arrows. The gain in
the converted beam as it propagates through the crystal is shown for no phase matching,
first-order, and third-order QPM (discussed below).
There are a few notable limitations to what wavelength conversions QPM can currently
3
Figure 1.1: A comparison of the converted wave amplitude as a function of propagation
distance through a nonlinear crystal for first-order QPM, third-order QPM, and no phase-
matching. The periodically poled QPM crystal is shown along the bottom with arrows
indicating the direction of polarization, and input and output beams. Lcoh is the coherence
length that corresponds to the distance over which the fundamental and converted waves
de-phase.
achieve. While many nonlinear conversion processes can be achieved with periodically poled
domain periods on the micron-scale, they are generally limited to forward propagating
interactions in materials with relatively low dispersion. It would be extremely useful in many
applications to achieve backwards wave interactions [9–11], allowing separation of the input
and output beams by direction of propagation. Micron-scale domain lengths are generally
too large to utilize the grating’s first-order periodicity in order to achieve the backwards wave
interactions. Using a higher order (Nth-order) of the periodically poled grating can achieve
periodic phase matching, but at the cost of decreased conversion efficiency. The effective
nonlinear coefficient for the Nth-order is lowered by a factor N, which reduces the output
intensity by a factor of N2 for a given length of nonlinear material. This is illustrated for
third-order phase matching in figure 1.1. In scenarios with a weak input signal, the extremely
low conversion efficiency of using a high grating order is unacceptable. Furthermore, because
a lower grating order achieves higher frequency conversion per length, the required length
4
of NLO material is shorter. This is critical as most fabrication methods limit the length of
periodic poled regions to the scale of a few centimeters. Thus, developing a means of creating
reliable first-order QPM gratings would be a great technological improvement that would
make new and desirable wavelength conversion processes and their related devices possible.
1.3 Requirement of sub-micron periodic poling
To achieve first (or lowest) order QPM for NLO conversion in backwards propagating
geometries often requires fabrication of domains on the sub-micron scale. The following
example of a backwards propagating interaction illustrates the advantages of sub-micron
domain periods. To achieve QPM it is necessary to maintain momentum conservation. For
three-wave interactions, momentum conservation in QPM is determined by the following
equation relating the process’s different wave vectors,
k⃗p + k⃗i = k⃗s + K⃗G, (1.1)
where p, i, s, and G indicate pump, idler, signal, and grating respectively. The grating
vector is related to the poling period, Λ and grating order, N ,
2πN
KG = . (1.2)
Λ
For the case of co linear sum-frequency generation (SFG) with backwards propagating
signal and idler, equation 1.1 becomes
kp − ki + ks = KG, (1.3)
where the k’s here are the wavenumbers equal to 2πn(λ) > 0, for wavelengths λ, in a
λ
material with refractive index n.
5
A computer program was developed to determine the required grating period for first-
order (or higher) QPM, given an NLO material, and pump and signal wavelengths of
interest. The program uses bulk crystal properties (dispersion and absorption) and allows
calculations for a variety of nonlinear crystals (KTiOPO4, LiNbO3, LiTaO3, MgO2:LiTaO3,
MgO2:LiNbO3, or RbTiOPO4) and interaction types (parametric down conversion, sum-
frequency generation, or second-harmonic generation (SHG)). A plot of the required grating
period vs idler wavelength is shown in figure 1.2 for several common NLO materials. The
signal wavelength is chosen to span from∼0.5–1.0 µm, the range where readily accessible, and
highly efficient silicon detectors function. This range then determines the idler’s wavelength
range. For the scenario plotted, first-order QPM requires a grating period between 200-500
nm. This nano-scale grating period requirement underscores the need for nano-scale poling
methods, as well as the need for characterization methods for nano-scale poled materials.
Figure 1.2: Plots of the grating period vs. input (idler) signal wavelength with backward
propagating signal and idler configuration. The idler range for the 1550 nm pump is limited
by the corresponding signal range of .5–1.0 µm
6
1.4 Overview of Ferroelectric Surface Domain Characterization Methods
1.4.1 Need for characterization methods
Some of the difficulties in the fabrication of periodically poled nonlinear crystals used
for QPM are obtaining uniform poled feature widths, desired poled grating duty cycles,
and suitable depths of poling in desired regions. As the practical applications of periodic
poling grow, and new poling technologies enable smaller poling periods, characterization
methods for the poled domains become more important. It is necessary to accurately and
non-destructively characterize the +/-Z domains to check for proper poling width, duty
cycle, uniformity, and domain wall roughness, in order to ensure efficient conversion at the
desired wavelengths. Further characterization tools are also needed in order to study the
physics of ferroelectric surfaces such as LN, a wide band-gap ferroelectric semiconductor
[12]. An ideal characterization method would allow unambiguous, non-destructive, rapid
domain characterization at sizes ranging from millimeter to nano-scales, and would also give
information about the surface potential and its origin.
Several established characterization methods are available, with each one having unique
limitations such as long scan times, probe degradation from contact, being destructive
to the crystal surface, ambiguous domain polarization, complicated or unknown imaging
mechanisms, or optically limited resolution. The largest contribution of this dissertation is
the development of a new characterization techniques for LN ferroelectric domains using
Auger electron spectroscopy (AES). This new method is non-contact, non-destructive,
unambiguous, and can operate at size-scales from hundreds of microns down to the nano-
scale. In order to understand these advantages and the gaps they fill, an overview of
established methods for characterization of ferroelectric polarization domains is first given.
7
1.4.2 PFM
A common and well-established method for visualizing ferroelectric domains is piezore-
sponse force microscopy (PFM), a modified scanning probe microscopy (SPM) technique.
The discussion of PFM will be more in depth compared to other methods because it touches
on many important properties of ferroelectric crystals. PFM takes advantage of the fact that
ferroelectric domains are also piezoelectric and detects the converse piezoelectric response of
a material to infer the ferroelectric domain structure. A review of topics relevant to PFM
for characterization of nonlinear crystals (such as LN and KTP) and their domains is given
here, including an overview of PFM operation, what the PFM response indicates, lateral
and depth resolution, as well as other relevant SPM techniques. The focus will be on PFM
for detecting 180° domain walls in uniaxial crystals, such as LN and KTiOPO4 (KTP).
1.4.2.1 Ferroelectric and Piezoelectric Coupling Piezoelectricity is the property of some
crystal classes to acquire electric charge under applied strain. Similarly, the converse
piezoelectric effect is when an applied electric field creates a strain in the piezoelectric crystal.
For a material with piezoelectric tensor components dij, an applied field Ei will result in a
strain
Sj = dijEi. (1.4)
The piezoelectric coefficient is related to a material’s spontaneous polarization Psk,
dij = ϵIQjmkPsk. (1.5)
with dielectric constant ϵI , and electrostriction coefficient Qjmk [13, 14] (Q does not depend
on the direction of the applied field). Because there is a linear relationship between the
piezoelectric coefficient and ferroelectric polarization, we can infer ferroelectric domains by
detecting piezoelectricity. More specifically, an applied uniform field along the crystal’s
8
polar axis will cause a ferroelectric domain with polarization parallel to the applied field
to expand. Similarly, polarization antiparallel to the applied field will cause the domain to
contract [15, 16].
1.4.2.2 PFM Overview PFM utilizes an oscillating voltage applied to a conducting
AFM probe operated in contact with the sample. The piezoelectric deformations of the
sample due to the applied AC voltage are transmitted to the tip and the tip oscillations
are compared to the applied AC voltage using a lock-in amplifier. The difference in
phase between the tip oscillation and applied AC voltage generate the image contrast
[15]. As discussed in the previous section, because ferroelectricity is linearly coupled with
piezoelectricity, the piezoelectric deformation from the applied voltage allows us to visualize
the ferroelectric domain pattern. It is important to note however, that the domain contrast
mechanism is still an active area of research due to frequent inconsistencies between the
theoretical and observed signals [17]. Still, in theory the phase of the piezoelectric oscillation
(or corresponding tip oscillation) with respect to the applied oscillating voltage tells us
the ferroelectric state, “With the modulation voltage applied to the probing tip, positive
domains (polarization vector oriented downward) will vibrate in phase with the applied
voltage so that φ(+) = 0◦, while vibration of negative domains (polarization vector oriented
upward) will occur in counter phase: φ(–) = 180◦ [13].” Because PFM is operated in contact
mode, the topography of the scanned area is also recorded simultaneously with the phase
of the piezoelectric response. Sensitivity to surface displacement in most PFM systems is
limited to about 0.1pm [15]. While the amplitude of the tip oscillation can be used to gain
some information, it is complicated by material properties, the tip and surface geometry,
electrostatic forces due to the non-uniform field of the cantilever and other effects such as
the sample’s resonant mechanical response [16, 18]. Furthermore, in practical application
the measured phase can be complicated and offset by instrument electronics, sample-tip
9
contact, and other causes including “unknown contributions” [19]. For an ideal scenario
however, the difference between the tip’s phase, magnitude, and amplitude visualization is
illustrated in figure 1.3 [15] that shows the theoretical detected signal as the tip is scanned
across ideal domains of opposite orientations. Our focus on crystals such as LN, KTP, and
their isomorphs is due to their common use as periodically poled nonlinear materials. While
some materials can have multi-directional ferroelectric polarization, a benefit of our materials
of interest is that they are uniaxial and only exhibit polarization in two polar states along
the +/-Z axis.
Figure 1.3: A PFM tip scans across a periodically poled sample, with uniaxial polarization,
and the PFM response is shown for phase, φ, and amplitude, R. note that the amplitude
signal goes to zero at the domain boundaries, as the responses from the +/-Z domains on
either side of the tip’s position cancel out.
1.4.2.3 PFM Resolution Two important parameters of interest when utilizing PFM are
depth and lateral resolution. The PFM signal originates from the piezoelectric response of
10
the material, but it is not exactly clear to what degree the PFM probes the surface or bulk
material. One source notes that bulk samples are much thicker than the tip-sample contact
area and that because the probing field is very inhomogeneous, and decays quickly away
from the tip, they conclude the piezoresponse comes from only a small probing volume near
the tip-sample contact area [20]. In a different study, PFM was performed on a surface
domain created by UV-laser inhibited poling and wedge polished to create a linearly sloped
domain depth. The study found a domain depth of 1.7 µm was indistinguishable from the
bulk domain using PFM [21], and results were verified with HF etching. The probe tip in
the study had a radius of 50-70 nm, and they note that a larger radius increases the depth
of probed volume. Similarly, the lateral resolution achievable with PFM is limited by tip
radius [15, 22, 23], and the resolution is closely related to the tip-sample contact area. In
theory a perfect tip with point contact would allow infinite resolution [24] but practically,
material considerations, cantilever loading force, tip radius, and tip-sample contact area give
a finite resolution on the order of 10-20 nm [21, 22]. Furthermore, any practical scan will
results in tip degradation with more tip-sample contact lowering the resolution over time.
In one source, PFM resolution is defined as the FWHM when scanning across a LN domain
wall and is investigated as a function of tip radius. They determine a relationship of FWHM
= 1.2*r where r is the conducting tip radius. This allows measurement of apparent domain
wall widths of ∼18 nm with a 15 nm tip [25]. It should be noted that the question of PFM
resolution is naturally related to the question of ferroelectric domain wall width; in theory
the wall width may be a single lattice wide [26] but has been shown to be at least several
lattice widths in certain thin film materials [27]. A ferroelectric domain wall’s location may
be determined with high precision with PFM (assuming background-free) but for actual
domain wall widths that are on the order of lattice widths, the apparent domain wall width
measured with PFM will be much larger [15].
11
1.4.2.4 Other SPM Techniques for Ferroelectrics While PFM might now be the most
common method of SPM for imaging domains in nonlinear optical ferroelectric materials,
other methods have been reported that rely on different interactions than the piezoresponse
and its coupling to the ferroelectric domains. Electrostatic force microscopy (EFM) is a
non-contact SPM method where the tip is driven at the resonant frequency while located
10–100 nm above the sample. For materials with spontaneous polarization or charges, the
electrostatic force modifies the tip’s amplitude and phase allowing detection of charge density
and polarity [20]. However, this method suffers from the effect of leakage. In a study
comparing EFM and PFM, the EFM signal decayed by an order of magnitude over the course
of months while the PFM was unchanged [28]. EFM also has a relatively low resolution when
compared to PFM [29].
Kelvin probe force microscopy (KPFM) is another non-contact mode that utilizes both
an AC and a DC bias applied to the tip. The sample’s effect on the tip’s phase and amplitude
are well approximated by simple harmonic oscillator models. Nullifying the sample’s effects
on the tip by modulating the DC bias can then be used to map the surface potential,
which can be used to imply the ferroelectric domains [16]. Some disadvantages of non-
contact methods (EFM and KPFM) include lower resolution and high sensitivity to surface
conditions [20], such as debris, roughness, or sample history.
One contact method for ferroelectric domain characterization is lateral friction mi-
croscopy (LFM). “The structural differences between the surfaces of oppositely polarized
domains modify the surface potential and chemical properties of the surface and results in
the variation of friction coefficients imaged by LFM ... While much of the early work on
ferroelectric domain imaging has been based on the friction signal, the dearth of information
on contrast formation mechanism, and emergence of piezoresponse force microscopy (PFM)
resulted in decrease of the effort in this direction [20].” Another possible contact method is
standard atomic force microscopy (AFM) but with a DC bias applied to the tip. The static
12
piezoelectric deformation from the DC bias could be detected but this method also suffers
from extreme sensitivity to surface conditions [20]. Therefore, while other methods have
been experimented with, PFM has become the most often used form of SPM for ferroelectric
domain determination.
PFM is considered advantageous to other mentioned SPM methods for several reasons.
First, PFM contrast is only weakly dependent on tip-sample contact area. Because of this,
it is less sensitive to the topographic defects that strongly affects EFM, KPFM, LPM,
and AFM. Secondly, its implementation and sample preparation are relatively simple for
many standard SPM setups. Lastly, the contrast mechanism, namely phase of the sample’s
piezoresponse, is much greater than other methods, [13].
1.4.3 SEM
Scanning electron microscopy (SEM) is a useful surface imaging and characterization
method that can detect the contrast in secondary electron yield between LN domains of
opposite polarization with certain primary beam parameters [30–32]. There are several ways
that SEM can be operated to detect domains although each has distinct drawbacks. In one
method, surface charging under the electron beam causes a topographic relief due to the
converse piezoelectric effect, where the positive domains are displaced in the +z direction
[33]. A different method of heating or cooling a ferroelectric sample can be used to create
a temporary voltage contrast between domains via the pyroelectric effect (all ferroelectric
crystals are also pyroelectric), which can then be imaged as a differential amount of detected
secondary electrons [34]. Furthermore, detection of secondary electrons (SE) using a through-
the-lens (inLens) detector can provide information about the material’s work function,
though the work function and image contrast are complicated by “beam parameters, beam-
induced contamination, specimen electric potential, SE collection efficiency, etc.” [35]. Still,
this method is more sensitive to the surface potential, including the difference in potential
13
between +/-Z domains.
The image contrast mechanism between ferroelectric domains in SEM is not well
understood. Due to the complexity and competition of insulating, pyroelectric, piezoelectric,
and ferroelectric effects under the beam irradiation, the mechanism of image formation is
complicated [36]. There are different theories on whether the domain contrast originates
primarily from the pyroelectric effect due to beam-induced heating [33, 37], the piezoelectric
effect creating a topographic relief between domains [33], or differential charging when the
SEM is operated in the regime where primary beam current and surface emitted current are
roughly equal [30] or create a positive surface [31]. Another issue is the domain contrast is
often short lived under primary beam irradiation as the crystal’s insulating surface becomes
charged [33, 38, 39]. Similar effects have been observed in-house and are shown in figure 1.4,
SEM images of a roughly hexagonal poled MgO3:LiNbO3 (MgLN) domain. The left and right
images are both taken with the inLens detector, but the domain edges have opposite contrast.
The middle image is taken with the SE2 detector and the domains edges are not apparent,
only the damage/debris from the poling process (more details in section 3.5). Lastly, with
charging being a consistent problem on insulating ferroelectric samples, the resolution may
be limited to the micron-scale [40]; large when compared with SEM’s nanometer resolution
on conducting samples.
One way to compensate for sample charging from the electron beam is to perform
environmental SEM (ESEM) [38, 39, 41]. By operating the SEM’s chamber pressure
approximately 10000 times higher than normal SEM chamber conditions, enough neutral
gas particles can be ionized by the electron beam such that a conducting path is created
for the charge that would otherwise accumulate on the sample. ESEM generally requires
a dedicated instrument and often uses water vapor as the chamber’s neutral gas that is
ionized to carry charge away from the sample surface. In this special method, positive
ferroelectric domains appear brighter than negative domains, although the image contrast
14
Figure 1.4: Three successively taken SEM images of a poled domain in MgLN. Left:
InLens detector image taken immediately upon finding domain. Middle: SE2 image taken
immediately after the left image. Domain edges not detected with SE2. Right: InLens
image taken immediately after the middle image. Domain edge contrast flips compared to
left image.
mechanism here is still not completely understood. However, imaging is stable over time
and sub-micron resolution is claimed [38].
1.4.4 UV-PEEM
Another method to visualize ferroelectric domains in lithium niobate (or similar
uniaxial ferroelectrics) is UV-photoelectron emission microscopy (UV-PEEM). Under UV
laser illumination, domain contrast arises from the enhanced electron emission from negative
ferroelectric domains due to their lower surface electron affinity. This lower surface
electron affinity, thought to stem from preferential surface adsorption, leads to the lower
photothreshold and thus negative/positive domain contrast [12]. This method provides
interesting information about the surface potential and relatedly, the work function difference
between domains. However, the resolution of UV-PEEM is limited by the optical diffraction
limit of the probing laser. Furthermore, due to the proposal that differential surface
adsorption causes the variation in photothreshold between +/-Z domains, sample history
of atmosphere exposure and heating affect the UV-PEEM domain contrast.
15
1.4.5 Optical Methods
The methods outlined thus far (PFM, SEM, UV-PEEM) have focused on characterizing
ferroelectric domains on the surface of the sample. However, both Raman spectroscopy
and Cherenkov second harmonic generation are scanning optical methods capable of non-
destructively imaging the volume of lithium niobate.
Raman Spectroscopy (RS) works by probing atomic vibrational modes in a material.
An input scanning laser’s photons are inelastically scattered up or down in energy by
the vibrational modes which can then be spectroscopically detected. Lithium niobate
domain walls and domains have been separated in the RS signal, using different vibrational
modes. For the domain walls, both the A1(LO4) phonon mode intensity and the E(TO8)
phonon mode frequency can be used [42–46]. For the domains, both the E(TO1) phonon
mode intensity and the frequency of the A1(LO4) phonon mode can be used to relatively
differentiate between domains [43–45, 47]. This technique operates on the interaction of
approximately 1 out of 10 million incident photons.
Cherenkov second harmonic generation (CSHG) has been demonstrated to occur only
at the domain walls in LN. Scanning the probe laser through the volume of a crystal and
creating a 3D map of the CSHG signal can then give a 3D map of the domain wall in
the crystal volume [48–52]. This has allowed important observations of the domain wall’s
position through the crystal volume [52] without the need to destructively etch the crystal
(see next section, 1.4.6). The amplification of the CSHG signal at the domain walls is not
understood and there are competing theories [51, 53]. Both RS and CSHG methods are
limited by the optical diffraction limit of the probe laser, about 0.7 µm for CSHG [51] and
RS experiments [44], though this could be improved somewhat by changing the probe laser
wavelength and numerical aperture of the system. In theory however, both methods can
determine the position of the domain wall with higher precision [52], and the ability to
probe the crystal volume is an important addition to available characterization methods.
16
1.4.6 Domain Selective Chemical Etching
1.4.6.1 LN One common method of domain visualization in lithium niobate is selective
chemical etching of the –Z crystal domains’ surface using hydrofluoric acid (HF). HF will
selectively etch only the -Z surface [54–61] and the +z domain surfaces are left essentially
untouched [62–65], even after 600 hours in HF [66]. The domain structure is thus transferred
to a topographic relief pattern and can then be visualized with an optical microscope, SEM,
AFM, etc. A key mechanism of the etching process is thought to be an initial protonation of
the surface, which is significantly slower or inhibited on the +z surface and accounts for the
differential etch rates [67]. While selective etching is commonly used to visualize domains,
one of the drawbacks is that it is a destructive technique, especially for surface applications
such as waveguides.
While etching LN in HF does not directly impact the +z domains, it does slowly attack
the sidewalls [59, 65, 68], and in this way can affect how the etched relief maps to the poled
domains. According to one source [54] an etch depth of 200 µm reveals the existing domains
with <0.1 µm accuracy. Another source [65] reveals that the sideways etching occurs along
the threefold symmetrical y-directions of the crystal.
1.4.6.2 KTP Another important ferroelectric crystal commonly used for periodic poling
is KTP (potassium titanyl phosphate). Fortunately, selective chemical etching of the –Z
domain surface of KTP crystals and related isomorphs is also possible [69–76]. The most
common etchant is a solution of potassium hydroxide and potassium nitrate (KOH and
KNO3), and it is claimed (like with LN) that the +z faces of the crystal are “essentially
untouched” [73, 75]. The HF based solutions used to selectively etch LN were not found to
work on KTP [71], but overall, the process is similar.
17
1.4.6.3 Characterization of Domains after Etching Several methods can be employed to
characterize the domain structure in the crystal relief after selective chemical etching. The
etched samples (LN, KTP, and their isomorphs) can be inspected with low magnification
SEM (care must be taken to avoid charging the insulating samples). AFM can also be used,
especially when it is necessary to classify the etch depth. If the structures of interest are
large enough, they can also simply be inspected with an optical microscope. One interesting
aspect of selective chemical etching is that when the poled regions extend through the crystal,
etching will also reveal a relief on the bottom of the crystal because the bottom of the un-
etched +z surface will be a –z surface. This can be used to estimate the degree of domain
broadening and propagation through the thickness of the crystal.
1.5 Summary
In order to motivate the characterization technique developed in this dissertation, we
began with a general understanding of the QPM method. QPM’s important wavelength
conversion applications were discussed, followed by the underlying NLO principles, and
finally the requirement for sub-micron periodic poling for first-order QPM. Characterization
is a necessary precursor for the development of sub-micron periodic poling, therefore an
overview was then given of the relevant and currently available characterization methods.
This included PFM and other SPM techniques, SEM, UV-PEEM, RS, CSHG, and HF etch-
ing. The advantages and drawbacks of each method were explored. The drawbacks include
sample-probe contact and probe degradation, long scan times, unknown imaging contrast
mechanisms, ambiguous domain determination, destructiveness, and limited resolution due
to either probe size, optical diffraction, or sample charging. These limitations were discussed
in order to illustrate the need for a new characterization method which complements them.
18
1.6 Overview of dissertation
Chapter 2 presents the electrode fabrication processes developed, shows several micron
and sub-micron scale grating electrodes, and discusses some obstacles in the fabrication
processes. Chapter 3 is an overview of the poling performed, including constructing an in-
house poling system, and three poling methods: contact poling with external bulk or grating
electrodes, poling with electrodes deposited on the crystal surface, and poling single spots
with a probe tip. Chapter 4 begins the major contribution of this dissertation, characterizing
MgLN ferroelectric domains using Auger electron spectroscopy. Chapter 5 alters the AES
characterization method and expands its utility to include periodically poled lithium niobate
with nano-scale spatial resolution. Chapter 6 adapts the principle of the AES method in
chapter 5 to perform mapping, so that +/-Z domains in PPLN can be fully imaged with AES.
This method is non-destructive, unambiguous, and demonstrates resolution down to 67 nm.
Chapter 7 discusses other preliminary findings with AES including some of the limitations
and unknown effects seen using the characterization method in chapter 4, an investigation
into whether the AES characterization method was dependent on prior fabrication on the
LN, which it was not, and an AES experiment on MgLN with a thin gold film for charge
mitigation. Chapter 8 concludes the dissertation and offers suggestions for areas of further
research.
19
CHAPTER TWO
ELECTRODE FABRICATION
A stated purpose of this project was the fabrication of contact electrodes for periodically
poling lithium niobate or similar nonlinear optical crystals. As described in section 1.3, first
order QPM can be achieved with periodically poled gratings with sub-micron poled periods,
therefore the poling electrodes generally require a grating with sub-micron periodicity as
well. In the following sections the fabrication processes for several types of electrodes will be
described, focusing on photolithographic techniques for micron-scale electrodes and electron
beam lithography (EBL) for sub-micron electrodes. Images and descriptions of fabricated
electrodes will be given. The final sections will detail some of the specific challenges of the
fabrication processes.
2.1 Photolithography
Photolithography is a process using a photoresist that is UV sensitive to pattern thin
films. It is generally performed in a cleanroom on chips or wafers of silicon, silica or similar
materials. A photolithographic mask is used to shield selected areas of the photoresist from
a UV light. For a positive photoresist the UV-exposed regions are dissolved in a developer
solution, leaving the opaque regions of the mask printed in the photoresist. For a negative
resist, UV exposed resist is crosslinked or strengthened, and that portion is un-dissolved
by the resist. In either case, the patterned photoresist then allows the chip or wafer to
be selectively etched, coated, or other similar processes. Here, photolithography was used
to directly pattern micron-scale periodic grating electrodes with features as small as 2 µm
grating fingers with 4 µm periods. While the photolithography system could in theory achieve
2–3× better resolution, the photolithographic masks we designed and purchased were the
20
limiting factor.
2.1.1 Contact electrodes on insulating substrate
The first type of electrode fabricated for this project were periodic gratings of chrome
or chrome/gold films on fused silica wafers. The periodic gratings were to be used for
periodic poling of lithium niobate (or similar materials such as KTP, MgLN, etc.). Gold’s
high conductivity and softness (good for contact poling) are beneficial, but it suffers from
low adhesion. Chrome can be used to overcome gold’s low adhesion and is often used as a
base and protection layer (chrome on gold on chrome on substrate) or simply chrome alone.
Fused silica wafers are used because they are electrically insulating for the poling process.
Figure 2.1: Photolithography process for electrode patterning. This block diagram shows a
simple chrome-only layer. For electrodes with a Cr-Au-Cr stack two gold wet-etching steps
are required.
A high-level block diagram of the initial optimized photolithography process is shown
in figure 2.1. The fabrication process begins with blank fused silica wafers. The wafers
were coated in MMF with different metal thin films that will be patterned into micron-scale
electrodes. Most often at this stage, either Cr-only (50 nm) or Cr-Au-Cr (2 nm, 45 nm, 2nm)
21
were the common metal films of choice. The metal-coated wafer is spin coated with a thin
film of UV photoresist (AZ1512). The wafer is placed in the contact aligner (the UV exposure
tool) and brought into contact with the mask and exposed to the UV lamp. After exposure,
the positive resist is developed such that the UV-exposed regions are dissolved. The pattern
on the lithographic mask is now written into the photoresist that then functions as a mask
for the underlying chrome. The wafer is placed in the appropriate wet metal etchant(s)
removing the unprotected metal(s). The remaining photoresist is removed with solvents and
the wafer is left with several individual metal electrodes. A coating of photoresist is applied
as a protective layer before being diced into individual chips by an external vendor. After
dicing, each chip is cleaned of remaining photoresist and dicing debris. At this stage, chips
that were patterned for micron-scale poling periods are completed and ready for testing.
The photolithography process details are given in table 2.1.
The first photolithographic mask we designed in L-edit and ordered had numerous
10×10 mm2 die with different finger widths, grating periods, and electrode shapes. The
mask also had several electrodes without gratings designed to be used for EBL, detailed in
section 2.2.1. An example of a common micron-scale grating pattern is shown in figure 2.2.
Figure 2.2: CAD drawing in L-edit of electrode design. Left side of electrode is periodic
grating for contact periodic poling, and right side is the partially trapezoidal shaped contact
pad for connecting to the poling voltage.
A photolithographically patterned periodic grating electrode is shown in figure 2.3. The
grating has an 8 µm period and a 40/60 duty cycle (3.2 µm finger widths). There is a vertical
22
bus bar on the left side of the grating to help electrically connect any broken fingers, and
part of the electrical contact pad for applying high voltage (HV) for poling is seen on the
right. The rest of the contact pad mirrors the shape shown in figure 2.2.
Figure 2.3: Periodic grating electrode for periodic contact-poling. Electrode is 100 nm thick
Cr on silica substrate. Four images were stitched together due to microscope limitations.
Shown in figure 2.4 is a close-up image taken with the SEM of the electrode from figure
2.3. The metal boasts fairly clean edges, and the 40/60 duty cycle is apparent. Figure 2.5
is an even further zoomed in image. There is a small amount of residue noticeable on the
substrate, which is likely undissolved or redeposited photoresist. The photoresist, if on top
23
Figure 2.4: SEM image of micron-scale electrode patterned with photolithography. Electrode
fingers are 3.2 µm wide with a period of 8 µm. Electrode is 100 nm thick Cr on silica.
of the electrode, could affect the electrode’s contact with an LN sample when poling. Based
on analysis of this SEM image in ImageJ, the scale of the edge roughness was measured to
be less than 50 nm.
While an ideal periodically poled device has a poled period duty cycle of 50/50, the
fabricated duty cycle of many electrodes is deliberately less than 50/50. This is to account
for ‘over-poling’, the tendency of domains to nucleate and spread out from underneath the
electrode area, such that the final domain is larger than the electrode [77]. By allowing
electrodes with less than 50/50 duty cycle to over-pole, and with correct calibration and
24
Figure 2.5: Further zoomed in SEM image of 3.2 µm wide, chrome, grating finger. Electrode
is 100 nm thick Cr on silica.
25
poling control, a 50/50 duty cycle in the periodically poled crystal can still be achieved.
Initial electrodes were patterned with Cr or Cr-Au-Cr using an electron-beam evapo-
rator tool. The Cr came from thermally evaporated Cr rods, and the Au from an electron
beam physical vapor deposition (EBPVD) source. While table 2.1 states the deposition
as Cr or Cr-Au-Cr, several deposition methods and recipes were being tested. We learned
that a sputtering deposition tool might produce smaller grain structure so preliminary tests
were performed with sputtered chrome film deposition using MMF’s Angstrom Engineering
sputtering system. The chrome films were inspected with SEM similar to figure 2.5 and
sputtering was found to produce smaller grain structure than films produced with EBPVD;
this is possibly due to the sputtered atoms having a higher energy and ability to diffuse once
adsorbed onto the substrate leading to tighter grain structure [78–80]. An optimized recipe
for sputtered Cr deposition was determined; 40% power at 25 mTorr Ar pressure for 580 s,
with a measured deposition rate of 7.7 nm/min.
2.1.2 Electrodes on MgO:LiNbO3 substrate
Despite much effort fabricating electrodes for contact poling, imperfect contact was
deemed one of the main problems leading to incomplete poling and micro-domains, detailed
further in section 3.3. Fabricating the electrode on the crystal surface would remove the
uncertainty of bringing the electrode into mechanical contact with the crystal. However,
there are several differences that need to be addressed for fabrication on MgLN instead
of fused silica. LN and MgLN are pyroelectric crystals such that charge accumulates on
the surface with temperature change. Changing the crystal temperature too rapidly can
lead to crystal sparking and fracturing due to charge accumulation. Additionally, due to
MgLN’s cost, electrodes would be fabricated on single chips, which requires special attention
compared to whole wafers. Furthermore, poling efforts at the time were focused on the
sub-mm scale, for which a new mask had been designed. Because of the relaxation on the
26
Table 2.1: Photolithography process details for chrome or chrome-gold-chrome electrodes on
fused silica substrate wafer.
Step Equipment Details
Deposition Amod Evaporator Thermal Cr rod deposition, 50 nm, 2 Å/s.
Surface prep YES HMDS Oven Standard HMDS Run
Spin coat Laurell spinner AZ1512 photoresist. Ramp: 5000 rpm/s. Spin:
5000 rpm. Time: 30 s
Soft Bake Triotech Hotplate 84 s on 100 C hotplate
Exposure ABM Contact Aligner 33.2 mJ/cm2
Develop Development bench AZ300MIF. 60 s with agitation. DI Rinse, N2
dry
Hard Bake Triotech Hotplate 5 minutes on 115 C hotplate
Metal etch Acid bench Depending on Metal film: 1020AC Cr etchant,
8 Å/s etch rate. TFA Au etchant, 2.8 Å/s.
Strip Solvent bench Acetone + IPA rinse. AZ400T: 2× 15 minute
baths on 85 C hotplates. DI Rinse, N2 dry
Spin coat Laurell spinner AZ1512 photoresist. Ramp: 5000 rpm/s. Spin:
5000 rpm. Time: 30 s
Dicing External vendor
Strip Solvent bench Acetone + IPA rinse. AZ400T: 2× 15 minute
baths on 85 C hotplates. DI Rinse, N2 dry
27
mask resolution, a more cost-efficient mask was ordered on pliable Fuji .007” thick film,
with 10 µm minimum feature size. The pliable film mask creates specific challenges in the
contact aligner when applying the vacuum between the substrate stage and mask. Lastly,
adhesion can be an issue for metal deposition on lithium niobate. For these reasons our
photolithography process required modifications.
Figure 2.6: Film mask in contact aligner. (a) Mask vacuum pulls the mask over chip edges
so that mask bends out of contact with chip. (b) Addition of dummy chips minimizes film
mask bending and improves mask-chip contact.
For the first issue of MgLN’s pyroelectric properties, the baking stages of photolithog-
raphy were applied with slow ramps. Fortunately, the first attempt of ramping the soft
bake temperature from 50 C to 100 C at the hotplate’s natural pace was sufficient. The
hard bake step after development was found unnecessary and removed from the process.
Photoresist application onto the 10 mm square chips was also more challenging and requires
a special chuck in the Laurell spin coater with a <10 mm diameter gasket to apply vacuum
to hold the chip in place. These small chips also introduce the issue of edge beads such
that the photoresist thickness is not uniform across the chip. Fortunately, since our patterns
of interest are centered on the chip this was not an issue. The next issue was how to use
the pliable film mask in the contact aligner. Because the mask is flimsy it was initially
taped to a glass plate and placed in the normal mask position. This method was flawed
28
Figure 2.7: Sub-mm scale electrode on MgLN with 50 µm wide fingers and varied spacing
for over-poling investigation. Electrode is 75 nm thick Al. Inset image is photomask pattern
used to pattern electrode on MgLN.
29
Table 2.2: Photolithography process details for aluminum electrodes on MgLN.
Step Equipment Details
Deposition Angstrom Sputterer Glow discharge, O2, 60% power, 30 s. Al
deposition, 1 Å/s.
Surface prep YES HMDS Oven Standard HMDS Run
Spin coat Laurell spinner AZ1512 photoresist (no lid). Ramp: 5000
rpm/s. Spin: 5000 rpm. Time: 30 s
Bake Triotech Hotplate 50 C ramp to 100 C and back to 50 C at
hotplate’s natural rate.
Exposure AB-M Contact Aligner 33.2 mJ/cm2
Develop Development bench AZ300MiF. 1 min with agitation. DI Rinse, N2
dry
Metal etch Acid bench Al etchant type A on 60 C hotplate, ∼80 Å/s.
Strip Solvent bench Acetone + IPA rinse. AZ400T: 2× 15 minute
baths on 85C hotplates. DI Rinse, N2 dry
Spin coat Laurell spinner (none)
Dicing External vendor (none)
Strip Solvent bench (none)
Metal etch Acid bench (After poling) Al etchant type A on 60 C
hotplate, ∼80 Å/s.
30
because the film mask’s edges and tape interfered with the mask vacuum so it could not
be pressed into contact with the chip. It was decided not to tape the film mask to the
glass plate but to place the film mask over the chip. Then, the mask vacuum could pull
the film mask against the chip. However, the vacuum pulled the mask over the edges of the
chip such that it was curved over the top and not in contact. This is illustrated in figure
2.6 (a). A work around was found that by adding ‘dummy’ chips to the contact aligner
stage surrounding the MgLN chip, the mask was made to lie flat on the chips, shown in
figure 2.6 (b). With regards to metal adhesion on lithium niobate, it was desired to apply a
glow discharge plasma cleaning before metal deposition. While not available in the sputter
system used on previous depositions, MMF had by then acquired an Angstrom Engineering
EvoVac Evaporator, an improved version of the EBPVD system previously used. It was also
decided to switch to aluminum films since it is softer than chrome and deposition recipes
were readily available. After applying an O2 plasma cleaning in the Evovac chamber, films
were sufficient with standard deposition recipes. Note that in this configuration, after poling
the underlying substrate, the aluminum electrode would be removed with aluminum wet-
etch. With these modifications, electrodes were fabricated directly on the MgLN surface,
removing the uncertainty of contact poling. An electrode image is shown in figure 2.7. The
grating fingers are 50 µm wide and have a varied spacing to allow over-poling and investigate
the resulting duty cycle. The measured line across the top in the image is 1450 µm. The
inset image is a zoomed-out view of the film mask used to pattern the grating (note the
square collar designed to support a MgLN chip if used as a contact electrode, section 3.3).
Updated fabrication details for electrodes on MgLN are given in table 2.2.
2.2 Electron-beam lithography
Contact poling on the nano-scale requires nano-scale metal electrodes. In this section,
the reason for using EBL to fabricate nano-scale electrodes is first explained. Then the two
31
methods of fabricating nano-scale grating electrodes with EBL are discussed, and several
images of final nano-scale structures are shown.
Analogous to the UV photoresist used in section 2.1, electron beam resists are sensitive
to electron beam exposure. For the contact aligner available in MMF, the diffraction limit of
the incoherent UV light, along with mask printing limitations, generally limits the resolution
of photolithography features to the ∼1 µm scale. For electron beams however, resolution
is dependent on the accelerating voltage, though realistically system factors will affect this
too. This can be illustrated by looking at the de Broglie wavelength for matter,
h
λ = , (2.1)
mv
where h is the Planck constant, m is the electron mass, and v is the velocity. Substituting
in the electron’s velocity gives the electron beam’s de Broglie wavelength as a function of
accelerating voltage V ,
√ h 1.2√× 10
−9
λ = = , (2.2)
2meV V
where m and e are the electron mass and elementary charge. Lastly, solving for the radius of
the beam’s Airy disk gives the approximate beam diameter (sometime called its diffraction
limit),
.61λ .61λ
rAiry = ≈ (2.3)
sin(α) α
where α is the electron beam’s convergence angle and is generally on the order of mrad,
such that sin(α) ≈ α. EBL is performed at 30 kV (still in the non-relativistic realm), so for
an example beam angle of α = 10mrad, the electron beam’s diameter rAiry = 4Å. This low
beam diameter illustrates the capabilities gained by EBL.
32
2.2.1 Wet etch method for nano-scale contact electrodes
The wet etch method was the first method tested using EBL to fabricate nano-scale
contact electrode gratings. A block diagram of the fabrication process of electrodes down
to the nano-scale is shown in figure 2.8. First, several photolithography steps (described in
section 2.1.1) are used to pattern the macro-scale metal on a silica substrate upon which
nano-gratings will be fabricated. The EBL chips are spin coated with e-beam resist (PMMA)
that will become the mask for the underlying chrome. A gold film thinner than 10 nm is
sputter coated onto each chip to serve as a conducting layer during EBL. The EBL pattern
is programmed in lithography specific software, Nanometer Pattern Generation System
(NPGS). Using the Field Emission Scanning Electron Microscope (FE-SEM) in ICAL, NPGS
writes the programmed electron beam pattern into the PMMA. Chips are then developed
such that the unexposed e-beam resist remains and becomes the nano-scale mask for the
underlying chrome. The chips are wet etched again and the un-masked chrome is etched
away. The remaining PMMA is then removed in a solvent bath. Ideally, each chip is now a
nano-scale patterned grating electrode.
An important part of the EBL process is how the photolithographically patterned chips
are connected to the EBL patterning. The FE-SEM is limited to patterns of under ∼150 µm
×150 µm. Therefore, the EBL must be accurately positioned and performed on a comparably
sized metal pad. The photolithography patterned part of the electrode can be seen in figure
2.9. The central metal square is 100 µm wide and is designed to be patterned into a grating
with EBL. The arm to the left connects the grating pad to a larger trapezoidal contact
pad (out of view). The e-beam resist cannot be exposed prior to beginning EBL writing,
therefore finding the central metal pad to be written on must be accomplished without
viewing it in the SEM. In the SEM image on the left in figure 2.9, seven alignment marks
can be seen. Beginning with the positions of the chip’s outer corners, the stage can be moved
fairly accurately to these alignment marks with known location, which are okay to expose,
33
Figure 2.8: Fabrication process for nano-scale metal electrodes patterned in PMMA with
EBL.
34
without exposing the central square. Here, the beam is ramped up to 30 kV for EBL, and
the beam is focused to a tight spot of less than 20 nm. Next the NPGS software takes over,
moving the beam to the central metal square and writes the programmed pattern in the
PMMA. The optical image on the right in figure 2.9 was taken after EBL and development,
and the inset image shows the resulting grating after wet etching the un-masked chrome.
The grating shown was programmed with a 1 µm period and 250 nm lines.
Figure 2.9: Seven alignment marks surrounding the photolithographically patterned, 100 µm
square chrome pad and its connected metal line connecting to the contact pad. Left: SEM
image before performing EBL. Right: optical image after performing EBL and developing
the e-beam resist. Inset image is after wet etching the chrome.
Figure 2.10 is an SEM image of an EBL patterned grating programmed with a 1 µm
period and 250 nm lines. There is minimal debris, and lines are fairly clean. The connecting
bus bar and arm that connect to the contact pad can be seen on the right. The next
image, figure 2.11, is 70 kX magnification image of the same grating. A measurement of
the linewidth shows 251.5 nm linewidth, a deviation from the programmed linewidth of less
than 1%. At this scale, the degree of edge roughness is visible.
Several structures such as those shown in figure 2.10 and 2.11 were produced but it must
be noted that much calibration went into each one. Calibration was a constantly moving
35
Figure 2.10: SEM image of a nano-scale chrome grating electrode structure. The chrome
lines are 251 nm wide with a 1 µm period.
36
Figure 2.11: Nano-scale chrome grating fingers, measured to be 251 nm wide with a 1 µm
period.
37
Figure 2.12: Incomplete PMMA stripping after final metal wet etch. PMMA has grating
written into it and is only partially stripped but appears to have redeposited.
38
target as numerous factors including electron beam stability, achievable focused spot size,
age of PMMA, both in the bottle and the spun-on resist on a chip, PMMA thickness, and
likely many other factors seemed to affect the outcome of these chips. Figure 2.10 was a
success but realistically there were more failures than successes.
One issue that became clear as more electrodes were generated was during the very
last fabrication step: PMMA removal after wet etching the chrome. An example is shown in
figure 2.12, where PMMA can be seen with its written EBL grating lines and is only partially
removed and redeposited. This will be expanded upon in section 2.3 but ultimately it was
an important factor in the decision to test another EBL nano-scale fabrication method.
2.2.2 Liftoff method for nano-scale contact electrodes
A new method of fabrication was devised to work around the issue of hard-to-remove
PMMA and still allow a macro-scale patterned contact pad to connect to an EBL patterned
nano-grating. Rather than wet etching the exposed chrome in a developed PMMA resist
pattern, a liftoff method was used. Liftoff is a microfabrication process where a bilayer of
resist is patterned and developed and then used as a mask for deposition. The top row of
figure 2.13 shows the liftoff process, with the shape of the developed bilayer being important.
After developing top layer of resist (PMMA in this process), the bottom layer resist (PMGI)
is developed to create an undercut relative to the top layer. This undercut allows for the
metal deposited on the substrate to be unconnected to the metal on top of the resist so that
the final stripping step leaves only the metal on the substrate. It must be noted that liftoff
is more successful with directional deposition so that the metal films on the substrate and
on top of the resist bilayer are not connected. The best option for direction deposition is
EBPVD where our chips can be mounted directly above the thermally evaporated source.
For several liftoff chips, nickel was used for the EBPVD material as one reference states it
leads to a higher density of domain reversal nucleation sites [81].
39
Figure 2.13: Fabrication process for nano-scale metal electrodes patterned with a combina-
tion of liftoff using EBL and photolithography. *Either a negative resist or a dark field mask
(electrode pattern is clear) can be used.
40
Because liftoff requires metal deposition after EBL, the photolithography process to
create the contact pad is performed after the EBL/nano-scale patterning. This is shown in
the rest of the block diagram in figure 2.13, where after liftoff, photolithography is performed
to partially overlay a connected contact pad layer on top of the nano-gratings. Note that
this process requires a second metal deposition for the photolithography. The left image in
figure 2.14 shows an example of a finished electrode. The large metal square is the contact
pad, from which several arms extend down and connect to the lifted off nano-gratings (100
µm wide). The image on the right shows three of the nano-gratings, the scale bar is 200 µm.
Figure 2.14: Liftoff method electrode with nano-scale gratings patterned with liftoff and
contact pad and connecting arms patterned with photolithography. Left: Whole electrode
structure, with 3 mm wide contact pad. Right: Three individual lifted off nano-gratings
with connecting arms.
Zoomed in SEM images of the lifted off, 600 nm period, nano-scale gratings from figure
2.14 are shown in figure 2.15. The top row shows the general structure of the pattern with
the contact pad in the top of the left image and the bus bar in the bottom of the right image.
The programmed linewidths are 150 nm (top row) and 200 (bottom row), and the measured
values are 176 nm and 247 nm, respectively. There are a handful of broken metal lines,
between 15–20%. In theory, the bus bar will mitigate the impact of broken lines on poling
41
(unless a line is broken twice). Broken lines were likely due to inadequate chip cleaning
that relied on solvents alone and left debris on the surface. An improved substrate cleaning
process is described in section 2.4.
Figure 2.15: SEM images of 600 nm period, nickel, grating electrodes fabricated with the
liftoff method. Top row electrodes have 176 nm linewidths, bottom row have 247 nm
linewidths.
In order to investigate the liftoff process, chips were cold-cleaved after bilayer
development, and after liftoff. Because the chip substrate is fused silica, which is not
crystalline but glass-like, chips do not necessarily break along a plane (such as silicon
chips/wafers). Cold-cleaving refers to a process of scribing the chips and breaking them
while cold in order to increase the chance of breaking along the desired line. In practice, a
42
metal block is cooled in liquid nitrogen and the chip is cleaved on top of the block. This
method was more than 50% successful with an accuracy of the cleave of about 2 mm. Shown
in figure 2.16 are SEM images of the profiles after cold-cleaving of the bilayer (top row) and
of a chrome metal grating (bottom row). Chips were coated with gold prior to imaging due
to charging effects from the rough insulating chip’s edge. In the top right image, the desired
undercutting in the bilayer profile can be seen. In the bottom right image, chrome grating
lines are ∼240 nm wide, and there is a noticeable amount of fencing of unknown origin on the
top edges of some of the grating lines (the lines on the left side of the image appear clean).
The fencing was zoomed in on at high magnification with the SEM beam to >300kX, to test
if the electron beam would burn, melt or alter the material but it did not have an effect.
This may suggest that the fencing is not organic but chrome.
Many electrodes were fabricated with the liftoff method, but like the wet etch method,
calibration was a nonstop endeavor. On top of the issues seen when calibrating processes for
the wet etch method, the liftoff method adds the complexity of a bilayer. The second resist
in the bilayer adds the need for calibration of film thickness and development duration as
well as the age of the PMGI, both on the chip and in the bottle. Again it must be noted,
as with the wet etch method, that there were many more failed chips than successful. An
example of a failed grating is shown in figure 2.17, where a large portion of the contact pad
is missing, and all grating lines are broken or missing sections. This type of failure may have
been caused by any or all of the following factors: insufficiently cleaned substrate lowering
the metal adhesion, incorrect thickness of either resist layer, developing top PMMA layer
too long so such that there is not or not enough undercutting, developing PMGI too short
again limiting the undercut, depositing metal too thick so that it is connected to metal on
top of bilayer, incorrect EBL dose for desired linewidth/period, or poorly focused electron
beam. In practice many of these factors are connected, for example, a slightly out of focus
electron beam will lower the incident dose, which affects the required development time for
43
Figure 2.16: SEM images of the profile of the liftoff bilayer after development, and the
chrome metal grating after liftoff. Chips were cold-cleaved to obtain images.
44
both resists. Overall however, the new process developed in this section produced contact
electrodes with nano-scale gratings patterned with liftoff, and mm-scale electrode features
patterned partially over the nano gratings to connect the features. With this method, a
handful of successful electrodes were fabricated.
Figure 2.17: SEM image of failed liftoff grating. A portion of the contact pad is missing as
are portions of the grating.
2.3 PMMA removal
A problem that arose at the tail-end of the wet etch fabrication method was difficulty
removing the PMMA resist after wet etching the chrome layer. Without successful PMMA
45
removal, electrodes would be left with a dielectric layer between the chrome and the crystal
during poling tests. Several solvents were tested to remove the PMMA and a summary of
the tests is given in table 2.3. None of the tests appeared to completely remove the PMMA
under SEM inspection, though some were more successful than others. An example of the
remaining PMMA residue after 24 hours in a toluene bath is shown in figure 2.18. A partially
removed film of resist is observable and, in some places, appears folded on the surface as if
redeposited. The EBL written grating lines are also visible in some parts of the film. These
ineffective removal results hastened our switch to the liftoff method, detailed in the previous
section.
To investigate the difficult-to-remove PMMA, Remover PG was tested on a chip that
was developed but not chrome etched and was found to be successful based on SEM
investigation. This test connoted that the chrome etchant’s effect on the PMMA caused
it to be difficult to remove. The chrome etchant (1020AC from Transene) manufacturer
noted that the chrome etchant is strongly oxidative, which might make the resist difficult
for solvents to remove. Investigating the idea that the chrome etchant is strongly oxidative,
PMMA was spun on silicon chips and were prepared with and without being exposed to the
chrome etchant, the latter being the control group. The chips had XPS (x-ray photoelectron
spectroscopy) analysis performed on them in ICAL. The basic principle of XPS is that a
sample is illuminated by an x-ray source and the ejected electrons are counted and their
kinetic energies measured. Comparing this with the energy of the incident x-rays allows the
binding energy to be calculated, which is characteristic of the atomic configuration and the
relative amounts of the elements can then be determined. Mean concentration of O1s was
measured with XPS and found to be 27% and 30% for the control and Cr-etchant-exposed
chip, respectively. Therefore, this data did somewhat indicate that the PMMA sample
exposed to Cr etchant has a nominally higher level of O1s concentration.
The chrome etchant manufacturer recommended piranha solution to remove the PMMA
46
because piranha is a strong oxidizer and will dissolve most organic compounds. However,
piranha solution will also impact chrome and nickel gratings due to their native oxides, as
well as the fused silica substrate so this must be considered. This would be a useful subject
for further research.
Table 2.3: PMMA removal solvent tests.
Solvent Time Temperature Notes
Acetone 48 hr 20 C
Methylene Chloride 24 hr 20 C
Toluene 24 hr 20 C
Anisole 24 hr 20 C
Remover PG 60 min 70 C 30 min in each primary and secondary
baths. Tested on two chips, one which
had not been Cr etched.
Epoxy Dissolver 4 hr 150 C From Allied High Tech, (mix of DMSO
and 1-Phenoxy-2-propanol)
2.4 Substrate Cleaning
For much of the processing detailed in this chapter, substrates were cleaned using only
solvents. Solvents-only cleaning was found to be insufficient due to persistent debris that
sometimes interfered with the nano structures. Dicing is a messy process and despite applying
a protective layer of photoresist, debris was persistent even after cleaning the chips with an
acetone ultrasonic bath, followed by three solvent clean, followed by DI water rinse and
N2 dry. A search through literature and other facility processes found that glass cleaning
generally includes at least one step of acid such as hydrochloric acid (HCl), nitric acid
47
Figure 2.18: SEM images of PMMA residue surrounding the metal grating.
48
(HNO3), or sulfuric acid (H2SO4) [82–85].
The initial cleaning test was a simple hydrochloric acid (HCl) and water solution.
Because HCl mildly etches glass, debris that was not loosened by the ultrasonic solvent
bath might be freed by slightly etching the glass surface. A mixture of 1:1 HCl:H2O was
prepared in a teflon dish (note to always add acid to water and not the other way around).
After undergoing the standard acetone ultrasonic bath and three solvent clean, chips were
placed in the HCl acid bath and removed after 10 minutes. This was followed immediately
by immersion in a large DI water bath, followed by rinse and N2 drying.
Compared to the control group (chips that only had the solvents cleaning process
applied), chips cleaned with HCl had noticeably less debris. This is shown in figure 2.19,
where optical images of the control chip and the acid-cleaned chip are on the left and right,
respectively. The control chip has several small, micron-scale debris spots, and the chip
cleaned with acid is free of these defects. This decrease in debris lowered the amount of
damage to our electrodes throughout processing. Both this finding and the potential need
for piranha solution in the previous section 2.3, indicate a need for students performing
microfabrication processes to be well acquainted with acids and their usage.
2.5 Summary
In this chapter, processes were developed for fabrication of periodic poling electrodes
down to the sub-micron level. First, photolithography was used to create contact poling
electrodes with micron-scale grating periods of various metals on silica substrates. Next,
metal grating electrodes were fabricated on the MgLN substrate in order to overcome the
limitations of contact poling (more info in section 3.3). In order to fabricate nano-scale
grating contact electrodes, EBL was used in conjunction with photolithography in two
separate methods. First was a wet-etch method using a single PMMA layer as an etch mask
after EBL. Due to problems removing the PMMA layer after exposure to chrome-etchant,
49
Figure 2.19: Fused silica chips that are solvent-only cleaned (left) and cleaned with HCl
(right) after solvents. Solvent-only cleaned chip has several specs of debris that are absent
on the HCl-cleaned chip.
a second method was developed utilizing EBL liftoff to fabricate nano-scale grating contact
electrodes. Various examples of final structures were shown, including a liftoff method
electrode with 600 nm grating period. PMMA dissolution was tested in a variety of solvents,
none of which were completely successful. Finally, an improved chip cleaning process was
developed using HCl that greatly limits the amount of chip debris.
50
CHAPTER THREE
POLING
3.1 Overview of poling process
QPM relies on a switching of the nonlinear susceptibility at a regular interval. The
most common method of achieving this is fabrication of periodically poled crystals. Poling
is achieved by applying an external electric field to the crystal surface, parallel to the
ferroelectric axis, that is high enough to reverse the direction of the built-in ferroelectric
polarization. The magnitude of the external field required for polarization reversal is defined
as the coercive field. The coercive field for several crystals of interest for periodic poling
and QPM can be found in table 3.1. We define the coercive voltage as Vcoerc = Ecoerc ∗ t,
where t is the crystal thickness in the direction of the applied field. The intricacies of
the poling process, such as how to create good electrode-to-crystal contact, pressure, and
applied voltage parameters, led to the design and creation of a poling system built in-house.
With this system, three methods of poling were explored in this work; contact poling with
an electrode pressed into contact with the ferroelectric crystal, poling with an electrode
fabricated on the crystal surface, and contact poling with a single probe tip.
3.1.1 Determining Crystal Orientation
Prior to poling, it is necessary to determine the orientation of a crystal’s +/-Z axes, a
property usually ingrained during the crystal growth process. There is a linear relationship
between a crystal’s piezoelectricity and ferroelectric polarization and this coupling is used
to infer ferroelectric domains in LN (or MgLN). By applying a mechanical stress to the
crystal and observing the bias response, the direction of ferroelectric polarization can be
determined. This is accomplished by pressing a conducting probe attached to a voltmeter
51
Table 3.1: Coercive fields of common ferroelectric crystals used for periodic poling. c and s
prefixes denote congruent and stoichiometric configurations, respectively.
Ferroelectric Crystal Abbr. Coercive References
Field
(kV/mm)
cLiNbO3 LN 21 [86–93]
sLiNbO3 SLN 4–5 [86, 88, 90, 91, 94]
MgO:LiNbO3 (5mol%) MgLN 4.5 [86, 87, 91–93, 95, 96]
cLiTaO3 LT 16, 21, 22.1 [97], [86, 87, 98], [94, 99]
sLiTaO3 SLT 1.7 [86, 94, 99–101]
KTiOPO4 KTP 2 [86, 90, 93, 99, 101–106]
RbTiOAsO4 RTA 1.76 [86, 89, 90, 99, 101]
against the crystal, and the piezoelectric effect leads to a buildup of negative charge on the
positive face (+Z) of the crystal, as shown in figure 3.1. Physically this is due to the crystal’s
ions being moved closer to the center of the oxygen layers, or closer to the paraelectric phase,
which reduces the crystal’s net polarization and forms a negative compensating charge on
the +Z face [107]. Note that the crystal face’s designation (+Z or -Z) is opposite to the sign
of the change in voltage; a negative change in voltage indicates the +Z face and vice versa.
A simple system to test the bulk polarity of LN/MgLN samples was built. A sharp
metal probe is held in a spring-like metal arm, and the probe connected to a voltmeter. The
crystal sits under the probe in a mount machined to fit the sample. The sample mount is
on an XY stage in order to allow polarity measurements in precise locations. The sample
mount is conducting and connected to the grounded optical table. The probe tip’s spring
arm is pressed down to press the probe into the crystal, and the bias response is read from
the voltmeter. While either crystal face may give either a positive or negative initial voltage,
52
putting pressure on the probe will change the voltage up or down and this change, whether
positive or negative, implies the ferroelectric polarization.
Figure 3.1: Process for determining crystal polarity from piezoelectric effect. Note that this
measurement on the positive (+Z) face returns a negative bias on the voltmeter. Figure
adapted from G&H’s LN Application Notes [108]
3.2 Creation of Poling System
Initial contact poling tests were performed at AdvR Inc. in their poling system;
however, a poling system was subsequently built in-house. There are several considerations
for building a poling system that include generating a high enough voltage to overcome the
crystal’s coercive voltage (coercive field multiplied by the crystal thickness), achieving good
contact between the electrode and crystal, electrical arcing due to the high voltage (HV),
properly aligning the electrode with the crystal axis, and choosing the contact pressure, pulse
shape, duration, voltage, and temperature. A significant motivation for the in-house poling
system was to fabricate unambiguous bulk poled domains that could be used for testing
characterization methods. For this reason, the initial setup was a relatively simple option
made for poling with contact electrodes and did not include complications such as the use
of liquid electrodes, electrodes fabricated on the crystal surface, temperature control, etc.
53
A diagram of the in-house built system, configured for contact poling, is shown in figure
3.2. The necessary poling system components were determined mainly from two literature
sources [109, 110].
Magnesium doped lithium niobate (5% mol) is a popular crystal for periodic poling
applications as it has a higher photorefractive damage threshold than the un-doped material
[86, 96, 101, 111–114]. More importantly, MgLN has a coercive field nearly 5 times lower
than congruent LN and so would not require as large of a voltage to pole (see table 3.1).
Poling crystals with lower coercive fields is more feasible in practice as there is less chance
of arcing. A high voltage (HV) amplifier was deemed essential for pulse control and safe
HV application and a Trek 610B High Voltage Power Supply Amplifier capable of up to 10
kV was purchased. The 10 kV limit can pole up to 1.0 mm thick MgLN chips, or even 0.3
mm thick LN chips. The HV amplifier is driven by an arbitrary waveform generator (AWG,
Agilent 33220A) to control the applied voltage waveform. In the first iteration of the system
designed for contact poling, the stage where the crystal sits has four springs pulling upwards
on it such that when the ground electrode is pressed down against the crystal, the stage
is pulled up against it. The springs’ top mount can be moved up and down in order to
stretch the springs and control the contact pressure. The ground electrode is machined in a
shape that will maximize its distance from the HV external electrode and mitigate electrical
arcing. Furthermore, silicon oil with a very high dielectric breakdown strength is (sometimes)
applied to the setup in order to further help limit arcing. The ground electrode was covered
in highly malleable indium foil in order to improve crystal contact under pressure. A notable
limitation of the system when setup for contact poling was that the positioning and alignment
of the external electrode was confined by the ability to physically position the crystal on top
of the electrode. The poling process generally utilized a single square voltage pulse, though
it should be noted that several sources state triangular or other shapes with sub-coercive
field seed pulses are beneficial [109, 115, 116]. The pulse’s duration and voltage were both
54
explored as parameters to optimize.
Achieving repeatable, good electrode to crystal contact proved to be a major obstacle
and eventually led to us considering fabrication of the electrode on the crystal surface,
which would be removed after poling. Still, the electrodes on the crystal surface did not
achieve clean repeatable poled regions and debris and residue from electrode fabrication on
the crystal surface was a source of confusion for domain characterization. As the project’s
focus turned further towards characterization with AES, it was decided to test poling with a
probe tip. This allowed for numerous domains to be poled on a single 10 mm square chip and
faster testing of poling parameters. The shift from contact poling, to electrodes fabricated
on MgLN, to probe-tip poling (PTP) is detailed in the following sections.
Figure 3.2: Schematic of the poling system built in-house.
3.3 Contact poling
Contact poling utilizes an external electrode pressed into mechanical contact with the
surface of the ferroelectric crystal to be poled, with a ground electrode on the opposite side.
55
A voltage is applied to the electrode that is larger than the crystal’s coercive voltage in
order to switch the domains underneath the electrode. The key features of this process are
fabrication of the external electrode (discussed in chapter 2), and the system that presses
the external electrode into contact with the crystal.
Electrode fabrication consists of several fabrication steps performed in the MMF
cleanroom. Different processes are required depending on the size-scale of electrode features
required, with photolithography used for micron-scale grating electrodes, and electron beam
lithography (EBL) used for sub-micron grating electrodes. Both processes have multiple
variations but generally consist of a metal thin film deposition onto an insulating silica
substrate, electron beam or photo-lithographic patterning the electrode, and a metal wet
etch or liftoff process to remove the unneeded metal.
Initial poling tests were performed with relatively large, mm-scale, machined aluminum
electrodes that were 2 mm × 4 mm. The samples being poled were 10 mm square, 0.5 mm
thick, double side polished, z-cut MgLN purchased from MTI Corp. In theory these crystals
would have a coercive voltage of 2.25 kV (4.5 kV/mm × 0.5 mm). Early tests determined
that voltages over 5 kV tended to break the chips. The applied voltage was further narrowed
down based first on bulk polarity measurements of domain reversal, and later by HF etching.
A nominal applied voltage of 3.9 kV for 1–5 s was determined to be sufficient for poling large
areas with machined contact electrodes. After poling tests with the machined mm-scale
contact electrodes, we began testing micron-scale contact grating electrodes fabricated in
section 2.1.1. One such periodic grating electrode is shown in figure 3.3. The chrome grating
was patterned with photolithography on an insulating silica chip. The grating had 5 µm
wide grating fingers in a 20 µm grating period with a 25/75 duty cycle, in order to allow
over-poling. A bus bar connecting all the fingers can be seen on the left side of the image
in order to help compensate for any broken fingers. The larger metal area on the right is
a small portion of the metal pad used for electrical contact. The HV amplifier source is
56
connected to a copper clip that is clamped down onto the metal contact pad to complete
the circuit.
Once electrode-to-LN contact was determined to be an issue due to the unpredictable
poling, a new type of electrode was devised. The next photolithography mask had electrodes
with a square collar surrounding the grating so that the chips would squarely sit on the
collar and pressure from the ground electrode would ensure contact between the grating and
crystal. An example of this type of electrode fabricated with chrome on a silica substrate
with larger 100 µm features is shown in figure 3.4. The silica substrate chip is 15 mm wide
to allow the HV system to clip onto the contact pad while in contact and centered on a 10
mm MgLN chip. The square collar is therefore 10 mm wide to support the chip and ensure
level electrode to crystal contact. The outer edges of the chip have some chrome remaining
from photolithography that was removed by carefully dipping each edge in chrome etchant.
Contact poling was partially successful in that there were generally reversed domains
beneath the electrodes. However, poling often failed in both of the following ways: incomplete
poling in the desired area beneath the electrode, and over-poling out from beneath the
electrode. Furthermore, poling regions were often characterized by numerous micro-domains.
These features were characterized by selective HF etching that only etches the -Z domains
(discussed in section 1.4.6). An example of these characteristics on a chip after HF etching
is shown in the SEM images in figure 3.5. The images on the left are taken with the SE2
detector that is more sensitive to topography, and the images on the right are taken with the
inLens detector that is more sensitive to surface potential. The images are of the ‘bottom’
surface of the chip, which had the ground electrode applied to it, and was initially uniformly
in the -Z state. Thus, the un-etched ‘islands’ are +Z domains that were through-poled.
There is a large hexagon shape, characteristic of LN domains [117]. The hexagon’s edges are
rounded, likely due to domain walls not being perfectly vertical in the crystal volume. The
hexagonal shape was most likely not underneath the bulk electrode during poling but seeded
57
Figure 3.3: Periodic grating electrode used in poling test. Four images are stitched together
due to microscope limitations.
58
Figure 3.4: Grating electrode with collared support to ensure chip sits level. Grating fingers
are 100 µm wide with a range of spacing to investigate over-poling.
59
and propagated through the crystal regardless. Strings of micro-domains are also visible,
but it is unclear why they follow such paths.
Figure 3.5: Bulk through-poled hexagon in MgLN. Left images are taken with the SE2
detector, right images are taken with the inLens. Hexagon was poled despite not being
underneath electrode position during poling.
The micro-domains were prolific on our poled samples (for all types of electrodes
described above). A high magnification SEM image of the micro-domains is shown in figure
3.6. Their three-sided structure mirrors lithium niobate’s trigonal crystal structure. One
hypothesis was they were caused by imperfect chip-to-electrode contact. They were deemed
problematic specifically because their presence makes differentiating domains uncertain.
For example, in the development of AES as a characterization method, survey areas were
60
generally boxes spanning 200 × 200 pixels. If there are numerous unseen micro-domains of
opposite polarization from the rest of the region, the AES signal will be from a mix of +/-Z
domain. This concern led us to the next method of poling utilizing electrodes fabricated
directly on the MgLN surface.
Figure 3.6: High magnification SEM image of un-etched micro-domains on -Z MgLN surface.
3.4 Poling with electrode fabricated on crystal surface
In a similar manner to contact poling, applying a high voltage pulse to a ferroelectric
crystal though an electrode fabricated on the crystal surface can pole the material.
Fabrication of the electrodes on MgLN is detailed in section 2.1.2. The electrode is fabricated
61
in much the same manner as the external electrode but uses the ferroelectric chip as the
substrate. Although fabrication on MgLN introduces a few new challenges as they are
ferroelectric, piezoelectric, and pyroelectric, it removes the problem of poor or inconsistent
contact with an external electrode when poling.
An example of an electrode fabricated on the MgLN surface is shown in figure 3.7. Note
that the grating fingers’ spacing is varied to allow calibration of over-poling. The fingers are
100 µm wide, and 2 mm long. The top edge of the contact pad can be seen in the bottom
of the image. This electrode was fabricated with aluminum in order to avoid any potential
effect from chrome etchant on MgLN. A square aluminum back plane was also deposited on
the crystal face opposite the structured electrode, to be used as the ground in the poling
process. The sample was poled with 3.9 kV for 4.0 s.
After poling, the sample had the electrode removed in aluminum wet etchant, and this
sample was also HF etched to approximately 50 nm depth (10 min in 40% HF) in order to
examine the quality and fidelity of the poling. Figure 3.8 shows the MgLN surface after HF
etching. Comparing with figure 3.7 to see where the electrodes were located, we see that the
chip is significantly over-poled. The over-poled domains from the fingers that were closest
together completely merged. Note the hexagonal tips to each finger, characteristic of poled
domains in LN [117].
The poling in figure 3.8 is not very uniform, and a closer look at the domain edges
is shown in figure 3.9. The striations or feathered appearance are stripes of un-poled (or
perhaps backswitched [117–119]) micro-domains in the region of over-poling that stretches
out from under the electrode’s former position. Figure 3.9 shows the original -Z face of the
crystal, where poled regions are +Z, such that the original -Z crystal face is etched down,
and the poled regions are un-etched.
There were three persistent challenges that led to the decision to switch poling methods
again: the persistence of micro-domains even when the electrode was fabricated on the MgLN
62
Figure 3.7: Grating electrode on MgLN substrate, with varied distance between fingers.
63
Figure 3.8: Poled MgLN after removal of electrode and HF etching the surface. Poled
fingers are much wider than electrode finger widths and poling from fingers nearest each
other merged.
64
Figure 3.9: Micro-domains in over-poled region.
65
surface, the effort and cost of having a single shot at poling per MgLN chip, and the need
for unambiguously poled domains for use in developing AES characterization. These factors
motivated the decision to test poling single spots with a probe-tip.
3.5 Probe-tip poling
Probe-tip poling (PTP) is a form of contact poling we developed that utilizes a single
probe-tip as the HV poling electrode, as opposed to the structured electrodes in the previous
sections. Poling with a probe tip has the advantage of ensuring contact to the crystal, without
any fabrication steps, and allows many poled spots with different parameters for each, on a
single chip. The probe tip is a sharp microprobe and is shown in figure 3.10, with the inset
image a zoomed-out view. The tip is roughly 65 µm wide as shown in the measurement in
figure 3.10.
Figure 3.11 shows an SEM image of a chip poled with 5 rows of 5 spots with each row
poled with a different applied voltage and duration. The image was taken with the Auger
Nanoprobe’s SEM while the chip is tilted 75◦ and part of the clip holding the chip can be
seen in the top left of the image. The colored boxes around each row of five PTP spots
indicates a different poling voltage and duration, according to the legend above. 0 kV spots
where the probe tip was brought into contact with the surface, but no voltage applied were
used to check for damage from the probe contact. In the region of the top row of spots in the
purple box that were poled with the largest voltage, 4.7 kV, there are numerous secondary
domains that poled despite being unconnected to the spot poled with the probe tip. 3.9 kV
was affirmed as a nominal working applied poling voltage because the number of secondary
poled domains was limited, yet the poled spot was large enough to be used for AES testing.
An example of the poling from the probe tip is shown in figure 3.12, where the spot
was poled with 3.9 kV for 4.0 s. The top row images are before HF etching and the bottom
is after (note the round residue spot after HF etching). The images on the left are taken
66
Figure 3.10: Optical images of the probe tip. Tip diameter measurement is approximately
65 µm.
67
Figure 3.11: SEM image of chip with five rows of 5 spots poled with a probe tip. Each row
is poled with a different applied voltage and duration, denoted by the colored boxes. Chip
is tilted 75◦.
68
with the SE2 detector, and the right are inLens detector. Again, the roughly hexagonal
shape is characteristic of MgLN domains. There are a few secondary domains surrounding
the main hexagonal domain that are not connected to each other. In the top right inLens
SEM image, ten raster-burned boxes from AES survey areas can be seen on the left side of
the image, with five outside the poled domain and five inside. More details on AES on these
PTP domains is given in section 7.3. There is damage in the center of the poled region from
the probe tip but there are regions of uniformly inverted poling surrounding this damaged
area in most cases that was useful for AES characterization.
The damage to the crystal from the probe tip was caused by the applied voltage because
damage was not present when the tip was brought into contact, but no voltage was applied
(see black box in figure 3.11). Figure 3.13 shows SEM images of the probe tip damage. The
triangular damage was characteristic of each damaged spot and mirrors the MgLN’s trigonal
structure. While this damage was a drawback to the PTP method, the poled domains still
allowed AES testing in the region away from the voltage-contact damage.
An important finding from PTP was that the use of Silicon (Si) oil severely inhibited
over-poling. While applied voltage pulses with longer duration led to larger poled spots,
when Si oil was placed on the chip during poling, the poled region was almost completely
confined to the region of tip contact. This difference is shown in figure 3.14. These SEM
images were taken in the Auger Nanoprobe and the chip is tilted 75◦. The domains in
both images were poled with 3.9 kV, but the left image had Si oil applied and was poled
for 45 s, while the right image had no Si oil and was poled for 1.5 s. The domain poled
without Si oil is significantly larger than the domain poled with Si oil, despite the poling
being 30 times shorter in duration. The ability of Si oil to suppress over-poling when used
during the poling process could greatly improve the fidelity of domains poled with structured
electrodes, and further research could greatly improve periodic poling efforts. At this time in
the project however, un-etched PPLN scrap wafers were acquired from G&H, which greatly
69
Figure 3.12: SEM images of a hexagonal PTP domain. Top and bottom images are before
and after HF etching, respectively. Left and right images are SE2 detector and inLens
detector, respectively. Hexagonal Domain was poled at 3.9 kV for 4.0 s and is approximately
240 µm wide.
70
Figure 3.13: SEM images of damage to MgLN from PTP.
facilitated AES characterization of periodically poled structures and became the primary
priority (chapters 5 and 6).
3.6 Summary
In this chapter I described the poling efforts on MgLN. We started out trying to
explore poling on small size scales, then switched to trying to create samples useful for
characterization. First, an overview of the poling process was given. Next, an in-house poling
system was built, and setup for contact electrodes. Several electrode sizes/shapes were tested,
from mm-scale machined electrodes, to micron-scale grating electrodes. Ultimately, achieving
good electrode-to-crystal contact was deemed an issue, and electrodes were fabricated on
the MgLN surface. Over-poling and micro-domains remained an issue. In the interest of
creating samples with unambiguous domains in order to develop characterization methods,
chips were then poled with a probe-tip. This ensured electrode-to-crystal contact and allowed
many more poling tests per chip. Except for the region in contact with the probe-tip that
was damaged, this method produced cleaner poled domains than previous efforts. The
application of Si oil during probe-tip poling was found to significantly limit the degree of
71
Figure 3.14: SEM images of the effect of Si oil on PTP over-poling domains. Both domains
were poled at 3.9 kV, but left image was poled with Si oil for 45 s and right image was poled
without Si oil for 1.5 s and is much larger. Images were taken in the Auger Nanoprobe and
the chip is tilted 75◦.
72
over-poling on MgLN.
73
CHAPTER FOUR
AUGER ELECTRON SPECTROSCOPY FOR SURFACE
FERROELECTRIC DOMAIN DIFFERENTIATION IN
SELECTIVELY POLED MgO:LiNbO3
4.1 Contribution of Authors and Co-Authors
Manuscript in Chapter 4
Author: Torrey McLoughlin
Contributions: Performed experiments. Developed the analysis code. Performed the
analysis. Wrote first draft of manuscript. Made edits based on comments from other authors.
Managed the submission and review process.
Author: Dr. Wm. Randall Babbitt
Contributions: Developed the analysis code. Provided feedback/guidance for the analysis.
Provided comments and edits for the manuscript.
Author: Dr. Phillip A. Himmer
Contributions: Conceived the study idea.
Author: Dr. Wataru Nakagawa
Contributions: Provided feedback/guidance for the analysis. Provided comments and edits
for the manuscript.
74
4.2 Manuscript Information Page
Torrey McLoughlin, Wm. Randall Babbitt, Phillip A. Himmer, Wataru Nakagawa
Optical Materials Express
Status of Manuscript:
Prepared for submission to a peer-reviewed journal
Officially submitted to a peer-reviewed journal
Accepted by a peer-reviewed journal
X Published in a peer-reviewed journal
Published by OSA Publishing
Published October 2020, Opt. Mater. Express 10, 401938
75
4.3 Abstract
Auger electron spectroscopy (AES) as a method to characterize the ferroelectric polar-
ization domains in magnesium-doped lithium niobate crystals is demonstrated. Preliminary
measurements on a test sample show a clearly identifiable relative shift in the energy of
the Auger oxygen KLL transition peak between poled (inverted) and un-poled domains.
Auger electrons detected from the negative polarization domains (-Z) have a higher energy
than those from the positive domains indicating a lower ionization energy at the -Z domain
surface. The degree of electron energy separation between the –Z and +Z domains was found
to be dependent on proximity to the domain boundary and was potentially diminished by
the accumulated charge under the incident primary beam. Polarization domain resolution
is demonstrated on both the micron and millimeter scale, suggesting potential applicability
of this technique to surface investigation and domain structure characterization of nonlinear
optical devices such as periodically poled lithium niobate.
4.4 Introduction
Ferroelectricity is a unique material property of a built-in electric polarization field
that is maintained in the absence of an external field. The orientation of this spontaneous
polarization can be switched by the application of a sufficiently strong external electric
field that overcomes the material’s coercive field [120]. Lithium niobate (LN) is one
such ferroelectric crystal of interest due to its nonlinear optical properties and uniaxial
ferroelectricity [86]. Ferroelectric crystals have found numerous applications in nonlinear
optics [72, 74, 86, 121–123]. Periodically poled lithium niobate is especially noteworthy
for its ability to maintain quasi-phase matching and therefore high nonlinear conversion
efficiency [58, 124–132]. Furthermore, by adding magnesium doping to lithium niobate
(MgLN), the coercive field can be substantially lowered to facilitate polarization switching
76
[95, 133], and the threshold for optical damage due to the photo-refractive effect substantially
increased [115], thus making magnesium-doped lithium niobate crystals of particular interest
for nonlinear optics applications. However, the periodic poling process requires precise
manipulation of the crystal’s polarization orientation.
In order to form a stable state, a depolarization field is present on LN’s ferroelectric
surface [55, 58, 134]. The depolarization field is polarization dependent and is accomplished
by spontaneous atomic desorption, adsorption of environmental contaminants, different
surface terminations, or other processes which are not completely understood [12, 134, 135].
These complex surface interactions require further investigation in order to enable the
growing number of novel applications of LN. Developing new techniques for studying the
crystal’s surface properties and characterizing the polarization and crystal domains of LN and
MgLN will also help to better understand and mitigate the challenges of poling [136]. In the
present work a new characterization method is developed using Auger electron spectroscopy
(AES). Polarization domains in MgLN crystals are differentiated by a relative shift in the
energy of Auger spectral peaks. While AES has been used to depth profile ion-exchanged
waveguides in LN and LiTaO3 [137], AES has not been used to study the surface potential or
characterize polarization domains. This novel technique is non-destructive and offers unique
insight into the crystal surface when compared with other methods of characterization.
Currently, polarization selective domain etching in hydrofluoric acid (HF) is the most
dependable method for ferroelectric domain characterization in MgLN. HF selectively etches
-Z domains of lithium niobate crystal species and leaves the +Z face essentially unaffected
[62–66, 68, 138], so that after a sufficiently long immersion in HF, the domain pattern
is topographically etched into the crystal’s surface. The domain pattern can then be
characterized as a relief image with scanning electron microscopy (SEM), scanning probe
microscopy (SPM), or optically. While selective etching in HF is a common method [139, 140]
and is useful for characterizing large areas of poled materials, it has the disadvantage of being
77
destructive to LN and MgLN surface applications.
Other methods of characterizing MgLN ferroelectric domains include piezo-response
force microscopy (PFM), SEM imaging of secondary electrons (SE), and Raman spectroscopy
(RS) [134, 141]. PFM is a novel SPM technique where a conducting probe is brought
into contact with a ferroelectric surface and an AC voltage applied to the tip. The
sample’s oscillating response from the converse piezoelectric effect will either be in phase
or out of phase with the applied voltage and, because there is a linear relationship
between piezoelectricity and ferroelectricity, the direction of the ferroelectric domains can
be determined [13, 15]. It has been noted that the domain contrast mechanism with PFM
is not well understood due to the complexity of the tip-surface interaction [16, 142].
SEM imaging is a useful method of characterization that allows for imaging large areas.
SEM images detect the contrast in secondary electron yield between MgLN domains of
opposite polarization with certain primary beam parameters [30–32]. Focusing on detection
of secondary electrons through the use of an in-lens detector also gives information about
the material’s work function, although the information about the work function and image
contrast can be complicated by “beam parameters, beam-induced contamination, specimen
electric potential, SE collection efficiency, etc.” [35]. Similar to PFM, the contrast mechanism
from SEM is not well understood and there are different theories on whether the domain
contrast originates primarily from beam-induced pyroelectric effects [37] or differential
charging when the SEM is operated in the regime where primary beam current and surface
emitted current are roughly equal [30] or create a positive surface [31].
Raman spectroscopy has proven to be a useful technique for probing lithium niobate
domain walls, internal fields, and defect structures. By exciting atomic vibrational modes
with an incident laser, the inelastically scattered laser photons are shifted up or down in
energy which can then be detected by a spectrometer to probe material properties. Domain
walls in lithium niobate can be detected by a modulation in the intensity of the A1(LO4)
78
phonon mode [42, 43], or a shift in frequency of the E(TO8) phonon mode [44–46]. Similarly,
opposite domains can be detected relative to each other by a modulation in the E(TO1)
intensity, or a shift in the frequency of the A1(LO4) phonon mode [43–45, 47]. RS also
has the especially useful ability to non-destructively probe dopant distribution and the ion-
implanted waveguide structure in the bulk of nonlinear optical materials [44, 143]. As an
optical technique, spatial resolution of RS is set by the diffraction limit of the laser and
generally on the order of 0.3 µm, a relatively small value but large when compared with
the size of domain walls which are only a few lattice constants wide in LN [144]. In non-
stoichiometric MgLN or LN samples, a single phonon mode’s detected shift in intensity,
frequency, or width may be affected by a combination of factors such as impurities, lattice
inhomogeneity, strain, point defects, and internal fields [44, 134, 145, 146], though it has been
demonstrated that domain walls can be studied independent of these factors in annealed,
defect free, stoichiometric LiTaO3 and near-stoichiometric LiNbO3 [144].
Cherenkov second harmonic generation (CSHG) microscopy is another demonstrated
technique for imaging domain walls in ferroelectric crystals, including LN [48–50]. It has
been demonstrated that a laser focused on a ferroelectric domain wall generates a CSHG
signal that is absent in the homogeneous bulk of the crystal [51, 52]. Thus, by monitoring the
CSHG signal while scanning the sample with a laser beam, domain walls can be mapped out.
This non-destructive optical method has the added benefit of being able to probe the crystal
volume and has been employed in-situ while poling in order to probe the 3D domain growth
process in a 40×60×60 µm3 volume of strontium barium niobate crystals [147]. The origin
of the CSHG signal at the domain walls is not fully understood, but has been attributed to
either the broad range of reciprocal vectors perpendicular to the domain wall, a strong local
electric field at the domain wall [51], or to lattice distortions from the poling process [53].
While super-resolution methods can be employed to precisely determine domain wall position
with sub-micron resolution [52], CSHG microscopy’s imaging resolution is still restrained by
79
the primary laser beam’s diffraction limit. Furthermore, a domain’s polarization direction
cannot be determined from CSHG without prior knowledge.
In order to develop a non-destructive technique for characterizing ferroelectric domain
patterns and study the surface of MgLN crystals, Auger electron spectroscopy (AES) is used
to characterize the domains of in-house poled magnesium-doped lithium niobate crystals.
AES is a technique that analyzes the characteristic Auger electrons ejected from a sample
under primary electron beam irradiation. The kinetic energies of the detected Auger electrons
are specific to the atomic species that emits them, enabling material characterization. AES
is a highly surface sensitive method due to the short mean free path of the low energy Auger
electrons in solids and thus only probes to depths of approximately 1–5 nm of a material
[148], much less than other techniques such as SEM and CSHG. Oppositely polarized MgLN
domains are found to be differentiable by a relative shift in the energy of the Auger O
KLL peak, with -Z domains having a higher peak energy, in agreement with the literature
where -Z domains have a lower electron affinity [12, 149, 150]. This technique has the
potential to non-destructively characterize polarization domains on both the sub-micron and
millimeter scale, benefits of PFM and SEM, respectively; have higher spatial resolution than
diffraction limited RS and CSHG techniques; and study the material’s surface and origin of
its work function without the complexity of PFM’s tip-surface interaction. We demonstrate
the feasibility of AES characterization of ferroelectric domain with samples that have been
etched in HF first in order to facilitate finding domains of opposite polarization. However,
the AES domain characterization technique could in theory be employed on samples with
unknown polarization domains as a non-destructive means of characterizing poled crystals.
4.5 Experiment
Samples of single crystal, Z-cut, optical grade magnesium doped (5% molar) lithium
niobate with dimensions of 10×10×0.5 mm3 were obtained from MTI Corp. MgLN was
80
Figure 4.1: Contact poling system. Ferroelectric MgLN is placed between a flat HV electrode
and a smaller ground electrode, the shape and size of which spatially defines the poled area.
The HV amplifier is controlled by the AWG, which sets the pulse duration and voltage. The
force applied by the springs is controlled by vertical position of XYZ stage.
chosen in order to facilitate poling due to its lower coercive field of ∼4.45 kV/mm [96],
compared to ∼21 kV/mm for congruent un-doped lithium niobate [99, 100] and because of
its preferred use for nonlinear optics applications. Upon receiving the crystal, the +Z face
is determined using the compression method from Weis [107], which detects differences in
the piezoelectric voltage generated when an electrode is pressed on the crystal surface due
to the converse piezoelectric effect. Poling is performed using an in-house contact poling
system shown in figure 4.1, which consists of millimeter-scale metal contact electrodes, and
a Trek 610B High Voltage Power Supply Amplifier coupled with an Agilent 33220A arbitrary
waveform generator (AWG) for controlling the poling voltage amplitude as a function of time.
Samples were poled using a +3.9 kV, 4.0 s square pulse applied to the crystal’s original +Z
face. After poling, samples are placed in room temperature 40% HF acid for 100 min, which
results in etching the sample to a depth of approximately 0.5 µm in the -Z regions (the
switched regions on the original +Z face). An SEM image of the MgLN sample is shown in
figure 4.2. A roughly triangular poled area, corresponding to a region of etched –Z domain,
is visible in the left side of the image. Examining the other side of the MgLN sample shows a
complementary set of domains were etched, indicating that the samples were through-poled.
81
Figure 4.2: SEM image of MgLN sample. The roughly triangular region on the left side of
the image corresponds to the poled region, made visible under SEM by HF-etching of the
–Z domain. The FOV is 500 µm and primary beam voltage and current are 1 kV and 1
nA, respectively. The sample is tilted 75° in the SEM, with the bottom of the image being
closest to the observer.
The Auger electron spectroscopy analysis process begins by placing a sample of interest
on a metal stage with a 30° tilt and securing it with a copper clip. The stage is transferred into
the UHV main chamber of the Scanning Auger Electron Nanoprobe (Physical Electronics
710) and tilted 45° such that the sample’s total tilt is 75°. The use of a high tilt is an
important technique to mitigate charging of the insulating sample because the primary beam
is more likely to be reflected off of the sample surface rather than penetrating deeper into
the material, lowering the amount of embedded charge [151]. The area of interest is located
using the Nanoprobe’s SEM at low magnification and the primary beam at 1 kV and 1 nA.
In order to calibrate the spectrometer, the SEM’s field of view (FOV) is first moved
82
Figure 4.3: An example of an AES survey of MgLN with six survey areas. a) 500 µm wide
SEM image of MgLN sample with the roughly triangular area on the left side of the image
corresponding to the poled and etched -Z domain. The labeling of the six AES survey areas
corresponds to the order in which the surveys were performed. Odd numbered AES survey
areas are located inside the poled (–Z) domain region. b) Auger spectra for the 6 survey
areas taken with a relatively wide energy range in order to find the O KLL peaks’ initial
position (highlighted by the green stripe), located at approximately 525 eV in this instance.
Niobium MNN and carbon KLL transitions are also noted. These surveys were taken on a
separate day from the rest of the surveys shown in figures 3-8.
laterally to the side (away from the area of interest) to an area that is roughly the same height
on the tilted sample in order to avoid excess beam exposure to the area of interest during
calibration. The FOV is reduced to 50 µm and the spectrometer scans a narrow range of
electron energies surrounding 1 kV, the primary beam voltage. The sample generally needs
to be raised up in order to ensure electrons at the elastic peak energy (1 kV) are focused
onto the detector. This process brings the sample into the focal point of the detector, which
also ensures the strongest signal. After spectrometer calibration, the primary beam analysis
83
parameters are set to 5 kV and 10 nA, and the focus and stigmation are adjusted at high
magnification to give the clearest image. The SEM magnification is decreased and the SEM’s
FOV is moved back to the area of interest.
The Auger peak from the oxygen KLL transition (O KLL, nominally at 531 eV) is
chosen to differentiate domains due to the larger peak amplitude when compared with other
peaks. For insulating ferroelectric samples, the initial position of various Auger spectral
peaks is often shifted by tens of eV due to charging and sample history. In order to focus a
survey on the Auger O KLL transition, an initial survey is performed with a wider range of
5-605 eV in order to find the O KLL peak’s initial position. An example of this is given in
figure 4.3, where the O KLL peak is seen around 525 eV. This AES survey consisted of scans
of survey areas 1-6 in figure 4.3-a, which places odd areas inside the etched –Z domain (on
the left side of the image) and even numbered areas in the original +Z domain. The survey
was a single cycle from 5-605 eV with 1.0 eV steps, 20 ms/step, and primary beam settings of
5 kV and 10 nA. Survey areas are scanned in an alternating manner inside and outside of the
poled area (figure 4.3-a) to increase spacing between most subsequent scans and allow more
charge relaxation. Once the O KLL transition’s initial peak position is identified, subsequent
scans can use a narrower survey energy range.
4.6 Results and discussion
AES scans are performed with a narrower range of energies to demonstrate how a
relative shift in the O KLL transition’s peak energy can be used to differentiate polarization
domains in MgLN. An SEM image of an MgLN sample is shown in figure 4.4-a, with a
roughly triangular poled and etched –Z domain on the left side of the image. An AES
survey is performed on the ten areas indicated by the numbered and colored boxes in Figure
4.4-a. The scans are performed sequentially in numerical order. The odd-numbered areas
are in the -Z domain and even-numbered areas are in the unmodified +Z domain. Each area
84
has a side dimension of approximately 24.4 µm.
Figure 4.4: a) 500 µm wide SEM image of MgLN sample with poled/etched –Z domain
on left side of image. Selected AES survey areas are located across middle of image with
numbering showing their sequential order. Note the alternating order in and out of the poled
region. b) AES spectra corresponding to color-coded survey areas in SEM image. Survey
has relatively low time/step (2 ms) but poled/un-poled spectra are still clearly differentiable.
c) Spectra with higher time/step (100 ms) to improve signal to noise. d) Gaussian fit curves
corresponding to spectra in 4-b (solid lines) and 4-c (dashed lines) shown above. Blue colored
spectra are in –Z domain and are well separated due to their higher peak energies, even for
the two survey areas closest to the domain boundary. 4-b and 4-c are surveys #39 and #40
in figures 4.6 and 4.7.
Figures 4.4-b and 4.4-c shows the results of two individual surveys, consisting of ten
color-coded spectra corresponding to each of the areas. There is a clear trend with the
85
odd-numbered spectra from survey areas inside the -Z poled region (various shades of blue)
having a higher peak energy than the even-numbered spectra positioned outside in the +Z
region (colored red or orange). This shift in the spectral peak demonstrates that domains
of opposite polarization can be differentiated using this method. Figure 4.4-c utilized a
relatively long 100 ms per step, however the much faster measurement of 2 ms per step
shown in 4.4-b still provides the differentiable shift in peak energies between domains of
opposite polarization.
Auger electron spectroscopy data is analyzed in MATLAB. The Auger O KLL energy
peaks are treated as approximately following a normal distribution. The uneven background
is ignored, and the top half of each peak is fit with a Gaussian in order to determine the peak
energy. Figure 4.4-d shows the Gaussian curves fit to the spectra in 4.4-b and 4.4-c (dashed
lines), and emphasizes the separation between spectra inside and outside of the poled and
etched –Z domain.
The degree of relative peak separation is partially dependent on the distance from the
domain boundary. This is notable in figures 4.4-b and 4.4-c when looking at the two areas
located nearest to the domain boundary, areas 9 and 10 are just inside and outside of the
poled region, respectively. Their corresponding spectra have a smaller relative separation
when compared with the rest of the areas that lie further from the domain boundary.
Next, a series of AES scans is performed at higher magnification, resulting in smaller
scan areas with side dimensions of roughly 4.9 µm. Figure 4.5-a shows an SEM image with
survey areas and the corresponding spectra similar to figure 4.4-a, except that the FOV is
lowered to 100 µm. Figures 4.5-b and 4.5-c are surveys with relatively low 2ms per step,
and higher 100 ms per step (which increases survey time), respectively. The fitted Gaussian
curves shown in 4.5-d correspond to the spectra in 4.5-b and 4.5-c (dashed curves) and show
the clear separation and differentiation of peak energies between poled (blue colors, left side
of SEM image) and un-poled (red/orange colors, right side of SEM image) regions. At this
86
smaller FOV the relative peak separation is more apparent for faster surveys (4.5-b). The
measured peak energy of areas 9 and 10 both exhibit unique behavior, likely due to their
Figure 4.5: a) 100 µm wide SEM image with triangular poled/etched –Z domain on left
side of image. AES survey areas across center of image with alternating order in and out of
poled region, with numbering showing temporal order of survey areas. b) Fast AES spectra
with 0.1 eV steps and low time/step (2 ms); blue spectra for areas that are located in poled
–Z domain have higher energies and are well separated from red spectra for areas that are
located in original +Z domain. c) Slower AES spectra with 0.1 eV steps and relatively high
dwell time (100 ms/step) with higher signal to noise. Peak energies of poled and un-poled
regions are still separable but energy separation is smaller relative to larger FOV shown in
figure 4.4. d) Gaussian fit curves for surveys from 5-b and 5-c (dashed lines), demonstrating
peaks are well separable and domain resolution of approximately 9 µm. 5-b and 5-c are
surveys #24 and #29 in figures 4.6 and 4.7.
87
proximity to the domain boundary, and it must also be noted that these areas may not
be wholly inside/outside of the domain due to both image drift at this smaller FOV and
imprecision in positioning the FOV.
Figure 4.6: Peak energies for numerous surveys with changing FOVs. The studied FOVs
are indicated by color-coded background, with FOVs given in microns. Surveys 1-40 have
odd areas [1,3,5,7,9] in poled –Z domain, which have higher average energy. Surveys 41-70
are located in a control region on the MgLN sample and all 10 survey areas are located in a
region of un-poled original +Z domain. The figure shows why relative peak energy shifts in
an individual survey are used for domain characterization, because the average peak energy
of an individual survey drifts more than the relative peak separation on any single survey.
Surveys 29 and 40 were taken with 100ms/step, all other surveys utilize a faster 2ms/step.
88
To confirm these results, a number of additional surveys are conducted and the peak
energies of all 10 survey areas are plotted for each survey in figure 4.6. Surveys are performed
comparing poled/etched and un-poled regions of the sample at three different FOV settings:
500 µm FOV (surveys 1–10, 31–40), 200 µm FOV (surveys 11–20), and 100 µm FOV (surveys
21–30). Additional surveys are conducted in an un-poled, un-etched region of the sample
nearby (meaning all 10 scan areas are in un-poled regions) at each FOV setting: 500 µm
FOV (surveys 41–50), 200 µm FOV (surveys 51–60), and 100 µm FOV (surveys 61–70).
As described above, a curve-fitting process is performed on the results of each survey to
determine the peak energy value. For surveys 1–40, odd-numbered survey areas are located
in -Z domains and are color-coded shades of blue, while even-numbered survey areas are
located in +Z domains and are color-coded shades of red or orange. For surveys 41–70, the
same color code is used, but all areas are located in +Z domains. In Figure 4.6, the results
shown in Figures 4.4-b and 4.4-c correspond to surveys 39 and 40, respectively, while the
results shown in Figures 4.5-b and 4.5-c correspond to surveys 24 and 29, respectively.
For each survey including both poled and un-poled areas (1–40), in almost all cases there
is a separation between the measured peak energy value for poled areas and un-poled areas.
Surveys 21, 22 and 30 do show the measured peak energy in area 10 being higher than for
area 7. This is likely due to its proximity to the domain boundary and image drift such that
area 10 may be partially straddling the domain boundary. In the remaining surveys, wholly
in un-poled regions, no notable distinction is seen. Thus, Figure 4.6 provides additional data
supporting the conclusion that a relative shift in the O KLL transition energy peak can be
used to differentiate polarization domains in MgLN.
Over time, as more surveys are performed the mean energy of individual surveys can
drift farther than the relative shift in peak energies from opposite domains in a single survey.
This necessitates a relative measurement in order to characterize domain polarization and
has so far been one notable limitations of using AES for characterizing MgLN domains.
89
Figure 4.7: Peak energies’ deviation from the mean of each individual survey’s peaks for the
surveys shown in figure 4.6. Changing FOV shows the relative peak shift between domains is
consistent but decreases with FOV. At 100 µm FOV the deviation from the mean decreases
with more surveys indicating an effect from cumulative current density incident on a survey
area. Survey areas 9 and 10 are nearest to the domain boundary and consistently have less
relative peak shift due to this proximity. FOVs are in microns.
Because the drift of mean energy after a number of surveys can be larger than the
peak separation on an individual survey, experiments have so far been limited to selecting a
cluster of survey areas rather than employing mapping across the whole SEM image. Another
challenge to mapping the whole FOV is that the MgLN sample is highly tilted in the AES
90
instrument in order to mitigate charging. Thus the top (bottom) of the image is farther
away from (closer to) the Auger electron analyzer, which alters the detected Auger electron
energy. The mean survey energy generally drifts faster at smaller FOV, which is at least
partially caused by the higher incident charge density at smaller FOVs. As noted, survey
areas closer to the domain boundary have smaller deviations from the mean peak energy and
this can be seen more broadly by tracking the positions of areas 9 and 10 (cyan and light
orange) across surveys 1-40 in figure 4.6.
Due to the drift of the mean peak energy between individual surveys, figure 4.7 shows
the peak energy of all 10 survey areas for the surveys shown in figure 4.6, with each individual
survey’s mean peak energy subtracted (deviation from the mean). The reliable separation of
peak energies from poled (blue colors) and un-poled areas (red/orange colors) is more clear
with each survey’s mean energy subtracted.
In order to quantify the peak energy separation from opposite polarization domains,
it is useful to look at the average difference of the +Z and -Z domains’ deviations from
the mean, in particular when looking at the decreasing relative shift between polarization
domains with decreasing FOV. For survey numbers 1-40 at FOVs of 500 µm, 200 µm, and
100 µm shown in figure 4.6 the average difference of the poled and un-poled areas’ mean
deviations is 6.6 eV (surveys numbers 1-10, 31-40), 4.5 eV (surveys numbers 11-20), and 1.3
eV (surveys numbers 21-30), respectively. The average difference of the deviation from the
mean for the control group surveys (areas only in the unmodified +Z region, surveys 41-70)
for FOVs of 500 µm, 200 µm, and 100 µm are -0.07 eV (surveys 41-50), 0.01 eV (surveys
51-60), and 0.05 eV (surveys 61-70), respectively, with their mean standard deviation being
0.15 eV. The average difference of the deviation from the mean for the surveys with both
+Z and -Z domains are all significantly larger than both those of the control group and the
control group’s standard deviation, though there are still not well understood positional and
charge dependent effects.
91
Figure 4.8: Peak energies across numerous surveys and FOVs; similar to figure 4.6 except
taken on a different day and AES session. Peak energies from areas located in the poled -Z
domain [1,3,5,7,9] in surveys 1-53 have consistently higher energy than those in the un-poled
+Z domain [2,4,6,8,10]. This indicates characterization experiments are repeatable. The
initial surveys in the un-poled region (#54 and #55) have an uncharacteristic shift of the
survey areas on the right hand side of the image [10,8,6,4,2], which quickly fades after a few
surveys or cumulative dwell time under the SEM beam. This is likely due to initial charge
relaxation effects of the area when initially exposed to the e-beam. Surveys 42 and 53 were
taken with 100 ms/step, all other surveys used a faster 2 ms/step.
92
Figure 4.9: Peak energy deviation from the mean energy of individual survey for a large
number of surveys, similar to figure 4.7 but data was taken on a different day (raw data
already given in figure 4.8). Once the survey areas were moved to the unmodified region of
the sample there is an initial uncharacteristic shift in approximately surveys #54-59. The
relative peak separation between odd and even survey areas is initially opposite to that seen
when odd survey areas are in the poled –Z domain. We attribute this to charging effects due
to the new survey area being in close proximity to the previous area which has accumulated
substantial charge. The dissipation of the relative shift after about 5 surveys at the larger
500 µm FOV supports this notion.
A second data set is shown in figure 4.8, taken from the same MgLN sample and in
the same region as figure 4.6, but on a different day. Surveys #1-53 have odd-numbered
areas located within the poled –Z domain, and even-numbered areas located in the un-poled
original +Z domain, while surveys #54-82 have all 10 survey areas located in the original
93
un-poled +Z domain. In surveys #1-53 blue colors (corresponding to poled –Z domain)
have consistently higher peak energies than red/orange colors (corresponding to un-poled –Z
domain).The data in figure 4.8 is also given in figure 4.9 but shown as the deviation from the
mean. The peak separation between poled and un-poled survey areas is quite clear across
a number of surveys and FOVs. As stated previously, survey areas closest to the domain
boundary (areas 9 and 10 corresponding to cyan and light orange) have consistently smaller
peak separation. When comparing the 200 µm FOV surveys in the two data sets of figures
4.6 or 4.8, we see that the survey-to-survey change in mean energy is positive or negative,
respectively. Even though the survey-to-survey average peak energy drift is somewhat
unpredictable, Figures 4.7 and 4.9 show that the AES domain characterization experiments
using relative peak shifts are repeatable. This also again emphasizes the requirement of a
relative measurement of peak position rather than an absolute peak energy measurement:
the energy of the O KLL transition from MgLN samples is modified by multiple factors
which may include MgLN’s insulating, pyroelectric, and ferroelectric properties.
The limits of the AES system’s (Physical Electronics 710) spatial resolution is reported
to be 8 nm (20 kV, 1 nA primary beam)[152], although we expect that this resolution likely
cannot be achieved with insulating ferroelectric samples due to charging effects. The results
with 100 µm FOV presented in figure 4.7 (surveys 31-40) and in figure 4.9 (surveys 32-42)
have survey areas in -Z and +Z domains which are separated by 9 µm, center-to-center. For
this set of measurements, most but not all are separable, suggesting that 9 um is close to
the resolution limit for this experimental configuration (5 kV, 10 nA primary beam), likely
due to charging effects. Preliminary results testing lower current (1 nA) at a FOV of 10
µm have shown reduced peak energy drift and promising improvements in spatial resolution,
with 0.9 µm spatial resolution between survey areas potentially achievable, however with
reduced signal-to-noise ratio. Ongoing investigations are exploring methods for reducing
sample charging to achieve higher spatial resolution while maintaining sufficient signal-to-
94
noise ratio.
The shift in the Auger O KLL transition energy peak between the +Z and -Z domains
of LN surfaces could stem from a difference in electron affinity between the two domains.
Other research has seen a difference in the electron affinity between domains of opposite
polarization in LN using UV-photoelectron emission microscopy (PEEM) with -Z domains
having a lower electron affinity [12]. The difference between domains of 1.6eV in the
electron photo-threshold was attributed to differential adsorption on the domain surfaces
with exposure to atmosphere [12]. In AES, the kinetic energy of the Auger electrons is
decreased by the electron affinity [153]. Thus, the lower electron affinity of -Z domains would
be expected to result in a shift to higher energy Auger peaks for -Z domains compared to
+Z domains, consistent with the results presented in this paper. Similar differential effects
between different domains were observed in the selective deposition of charged polystyrene
microspheres on domain-patterned LN [154]. Electrostatic force microscopy and scanning
surface potential microscopy studies of ferroelectric barium titanate measured a potential
difference between domains of opposite polarity and concluded that surface adsorbates
played a role in the shift [142]. However, the electron affinity can also be affected by
surface reconstructions and atomic steps [134]. Density functional theory calculations of the
ionization energy differences between domains suggest the differences exist in clean surfaces,
and originate largely from different surface terminations [134]. Regardless of the sources of
the differences in electron affinity between domains, AES can play a role in its detection
and mapping domain structures. Furthermore, studying the origin of the relative peak shift
between domains may lead to a method to enhance the dipole field and further improve
spatial resolution.
There are two main advantages to AES for studying the contribution of adsorbates to
the observed shift between the +Z and -Z domains. First, AES is operated in UHV such
that once a surface is clean, atmospheric contamination can be negated. Second, the AES
95
instrument contains an in situ Ar-ion etching capability that can be used to clean the surface
and perform depth profiling. Initial tests utilizing the Ar-ion gun showed the relative peak
shift between domains vanished after light Ar-ion etching. This suggests that the detected
peak energy shifts between domains may be related to chemical adsorbates on the sample
surface, however the shift may also be affected by differential surface charging. Further
investigation into the origin of the ionization energy gap between different polarization
domains is ongoing.
4.7 Summary
This work has demonstrated a new method of characterization for polarization domains
in MgLN. Auger electron spectroscopy was used to differentiate polar +/- Z domains in
MgLN by observing a relative shift in the energy of the O KLL transition energy peak
between the two domains. Auger electrons emitted from the –Z domain had a consistently
higher energy; the magnitude of the peak separation was found to be partially dependent on
the proximity to the domain boundary and the amount of accumulated incident charge at
smaller FOVs. Currently, domain differentiation is determined relatively, because the survey-
to-survey mean peak energy drift can be greater than the per-survey peak energy separation.
Spatial resolution down to approximately 9 µm has been demonstrated alongside mm-scale
characterization, a wide range of scales when compared with other non-destructive methods
of characterization. Recent preliminary results exploring lower incident currents have shown
some promise of improved spatial domain resolution (to below 1 µm) and reduced peak energy
drift (potentially negating the need for a relative measurement). Additionally, incorporating
AES’s in situ Ar-ion sputtering will allow us to investigate the origin of the peak shift on
various ferroelectric surfaces cleaned of contamination.
96
Funding
This material is based upon work supported by the National Science Foundation under
grant number 1710128.
Acknowledgments
The authors gratefully acknowledge Nathaniel Rieders for technical assistance and
useful discussion. This work was performed in part at the Montana Nanotechnology Facility
(MONT), a member of the National Nanotechnology Coordinated Infrastructure (NNCI),
which is supported by the National Science Foundation under grant number ECCS-1542210.
Disclosures
The authors declare no conflicts of interest.
97
CHAPTER FIVE
NANO-SCALE FERROELECTRIC DOMAIN DIFFERENTIATION IN PERIODICALLY
POLED LITHIUM NIOBATE WITH AUGER ELECTRON SPECTROSCOPY
5.1 Contribution of Authors and Co-Authors
Manuscript in Chapter 5
Author: Torrey McLoughlin
Contributions: Performed experiments. Developed the analysis code. Performed the
analysis. Wrote first draft of manuscript. Made edits based on comments from other authors.
Managed the submission and review process.
Author: Dr. Wm. Randall Babbitt
Contributions: Developed the analysis code. Provided feedback/guidance for the analysis.
Provided comments and edits for the manuscript.
Author: Dr. Wataru Nakagawa
Contributions: Provided feedback/guidance for the analysis. Provided comments and edits
for the manuscript.
98
5.2 Manuscript Information Page
Torrey McLoughlin, Wm. Randall Babbitt, Wataru Nakagawa
Optics Continuum
Status of Manuscript:
Prepared for submission to a peer-reviewed journal
Officially submitted to a peer-reviewed journal
Accepted by a peer-reviewed journal
X Published in a peer-reviewed journal
Published by Optica Publishing Group
Published April 2022, Opt. Continuum 1 (4), 649-659.
99
5.3 Abstract
A new method for characterizing lithium niobate +/-Z ferroelectric polarization
domains using Auger electron spectroscopy (AES) is presented. The domains of periodically
poled samples are found to be differentiable using the peak amplitude of the Auger oxygen
KLL transition, with -Z domains having a larger peak-amplitude as compared to +Z
domains. The peak amplitude separation between domains is found to be dependent on
the primary beam current, necessitating a balance between the insulating samples charging
under the primary beam and achieving sufficient signal to noise in amplitude separation. AES
amplitude-based domain characterization is demonstrated for fields of view (FOV) ranging
from 1 µm to 78 µm. Domain spatial resolution of 91 nm is demonstrated at 1 µm FOV.
5.4 Introduction
Ferroelectric materials, with their built-in electric polarization fields, have found nu-
merous applications in optics due to their wavelength conversion capabilities [72, 74, 86, 121–
123, 128]. One material of particular interest is periodically poled lithium niobate (PPLN),
for its numerous nonlinear optical applications [58, 127, 128, 130, 132]. Periodically switching
the direction of the ferroelectric polarization enables quasi-phase matching (QPM) with
optical beams polarized along the largest element of the nonlinear tensor, allowing high
nonlinear conversion efficiency [124–126, 129, 131]. One limitation to PPLN is that the most
efficient first-order QPM processes often require sub-micron poling periods [58, 121, 127, 155].
A new non-destructive domain characterization method with sub-micron resolution that
complements existing methods would be useful for the study and understanding of sub-
micron ferroelectric domains [121].
There are several established methods for ferroelectric domain characterization, though
they each have their individual limitations. Perhaps the most common method is visualizing
100
the domains after selective etching in HF acid [139, 140], which has the effect of only
etching the -Z domain surface [62–66, 68, 138], however this method is destructive for
surface applications. Piezoresponse force microscopy (PFM) is a scanning probe microscopy
technique that images domains based on the phase of the sample’s piezoelectric response to
an AC voltage applied to the probe [13, 15]. The method has nano-scale resolution, but
measuring domain wall width is limited by probe tip size [25, 156], and the origin of the
image contrast is complicated [157–161], as is data interpretation [17, 162–164]. Scanning
electron microscopy (SEM) is a useful technique for visualizing domain boundaries. However,
due to the complexity of pyroelectric and ferroelectric lithium niobate (LN), +/-Z domain
identification is often ambiguous due to variations in domain shading with changes in beam
exposure and beam parameters [30, 31, 35]. Two optical methods, Raman spectroscopy (RS)
and Cherenkov second harmonic generation (CSHG) microscopy, have been demonstrated.
While these methods are advantageous for their ability to probe the volume of LN crystals
[44, 51, 52, 143, 147], the imaging resolution for both methods is limited to ≈ 500 nm by
the probing laser’s diffraction limit [52, 144, 165, 166].
Ferroelectric domain characterization in magnesium doped lithium niobate (MgLN)
using the oxygen KLL (O KLL) peak energy shifts in the Auger electron spectroscopy (AES)
spectra was recently demonstrated [167]. However, insufficient peak energy separation at
smaller fields of view (FOV) limited the spatial resolution to a few microns. In the present
work, a new Auger electron spectroscopy method for ferroelectric domain characterization
in lithium niobate species is reported, which uses peak amplitude separation of the oxygen
KLL peak. We demonstrate non-destructive surface domain characterization of PPLN with
unambiguous determination of +/-Z domains with sub-micron spatial resolution.
101
5.5 Experiment
Wafers of un-doped, congruent PPLN were obtained from G&H with a 15 µm poling
period and were diced into square 10 mm chips. The PHI 710 Auger Nanoprobe loading
process is described in a previous work [167]. Due to the insulating property of LN, the
sample is tilted to 75° relative to the electron beam axis in order to limit sample charging
by increasing the portion of the beam that is reflected off the sample surface and decreasing
embedded charge [151, 168]. The sample is raised up until the region of interest is brought
into the focal point of the Auger electron detector, a cylindrical mirror analyzer (CMA), and
the SEM image is focused with the primary beam set to 5 kV. The AES energy range is
chosen to include the oxygen KLL transition (O KLL, nominally at 531 eV [169]), the most
prominent AES elemental peak on LN samples. A fitting routine in MATLAB is applied
to each AES spectrum in order to determine the peak energy and amplitude of the O KLL
peak. The fit function is a Gaussian with a linear background and a non-zero y-intercept.
The sloped background accounts for the contribution due to inelastically scattered Auger
electrons at energies below the peak that are not present at high energies.
5.6 Results
The SEM image in figure 5.1(a) shows the approximate location of ten sequential AES
survey areas on the PPLN sample. With an FOV of 78 µm, survey areas are spaced half
of a poling period apart, such that adjacent scan areas alternate between +/-Z domains.
Light/dark blue colored survey areas are in -Z domains and red/orange colored survey areas
are in +Z domains. The +/-Z domains were confirmed independently using HF etching on
a nominally identical die diced from the same wafer. The boxes are numbered according to
their sequential scan order and are placed in such a configuration that subsequent survey
areas are farther away from each other in order to help mitigate charging effects. The ten
102
Figure 5.1: (a) 78 µm FOV SEM image of PPLN with associated AES survey areas, numbered
and color-coded to spectra in (b). (b) Auger O KLL transition spectra from PPLN. Colored
spectra correspond to the 10 colored survey areas in the SEM image in (a). Light/dark blue
curves correspond to -Z domains and red/orange curves correspond to +Z domains. The
peak of each spectrum, as determined by the fitting routine, is plotted with a circle/triangle
for -/+Z domains, respectively. 78 µm FOV was chosen such that 10 evenly spaced survey
areas alternate +/-Z domains.
AES spectra shown in figure 5.1(b) are color-coded so that each spectrum can be matched
with the survey area in which the spectrum was taken, as indicated by the coloring of the
boxes in the SEM image in figure 5.1(a). Figure 5.1(b) demonstrates that the -Z domains
(blue colors) and +Z domains (red/orange colors) are distinguishable by their separation
in both peak energy and peak amplitude: a comparable shift is observed in the Auger
electron peak energy and the Auger electron count rate (amplitude). This is consistent with
the previously published results of peak energy separation at 100 µm FOV on MgLN [167]
although at slightly smaller size scale. The domain differentiation demonstrated in figure
5.1 is non-destructive and did not involve HF etching as was done in previous work [167]. It
is important to note that the domain orientation in the SEM image is ambiguous, but can
be determined when performed in conjunction with the differentiated AES spectra.
The Auger O KLL transition on LN samples was observed to drift by tens of eV across
an Auger session containing many surveys, likely due to charging issues [167]. Because of
this, relative measurements of peak energy separation were made for each survey, rather than
103
using absolute energy values. This is accomplished by subtracting each individual survey’s
mean peak energy from the peak energy of each survey area (in this case 10, 5 from each
+/-Z domain), such that peak energies can be visualized more clearly as a deviation from
the survey mean. The same is done for peak amplitude, though the variations in the mean
amplitudes were relatively constant within each set of surveys at a given primary beam
current. In figures 5.2(a) and 5.2(b), the deviation from the mean for peak energy and
amplitude, respectively, are plotted for 31 surveys at different primary beam currents. The
survey areas in survey numbers 5–27 correspond to the 10 color-coded AES survey areas
in the 78 µm FOV SEM image shown in figure 5.1(a). As a control group, the first and
last sets of surveys (surveys 1–4 and 28–31) were taken by moving the sample such that all
ten different survey areas on the crystal were on a -Z domain. These control surveys are
denoted ’All -Z’ in figures 5.2(a) and 5.2(b). In the ’All -Z’ surveys blue/red separation is
Figure 5.2: (a) Peak energies (deviation from the mean) of all 10 survey areas, for 31 separate
AES surveys with 78 µm FOV and different primary beam currents (indicated by labels at top
of graph and changing background color). (b) Peak amplitudes (deviation from the mean)
for the 31 surveys with 78 µm FOV and different primary beam currents. Ordinate axis
values are in kilocounts/second. In surveys 5–27, the ten numbered survey areas correspond
to alternating -/+Z domains as shown in the SEM image in figure 1(a), while in the control
group (surveys 1–4 and 28–31, denoted ‘All -Z’) all ten survey areas were placed within a
region of the crystal with only -Z domain.
104
not observed, indicating that the separation seen between +/- Z domains is not a function of
the procedure used to perform the surveys. The different background colors used in figures
5.2(a) and 5.2(b) denote changes in the primary beam current, as indicated in the second row
of labels at the top of the figures. The spectra shown in figure 5.1(b) correspond to survey
#9 in figure 5.2. As previously noted, the light/dark blue circles are in -Z domain and the
Figure 5.3: (a–c) SEM images of 9 µm, 3 µm, and 1 µm FOVs at 5 nA primary beam
current, with associated AES survey areas. At 1 µm FOV, the separation between two
adjacent survey areas is 91 nm, center to center. (d) Peak amplitude (deviation from the
mean) for several sets of surveys for different FOVs and primary beam currents (as indicated
at the top of the figure). The surveys labelled “All -Z” were performed with the survey areas
in a different part of the crystal, with all areas in -Z domain.
105
red/orange triangles are in +Z domain, in order to better see the separation of the plotted
peak energy and amplitude values. In the deviation from the mean peak energy plot in figure
5.2(a), -/+Z domains (blue/red colors) are separated at 10 nA, and mostly separated at 5
nA, however at lower primary beam currents, the domains are not well differentiated by their
measured peak energies. In contrast, in the deviation from the mean peak amplitude plot
Figure 5.4: A second data set from a separate sample nominally identical to the one shown in
figure 5.3. (a–c) SEM images of 9 µm, 3 µm, and 1 µm FOVs, respectively, at 1 nA primary
beam current and associated AES survey areas. (d) Peak amplitude (deviation from the
mean) for a number of surveys at 9 µm, 3 µm, and 1 µm FOVs and 10 nA and 1 nA primary
beam currents. The surveys labeled “All -Z” were performed with the survey areas in a
different part of the crystal, with all areas in -Z domain.
106
in figure 5.2(b), -/+Z domains (blue/red colors) are separated at all primary beam currents.
There is a strong correlation between peak amplitude separation and primary beam current
(though too high beam current causes problematic sample charging). This demonstrates
that peak amplitude separation is a more pronounced method for domain characterization
at this FOV over a wider range of beam currents compared to the peak energy separation
method [167].
Consequently, the ability of peak amplitude separation to characterize +/-Z domains
at smaller FOV is investigated. In figure 5.3(a–c) the SEM images with AES survey areas at
9 µm, 3 µm, and 1 µm FOV are shown, respectively. For these smaller FOVs, the five survey
areas on the left hand side of the image are all located in -Z domain, while the survey areas
on the right are in +Z domain, and the color-coding has been adjusted accordingly so that
light/dark blue circles (red/orange triangles) denote data derived from -Z (+Z) domains.
Figure 5.3(d) shows the deviation from the mean peak amplitude for sets of surveys with
different FOVs and primary beam currents, as indicated along the top of the plot and by
the changing background colors. The initial 8 and final 12 surveys are a control group,
performed with all ten survey areas located in a region of the crystal with only -Z domain,
as described above. They are labeled ’All -Z’ in figure 5.3(d). There are a few instances
where area 9 (cyan color), which is located closest to the domain wall on the -Z side, crosses
over or nearly crosses into the +Z domains’ amplitudes. We speculate this is due to sample
charging causing the electron beam location to drift slightly during scanning such that area
9 is sometimes straddling the domain wall. Furthermore, at 1 µm FOV and 1nA primary
beam current, peak amplitude separation is poor, also thought to be due to image drift. The
level of peak amplitude separation is well correlated with primary beam current, however
the use of large primary beam current must be balanced against increased sample charging
and the associated image drift, which limit spatial resolution. Furthermore, peak amplitude
separation between domains is not strongly dependent on FOV, as was observed with peak
107
energy separation [167]. The amplitude differentiation at 1 µm FOV at 5nA and 2nA in
figure 5.3(c) (surveys 27–31, and 43–47) is especially interesting as it demonstrates 91 nm
spatial resolution between the central survey areas 9 and 2, center to center.
To investigate the repeatability of the measurement, figure 5.4 shows a second set of
AES data taken using a different die diced from the same PPLN wafer as the sample shown
in Fig. 3. The two samples have nominally identical poling patterns. As with the previous
sample, as shown in Figure 4(a–c), in the SEM images at 9 µm, 3 µm, and 1 µm FOVs, the
five survey areas on the left hand side of the image are in -Z domain and the five surveys
on the right are in +Z domain. In figure 5.4(d), the initial 12 and final 17 surveys are
placed with all ten survey areas within the bulk -Z domain as a control group. In the
control surveys, no differentiation is observed, whereas in surveys with both +/-Z domains
there is good red/blue separation and thus good domain characterization based on the peak
amplitude differentiation method. In a few instances at 10 nA, the +Z area closest to the
domain wall (area 2, orange triangle) are higher than the other red/orange points, likely due
to image drift caused by surface charging. At 1 nA beam current, there are a few instances of
imperfect red/blue separation, likely due to insufficient signal to noise in the measured peak
amplitude. It should be noted that on this sample, domain differentiation was possible for
1 nA surveys at 1 µm FOV, whereas on the previous sample (used in figure 5.3), separation
at 1 nA, 1 µm FOV was poor, likely due to image drift. While domain differentiation using
the amplitude separation method was possible on both samples, there were slight differences
in the optimum configuration between the two nominally identical samples. Identifying and
controlling the causes of these inconsistencies could lead to better domain characterization
and finer spatial resolution. For underlying values for figures 5.1–5.4 see Data File 1–4,
respectively.
108
5.7 Discussion
The underlying physical reason for the peak amplitude separation between +/-Z
domains is not well understood. In our previous work, the peak energy separation between
domains was explained by the difference in potential due to polarization surface charges in
each domain, such that Auger electrons escaping from +Z domains would be slowed more
than those from -Z domains and thus their detected peak energy would be lower [167]. While
the peak energy separation is observed to be FOV dependent, the peak amplitude separation
is observed to be constant for different FOVs with the same primary beam current. Whether
the potential difference between domains also causes the separation in Auger electron count
rates (peak amplitude) is unclear.
One hypothesis is that the amplitude separation between domains originates from a
change in the amplitude of the background and not the Auger O KLL peak itself. This
hypothesis is discounted, since spectra from different domains are not separated by amplitude
outside the region of the peak. Another hypothesis is that the amplitude separation is a result
of the unique geometry of the experiment, in which the sample is tilted to 75° with respect
to the axis of the electron beam and CMA detector. The CMA detector has an angular cone
of acceptance of 42+/-6° and thus it is likely that the chip’s high tilt angle shadows some
portion of the cone of acceptance. Perhaps slight variations between the domains’ surface
potential vary this clipping. Consequently, the impact of incidence angle on count rate was
investigated using a conducting aluminum sample. An external bias voltage is applied to the
aluminum sample, ranging from 0 V to 12 V. As the applied bias is increased, a decrease in
amplitude (Auger electron count rate) was observed (this investigation is described in greater
detail below). This experiment was repeated at several lower incidence angles, including 0°
(normal incidence), and although a slight shift in the amplitude was observed, qualitatively
similar results were obtained for all incidence angles tested. Thus, although this effect
109
Figure 5.5: AES of the O KLL transition on an aluminum sample tilted at 75° with an
external applied bias (denoted along top of plot). (a) Peak energy values. A horizontal
line is drawn at the mean energy for all surveys with no applied voltage (507.2 eV) as well
as at offset energies corresponding to the applied voltage values [12, -3 -6 -9 -12]eV above
and below. (b) Mean peak amplitudes for all ten spectra in all four surveys at each voltage
setting. Black horizontal lines show the average amplitude of adjacent measurements with
no applied voltage; these values are used as a baseline to determine the amplitude shift (red
lines) for each measurement with an applied voltage. Nearest-neighbor baseline averages were
used in this way in order to account for the overall upward drift in mean peak amplitude.
See Data File 5 for underlying values
could not be tested with an LN sample, these results suggest that the incidence angle to
the sample does not have a significant impact on the AES signal amplitude for surfaces of
110
different potential.
Another hypothesis is that the negative polarization charge of the -Z domain may slow
the incident primary electrons more than the +Z domain, causing a higher output signal
because the lower energy electrons are more likely to interact near the surface of the sample,
1–5 nm deep where Auger electrons are able to escape [148] and be detected. Experimentally,
when the beam voltage (energy) of the primary electron beam is decreased from 5 kV to 3
kV with the same primary beam current on PPLN samples, the Auger signal increased. This
may be evidence that lower energy primary electrons have a higher probability of interacting
in the region where most detected Auger electrons are produced.
To further explore the possible effects of surface potential on Auger O KLL peak
amplitude, AES was performed on a conducting aluminum sample. The Al sample was
tilted to 75° and the sample’s potential was varied with an external applied bias voltage.
The results are shown in figure 5.5(a), with applied voltages listed along the top. The O KLL
peak energy of the aluminum sample shifts an amount equal to the applied voltage (within
+/-0.08 V). The stability of the peak energy at a given voltage on aluminum supports the
idea that the peak energy is unstable on LN due to its insulating properties. Similar data
to that in figure 5.5(a) was obtained for the peak amplitudes. The average amplitudes from
40 spectra (4 surveys of 10 different survey areas) at each voltage setting were averaged
and plotted in figure 5.5(b). Due to issues with a varying background amplitude, the shift
in amplitude (the red lines in figure 5.5(b)) was measured by subtracting the amplitudes
measured with an applied voltage from the mean 0V amplitude measured just before and
after (the black lines in figure 5.5(b)). The general trend of the amplitude increasing with
increasing number of surveys may be due the AES chamber vacuum improving over time.
As seen in figure 5.5(b), increasing (decreasing) the applied bias on the sample decreases
(increases) the peak amplitude, by 0.2% per volt on average. In one reference, photoemission
electron microscopy (PEEM) measurements of the +/-Z domains work functions found 6.2
111
eV and 4.6 eV, respectively, a separation of just 1.6 eV [12]. Applying this work function
shift to our results would predict a shift of 0.3% in amplitude between the +/-Z domains.
While the sign of the shift is consistent, the predicted value is approximately an order of
magnitude lower than our calculated shift of 3.6% between +/-Z domains on PPLN. This
may be due to material differences between aluminum and LN. These two experiments show
that increasing beam voltage and applying external voltage bias both lower the Auger signal,
which qualitatively support the hypothesis. However, a better understanding is needed to
explain the quantitative differences, and whether -Z domains’ higher Auger signal is due
to primary beam electrons losing more energy when incident upon the sample. While the
physical mechanism underlying the amplitude shifts between the +/-Z domains is not fully
understood, the empirical ability of the AES peak amplitude method to characterize PPLN
domains was confirmed using the established HF etch method.
5.8 Summary
A new method for using Auger electron spectroscopy to characterize ferroelectric
domains in periodically poled lithium niobate is demonstrated. It is found that in addition
to a relative peak energy shift of the Auger O KLL transition between +/- Z domains (as
demonstrated previously [167]), +/-Z domains can be reliably differentiated using the peak
amplitude separation of AES spectra. Advantages of this method include that it is non-
destructive, it allows unambiguous determination of the +/-Z domains, and it has been
demonstrated over a range of spatial resolutions spanning nearly two orders of magnitude,
down to 91 nm. This is an improvement of two orders of magnitude over the spatial
resolution achieved with peak AES energy separation in previous work on MgLN [167].
Further investigation is needed to understand the origins of AES peak amplitude shifts and
to demonstrate the applicability of AES peak amplitudes as a non-destructive method for
imaging ferroelectric domains with fine resolution over large fields of view.
112
Funding
This material is based upon work supported by the National Science Foundation under
grant number 1710128.
Acknowledgments
The authors gratefully acknowledge G&H for supplying the crystals, and Nathaniel
Rieders and Christopher Ebbers for technical assistance and useful discussions. This work
was performed in part at the Montana Nanotechnology Facility (MONT), a member of the
National Nanotechnology Coordinated Infrastructure (NNCI), which is supported by the
National Science Foundation under grant number ECCS-1542210. This material is based
upon work supported by the National Science Foundation under grant number 1710128.
Disclosures
The authors declare no conflicts of interest.
113
CHAPTER SIX
LITHIUM NIOBATE DOMAIN MAPPING WITH AUGER ELECTRON
SPECTROSCOPY
6.1 Overview
Chapter 5 demonstrated that lithium niobate’s (LN) polar ferroelectric +/-Z domains
are separable by the difference in magnitude of the O-KLL Auger electron transition, with the
AES count rate being larger for the -Z domain. In this chapter, the +/-Z signal separation
is used to perform AES mapping on a periodically poled LN sample, allowing full imaging
of the domains with sub-micron resolution.
6.2 Background
Auger electron spectroscopy (AES) was previously demonstrated as having the ability
to differentiate ferroelectric domains in magnesium doped lithium niobate (MgLN) using the
separation of the O-KLL peak energy [167]. Subsequently, periodically poled lithium niobate
(PPLN) domains were differentiated using the Auger O-KLL transition peak amplitude [170].
The energy differentiation method was found to be unreliable at smaller fields of view (FOV),
but the amplitude differentiation method was more robust, demonstrating domain resolution
of 91 nm. Building upon the phenomenon of the Auger O-KLL transition amplitude being
larger in the -Z than the +Z domains in PPLN, full imaging of PPLN domains using AES
mapping is now demonstrated. AES mapping is an imaging method where each pixel’s gray-
scale represents the AES amplitude at a specified energy. This method is generally used to
characterize elemental differences in a material but here is adapted to determine ferroelectric
domain orientations in PPLN by the two different O-KLL signal levels.
114
6.3 Experiment
PPLN ferroelectric domains are imaged using AES mapping with a PHI 710 Auger
Nanoprobe. We obtained a double-side polished PPLN wafer with a 15 µm poling period
from G&H. The wafer was diced into 10 × 10 mm2 chips. The PPLN samples are tilted to
75◦ relative to the axis of the electron beam and the cylindrical mirror analyzer. Tilting the
samples is necessary in order to prevent the insulating lithium niobate (LN) from charging by
decreasing the number of embedded incident primary electrons [151, 168]. Oxygen’s Auger
KLL transition (nominally at 531 eV [169]) is found to be the most prominent peak in the
LN AES spectrum and is used for this analysis. The rest of the loading process and system
setup is described previously in chapter 4.
Mapping on LN is acquired in the following way. The region of interest is first focused
in the SEM. An AES spectrum is next taken that covers an energy range generally at least
30 eV above the nominal peak position due to the O-KLL transition energy’s potential to be
shifted by sample charging. Once the current O-KLL peak energy is determined, a map is
taken with the mapping energy chosen on the high side of the O-KLL energy peak position.
Relatively low-resolution maps are initially taken, with consecutive maps stepping down in
energy from above the O-KLL transition energy until domain contrast appears. The optimal
mapping energy is determined based on best contrast, at which point the resolution and
time per step are increased as needed. If domain contrast decreases during map acquisition,
it is likely due to drift in the O-KLL peak energy and can occur if the FOV is too large,
or primary beam current too low. Counter-intuitively, selecting smaller FOV or increasing
beam energy/current, although increasing the incident charge density, tends to stabilize the
charging and peak energy. Separately, if the FOV is too small or current too high, the
charging may cause the center of the FOV to drift. Therefore, FOV and beam current must
be balanced to some extent so that both the O-KLL peak position and the FOV window are
115
stable. The successful mapping results presented here were taken with 5 kV primary beam
at either 5 nA or 10 nA, and FOVs of [7.45, 18.3, 41.4, 200] µm .
6.4 Results and image processing
The optimal energy for domain contrast was determined with the process described
above and a set of SEM and mapping images at a 20 µm FOV are shown in figure 6.1. The
SEM image of the region is shown in figure 6.1(a). Figure 6.1(b) was acquired at 520eV,
which was above the optimal energy, and demonstrates that the mapping image contrast
disappears when the acquisition energy is too far from the O-KLL peak that is nominally at
510 eV. Figure 6.1(c) was acquired at the optimal 510 eV, and a relatively fast 1.0 ms/pixel
and 256 × 256 pixels (1.1 minute scan time). Figure 6.1(d) is taken at the same energy as
6.1(c), 510 eV, with a slower 5.0 ms/pixel, and higher resolution of 1024 × 1024 pixels, a
pixel-to-pixel distance of 19 nm. For maps, brighter pixels correspond to higher count rate
and based on previous work [170], -Z domains have larger amplitudes than the +Z domains.
Thus, mapping allows immediate determination of relative domain polarity. Figure 6.1(c)
and (d) also show the variation in contrast between ’slow’ and ’fast’ mapping images. All
maps were single-point mapping, meaning the map was acquired only at the noted energy,
with no reference energies. Next, image processing (as described below) is performed on
figure 6.1(d) in order to enhance domain contrast.
An algorithm was developed to process AES map images. The main challenge of
analyzing the map images is removing random noise in the data such that domain orientations
can be clearly and accurately determined, and doing so while minimizing loss of image
resolution. Image processing is performed on the raw data map shown in figure 6.1(d).
Several of the following steps can be performed because we know a priori that the domain
walls (DWs) are oriented close to vertical in the image. For ambiguous domain configurations,
several of the steps would require modifications. The initial map is shown on the top left
116
SEM image gnh03-09-145.map
0
2
4 50
6
8 100
10
12 150
14
16 200
18
20 250
0 5 10 15 20 50 100 150 200 250
µm (1024 pixels) pixels
(a) (b)
gnh03-09-151.map gnh03-09-154.map
100
50 200
300
100 400
500
150 600
700
200 800
900
250 1000
50 100 150 200 250 100 200 300 400 500 600 700 800 900 1000
pixels pixels
(c) (d)
Figure 6.1: (a) SEM image of PPLN with a 20 µm FOV. (b-d) Three AES maps at 20 µm
FOV on the same region with different parameters. (b) Map at 520 eV, above the O-KLL
peak, domains are not apparent. (c) Map at the optimal energy, 510 eV, with 256 × 256
pixels and a relatively fast 1.1 minute scan time. (d) Map also at the optimal energy, 510
eV, with 1024 × 1024 pixels and a relatively slow scan time of 88 minutes.
pixels µm (1024 pixels)
pixels pixels
117
gnh03-09-154.map Rows Normalized Map After Gauss Smoothing,  = 7
0 0 0
5 5 5
10 10 10
15 15 15
0 5 10 15 0 5 10 15 0 5 10 15
µm (936 pixels)
(a)
(b)
Figure 6.2: Top row: After cropping the map in figure 6.1(d), map data (left), map after
row-by-row normalization (center), and map after applying Gaussian smoothing filter with
standard deviation of 7 (right). Bottom row: Histograms of map data in row above. Map
FOVs are 936 pixels, 18.3 µm.
image in figure 6.2 with the only processing being that the edges are cropped to remove
edge effects from the AES system. The image in the center is after each row is normalized
(each row is divided by the mean value of that row). This is done in order to remove
fluctuations that likely comes from the instrument’s method of raster scanning left-to-right,
top-to-bottom. The removal of some horizontal stripes can be seen when comparing the left
and middle images.
The AES signal is not initially clearly separated between domains and so must be
smoothed. We employed a Gaussian image filter in MATLAB (imgaussfilt) and varied the
standard deviation (the half width of the blurring Gaussian). This can be seen for a single
row of data from the map in figure 6.3. Note the noise of the un-smoothed data (σ=0), and
µm (936 pixels)
118
Lineplot of Single Row for Different 
1.04  = 0
 = 1
1.03  = 2
 = 3
 = 4
1.02  = 5
 = 6
1.01  = 7
 = 8
 = 9
1  = 10
0.99
0.98
0.97
0.96
100 200 300 400 500 600 700 800 900
Figure 6.3: Lineplots of a single row from map for a range of sigma values, 0–10.
that the smoothed data begins to stratify into two levels. In order to further illustrate the
need for smoothing as well as show the effect of smoothing with different standard deviations,
lineplots of all rows of data are shown at different values of σ in figure 6.4. Again, note how
the data stratifies into two levels with smoothing. In order to find the DWs, we look for
peak and valleys in the differentiated smoothed data and ensure they are prominent enough
and have the proper spacing based on our knowledge of the poled period. The smoothed
data with the lowest standard deviation that finds the correct peaks is then chosen as the
‘optimal’ standard deviation.
At this point we want to find threshold values that properly divide the +/-Z domains
in the smoothed data. Two methods were tested to address this. In the first method, based
on the positions of the domain walls, the mean value of the signal for +/-Z domains can be
determined without including the DWs transition region, four total in this map. The mean
119
Figure 6.4: Lineplots of all rows from map (from figure 6.1(d)) for a range of sigma values,
[0, 1, 2, 3, 5, 7]. The reduction in noise for higher sigma underscores the need to smooth
until domains are well-separated in amplitude.
of those four values is taken as the baseline threshold. We also select thresholds above and
below that are 10% of the span of the max and min values of the domains’ mean signal. This
can be seen in figure 6.5, where the lineplots for all data rows (normalized and smoothed) are
plotted. The thresholds are also plotted as horizontal lines, where besides the DW transition
regions, most of the data lies outside the high and low thresholds.
The thresholding is first applied based on the baseline threshold (the middle horizontal
line in figure 6.5), and the binary image is shown in figure 6.6(a) where green/red data is
above/below the threshold, respectively. The image separates the domains as intended, with
no mixture of +/-Z domains. The remaining questions are how well the DWs are selected,
and how wide they are. To investigate this, the high and low thresholds are then applied to
the image, creating the tricolor image seen in figure 6.6(b), where black pixels represent data
120
Figure 6.5: Lineplots of all rows from map in top right image in figure 6.2. Lineplots are after
normalizing and Gaussian smoothing with σ=7. From the first thresholding method, the
middle horizontal line corresponds to binary thresholding, and the top and bottom horizontal
lines correspond to three-level thresholding.
that is between the high and low threshold (upper and lower horizontal lines in figure 6.5).
Like the binarized map, green/red data is above/below the high/low threshold, respectively.
There is some mixed data, especially in the left green domain, and the domain walls are
black pixels as expected.
To measure the apparent DW width the total number of black pixels along the DWs
are counted, then divided by the number of DWs (3), and by the number of rows (936). I
emphasize that this is the apparent width from the measurement, and not expected to be
121
Binary map,  = 7, Thresh = 1.0009 Tricolor map,  = 7, Thresh = [0.99954      1.0023 ]
0 0
2 2
4 4
6 6
8 8
10 10
12 12
14 14
16 16
18 18
0 5 10 15 0 5 10 15
µm (936 pixels) µm (936 pixels)
(a) (b)
Figure 6.6: Left: Binarized map data using the first thresholding method, middle horizontal
line in figure 6.5. Right: Tricolor map using the first thresholding method, top and bottom
lines in figure 6.5. Green is above the upper threshold, red is below the lower threshold, and
black are between the two. Map FOVs are 936 pixels, 18.3 µm.
the actual DW width which is expected to be single or a few lattices wide [26, 27]. For the
image processing in 6.6(b), the average apparent DW width is 6.25 pixels or 122 nm. This
is similar in order to the chosen standard deviation of σ=7, thus the measured DW here
is currently near process or system limited. The main drawback of this first method is the
somewhat arbitrary choice of thresholding values, although the apparent DW width and σ’s
similarity offers some validity to the chosen threshold.
A second method of thresholding was also tested. In the first method, the mean values
of the AES signals from the four domains were used to determine a baseline threshold. In the
second method, the baseline threshold is simply the mean of all the data. Because the rows
were individually normalized, the threshold is simply equal to 1. The high and low thresholds
µm (936 pixels)
µm (936 pixels)
122
Figure 6.7: Lineplots of all rows from map in top right image in figure 6.2. Lineplots are
after normalizing and Gaussian smoothing with σ=7. From the second thresholding method,
the middle horizontal line (1) corresponds to binary thresholding, and the top and bottom
horizontal lines correspond to three-level thresholding.
were chosen based on the standard deviation of all data (std2 in MATLAB). This method
still relies on an arbitrary choice of a factor multiplied by the standard deviation to decide
the high and low thresholds. For the sake of comparison with the first method, the high and
low thresholds were chosen such that the apparent DW width is approximately the same as
the first method. This meant the high and low thresholds were the baseline threshold (1),
plus or minus the standard deviation (5.1e-3) times a chosen factor of 0.255. This is shown
in figure 6.7, where the lineplots of all rows are again shown, and the baseline threshold is
123
Binary map,  = 7, Thresh = 1 Tricolor map,  = 7, Thresh = [0.9987      1.0013 ]
0 0
2 2
4 4
6 6
8 8
10 10
12 12
14 14
16 16
18 18
0 5 10 15 0 5 10 15
µm (936 pixels) µm (936 pixels)
(a) (b)
Figure 6.8: Left: Binarized map data using the second thresholding method, middle
horizontal line in figure 6.7. Right: Tricolor map using the first thresholding method, top
and bottom lines in figure 6.7. Green is above the upper threshold, red is below the lower
threshold, and black are between the two thresholds. Map FOVs are 936 pixels, 18.3 µm.
the horizontal line at 1 with high and low threshold above and below.
The binary image shown in figure 6.8(a) was created using the baseline threshold, 1, the
mean of all data. Similarly, in figure 6.8(b) a tricolor image was created using the high and
low thresholds. Compared to the first method shown in 6.6(b), this method finds 3% fewer
black pixels (18,635 compared to 19,215). Qualitatively, the thresholding does a better job
in the left -Z domain (green) but does not do as well around the third DW (farthest to the
right). To reiterate, in order to help compare the first and second thresholding methods, the
thresholding here was chosen so that the apparent DW width is the same as method 1, 6.25
pixels, 122 nm. One limitation of the second method is that it relies on a priori knowledge
of there being roughly equal amounts of +/-Z domains in the image. If the configuration
µm (936 pixels)
µm (936 pixels)
124
Figure 6.9: Lineplots of all rows from SEM image in 6.1 (a). Lineplots are after normalizing
and Gaussian smoothing with standard deviation of σ=7. Note the ambiguity of the curves,
disallowing domain determination in this configuration of SEM and sample.
of +/-Z domains were far from equal amounts, the mean value would need to be weighted
towards the domain with less area.
We aimed to compare AES mapping of +/-Z domains with an SEM image of the
same region. The AES amplitude separation method and mapping based on this concept
are beneficial techniques because they offer unambiguous relative determination of domain
orientation. SEM images are an important characterization tool for ferroelectric domains
but suffer from ambiguity, as domain and DW contrast can switch based on sample charging
and beam parameters [30, 31, 35, 36]. To investigate this, the same image processing applied
125
SEM Image SEM Image, Rows Normalized
0 0
2 2
4 4
6 6
8 8
10 10
12 12
14 14
16 16
18 18
0 5 10 15 0 5 10 15
µm (936 pixels) µm (936 pixels)
SEM Image Smoothed,  = 7 SEM image overlaid with Mapping DW
0 0
2 2
4 4
6 6
8 8
10 10
12 12
14 14
16 16
18 18
0 5 10 15 0 5 10 15
µm (936 pixels) µm (936 pixels)
Figure 6.10: SEM image processing for comparison with mapping image processing. Top
left: SEM image. Top right: SEM image after row-by-row normalization. Bottom left: SEM
image after normalization and Gaussian smoothing with σ=7 (same σ used for map image).
Bottom right: Normalized, smoothed SEM image with DWs calculated from map image
processing overlaid in orange. The mapping DWs are shifted 15 pixels to the right to try to
align with the center DW.
µm (936 pixels) µm (936 pixels)
µm (936 pixels) µm (936 pixels)
126
to the map image was applied to the SEM image in figure 6.1(a). The image was cropped by
the same amount as the map to remove edge effects, then normalized row-by-row, and then
a Gaussian smoothing filter with a standard deviation of σ=7 was applied. The lineplots for
all rows are shown in figure 6.9. Because there are not separable levels corresponding to the
+/-Z domains, domains are ambiguous. In figure 6.10, the SEM image processing steps are
shown in (a), (b), and (c), and in (d) the map image DWs calculated using the second image
processing method are overlaid on the SEM image in orange. The second image processing
method was used as it found 3% fewer black pixels than the first method. The SEM image
in (a) was taken just before the map, (un-processed map shown in the top left of figure 6.2).
The orange map image DWs overlaid in figure 6.10(d) are shifted to the right 15 pixels in
order to align the center DW with the SEM image. There are noticeable differences even
after aligning the center DW in both images namely, the left and right DWs in the SEM
image are slightly to the right of the DWs from the map.
In order to test reproducibility, we go back to the image processing of AES maps and
apply the method to a second map image from the same sample and AES session. The
data and analysis are condensed into figure 6.11, which shows the full image processing on
a map with a smaller FOV and only one DW in the center of the image. Figure 6.11(a)
is a 7.45 µm FOV map of the region before image processing. The map was acquired at
505 eV, with the primary electron beam at 5 kV and 10 nA and is 424 × 424 pixels, at
5.0 ms/pixel. The domain wall runs vertically in the center of the image. In Figure 6.11(b)
is the map after determining the optimal standard deviation and applying the Gaussian
smoothing. This map required much less smoothing, standard deviation of σ=2, than the
previous example. The lineplots in figure 6.11(c) show all rows of the data from (b), as well
as the thresholds used for three level thresholding. In figure 6.11(d) and (e) are histograms
of the data before and after smoothing, respectively, again with a standard deviation of σ=2.
Lastly, the thresholded image is shown in figure 6.11(f). Measurement of the apparent DW
127
gnh03-09-172.map Map After Gauss Smoothing,  = 2
0 0
1 1
2 2
3 3
4 4
5 5
6 6
7 7
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 (c)
µm (424 pixels) µm (424 pixels)
(a) (b)
Tricolor map,  = 2, Thresh = [0.9979      1.0021 ]
0
1
2
3
4
5
6
7
(d) (e) 0 1 2 3 4 5 6 7
µm (424 pixels)
(f)
Figure 6.11: (a) Uncorrected map of the region, with 7.45 µm FOV and 424 × 424 pixels.
(b) Map after smoothing with optimized standard deviation, σ=2. (c) Lineplot of all rows
of data after smoothing with σ=2. (d) Histogram of uncorrected data from (a) above. (e)
Histogram of smoothed data from (b) above. (f) Thresholded image based on thresholds
seen in (c) above. Measured apparent DW width is 3.79 pixels, or 66.7 nm.
width is performed as described above and found to be 3.79 pixels on average, or 66.7 nm,
which is roughly twice the smoothing’s standard deviation, σ=2. This suggests that the
apparent DW width may be system limited (again, LN’s DW is expected to be on the order
of lattice widths). This near factor of two improvement in the measured apparent DW width
is partially due to the higher beam current (10 nA compared to 5 nA in the previous image
µm (424 pixels)
µm (424 pixels)
µm (424 pixels)
128
SEM Image, Rows Normalized Map Image, Rows Normalized
0 0
5 5
10 10
15 15
20 20
25 25
30 30
35 35
40 40
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
µm (424 pixels) µm (424 pixels)
(a) (b)
Figure 6.12: (a) SEM image of PPLN. (b) AES mapping image of same area as SEM
image. Brighter/darker regions are higher/lower count rates and indicate -/+Z domains,
respectively. Both FOVs are 41.4 µm. This region with distorted domains was chosen in
order to show mapping on structures besides just perfect periodic domains.
analysis), which increases both the count rate and the SNR. FOVs of this scale and at this
relatively higher beam current of 10 nA will sometimes suffer from drifting due to charging
but this sample and AES map was well-behaved.
Finally, in order to demonstrate AES mapping works on domain configurations besides
perfectly straight periodic poling, a section of the sample was chosen with crooked DWs.
Figure 6.12(a) is an SEM image of the PPLN with crooked DWs with a 41.4 µm FOV.
Figure 6.12(b) shows the AES map of the region on the O-KLL peak at 515 eV. Figure
6.12(b) used single point mapping with 512 × 512 pixels, and 5.0 ms/pixel, and the only
image processing performed was row-byrow normalization, which was applied to both the
SEM and map images. The crooked domain walls can be correlated with the SEM image
µm (424 pixels)
µm (424 pixels)
129
both vertically and horizontally. Please note that because the chip is tilted 75◦, for a 41.4
µm FOV the vertical direction in the image spans 160 µm on the crystal surface, because
that is the direction of the chip’s tilt. Again, it is emphasized that determining the domain
polarization from the SEM image alone is not possible due to the ambiguous shading, where
in this example -Z domains on the right-hand side of the image are darker than the +Z
domains on the left-hand side. This issue has been discussed in the literature, with varying
theories on which domain is brighter and whether the contrast changes or reverses under
beam exposure and certain beam parameters [30, 31, 35].
6.5 Discussion
Generally, the AES signal is comprised of back-scattered primary electrons, secondary
electrons, inelastically scattered Auger electrons from higher energy Auger peaks, and finally
the zero loss Auger electrons [171]. It is then necessary to ask whether the peak amplitude
separation between +/-Z domains is due to discrepancies in one of these constituents of the
Auger signal. However, +/-Z domains are not separable in the Auger signal in regions of
the Auger spectra away from the Auger O-KLL transition peak. This is demonstrated in
figure 6.1(b) where the mapping fails when far enough from the O-KLL peak and so we
conclude that it is the zero loss Auger electrons that comprise the differential amplitude
between +/-Z domains. Some references note that the +/-Z domains have different atomic
surface terminations [172] and state “[in] thermodynamically stable conditions the only stable
stoichiometry at the positive z-cut is –Nb–O3–Li2, while at the negative z-cut the O–Li–
termination occurs under most conditions. Under strong O-rich conditions, the surface
may become O2– or O– terminated, though” [134]. Considering AES is a surface sensitive
technique that probes only 1–5 nm of material [148], this potential for higher surface oxygen
concentration after enough atmospheric exposure may explain the -Z domain’s larger Auger
O-KLL signal.
130
Figure 6.13: Left: SEM image with 200 µm FOV of PPLN. Right: AES map of same area as
SEM image (no image processing). The SEM image was taken immediately after acquiring
the map.
It is important to note the current limitations of AES mapping. Due to the sample tilt
of 75◦, when the center of the image is in focus, the top and bottom edges of the images
are out of focus and this effect is worse for larger FOVs. This is shown in 6.13, where the
left image is an SEM image at 200 µm FOV and the right image is a map of the same area
taken just before the SEM image. First, note that the SEM image fades from light to dark
from left to right, an artifact not present in the map. Second, notice that there appears
to be some debris or residue in the map that are not apparent in the SEM image. This is
likely due to the low energy Auger electrons being more sensitive to the residue than the
on-average higher energy electrons detected in the SEM image. Lastly, notice that in the top
and bottom of both map and SEM images, the domains are blurry due to the tilt of the chip
and the focus only being optimized in the center of the FOV. This factor adds to the balance
that must be struck between FOV and beam current to limit both peak energy drift and
131
FOV window drift. AES on LN has been tested at a lower tilt of 45◦ but was hampered by
problematic sample charging. Still, it is possible that a balance of beam parameters, FOV,
and tilt below 75◦, might be found that allow AES on LN.
In some maps there are anomalous horizontal lines such as in figure 6.12(b). They are
likely an instrument artifact due to the raster scanning from left-to-right, top-to-bottom. On
consecutive scans the lines either disappear or shift in the map image, suggesting they are
random as well. As noted in section 6.4, normalizing the maps row-by-row removed the lines.
However, this normalization is only valid when the +/-Z domains are distributed similarly
from top-to-bottom in the image.
While AES mapping allows full imaging, its pixel-by-pixel scanning is a relatively ’slow’
method, comparable to PFM. For an image of 1024 × 1024 pixels at 5 ms/pixel (figure 6.1(c)
for example), the scan time is 88 minutes. Minimizing the scan time should be performed
by determining the needed resolution and the tolerable noise level in a mapping image.
6.6 Conclusion
Based on the principle of the AES peak amplitude separation method, AES mapping
was performed on PPLN. Full imaging of LN’s +/-Z domains with AES was demonstrated,
elevating the amplitude separation method from a way to discriminate between +/-Z
domains, to a full imaging method. Mapping domains is unambiguous, although the presence
of opposite domains is required to determine the relative orientation. Image processing was
performed on a 7.45 µm FOV, 424 × 424 pixel PPLN map, and determined an apparent
DW width of 3.79 pixels or ∼67 nm. This is an upper limit for the DW width, not only
because the DW is expected to be on the order of lattice wide, but also because of the
smoothing performed. Resolution could be further improved by lowering the FOV, increasing
the number of pixels, or further refining the image processing. For the future, reducing the
noise in the obtained signals would reduce the need for smoothing, thus potentially increasing
132
the resolution or accuracy with which features such as DWs could be characterized. To the
best of our knowledge, this is the first demonstration of LN’s +/-Z domains being imaged
with AES.
133
CHAPTER SEVEN
ADDITIONAL AUGER ELECTRON SPECTROSCOPY RESEARCH
7.1 Introduction
Most of the contributions to the development of AES as a viable and effective
characterization method of LN ferroelectric domains were given in the previous three
chapters. First in chapter 4 AES was demonstrated as being able to separate domains using
the peak energy separation method, where the -Z domains O-KLL peak energy was higher
than the +Z domains. Those experiments were performed on MgLN poled in-house with
contact electrodes, and was selectively etched in HF after poling but before AES. While
the peak energy separation method was still being explored, a small separation in peak
amplitude was noted on PPLN samples, though at first seemingly less significant than the
peak energy separation. At one point in 2021 ICAL had to replace the objective lens in the
Auger Nanoprobe due to a short circuit in the coil. The issue with the Nanoprobe’s objective
lens that led to its replacement was only expected to affect Auger results at extremely high
magnification, FOVs below what was being tested on LN. For an unknown reason however,
the lens replacement led to not only an improvement in the magnitude of peak energy
separation, but a clear and unambiguous peak amplitude separation larger than previously
seen. While the mechanism that was obscuring the peak amplitude separation prior to
objective lens replacement is unknown, the peak amplitude separation quickly became
the focus for AES characterization research, explored in chapter 5. The peak amplitude
separation method was used on PPLN chips that had not been HF etched and demonstrated
resolutions down to 91 nm. With the peak amplitude separation method it became possible
to test full image mapping with AES. Therefore in chapter 6 AES mapping is used on PPLN
again, in order to show AES can reliably, and non-destructively, image LN’s +/-Z domains
134
with nano-scale resolution.
A few topics that did not fit into the previous papers regarding AES’s characterization
of LN and MgLN are given in this chapter. The first topic is the tendency of the Auger peak
energy separation between +/-Z domains to switch signs at low FOVs. This is an important
limitation to the peak energy separation method and could be an interesting area of further
research. It also enhances the relative benefits of the peak amplitude separation method at
low FOV. AES on PTP chips is also discussed. Other chips, PPLN, MgLN with electrodes
fabricated on the chip, and MgLN poled with contact electrodes and perhaps Si oil, all had
some processing or contact with external metal or chemicals that were deemed as a possible
source of the peak amplitude separation between +/-Z domains. PTP chips have only the
probe-tip’s contact with the chip in a small region of the poled domain and thus were used
to check whether the Auger domain differentiation was caused by an external factor. Finally,
a thin gold film was applied to the MgLN surface to test charge mitigation in order to allow
lower chip tilt when performing AES.
7.2 Energy inversion
In chapter 4 domain differentiation was demonstrated on MgLN using the energy
separation method. An important aspect of an AES characterization method would be
to allow sub-micron or nano-scale characterization. While investigating this possibility, an
interesting effect was observed, where at low enough FOV, or low enough primary beam
current, the energy separation inverts. Specifically, +Z domains have a higher energy O-
KLL peak transition than the -Z domains.
Figure 7.1 shows SEM images of the sample used in the following study of peak energy
inversion. The MgLN sample was contact poled with the same electrode geometry shown in
figure 3.7, that is, 50 µm wide grating fingers with various spacing. In the left image the wide
poled region has its left and right edges centered in the red boxes where the AES surveys
135
Figure 7.1: Left: 1000 µm FOV zoomed out image of poled region with red box on left
showing surveys region for ‘West’ side of poling, and right red box showing survey region for
all other surveys (’East’). Right: 100 µm FOV SEM image and 10 survey areas corresponding
to red box on right in left image. Note the domain wall meanders from top to bottom in
center of image, and divides survey areas in half.
136
were taken. The red box on the left side of the image is on the ‘West’ side of the poling, such
that the -Z domain would be on the right-hand side of the AES FOV. Conversely, the red
box on the right side of the image is on the ‘East’ side of the poling, such that the -Z domain
would be on the left-hand side of the AES FOV. The image on the right in figure 3.7 shows
the 10 AES survey areas on the ‘East’ side (in the right red box) at 100 µm FOV, with the
domain edge traveling roughly down the middle of the image, such that the -Z domain is on
the left side of the image. The sample is tilted 75◦ such that the top edge of the image is
farthest from the electron beam gun.
Inversion of the peak energy separation between domains can be observed in figure 7.2.
In the top plot, the peak energies are plotted for all surveys at a range of FOVs and beam
currents, again showing the peak energy drift and need to look at the deviation from the
mean. The middle plot shows the deviation from the mean peak energy for all surveys. In
the first two bands (surveys 1–11) of the middle plot the peak energy inversion is clear; at
100 µm FOV and 5 nA primary beam current, the blue colors (-Z domain) have a higher
peak energy but when primary beam current is decreased to 1nA the red colors (+Z domain)
have a higher peak energy. In order to make viewing the data simper, the bottom plot in
figure 7.2 shows the red and blue data averaged separately, such that each red/blue point
represents the average of the 5 red/blue colors in the middle plot. In this way, the degree and
direction of domain separation is more easily interpreted. In all three plots, there is a control
group of surveys (66–69) where all the survey areas are positioned in the same domain type
and labeled ‘Unpoled’ in a yellow box. There is also a group of 5 surveys (70–74) where
the FOV is moved to the opposite side of the poled region such that numbered survey areas
flip from -Z to +Z and vice versa. This switch from ‘East’ to ‘West’ was noted in the SEM
image in figure 7.1 and was done in order to ensure that there was not a left-to-right bias
influencing our results.
The range of surveys in figure 7.2 was designed to investigate the stability of the
137
Figure 7.2: Top: Peak energies across all surveys. Note FOV and beam current are denoted
at the top of each colored band. Middle: Deviation from the mean peak energy across all
surveys. Bottom: Mean of all +Z and -Z domains survey areas for each survey i.e. blue dots
represent the mean of odd-numbered survey areas and red dots represent the mean of the
even numbered survey areas. All surveys, except for 70–74, are on the ‘West’ side of poling.
Peak energy is inverted when red data are higher energy.
138
inversion in the energy separation between domains at low FOV and/or beam current. If
the energy inversion happens at a well-defined beam current, or more likely at a specific
current density, then the energy separation method could still be valid at all FOVs with
the addition of this caveat. However, it is clear in figure 7.2 that the inversion is not stable
and is somehow dependent on sample history. Looking specifically at the first two bands
(surveys 1-11) the energy inversion at 100 µm FOV would appear to occur in the range of
1–5 nA. Looking at bands 3 and 5 (surveys 12–15 and 20–23, respectively), both at 2 nA, it
might be said that the energy inversion occurs near 2 nA due to the low red/blue separation.
However, in the 18th and 19th, (surveys 82–89) the inversion would appear to occur in the
range of 0.1–1 nA. Thus the energy inversion does not occur at a well-defined current density
and so the energy separation method is unreliable at too low of an FOV or primary beam
current.
It is not clear why the peak energy separation between +/-Z domains inverts at too
low of FOV or primary beam current. One theory is that the inversion is caused by the
pyroelectric properties of the sample, similar to the reversal seen with SEM [33, 37]. Under
electron beam irradiation the sample’s temperature would be changed by Joule heating,
causing a charge to develop on the surface. The sample heating would be dependent on
beam focus, chamber vacuum, and other factors that would make the inversion somewhat
unpredictable and dependent on more than just FOV and primary beam current (or current
density). This theory could be tested with the additional capability of sample temperature
control in the AES main chamber, though this was not available at the time of this research.
The Auger Nanoprobe does have an electrical input for the main chamber however, so it
would be possible to add an appropriate heating element to the sample stage. After careful
calibration of the heating element’s temperature in ultra-high vacuum (UHV), it would be
possible to measure the peak separation vs change in temperature.
While investigating lower FOV and the energy inversion instability, a clear separation
139
Figure 7.3: Top: AES peak amplitudes across all surveys. Note FOV and beam current are
denoted at the top of each colored band. Bottom: Deviation from the mean peak energy
across all surveys.
140
in amplitude in the raw spectra became apparent, including when the peak energies were
not well separated. Shown in figure 7.3 is a plot of the peak amplitudes for the same data
from the previous plot, figure 7.2. The top plot shows the raw amplitude data that is heavily
dependent on the primary beam current, but very stable for a given current, especially when
compared with the peak energy data tendency to drift. The bottom plot shows the deviation
from the mean peak amplitude, calculated in the same manner as the deviation from the
mean peak energy. Calculating the deviation from the mean is still important since the
peak amplitude depends heavily on primary beam current. Notably, the separation in peak
amplitude between +/-Z domains is cleaner than the separation in peak energy, as there are
fewer crossovers of blue/red colored data. On this sample, the peak amplitude separation
is mostly lost at 10 µm FOV for all currents (0.1, 0.3, and 10nA), surveys 46–81, with the
exception of 5 nA. This gave initial doubt to this method’s ability to achieve sub-micron
resolution. However, at 10 µm FOV and 5 nA (surveys 42–45), the red/blue separation
is not only large but clean. Another initial observation that seemed important was that
the magnitude of red/blue separation was not as correlated to the color gradient as the
peak energy separation often is. For example, dark red and dark blue data points (surveys
areas separated by the largest distance) are often more separated than lighter red/blues,
but the separation is not roughly linear as indicated by the gap between the color groups
in almost all surveys 1–34. Thus, it appeared that the peak amplitude might allow domain
differentiation with peak amplitude separation between domains independent of the survey
area’s distance from the domain wall; this was not the case with energy separation in chapter
4. Furthermore, if peak amplitude separation between domains was dependent only on beam
current and not domain wall proximity, then nano-scale differentiation would be possible if
we could overcome the charging and image drift at low FOV. On this data set, there was
peak amplitude separation between domains at 100 µm FOV for beam currents of 1 nA and
greater, and at 10 µm FOV, 5 nA.
141
Energy inversion at low FOV or low primary beam current limits the ability of the
peak energy separation method at low FOVs, although it works consistently at large enough
FOVs on the order of 100 µm. The peak energy inversion does not occur at a stable current
density but appears to be dependent on sample history. Looking at the reliability of peak
amplitude separation when there was peak energy inversion solidified our focus on the peak
amplitude separation method.
7.3 AES on probe-tip poled chips
Initial AES testing began on MgLN chips that were bulk poled in house and HF
etched. In chapter 5, AES tests used un-etched PPLN chips from G&H fabricated with
photolithography. As Auger testing progressed, PPLN chips from the same wafer as the
chips used in the chapter 5 experiments began to have issues. Namely, the photoresist that
was applied as a protective coating for dicing was now difficult to remove. This was likely due
to the wafer being exposed to months of ambient office and window light. Furthermore, after
some attempts to dissolve the resist, it appeared to sometimes selectively remove from one
polarization domain more readily than the other, leaving a differential residue between the
+/-Z domains. This is shown in figure 7.4 where the residue is only partially removed after
a 24 hour acetone bath, and the remaining photoresist appears attached primarily to the
+Z domains. This caused concern that the Auger amplitude differentiation demonstrated in
the previous chapter was in part or in whole caused by some fabrication step. This could be
leftover residue, developer or metal etch affecting the crystal, embedded metal, embedded
chemicals, or something else. In other words it seemed possible that the Auger amplitude
differentiation method was detecting a proxy for domains leftover from a fabrication step,
rather than a differential signal inherent in the +/-Z domains. It must be noted that the
chips used in chapters 4, 5, and 6 did not have any visible residue in SEM images.
The possibility of the differential Auger signal originating from the chemical processing
142
Figure 7.4: Selectively removed photoresist on PPLN. Resist appears to remain primarily
on the +Z domains.
143
(or HF etching in the case of chapter 4) led us to look for samples with no fabrication involved
in their poling. Some of the MgLN chips poled with a probe-tip described in section 3.5 had
no fabrication or etching performed on them other than bringing the probe-tip into contact
with the crystal to apply the poling voltage. Therefore, AES was performed on these chips
in order to examine if the Auger O-KLL peak amplitudes from +/-Z domains were still
differentiable. First, SEM images of the chip are given in figures 7.5 and 7.6, where the red
boxes on (a) highlights the PTP domain used for AES. Note that the AES survey areas in
figures 7.5 and 7.6 switch the positions of the +/-Z domains in the scan area; ‘East’ refers
to when the -Z poled domain is on the left side of the image, and ‘West’ refers to when it
is on the right. In figure 7.5 (b), (c), and (d) are different FOVs, with only (c) and (d)
corresponding to the surveys taken (1–12) and have the AES survey areas shown across the
center of the image. In figure 7.6 (b), (c), and (d) are the three FOVs corresponding to
surveys 13–38.
The compiled AES amplitude results are given in figure 7.7, with the raw peak
amplitudes on the left and the deviation from the mean peak amplitude on the right. Note
that red and blue data from this chip with no processing are still very well separated, thus
no differential chemical residues or other residual differential effects on the +/- domains
from fabrication could be the source of the signal separation. Surveys 1–12 and 13–38 have
opposite red/blue separation because the survey area switches from ‘East’ to ‘West’ side of
a poled domain, as explained above and shown in figures 7.5 and 7.6. The AES peak energy
data is given in figure 7.8, with raw peak energies on the left and the deviation from the
mean peak energy on the right. It is again worth pointing out that the energy separation
between domains fails at too low of FOV as discussed in section 7.2. Furthermore, the
energy separation at too low of FOV and/or beam current inverts, as can be seen when
comparing the first and last columns (100 µm FOV, 5nA and 20 µm FOV, 1nA); these
surveys are performed on opposite sides of the domains (‘East’ and ‘West’) so that one of
144
Figure 7.5: (a) 1000 µm FOV of PTP MgLN with red box indicating poled spot used for
AES surveys on ‘East’ side of poled region, surveys 1–12. (b) 200 µm FOV showing entire
poled region. (c) 100 µm FOV and AES survey areas. (d) 50 µm FOV and AES survey areas.
Note that chip is tilted 75◦, such that the top of the image is farthest from the electron gun.
145
Figure 7.6: (a) 1000 µm FOV of PTP MgLN with red box indicating poled spot used for
AES surveys on ‘West’ side of poled region, surveys 13–38. (b) 100 µm FOV showing entire
poled spot. (c) 50 µm FOV and AES survey areas. (d) 20 µm FOV and AES survey areas.
Note that chip is tilted 75◦, such that the top of the image is farthest from the electron gun.
146
(a) (b)
Figure 7.7: Peak amplitudes across all surveys on PTP MgLN. Note FOV and beam current
are denoted at the top of each colored band. (a) Peak amplitude across all surveys. (b)
Deviation from the mean peak amplitude across all surveys. Surveys 1–12 are on the ‘East’
side of poling (blue data in -Z domain) and surveys 13-38 are on the ‘West’ side (red data
in -Z domain).
(a) (b)
Figure 7.8: Peak energies across all surveys on PTP MgLN. Note FOV and beam current are
denoted at the top of each colored band. (a) Peak energy across all surveys. (b) Deviation
from the mean peak energy across all surveys. Surveys 1–12 are on the ‘East’ side of poling
(blue data in -Z domain) and surveys 13-38 are on the ‘West’ side (red data in -Z domain).
147
them should have red data with higher energy, but both have blue data points with higher
energy indicating an inversion with respect to the expected outcome. Area01 in surveys 1-7
is a notable outlier with a peak energy generally below the red data. The damage from the
tip is in the center of the poled region (see figures 3.13 and 3.14), and thus survey Area01
is sometimes located on top of the debris from the tip damage, which impacts the AES
signal in both amplitude and energy, though the amplitude data is still separated. This is
not unexpected, but does point towards the need for clean LN surfaces in order to obtain
meaningful AES data.
Overall, the initial concern that the peak amplitude differentiation method was
detecting a selective residue or other processing effect from a fabrication or poling process was
proven not to be the case. Since PTP chips have nothing applied to the surface outside of the
actual tip contact region, they would only have a differential +/-Z domain AES amplitude
if the signal was integral to the domain itself. AES tests confirmed that +/-Z domains were
still separable on PTP chips, so we conclude no processing step was responsible for the peak
amplitude separation.
7.4 AES on MgLN with thin gold film
Charge mitigation is a pressing challenge for AES on LN and MgLN. Due to LN being
an insulator, charge cannot always dissipate quick enough under the incident electron beam.
This leads to a buildup of charge that can deflect both the electron beam and secondary
electrons. In the experiments presented in the previous sections and chapters 4, 5, and 6,
charge mitigation was accomplished by tilting the samples 75◦ relative to the axis of the
electron beam in the UHV Auger main chamber. Tilting the samples mitigates charging
by lowering the amount of embedded charge, or more specifically “enhancing secondary
electron emission by matching electron penetration to secondary-electron escape distance”
[151]. While the tilting method is successful and has mitigated charging to a degree that
148
allowed nano-scale AES resolution, the loss of resolution in the direction of the tilt is a
drawback.
A thin film of gold was applied to MgLN bulk-poled samples in order to test its ability
to mitigate charging. Ideally, the gold film would still allow AES on the substrate material,
but also mitigate charging enough to allow the sample’s tilt to be lowered or set to 0◦. Auger
electrons have a relatively low kinetic energy and thus their escape depth is small, only about
1–5 nm [148]. AES is therefore a surface sensitive characterization method leading to the
concern that applying the gold film, even a very thin film, would inhibit Auger electrons
from the underlying MgLN material. Another lesser concern is that applying a thin gold
film to MgLN and subsequently removing it in a wet gold etchant may have some effect on
the crystal, as well as adding steps to the AES characterization method. Still, if the gold
thin film had negligible effects on the crystal’s properties and allowed AES at lower tilts it
would be a valuable technique.
Using ICAL’s sputter coater, gold was sputtered for 5 s at 35 mA. The gold film
thickness was measured on a silicon companion chip using AFM and found to be ∼6.5 nm
thick, a rate of 1.3 nm/s. Previous measurement of the sputter rate from nearly 3 years
prior gave a sputter rate of 0.3 nm/s. This difference in measured sputter rate could be
due to several factors; the chips’ distance from the sputter source may have changed, the
gold source may have been replaced, which increased the rate, the sputter current may have
fluctuated over time, the two measurements were made with different AFMs so one of the
measurements could be incorrect. Regardless of the reason, it must be emphasized that the
gold thin film may be less than 6.5 nm thick.
AES was performed on the MgLN chip with the thin film gold, and the results are given
in figure 7.9. There are several interesting results. In figure 7.9 (b) the energy separation
method works for FOVs of 100 µm or larger; blue (-Z domain) and red (+Z domain) data
points are separated with -Z domains having higher peak energies. The method mostly
149
(a) (b)
(c) (d)
Figure 7.9: AES peak energies and amplitudes across all surveys on MgLN with a thin gold
film. Note FOV and beam current are denoted at the top of each colored band, and the
last four bands are surveys in an un-poled region such that all 10 survey areas are in -Z
domain. (a) Peak energy. (b) Deviation from the mean peak energy. (a) Peak amplitude.
(b) Deviation from the mean peak amplitude. Blue data points are in HF etched -Z domain,
red data points are in +Z domain. 100 µm FOV SEM image showing survey location is
shown in figure 7.10
fails at 10 µm FOV and below, consistent with previous results. In the un-poled region
(surveys taken with all 10 survey areas located in bulk -Z domain) at 500 µm FOV, there
is peak separation between red and blue data in surveys 27–31. While this result is initially
150
troubling, it has been noted before and addressed in chapter 4, figure 4.9. This is attributed
to charging due to the new FOV in the un-poled region being too close in proximity to
the region of surveys taken just beforehand and where significant charge has accumulated.
Therefore one side of the FOV, and the surveys on that side, are closer to this potential
source. The decreasing energy separation between surveys 27 and 31 support this idea as
charge dissipates. The next surveys at 100 µm FOV in the unpoled region have no obvious
red/blue energy separation. In the example in chapter 4, figure 4.9, separation in the un-
poled region is initially opposite (red peaks higher) the previous data where blue peaks are
higher, indicating the FOV was moved to an un-poled region on the opposite side of the
charged area.
In the deviation from the mean peak amplitude data in figure 7.9 (d), red/blue
amplitude separation is initially inverted; survey areas in the +Z domain (red) have a larger
amplitude compared to the -Z domain (blue). This is explained by the surface roughness,
and can be seen in figure 7.10. The -Z domain on the left-hand side of the image is noticeably
rough from the HF etch, and can be expected to lower the Auger amplitude from that region.
This is an important finding that explains why the amplitude separation method is successful
on the smooth, un-etched, PPLN used in chapter 5, but went un-noticed on the HF etched
MgLN used in chapter 4. Furthermore, the amplitude separation method works as expected
at 2 µm FOV, surveys 22–26, except for areas 9 and 2, which likely overlap the domain wall.
This can be understood by noting that the scale of roughness of the +/-Z domains at 2 µm
FOV are comparable, as seen in figure 7.11, due to the specific selected portion of the -Z
domain surveyed. The two surveys areas straddling the domain wall, areas 9 and 2 (cyan
and orange data points), have a higher amplitude than the rest of the data. Their proximity
to the domain wall also makes their peak energy shift unpredictable in figure 7.9 (b). Figure
7.11 shows the six images of the 2 µm FOV to demonstrate the lack of image drift. The top
left image was taken before the first survey at this FOV, and the next five taken immediately
151
Figure 7.10: 100 µm FOV SEM image and AES survey areas corresponding to surveys 11–16
in figure 7.9. HF etched -Z domain on left-hand side of image is visibly rougher than +Z
domain on right.
after each subsequent AES survey. The FOVs are nearly identically, demonstrating the thin
gold film mitigates charging enough to stabilize the 2 µm FOV.
Overall, applying a thin gold film to MgLN samples for AES seems to be a promising
152
Figure 7.11: Six 2 µm FOV SEM images of gold-coated MgLN showing minimal image drift.
Images were taken before and in-between AES surveys 22–26 in figure 7.9, and have the
etched -Z domain on left-hand side.
technique for charge mitigation. AES was able to ‘see-through’ the thin gold film and
still perform the peak energy separation method at large enough FOVs. When the surface
roughness is comparable in both +/-Z domains, the amplitude separation method also
appears to work, even at 2 µm FOV. Image drift at 2 µm FOV is minimal, especially when
compared with un-coated samples. Unfortunately, this topic was not explored further and
thus may be a fruitful area of research. Questions to be investigated include the following:
How thick of a gold film optimizes both charge mitigation and separation of +/-Z domain
signals? Do smooth, un-etched, PPLN samples show peak amplitude separation without
inversion at all FOVs? How small of a FOV is the image stable enough to allow the peak
153
amplitude separation on a sample without the large step etched domain wall? Does the gold
film allow use of lower sample tilt? Is there any remaining diffused-in gold after removal
with wet etching or does the wet etchant affect the crystal properties at all? Determining
to what degree a thin gold film facilitates better AES characterization of LN most certainly
requires further research.
7.5 Summary
The sections in this chapter covered some interesting findings of AES on LN and MgLN.
The first section shows that the peak energy separation inverts at low enough FOV or primary
beam current, limiting the ability of the peak energy separation method at low FOV. This
was an important step in realizing the possibility and need for the amplitude separation
method. The next section of AES on samples poled with a probe-tip, and therefore absent
any other processing, was important in order to rule out the amplitude separation originating
from a processing step. Lastly, preliminary results of AES on MgLN with a thin gold film
does not appear to destroy the Auger O-KLL signal and seems promising for aiding charge
mitigation. Each of these sections explored fundamental aspects of AES on LN that produced
interesting outcomes and each would be greatly strengthened by further investigation.
154
CHAPTER EIGHT
CONCLUSION
The use of a quasi-phase matching (QPM) technique was theorized six decades ago
[173, 174], but the technical difficulties of fabrication required for QPM took two more
decades to be realized [129, 175], and even longer for the development of photolithography’s
use in controlled periodic domain poling [1]. While advances have been numerous over the
past three decades, and QPM applications have been wide-ranging, there are still practical
limitations to achieving first-order phase matching for interactions of desired wavelengths,
particularly for backwards propagating configurations. Controlling the poled domains on
the nano-scale across a large area is a significant challenge that often involves fabrication
of nano-scale grating electrodes or masks. Despite these complications in fabricating the
QPM grating, the characterization of such structures offers its own unique challenges.
Ferroelectric domain detection and characterization techniques can rely on numerous
properties including the ferroelectric-coupled piezoelectric effect, domain surface potential
differences, domains’ differing susceptibility to selective etching, among other things. Several
common characterization methods suffer from domain contrast of an unknown origin. Others
suffer from limited spatial resolution, limited range of FOV, or destructiveness to the
crystal surface. Certainly, complementing the characterization capabilities offered by current
techniques would allow better understanding of the ferroelectric domains and with this
knowledge, nano-scale fabrication efforts can be bolstered and improved.
In this work, several steps are taken towards the understanding of small-scale poling
processes and characterizing small-scale ferroelectric domains. In chapter 2, several processes
were developed for fabrication of poling electrodes. Micron-scale period grating electrodes
for contact poling were optimized. Electrodes and fabrication processes on MgLN crystal
155
surface were established as well. And nano-scale period grating electrodes were fabricated
after creation and calibration of two separate EBL methods, wet-etch and liftoff.
In chapter 3, poling experiments were carried out in three distinct configurations.
Contact poling, using periodic electrodes at several size scales including micron-scale, was
investigated followed by poling with electrodes fabricated on the MgLN surface. Both
methods led to uncertainty in the poling uniformity. Finally, contact probe-tip poling (PTP)
was applied to MgLN. PTP allowed numerous poling experiments to be performed on single
crystals, and produced useful samples with uniform domains for characterization efforts.
In chapter 4, characterization efforts focused on the development and implementation
of Auger electron spectroscopy (AES) to differentiate ferroelectric domains in MgLN are
described. Initially, ferroelectric domains were characterized by the Auger O-KLL peak
energy in MgLN that was bulk-poled in-house and HF-etched before AES. AES capabilities
were expanded in chapter 5 by using the Auger O-KLL peak amplitude to discriminate
between domains, demonstrating nano-scale resolution on un-etched PPLN. Based on the
principles of the peak amplitude separation method, in chapter 6, AES was used to perform
full image mapping of +/-Z domains in PPLN. Image processing determined an apparent
domain wall width of 3.79 pixels, or ∼67 nm spatial resolution, which could be improved by
different map parameters or refined image processing. Finally, three additional preliminary
investigations with AES are detailed in chapter 7, including the odd phenomenon of inversion
of the peak energy separation method at small FOV, confirming the Auger peak amplitude
separation method is due to inherent properties in the ferroelectric domains and not a
vestige of processing, and an initial investigation of using a gold thin film as an AES charge
mitigation method.
The electrode fabrication and poling efforts offer insight into small-scale ferroelectric
domains in LN and MgLN. With regards to characterization, AES can be added as a new
technique to the field of LN and MgLN ferroelectric domain characterization. The peak
156
energy separation method appears to have a limited resolution on the order of slightly
less than 10 µm, but offers an interesting insight into the +/-Z domains’ surface, not seen
before. The peak amplitude separation method is robust at a range of size scales and
demonstrated nano-scale spatial resolution. Based on the principle of the peak amplitude
separation method, full image mapping of LN’s +/-Z domains was performed at FOVs
ranging from 7.5 µm to 200 µm. AES offers consistent determination of +/-Z domains
relative to each other. The technique is non-destructive, non-contact, unambiguous, with
nano-scale resolution demonstrated down to 67 nm. This new method complements the other
existing characterization methods’ limitations, and offers a combination of advantages not
offered by any of those methods. The totality of this research offers insight into fabricating,
poling, and characterizing small-scale ferroelectric domains in LN and MgLN and will be an
important step towards developing PPLN devices with smaller domain sizes.
The content of this research stands as a good foundation from which several interesting
investigations could be launched. Beginning with the content of section 3.5, we found the
application of Si oil to severely inhibit over-poling beyond the area of probe-tip contact.
This knowledge combined with the fabrication of electrodes on the MgLN surface in section
2.1.2 seems promising for the creation of poled domains with high fidelity with respect to
the electrode structure and reproducibility. Furthermore, if either of the nano-scale EBL
electrode fabrication methods could be adapted to an MgLN substrate (or other NLO
crystal), this may lead to a novel nano-scale poling method. One challenge that would
need to be addressed is the spatial area limitations of our current EBL tool, about 150 µm
× 150 µm. There are other EBL instruments with the ability to stitch patterns together
to write larger areas such that their use which would overcome our limitations. Combining
the electrode fabrication methods detailed here, with the knowledge of Si oil’s inhibition of
probe-tip over-poling, offers many paths for further research and development.
The bulk of this dissertation’s novel findings are related to AES characterization of
157
MgLN and PPLN, and there are many related topics remaining for exploration. First,
it would be good to determine a true resolution limit for the AES mapping method,
both by improving the image processing to include a better edge preserving adaptive
smoothing routine, and by optimizing mapping parameters such as pixel number, beam
current, and beam voltage. Next, the origin of the O-KLL peak amplitude separation
between domains is not well understood at this time and its investigation is of primary
importance. Another topic, perhaps most easily accomplished, would be testing the AES
peak amplitude separation method on other ferroelectric crystals of interest, especially those
that are periodically poled and probe-tip poled. KTP, LT, and stoichiometric LN, are all
obvious candidates. First, one could determine whether AES is significantly more or less
challenging due to the different material conductivities, followed by which elemental peaks
are most reliable for AES ferroelectric domain differentiation in these different crystals.
One could then verify if the peak amplitude method works as expected on all crystal species.
Compiling the elemental peaks used for domain separation using the peak amplitude method
in these crystals, and quantifying their separation may also give insight into the origin of
the peak separation.
Another area of research related more to the fundamental physical properties of LN
could be testing AES on samples etched in the Auger Nanoprobe’s UHV main chamber.
Using the Nanoprobe’s argon ion gun to etch the surface would reveal the bulk material,
which would remain unexposed to atmosphere while in the chamber. Investigating this
virgin material could answer questions about the origin of the peak amplitude and peak
energy separations, and offer insight about the different surface terminations of the +/-Z
domains, a prominent question about LN in both theoretical and experimental fields [134].
Initial tests of argon ion etching of MgLN offered mixed results and sometimes included the
disappearance of the peak energy separation between +/-Z domains, however time did not
allow further investigation.
158
Lastly, the limitation of AES characterization methods due to the insulating properties
of LN could be addressed. As discussed in section 7.4, a thin gold film did not completely
inhibit the Auger O-KLL signal. Further investigation of the thickness of the gold film’s
relation to the amplitude of the O-KLL signal would help inform the optimal thickness.
With an optimized conducting gold layer, one could determine what sample tilt below 75◦
still allows AES characterization. The usefulness of these experiments partly hinges on
the crystal surface being unaffected by the processes of gold application and its removal
in wet-etchant. This could be resolved with a combination of SEM imaging, and surface
characterization using AES, Energy-dispersive X-ray spectroscopy (EDS), and/or X-ray
photoelectron spectroscopy (XPS).
There are many fruitful opportunities for further research related to nano-scale
fabrication, poling, and characterization in LN and related materials. I note that in just
over a year from 2020 to 2021, with no expectations to do so, the resolution of our AES
characterization increased by two orders of magnitude. While there are physical limitations
to such things so that similar rates of development are not always possible, they are often
available to the determined and inquisitive researcher.
159
REFERENCES CITED
[1] David S. Hum and Martin M. Fejer. Quasi-phasematching. Comptes Rendus Physique,
8(2):180–198, March 2007.
[2] Dimitri Geskus, Jonas Jakutis-Neto, Helen M. Pask, and Niklaus U. Wetter. Ten deep
blue to cyan emission lines from an intracavity frequency converted Raman laser. In
Nonlinear Frequency Generation and Conversion: Materials, Devices, and Applications
XIV, volume 9347, pages 41–48. SPIE, February 2015.
[3] Soumya Sarang and Martin Richardson. Power Scaling of CW Crystalline OPOs
and Raman Lasers. Photonics, 8(12):565, December 2021. Number: 12 Publisher:
Multidisciplinary Digital Publishing Institute.
[4] Ole Bjarlin Jensen, Anders Kragh Hansen, André Müller, Bernd Sumpf, Angelika
Unterhuber, Wolfgang Drexler, Paul Michael Petersen, and Peter E. Andersen. Power
Scaling of Nonlinear Frequency Converted Tapered Diode Lasers for Biophotonics.
IEEE Journal of Selected Topics in Quantum Electronics, 20(2):307–321, March 2014.
Conference Name: IEEE Journal of Selected Topics in Quantum Electronics.
[5] Kevin S. Repasky, Jennifer Williams, John L. Carlsten, Elizabeth J. Noonan, and
Gregg Switzer. Tunable external-cavity diode laser based on integrated waveguide
structures. Optical Engineering, 42(8):2229–2234, August 2003. Publisher: SPIE.
[6] Galan Moody, Volker J. Sorger, Daniel J. Blumenthal, Paul W. Juodawlkis, William
Loh, Cheryl Sorace-Agaskar, Alex E. Jones, Krishna C. Balram, Jonathan C. F.
Matthews, Anthony Laing, Marcelo Davanco, Lin Chang, John E. Bowers, Niels Quack,
Christophe Galland, Igor Aharonovich, Martin A. Wolff, Carsten Schuck, Neil Sinclair,
Marko Lončar, Tin Komljenovic, David Weld, Shayan Mookherjea, Sonia Buckley,
Marina Radulaski, Stephan Reitzenstein, Benjamin Pingault, Bartholomeus Machielse,
Debsuvra Mukhopadhyay, Alexey Akimov, Aleksei Zheltikov, Girish S. Agarwal,
Kartik Srinivasan, Juanjuan Lu, Hong X. Tang, Wentao Jiang, Timothy P. McKenna,
Amir H. Safavi-Naeini, Stephan Steinhauer, Ali W. Elshaari, Val Zwiller, Paul S.
Davids, Nicholas Martinez, Michael Gehl, John Chiaverini, Karan K. Mehta, Jacquiline
Romero, Navin B. Lingaraju, Andrew M. Weiner, Daniel Peace, Robert Cernansky,
Mirko Lobino, Eleni Diamanti, Luis Trigo Vidarte, and Ryan M. Camacho. Roadmap
on Integrated Quantum Photonics. Journal of Physics: Photonics, 4(1):012501,
January 2022. arXiv: 2102.03323.
[7] M.M. Fejer, G.A. Magel, D.H. Jundt, and R.L. Byer. Quasi-phase-matched second
harmonic generation: tuning and tolerances. IEEE Journal of Quantum Electronics,
28(11):2631–2654, November 1992. Conference Name: IEEE Journal of Quantum
Electronics.
160
[8] M. J. Angell, R. M. Emerson, J. L. Hoyt, J. F. Gibbons, L. A. Eyres, M. L. Bortz, and
M. M. Fejer. Growth of alternating <100>/<111>-oriented II-VI regions for quasi-
phase-matched nonlinear optical devices on GaAs substrates. Applied Physics Letters,
64(23):3107–3109, June 1994. Publisher: American Institute of Physics.
[9] Xinhua Gu, Roman Y. Korotkov, Yujie J. Ding, Jin U. Kang, and Jacob B. Khurgin.
Backward second-harmonic generation in periodically poled lithium niobate. JOSA B,
15(5):1561–1566, May 1998. Publisher: Optica Publishing Group.
[10] Y.J. Ding and Wei Shi. From backward THz difference-frequency generation to
parametric oscillation. IEEE Journal of Selected Topics in Quantum Electronics,
12(3):352–359, May 2006. Conference Name: IEEE Journal of Selected Topics in
Quantum Electronics.
[11] Xinhua Gu, Roman Y. Korotkov, Yujie J. Ding, Jin U. Kang, and Jacob B.
Khurgin. Observation of backward sum-frequency generation in periodically-poled
lithium niobate. Optics Communications, 155(4):323–326, October 1998.
[12] Woo Yang, Brian Rodriguez, Alexei Gruverman, and Robert Nemanich. Polarization-
Dependent Electron Affinity of LiNbO3 Surfaces. Applied Physics Letters, 85:2316–
2318, September 2004.
[13] A. Gruverman and S. V. Kalinin. Piezoresponse force microscopy and recent advances
in nanoscale studies of ferroelectrics. Journal of Materials Science, 41(1):107–116,
January 2006.
[14] Dragan Damjanovic. Ferroelectric, dielectric and piezoelectric properties of ferro-
electric thin films and ceramics. Reports on Progress in Physics, 61(9):1267–1324,
September 1998. Publisher: IOP Publishing.
[15] Elisabeth Soergel. Piezoresponse force microscopy (PFM). Journal of Physics D:
Applied Physics, 44(46):464003, November 2011.
[16] Sergei V. Kalinin and Dawn A. Bonnell. Imaging mechanism of piezoresponse force
microscopy of ferroelectric surfaces. Physical Review B, 65(12):125408, March 2002.
[17] Tobias Jungk, Ákos Hoffmann, and Elisabeth Soergel. Quantitative analysis of
ferroelectric domain imaging with piezoresponse force microscopy. Applied Physics
Letters, 89(16):163507, October 2006. Publisher: American Institute of Physics.
[18] D. A. Bonnell, S. V. Kalinin, A. L. Kholkin, and A. Gruverman. Piezoresponse Force
Microscopy: A Window into Electromechanical Behavior at the Nanoscale. MRS
Bulletin, 34(9):648–657, September 2009. Publisher: Cambridge University Press.
[19] Sabine M. Neumayer, Sahar Saremi, Lane W. Martin, Liam Collins, Alexander Tselev,
Stephen Jesse, Sergei V. Kalinin, and Nina Balke. Piezoresponse amplitude and
161
phase quantified for electromechanical characterization. Journal of Applied Physics,
128(17):171105, November 2020. Publisher: American Institute of Physics.
[20] Nina Balke, Igor Bdikin, Sergei V. Kalinin, and Andrei L. Kholkin. Electromechanical
Imaging and Spectroscopy of Ferroelectric and Piezoelectric Materials: State of the Art
and Prospects for the Future. Journal of the American Ceramic Society, 92(8):1629–
1647, 2009. eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1111/j.1551-
2916.2009.03240.x.
[21] Florian Johann, Yongjun J. Ying, Tobias Jungk, Ákos Hoffmann, Collin L. Sones,
Robert W. Eason, Sakellaris Mailis, and Elisabeth Soergel. Depth resolution of
piezoresponse force microscopy. Applied Physics Letters, 94(17):172904, April 2009.
Publisher: American Institute of Physics.
[22] Lili Tian, Aravind Vasudevarao, Anna N. Morozovska, Eugene A. Eliseev, Sergei V.
Kalinin, and Venkatraman Gopalan. Nanoscale polarization profile across a 180°
ferroelectric domain wall extracted by quantitative piezoelectric force microscopy.
Journal of Applied Physics, 104(7):074110, October 2008. Publisher: American
Institute of Physics.
[23] A. Gruverman, O. Auciello, R. Ramesh, and H. Tokumoto. Scanning force microscopy
of domain structure in ferroelectric thin films: imaging and control. Nanotechnology,
8(3A):A38–A43, September 1997. Publisher: IOP Publishing.
[24] Sergei V. Kalinin, Andrei Rar, and Stephen Jesse. A decade of piezoresponse
force microscopy: progress, challenges, and opportunities. IEEE Transactions on
Ultrasonics, Ferroelectrics, and Frequency Control, 53(12):2226–2252, December 2006.
Conference Name: IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency
Control.
[25] T. Jungk, Á Hoffmann, and E. Soergel. Impact of the tip radius on the lateral resolution
in piezoresponse force microscopy. New Journal of Physics, 10(1):013019, January
2008. Publisher: IOP Publishing.
[26] Sungwon Kim, Venkatraman Gopalan, and Alexei Gruverman. Coercive fields in
ferroelectrics: A case study in lithium niobate and lithium tantalate. Applied Physics
Letters, 80(15):2740–2742, April 2002. Publisher: American Institute of Physics.
[27] Chun-Lin Jia, Shao-Bo Mi, Knut Urban, Ionela Vrejoiu, Marin Alexe, and Dietrich
Hesse. Atomic-scale study of electric dipoles near charged and uncharged domain walls
in ferroelectric films. Nature Materials, 7(1):57–61, January 2008. Bandiera abtest: a
Cg type: Nature Research Journals Number: 1 Primary atype: Research Publisher:
Nature Publishing Group.
[28] J. Guyonnet, H. Béa, F. Guy, S. Gariglio, S. Fusil, K. Bouzehouane, J.-M. Triscone, and
P. Paruch. Shear effects in lateral piezoresponse force microscopy at 180° ferroelectric
162
domain walls. Applied Physics Letters, 95(13):132902, September 2009. Publisher:
American Institute of Physics.
[29] J. Wittborn, C. Canalias, K. V. Rao, R. Clemens, H. Karlsson, and F. Laurell.
Nanoscale imaging of domains and domain walls in periodically poled ferroelectrics
using atomic force microscopy. Applied Physics Letters, 80(9):1622–1624, March 2002.
Publisher: American Institute of Physics.
[30] R. Le Bihan. Study of ferroelectric and ferroelastic domain structures by scanning
electron microscopy. Ferroelectrics, 97(1):19–46, September 1989.
[31] Vitaly Aristov, L. Kokhanchik, K.-P Meyer, and H. Blumtritt. Scanning electron
microscopic investigations of peculiarities of the BaTiO3 ferroelectric domain contrast.
Phys. Status Solidi (a), 78:229–236, July 1983.
[32] L. S. Kokhanchik. The use of surface charging in the SEM for lithium niobate domain
structure investigation. Micron, 40(1):41–45, January 2009.
[33] S. Zhu and W. Cao. Imaging of 180° Ferroelectric Domains in LiTaO3 by
Means of Scanning Electron Microscopy. physica status solidi (a), 173(2):495–502,
1999. eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1002/%28SICI%291521-
396X%28199906%29173%3A2%3C495%3A%3AAID-PSSA495%3E3.0.CO%3B2-Z.
[34] D. V. Roshchupkin and M. Brunel. Scanning Electron Microscopy Observation of the
Voltage Contrast Image of the Ferroelectric Domain Structure in the LiNbO3 Crystal.
Scanning Microscopy, 7(2), April 1993.
[35] Yangbo Zhou, Daniel S. Fox, Pierce Maguire, Robert O’Connell, Robert Masters,
Cornelia Rodenburg, Hanchun Wu, Maurizio Dapor, Ying Chen, and Hongzhou Zhang.
Quantitative secondary electron imaging for work function extraction at atomic level
and layer identification of graphene. Scientific Reports, 6(1):1–8, February 2016.
[36] K. A. Hunnestad, E. D. Roede, A. T. J. van Helvoort, and D. Meier. Characterization
of ferroelectric domain walls by scanning electron microscopy. Journal of Applied
Physics, 128(19):191102, November 2020. Publisher: American Institute of Physics.
[37] Gil Rosenman, Alex Skliar, I Lareah, Nahum Angert, M Tseitlin, and Michael Roth.
Observation of ferroelectric domain structures by secondary-electron microscopy in as-
grown KTiOPO4 crystals. Physical Review. B, Condensed Matter, 54(9):6222–6226,
September 1996.
[38] Shining Zhu and Wenwu Cao. Direct Observation of Ferroelectric Domains in
LiTaO3 Using Environmental Scanning Electron Microscopy. Physical Review Letters,
79(13):2558–2561, September 1997. Publisher: American Physical Society.
163
[39] Wenwu Cao and Shining Zhu. Observation of ferroelectric domains in LiTaO3.
Ferroelectrics, 226(1):27–35, April 1999. Publisher: Taylor & Francis eprint:
https://doi.org/10.1080/00150199908230287.
[40] K. Tanaka and T. Suhara. Non-destructive observation of ferroelectric domain inverted
structures in MgO:LiNbO3 by scanning electron microscope. Electronics Letters,
51(12):923–925, 2015. Conference Name: Electronics Letters.
[41] A. Reichmann, A. Zankel, H. Reingruber, P. Pölt, and K. Reichmann. Direct
observation of ferroelectric domain formation by environmental scanning electron
microscopy. Journal of the European Ceramic Society, 31(15):2939–2942, December
2011.
[42] P. S. Zelenovskiy, M. D. Fontana, V. Y. Shur, P. Bourson, and D. K. Kuznetsov.
Raman visualization of micro- and nanoscale domain structures in lithium niobate.
Applied Physics A, 99(4):741–744, June 2010.
[43] R. Hammoum, M.D. Fontana, P. Bourson, and V.Y. Shur. Characterization of PPLN-
microstructures by means of Raman spectroscopy. Applied Physics A, 91(1):65–67,
April 2008.
[44] Marc D. Fontana and Patrice Bourson. Microstructure and defects probed by Raman
spectroscopy in lithium niobate crystals and devices. Applied Physics Reviews,
2(4):040602, October 2015.
[45] G. Nataf, Mael Guennou, Alexander Haußmann, Nick Barrett, and Jens Kreisel.
Evolution of defect signatures at ferroelectric domain walls in Mg-doped LiNbO 3.
Phys. Status Solidi (RRL), 10(3):222–226, October 2015.
[46] Pavel Capek, Greg Stone, Volkmar Dierolf, Chad Althouse, and Venkat Gopolan.
Raman studies of ferroelectric domain walls in lithium tantalate and niobate. Phys.
Status Solidi (c), 4(3):830–833, 2007.
[47] Greg Stone, Brian Knorr, Venkatraman Gopalan, and Volkmar Dierolf. Frequency
shift of Raman modes due to an applied electric field and domain inversion in LiNbO3.
Physical Review B, 84:134303, October 2011.
[48] Bang Liu, Yuanlin Zheng, Xiaohui Zhao, Haigang Liu, and Xianfeng Chen. Probe of
symmetry reduction at domain walls by nonlinear Cherenkov measurement. Optics
Express, 24(26):29459–29464, December 2016.
[49] T. Kämpfe, P. Reichenbach, A. Haußmann, T. Woike, E. Soergel, and L. M. Eng.
Real-time three-dimensional profiling of ferroelectric domain walls. Applied Physics
Letters, 107(15):152905, October 2015.
164
[50] Yan Sheng, Vito Roppo, Mingliang Ren, Ksawery Kalinowski, Crina Cojocaru, Jose
Trull, Zhiyuan Li, Kaloian Koynov, and Wieslaw Krolikowski. Multi-directional
Čerenkov second harmonic generation in two-dimensional nonlinear photonic crystal.
Optics Express, 20(4):3948–3953, February 2012.
[51] Yan Sheng, Andreas Best, Hans-Jürgen Butt, Wieslaw Krolikowski, Ady Arie, and
Kaloian Koynov. Three-dimensional ferroelectric domain visualization by Čerenkov-
type second harmonic generation. Optics Express, 18(16):16539–16545, August 2010.
[52] Thomas Kämpfe, Philipp Reichenbach, Mathias Schröder, A Haußmann, Lukas Eng,
T. Woike, and E. Soergel. Optical three-dimensional profiling of charged domain
walls in ferroelectrics by Cherenkov second-harmonic generation. Physical Review B,
89:035314, January 2014.
[53] Xuewei Deng, Huaijin Ren, Huiying Lao, and Xianfeng Chen. Research on Cherenkov
second-harmonic generation in periodically poled lithium niobate by femtosecond
pulses. JOSA B, 27(7):1475–1480, July 2010.
[54] D. K. Kuznetsov, D. S. Chezganov, E. A. Mingaliev, M. S. Kosobokov, and V. Ya.
Shur. Visualization of nanodomain structures in lithium niobate and lithium tantalate
crystals by scanning electron microscopy. Ferroelectrics, 503(1):60–67, October 2016.
Publisher: Taylor & Francis eprint: https://doi.org/10.1080/00150193.2016.1217703.
[55] Vladimir Ya. Shur. Kinetics of ferroelectric domains: Application of general approach
to LiNbO3 and LiTaO3. In Sidney B. Lang and Helen L. W. Chan, editors, Frontiers
of Ferroelectricity: A Special Issue of the Journal of Materials Science, pages 199–210.
Springer, 2007.
[56] Vladimir Shur, Evgenii Rumyantsev, Robert Batchko, Gregory Miller, Martin Fejer,
and Robert Byer. Physical basis of the domain engineering in the bulk ferroelectrics.
Ferroelectrics, 221(1):157–167, January 1999. Publisher: Taylor & Francis eprint:
https://doi.org/10.1080/00150199908016450.
[57] Robert G. Batchko, Gregory D. Miller, Vladimir Ya Shur, Evgenii L. Rumyantsev,
Martin M. Fejer, and Robert L. Byer. Domain patterning in lithium niobate using
spontaneous backswitching. In Laser Material Crystal Growth and Nonlinear Materials
and Devices, volume 3610, pages 36–43. SPIE, May 1999.
[58] Robert G. Batchko, Vladimir Y. Shur, Martin M. Fejer, and Robert L. Byer.
Backswitch poling in lithium niobate for high-fidelity domain patterning and efficient
blue light generation. Applied Physics Letters, 75(12):1673–1675, September 1999.
[59] C. E. Valdivia, C. L. Sones, J. G. Scott, S. Mailis, R. W. Eason, D. A. Scrymgeour,
V. Gopalan, T. Jungk, E. Soergel, and I. Clark. Nanoscale surface domain formation on
the +z face of lithium niobate by pulsed ultraviolet laser illumination. Applied Physics
Letters, 86(2):022906, January 2005. Publisher: American Institute of Physics.
165
[60] A. Busacca, M. Cherchi, S.R. Sanseverino, A.C. Cino, A. Parisi, G. Assanto,
M. Cichoki, F. Caccavale, D. Calleyo, and A. Morbiato. Surface periodic poling in
lithium niobate and lithium tantalate. In Proceedings of 2005 IEEE/LEOS Workshop
on Fibres and Optical Passive Components, 2005., pages 126–130, June 2005.
[61] C. L. Sones, A. C. Muir, Y. J. Ying, S. Mailis, R. W. Eason, T. Jungk, Á. Hoffmann,
and E. Soergel. Precision nanoscale domain engineering of lithium niobate via UV
laser induced inhibition of poling. Applied Physics Letters, 92(7):072905, February
2008. Publisher: American Institute of Physics.
[62] Prashant R. Potnis, Nien-Ti Tsou, and John E. Huber. A Review of Domain Modelling
and Domain Imaging Techniques in Ferroelectric Crystals. Materials, 4(2):417–447,
February 2011.
[63] S. Mailis, C. L Sones, J. G. Scott, and R. W. Eason. UV laser-induced ordered surface
nanostructures in congruent lithium niobate single crystals. Applied Surface Science,
247(1):497–503, July 2005.
[64] V Bermúdez, F Caccavale, C Sada, F Segato, and E Diéguez. Etching effect on periodic
domain structures of lithium niobate crystals. Journal of Crystal Growth, 191(3):589–
593, July 1998.
[65] Charlie Y. J. Ying, Collin L. Sones, Anna C. Peacock, Florian Johann, Elisabeth
Soergel, Robert W. Eason, Mikhail N. Zervas, and Sakellaris Mailis. Ultra-smooth
lithium niobate photonic micro-structures by surface tension reshaping. Optics Express,
18(11):11508–11513, May 2010.
[66] Collin L. Sones, Sakellaris Mailis, William S. Brocklesby, Robert W. Eason, and
John R. Owen. Differential etch rates in z-cut LiNbO3 for variable HF/HNO3
concentrations. Journal of Materials Chemistry, 12(2):295–298, January 2002.
[67] Collin Lawrence Sones. Domain engineering techniques and devices in lithium niobate.
phd, University of Southampton, August 2003.
[68] William L. Holstein. Etching study of ferroelectric microdomains in LiNbO3 and
MgO:LiNbO3. Journal of Crystal Growth, 171(3):477–484, February 1997.
[69] C. Canalias, V. Pasiskevicius, R. Clemens, and F. Laurell. Submicron periodically
poled flux-grown KTiOPO4. Applied Physics Letters, 82(24):4233–4235, June 2003.
Publisher: American Institute of Physics.
[70] V. Ya. Shur, E. V. Pelegova, A. R. Akhmatkhanov, and I. S. Baturin. Periodically poled
crystals of KTP family: a review. Ferroelectrics, 496(1):49–69, May 2016. Publisher:
Taylor & Francis eprint: https://doi.org/10.1080/00150193.2016.1157437.
166
[71] F. Laurell, M. G. Roelofs, W. Bindloss, H. Hsiung, A. Suna, and J. D. Bierlein.
Detection of ferroelectric domain reversal in KTiOPO4 waveguides. Journal of Applied
Physics, 71(10):4664–4670, May 1992. Publisher: American Institute of Physics.
[72] C. Canalias, V. Pasiskevicius, M. Fokine, and F. Laurell. Backward quasi-phase-
matched second-harmonic generation in submicrometer periodically poled flux-grown
KTiOPO4. Applied Physics Letters, 86(18):181105, April 2005.
[73] V. Pasiskevicius, C. Canalias, G. Strömqvist, and F. Laurell. Mirrorless OPO: first
steps towards unlocking the potential of counter-propagating three-wave interactions.
In Nonlinear Frequency Generation and Conversion: Materials, Devices, and Applica-
tions VII, volume 6875, pages 50–62. SPIE, February 2008.
[74] Andrius Zukauskas, Gustav Strömqvist, Valdas Pasiskevicius, Fredrik Laurell, Michael
Fokine, and Carlota Canalias. Fabrication of submicrometer quasi-phase-matched
devices in KTP and RKTP [Invited]. Optical Materials Express, 1(7):1319–1325,
November 2011.
[75] C. Canalias, J. Hirohashi, V. Pasiskevicius, and F. Laurell. Polarization-switching
characteristics of flux-grown KTiOPO4 and RbTiOPO4 at room temperature. Journal
of Applied Physics, 97(12):124105, June 2005. Publisher: American Institute of
Physics.
[76] C. Canalias, A. Zukauskas, C. Liljestrand, V. Pasiskevicius, and F. Laurell. Sub-
wavelength domain engineering in KTP isomorphs: QPM devices with counter-
propagating photons. In 2013 Conference on Lasers and Electro-Optics Pacific Rim
(CLEOPR), pages 1–2, June 2013.
[77] Stacie Elizabeth Smith. Investigation of nanoscale etching and poling of lithium
niobate. Thesis, Montana State University - Bozeman, College of Engineering, 2014.
Accepted: 2014-11-10T19:28:55Z Pages: 1-124.
[78] John A. Thornton. The microstructure of sputter-deposited coatings. Journal
of Vacuum Science & Technology A, 4(6):3059–3065, November 1986. Publisher:
American Vacuum Society.
[79] C. R. M. Grovenor, H. T. G. Hentzell, and D. A. Smith. The development of grain
structure during growth of metallic films. Acta Metallurgica, 32(5):773–781, May 1984.
[80] Y. G. Yang, R. A. Johnson, and H. N. G. Wadley. A Monte Carlo simulation of the
physical vapor deposition of nickel. Acta Materialia, 45(4):1455–1468, April 1997.
[81] Gregory David Miller. Periodically Poled Lithium Niobate: Modeling, Fabrication, and
Nonlinear-Optical Performance. PhD Thesis, Stanford, July 1998.
167
[82] J. J Cras, C. A Rowe-Taitt, D. A Nivens, and F. S Ligler. Comparison of chemical
cleaning methods of glass in preparation for silanization. Biosensors and Bioelectronics,
14(8):683–688, December 1999.
[83] UC Irvine INRF. Cleaning procedures for glass substrates.
[84] Janet Tate. Substrates - Cleaning Procedures.
[85] Ankur Verma. Substrate Cleaning.
[86] David N. Nikogosyan. Nonlinear Optical Crystals: A Complete Survey. Springer-
Verlag, 2005.
[87] H. Ishizuki and T. Taira. Study on the field-poling dynamics in Mg-doped LiNbO3
and LiTaO3. In Nonlinear Optics: Materials, Fundamentals and Applications (2007),
paper WE35, page WE35. Optica Publishing Group, July 2007.
[88] Venkatraman Gopalan, Terence E. Mitchell, Y. Furukawa, and K. Kitamura. The role
of nonstoichiometry in 180° domain switching of LiNbO3 crystals. Applied Physics
Letters, 72(16):1981–1983, April 1998. Publisher: American Institute of Physics.
[89] H. Karlsson, M. Olson, G. Arvidsson, F. Laurell, U. Bäder, A. Borsutzky, R. Wal-
lenstein, S. Wickström, and M. Gustafsson. Nanosecond optical parametric oscillator
based on large-aperture periodically poled RbTiOAsO4. Optics Letters, 24(5):330–332,
March 1999. Publisher: Optica Publishing Group.
[90] Koichiro Nakamura, Takaaki Hatanaka, and Hiromasa Ito. High Output Energy Quasi-
Phase-Matched Optical Parametric Oscillator Using Diffusion-Bonded Periodically
Poled and Single Domain LiNbO3. Japanese Journal of Applied Physics, 40(4A):L337,
April 2001. Publisher: IOP Publishing.
[91] Fulei Wang, Dehui Sun, Qilu Liu, Yukun Song, Feng Zhang, Weijia Zhou, Yuanhua
Sang, Dongzhou Wang, and Hong Liu. Growth of large size near-stoichiometric
lithium niobate single crystals with low coercive field for manufacturing high quality
periodically poled lithium niobate. Optical Materials, 125:112058, March 2022.
[92] Longyue Liang, Fulei Wang, Yuanhua Sang, Fei Zhou, Xiuping Xie, Dehui Sun,
Mingyang Zheng, and Hong Liu. Facile approach for the periodic poling of MgO-doped
lithium niobate with liquid electrodes. CrystEngComm, 21(6):941–947, February 2019.
Publisher: The Royal Society of Chemistry.
[93] Hideki Ishizuki, Takunori Taira, Sunao Kurimura, Jung Hoon Ro, and Myoungsik Cha.
Periodic Poling in 3-mm-Thick MgO:LiNbO3 Crystals. Japanese Journal of Applied
Physics, 42(2A):L108, February 2003. Publisher: IOP Publishing.
168
[94] Kenji Kitamura, Yasunori Furukawa, Kazuo Niwa, Venkatraman Gopalan, and
Terence E. Mitchell. Crystal growth and low coercive field 180° domain switching
characteristics of stoichiometric LiTaO3. Applied Physics Letters, 73(21):3073–3075,
November 1998. Publisher: American Institute of Physics.
[95] Hideki Ishizuki, Ichiro Shoji, and Takunori Taira. Periodical poling characteristics
of congruent MgO:LiNbO3 crystals at elevated temperature. Applied Physics Letters,
82(23):4062–4064, June 2003.
[96] Atsuko Kuroda, Sunao Kurimura, and Yoshiaki Uesu. Domain inversion in ferroelectric
MgO:LiNbO3 by applying electric fields. Applied Physics Letters, 69(11):1565–1567,
September 1996.
[97] Venkatraman Gopalan and Mool C. Gupta. Observation of internal field in LiTaO3
single crystals: Its origin and time-temperature dependence. Applied Physics Letters,
68(7):888–890, February 1996. Publisher: American Institute of Physics.
[98] Cynthia Baron, Hsing Cheng, and Mool C. Gupta. Domain inversion in LiTaO3 and
LiNbO3 by electric field application on chemically patterned crystals. Applied Physics
Letters, 68(4):481–483, January 1996. Publisher: American Institute of Physics.
[99] Takaaki Hatanaka, Koichiro Nakamura, Tetsuo Taniuchi, Hiromasa Ito, Yasunori
Furukawa, and Kenji Kitamura. Quasi-phase-matched optical parametric oscillation
with periodically poled stoichiometric LiTaO3. Optics Letters, 25(9):651–653, May
2000.
[100] J.-P. Meyn, C. Laue, R. Knappe, R. Wallenstein, and M.M. Fejer. Fabrication of
periodically poled lithium tantalate for UV generation with diode lasers:. Applied
Physics B, 73(2):111–114, August 2001.
[101] M. Peltz, U. Bäder, A. Borsutzky, R. Wallenstein, J. Hellström, H. Karlsson,
V. Pasiskevicius, and F. Laurell. Optical parametric oscillators for high pulse energy
and high average power operation based on large aperture periodically poled KTP and
RTA. Applied Physics B, 73(7):663–670, November 2001.
[102] J. Hellström, V. Pasiskevicius, F. Laurell, and H. Karlsson. Efficient nanosecond optical
parametric oscillators based on periodically poled KTP emitting in the 1.8–2.5-µm
spectral region. Optics Letters, 24(17):1233–1235, September 1999. Publisher: Optica
Publishing Group.
[103] H. Karlsson and F. Laurell. Electric field poling of flux grown KTiOPO4. Applied
Physics Letters, 71(24):3474–3476, December 1997. Publisher: American Institute of
Physics.
[104] W. P. Risk and S. D. Lau. Periodic electric field poling of KTiOPO4 using chemical
patterning. Applied Physics Letters, 69(26):3999–4001, December 1996. Publisher:
American Institute of Physics.
169
[105] Qibiao Chen, Qibiao Chen, and W. P. Risk. Periodically Poled KTiOPO4: A “New”
Nonlinear Material for Bulk Interactions. In Advanced Solid State Lasers (1995), paper
SD4, page SD4. Optica Publishing Group, January 1995.
[106] Charlotte Liljestrand, Fredrik Laurell, and Carlota Canalias. Periodic poling of Rb-
doped KTiOPO4 by coercive field engineering. Optics Express, 24(13):14682–14689,
June 2016. Publisher: Optica Publishing Group.
[107] R. S. Weis and T. K. Gaylord. Lithium niobate: Summary of physical properties and
crystal structure. Applied Physics A, 37(4):191–203, August 1985.
[108] Gooch and Housego. Lithium Niobate Wafers.
[109] Toijam Sunder Meetei, Shanmugam Boomadevi, and Krishnamoorthy Pandiyan.
Design and development of low-cost room temperature electric field poling system
for the fabrication of quasi-phase matching devices. Pramana, 91(3):37, August 2018.
[110] Andrew Phillips. Creating a dynamic periodic poling system for x-cut lithium niobate
chip-scale integrated optics. Thesis, The Ohio State University, May 2016. Accepted:
2016-05-13T22:37:49Z.
[111] J L. Nightingale, W J. Silva, G E. Reade, A Rybicki, W J. Kozlovsky, and R L. Byer.
Fifty Percent Conversion Efficiency Second Harmonic Generation In Magnesium Oxide
Doped Lithium Niobate. In Larry G. DeShazer, editor, Proc. SPIE 0681, Laser and
Nonlinear Optical Materials, page 20, San Diego, March 1987.
[112] W.J. Kozlovsky, C.D. Nabors, and R.L. Byer. Efficient second harmonic generation of a
diode-laser-pumped CW Nd:YAG laser using monolithic MgO:LiNbO/sub 3/ external
resonant cavities. IEEE Journal of Quantum Electronics, 24(6):913–919, June 1988.
Conference Name: IEEE Journal of Quantum Electronics.
[113] Tomoya Sugita, Kiminori Mizuuchi, Yasuo Kitaoka, and Kazuhisa Yamamoto. 31%-
efficient blue second-harmonic generation in a periodically poled MgO:LiNbO 3
waveguide by frequency doubling of an AlGaAs laser diode. Optics Letters, 24(22):1590,
November 1999.
[114] Shinichiro Sonoda, Isao Tsuruma, and Masami Hatori. Second harmonic generation
in electric poled X-cut MgO-doped LiNbO3 waveguides. Applied Physics Letters,
70(23):3078–3080, June 1997. Publisher: American Institute of Physics.
[115] Koichiro Nakamura, Jonathan Kurz, Krishnan Parameswaran, and M. M. Fejer.
Periodic poling of magnesium-oxide-doped lithium niobate. Journal of Applied Physics,
91(7):4528–4534, March 2002.
[116] Joong Hyun Kim and Choon Sup Yoon. Domain switching characteristics and
fabrication of periodically poled potassium niobate for second-harmonic generation.
170
Applied Physics Letters, 81(18):3332–3334, October 2002. Publisher: American
Institute of Physics.
[117] V. Ya. Shur, E. L. Rumyantsev, R. G. Batchko, G. D. Miller, M. M. Fejer, and R. L.
Byer. Domain kinetics in the formation of a periodic domain structure in lithium
niobate. Physics of the Solid State, 41(10):1681–1687, October 1999.
[118] V. Ya. Shur, E. L. Rumyantsev, E. V. Nikolaeva, E. I. Shishkin, D. V. Fursov, R. G.
Batchko, L. A. Eyres, M. M. Fejer, and R. L. Byer. Nanoscale backswitched domain
patterning in lithium niobate. Applied Physics Letters, 76(2):143–145, January 2000.
Publisher: American Institute of Physics.
[119] V. Ya. Shur, E. L. Rumyantsev, E. V. Nikolaeva, E. I. Shishkin, D. V. Fursov, R. G.
Batchko, L. A. Eyres, M. M. Fejer, R. L. Byer, and J. Sindel. Formation of self-
organized nanodomain patterns during spontaneous backswitching in lithium niobate.
Ferroelectrics, 253(1):105–114, February 2001. Publisher: Taylor & Francis eprint:
https://doi.org/10.1080/00150190108008448.
[120] L.-H. Peng, Y.-C. Fang, and Y.-C. Lin. Polarization switching of lithium niobate with
giant internal field. Applied Physics Letters, 74(14):2070–2072, April 1999.
[121] Mark Bashkansky, Marcel W. Pruessner, Igor Vurgaftman, Mijin Kim, and J. Reintjes.
Higher-order spontaneous parametric down-conversion with back-propagating idler
using a submicron poled KTP waveguide. In Quantum Information and Computation
IX, volume 9873, page 987303. International Society for Optics and Photonics, May
2016.
[122] J. Hellström, V. Pasiskevicius, H. Karlsson, and F. Laurell. High-power optical
parametric oscillation in large-aperture periodically poled KTiOPO4. Opt. Lett., OL,
25(3):174–176, February 2000.
[123] H. Karlsson, F. Laurell, and L. K. Cheng. Periodic poling of RbTiOPO4 for quasi-
phase matched blue light generation. Applied Physics Letters, 74(11):1519–1521, March
1999.
[124] Maxim Neradovskiy, Elizaveta Neradovskaia, Dmitry Chezganov, Evgeny Vlasov,
Vladimir Ya Shur, Hervé Tronche, Florent Doutre, Getachew Ayenew, Pascal Baldi,
Marc De Micheli, and Carlos Montes. Second harmonic generation in periodically poled
lithium niobate waveguides with stitching errors. JOSA B, 35(2):331–336, February
2018.
[125] G. A. Magel, M. M. Fejer, and R. L. Byer. Quasi-phase-matched second-harmonic
generation of blue light in periodically poled LiNbO3. Applied Physics Letters,
56(2):108–110, January 1990.
171
[126] V. Bermúdez, J. Capmany, J. Garcıa Solé, and E. Diéguez. Growth and second
harmonic generation characterization of Er3+ doped bulk periodically poled LiNbO3.
Applied Physics Letters, 73(5):593–595, July 1998.
[127] L. E. Myers, R. C. Eckardt, M. M. Fejer, R. L. Byer, W. R. Bosenberg, and J. W.
Pierce. Quasi-phase-matched optical parametric oscillators in bulk periodically poled
LiNbO3. JOSA B, 12(11):2102–2116, November 1995.
[128] M. Yamada, N. Nada, M. Saitoh, and K. Watanabe. First-order quasi-phase matched
LiNbO3 waveguide periodically poled by applying an external field for efficient blue
second-harmonic generation. Applied Physics Letters, 62(5):435–436, February 1993.
[129] Duan Feng, Nai-Ben Ming, Jing-Fen Hong, Yong-Shun Yang, Jin-Song Zhu, Zhen Yang,
and Ye-Ning Wang. Enhancement of second-harmonic generation in LiNbO3 crystals
with periodic laminar ferroelectric domains. Applied Physics Letters, 37(7):607–609,
October 1980.
[130] D. H. Jundt, G. A. Magel, M. M. Fejer, and R. L. Byer. Periodically poled LiNbO3
for high-efficiency second-harmonic generation. Applied Physics Letters, 59(21):2657–
2659, November 1991.
[131] J. Webjorn, F. Laurell, and G. Arvidsson. Fabrication of periodically domain-inverted
channel waveguides in lithium niobate for second harmonic generation. Journal of
Lightwave Technology, 7(10):1597–1600, October 1989.
[132] J. Webjorn, F. Laurell, and G. Arvidsson. Blue light generated by frequency doubling
of laser diode light in a lithium niobate channel waveguide. IEEE Photonics Technology
Letters, 1(10):316–318, October 1989.
[133] Pranay Kumar Sen, N. Sisodia, R. Choubey, Sid Kar, and K. Bartwal. Effect of MgO
Doping on Coercive Field in LiNbO3 Crystals. Journal of Nonlinear Optical Physics
& Materials, 17:175–183, June 2008.
[134] Simone Sanna andWolf Gero Schmidt. LiNbO3surfaces from a microscopic perspective.
Journal of Physics: Condensed Matter, 29(41):413001, September 2017.
[135] Yang Yun and Eric I. Altman. Using Ferroelectric Poling to Change Adsorption
on Oxide Surfaces. Journal of the American Chemical Society, 129(50):15684–15689,
December 2007.
[136] C. Canalias, S. Wang, V. Pasiskevicius, and F. Laurell. Nucleation and growth of
periodic domains during electric field poling in flux-grown KTiOPO4 observed by
atomic force microscopy. Applied Physics Letters, 88(3):032905, January 2006.
[137] VSh Ivanov, VA Ganshin, and YuN Korkishko. Analysis of ion exchanged Me : LiNbO3
and Cu : LiTaO3 waveguides by AES, SAM, EPR and optical methods. Vacuum,
43(4):317–324, April 1992.
172
[138] K. Nassau, H. J. Levinstein, and G. M. Loiacono. The domain structure and etching
of ferroelectric lithium niobate. Applied Physics Letters, 6(11):228–229, June 1965.
[139] F. Laurell, J. Webjorn, G. Arvidsson, and J. Holmberg. Wet etching of proton-
exchanged lithium niobate-a novel processing technique. Journal of Lightwave
Technology, 10(11):1606–1609, November 1992.
[140] V. Ya. Shur, A. I. Lobov, A. G. Shur, S. Kurimura, Y. Nomura, K. Terabe, X. Y.
Liu, and K. Kitamura. Rearrangement of ferroelectric domain structure induced by
chemical etching. Applied Physics Letters, 87(2):022905, July 2005.
[141] V. Bermúdez, A. Gil, L. Arizmendi, J. Colchero, A. M. Baró, and E. Diéguez.
Techniques of observation and characterization of the domain structure in periodically
poled lithium niobate. Journal of Materials Research, 15(12):2814–2821, December
2000.
[142] Sergei V. Kalinin and Dawn A. Bonnell. Local potential and polarization screening on
ferroelectric surfaces. Physical Review B, 63(12):125411, March 2001.
[143] David Tuschel. Three-Dimensional Raman Imaging of Ion-Exchanged Waveguides.
Spectroscopy, 29:14, December 2014.
[144] Greg Stone and Volkmar Dierolf. Influence of ferroelectric domain walls on the Raman
scattering process in lithium tantalate and niobate. Optics Letters, 37(6):1032–1034,
March 2012. Publisher: Optical Society of America.
[145] J.G. Scott, S. Mailis, C.L. Sones, and R.W. Eason. A Raman study of single-
crystal congruent lithium niobate following electric-field repoling. Applied Physics
A, 79(3):691–696, August 2004.
[146] P. S. Bullen, H. C. Huang, H. Yang, J. I. Dadap, I. Kymissis, and R. M. Osgood.
Microscopy and microRaman study of periodically poled domains in deeply thinned
lithium niobate wafers. Optical Materials, 57:243–248, July 2016.
[147] Mousa Ayoub, Hannes Futterlieb, Jörg Imbrock, and Cornelia Denz. 3D Imaging of
Ferroelectric Kinetics during Electrically Driven Switching. Advanced Materials, 29(5),
February 2017.
[148] H.L. Marcus. Auger Electron Spectroscopy. In Encyclopedia of Materials: Science and
Technology, pages 393–398. Elsevier, 2001.
[149] J. Garra, J. M. Vohs, and D. A. Bonnell. The effect of ferroelectric polarization on the
interaction of water and methanol with the surface of LiNbO3(0001). Surface Science,
603(8):1106–1114, April 2009.
173
[150] Mosha Zhao, D. Bonnell, and John Vohs. Influence of ferroelectric polarization on the
energetics of the reaction of 2-fluoroethanol on BaTiO 3. Surface Science, 603:284–290,
January 2009.
[151] D. R. Baer, A. S. Lea, J. D. Geller, J. S. Hammond, L. Kover, C. J. Powell, M. P. Seah,
M. Suzuki, J. F. Watts, and J. Wolstenholme. Approaches to analyzing insulators with
Auger electron spectroscopy: Update and overview. Journal of Electron Spectroscopy
and Related Phenomena, 176(1):80–94, January 2010.
[152] PHI 710 l ULVAC-PHI, Inc. https://www.ulvac-phi.com/en/products/aes/phi-710/.
[153] Jong Lee. Kinetic Energy Distribution of Auger Electrons of MgO, CaO, SrO and BaO
Induced by Auger Neutralization of He+, Ne+, Ar+ and Xe+ Ions. New Physics: Sae
Mulli, 67:1168–1173, October 2017.
[154] S Habicht, Robert Nemanich, and Alexei Gruverman. Physical adsorption on
ferroelectric surfaces: Photoinduced and thermal effects. Nanotechnology, 19:495303,
December 2008.
[155] Ayed Al Sayem, Yubo Wang, Juanjuan Lu, Xianwen Liu, Alexander W. Bruch, and
Hong X. Tang. Efficient and tunable blue light generation using lithium niobate
nonlinear photonics. arXiv:2107.01171 [physics], July 2021. arXiv: 2107.01171.
[156] David A. Scrymgeour and Venkatraman Gopalan. Nanoscale piezoelectric response
across a single antiparallel ferroelectric domain wall. Physical Review B, 72(2):024103,
July 2005. Publisher: American Physical Society.
[157] M. Labardi, V. Likodimos, and M. Allegrini. Force-microscopy contrast mechanisms
in ferroelectric domain imaging. Physical Review B, 61(21):14390–14398, June 2000.
Publisher: American Physical Society.
[158] Seungbum Hong, Hyunjung Shin, Jungwon Woo, and Kwangsoo No. Effect of
cantilever–sample interaction on piezoelectric force microscopy. Applied Physics
Letters, 80(8):1453–1455, February 2002. Publisher: American Institute of Physics.
[159] C. Harnagea, M. Alexe, D. Hesse, and A. Pignolet. Contact resonances in voltage-
modulated force microscopy. Applied Physics Letters, 83(2):338–340, July 2003.
Publisher: American Institute of Physics.
[160] A. Agronin, M. Molotskii, Y. Rosenwaks, E. Strassburg, A. Boag, S. Mutchnik, and
G. Rosenman. Nanoscale piezoelectric coefficient measurements in ionic conducting
ferroelectrics. Journal of Applied Physics, 97(8):084312, April 2005. Publisher:
American Institute of Physics.
[161] Huimin Qiao, Owoong Kwon, and Yunseok Kim. Electrostatic effect on off-field
ferroelectric hysteresis loop in piezoresponse force microscopy. Applied Physics Letters,
116(17):172901, April 2020. Publisher: American Institute of Physics.
174
[162] Alexei Gruverman, Marin Alexe, and Dennis Meier. Piezoresponse force microscopy
and nanoferroic phenomena. Nature Communications, 10(1):1661, April 2019. Number:
1 Publisher: Nature Publishing Group.
[163] Liam Collins, Yongtao Liu, Olga S. Ovchinnikova, and Roger Proksch. Quantitative
Electromechanical Atomic Force Microscopy. ACS Nano, 13(7):8055–8066, July 2019.
Publisher: American Chemical Society.
[164] Qibin Zeng, Qicheng Huang, Hongli Wang, Caiwen Li, Zhen Fan, Deyang
Chen, Yuan Cheng, and Kaiyang Zeng. Breaking the Fundamental Limita-
tions of Nanoscale Ferroelectric Characterization: Non-Contact Heterodyne Elec-
trostrain Force Microscopy. Small Methods, 5(11):2100639, 2021. eprint:
https://onlinelibrary.wiley.com/doi/pdf/10.1002/smtd.202100639.
[165] M. Rüsing, S. Neufeld, J. Brockmeier, C. Eigner, P. Mackwitz, K. Spychala,
C. Silberhorn, W. G. Schmidt, G. Berth, A. Zrenner, and S. Sanna. Imaging of 180°
ferroelectric domain walls in uniaxial ferroelectrics by confocal Raman spectroscopy:
Unraveling the contrast mechanism. Physical Review Materials, 2(10):103801, October
2018. Publisher: American Physical Society.
[166] Luis Mateos, Luisa E. Bausá, and Mariola O. Ramı́rez. Micro-spectroscopic character-
ization of ferroelectric domain structures in Yb3+:LiNbO3 prepared by electron beam
writing. Optical Materials Express, 4(5):1077–1087, May 2014. Publisher: Optica
Publishing Group.
[167] Torrey McLoughlin, Wm Randall Babbitt, Phillip A. Himmer, and Wataru Naka-
gawa. Auger electron spectroscopy for surface ferroelectric domain differentiation in
selectively poled MgO:LiNbO3. Optical Materials Express, 10(10):2379–2393, October
2020. Publisher: Optical Society of America.
[168] K. Tsutsumi, Toshiaki Suzuki, and Y. Nagasawa. Effective Methods to Prevent
Charging in Auger Electron Spectroscopy. JEOL News, Vol. 36E(No.1):66–70, 2001.
[169] Y. Yun, M. Li, D. Liao, L. Kampschulte, and E. I. Altman. Geometric and electronic
structure of positively and negatively poled LiNbO3 (0001) surfaces. Surface Science,
601(19):4636–4647, October 2007.
[170] Torrey McLoughlin, Wm Randall Babbitt, and Wataru Nakagawa. Nano-scale
ferroelectric domain differentiation in periodically poled lithium niobate with auger
electron spectroscopy. Optics Continuum, 1(4):649–659, April 2022. Publisher: Optica
Publishing Group.
[171] H. Bishop. The Role of the Background in Auger Electron Spectroscopy. Scanning
Electron Microscopy, 1982(1), January 1982.
175
[172] Simone Sanna and Wolf Gero Schmidt. Lithium niobate X-cut, Y-cut, and Z-cut
surfaces from ab initio theory. Physical Review B, 81(21):214116, June 2010. Publisher:
American Physical Society.
[173] J. A. Armstrong, N. Bloembergen, J. Ducuing, and P. S. Pershan. Interactions between
Light Waves in a Nonlinear Dielectric. Physical Review, 127(6):1918–1939, September
1962. Publisher: American Physical Society.
[174] P. A. Franken and J. F. Ward. Optical Harmonics and Nonlinear Phenomena. Reviews
of Modern Physics, 35(1):23–39, January 1963. Publisher: American Physical Society.
[175] A. Feisst and P. Koidl. Current induced periodic ferroelectric domain structures in
LiNbO3 applied for efficient nonlinear optical frequency mixing. Applied Physics
Letters, 47(11):1125–1127, December 1985. Publisher: American Institute of Physics.