FABRICATION AND CHARACTERIZATION
OF QUASI-2D CORRELATED SYSTEMS
By
Tung-Wu Hsieh

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Physics—Doctor of Philosophy
2017

ABSTRACT
FABRICATION AND CHARACTERIZATION
OF QUASI-2D CORRELATED SYSTEMS
By
Tung-Wu Hsieh
The combination of nanostructure and correlated systems has been an interesting
field in the frontier of condensed matter physics. Exploring and understanding the
physical properties of correlated electron materials with reduced dimensionalities is the
focus of this study. For this purpose, we fabricated thin flakes of calcium ruthenate
(Ca3Ru2O7) and iron telluride (Fe1+yTe) with various thickness via mechanical exfoliation
and characterized their magneto-electronic transport properties. The two materials are
chosen for their rich phase transitions in structure, conductivity and magnetic orders.
The electronic transport properties of exfoliated Ca3Ru2O7 flakes have been
investigated as a function of thickness. We find that flakes thinner than 10 nm are more
insulating and show drastic changes in magnetoresistance and metal-insulator transitions.
A possible scenario might be enhanced RuO6 octahedral distortion due to the loss of
apical oxygen atoms on the cleaved surfaces. This may play a more significant role in
thinner flakes as the surface-to-volume ratio increases. The study establishes a control of
ruthenate properties by dimensionality and connects the fields of two-dimensional
materials and correlated electron systems.
In Fe1+yTe exfoliated-flake devices we find the first-order structural phase
transition in Fe1+yTe bulk becomes broadened in flakes with an intermediate thickness
accompanying a superconducting-like dip below 10K. Further reducing the thickness of
the flakes wipes out all features, and the flakes behaves like a highly-disordered two-

dimensional system exhibiting charge-localized states with largely enhanced MR ratio.
The behavior can be well described by Mott’s variable range hopping regime, with a
localization length ~10 nm which may originate from the excess iron atoms.
One of the future goals of this project is the carrier doping on exfoliated thin
flakes (gating effect) using either solid high-Îş dielectric or ionic liquid. Another goal will
be exploring more perovskite type materials since our unique exfoliation method may be
the only choice to mechanically cleave the layered structured crystals with strong
interlayer binding force to date. Characterization of the structure of exfoliated flakes will
be also an important work to further verify the origin of exotic behaviors of the flakes.

ACKNOWLEDGEMENTS

I would like to thank my advisor Prof. Xianglin Ke for his support after I switched
from the former group in the middle of my graduate school and giving me the
challenging task of exfoliating calcium ruthenate. And I really appreciate the help from
our staff scientist Dr. Reza Loloee, the project cannot be finished without the precious
experience and inspiring suggestions from Reza. On the measurement part Dr. Tao Zou is
the one helps me the most, a lot of my data is taken efficiently with his written LabVIEW
program. I have to thank Michael Gottschalk for spending enormous time in finding
ruthenate flakes and scanning AFM. When learning the e-beam lithography, I am grateful
that Dr. Eric Gingrich and Dr. William Martinez provided me a lot of useful tips helping
me quickly getting my first ruthenate flake device done. For the ionic liquid problems
and static charge problems, I appreciate Prof. Zhixian Zhou and Dr. Hsun-Jen Chuang in
the Wayne State University because they generously spent their time and used their
instrument working with me to find the solutions. I will remember that Prof. Scott Pratt
and Prof. Brage Golding encouraged me in continuing my PhD. There are still many
people generously helped me in my graduate student life I could not include all in this
acknowledgement but will be remembered forever. Finally I would like to thank my wife
Dr. Yung-Hsiu Tang for supporting me and bringing me Claire, the cutest baby in the
world. At the end, I have to say the work is so challenging that it cannot be done without
the help from everyone.

iii

TABLE OF CONTENTS

LIST OF TABLES……………………………………………..…………………………………..v
LIST OF FIGURES…………………………………………………………………………….…vi

Chapter 1 Introduction and Motivation .............................................................................. 1
1.1 Motivation:.................................................................................................................... 1
1.2 Formation of Mott-insulators........................................................................................ 5
1.3 Calcium ruthenate Ca3Ru2O7 ........................................................................................ 7
1.4 Antiferromagnetic ground state and transitions in Fe1+yTe .......................................... 8
1.5 Why electrical transport measurements? ...................................................................... 9
Chapter 2 Characterization methods................................................................................. 11
2.1 Visibility of exfoliated flakes on substrates:............................................................... 11
2.2 Atomic force microscopy............................................................................................ 17
2.3 Scanning electron microscopy .................................................................................... 21
2.4 Raman spectroscopy ................................................................................................... 23
Chapter 3 Preparation of devices ...................................................................................... 27
3.1 Preparation of substrates ............................................................................................. 29
3.2 Exfoliation (Gorilla glue method)............................................................................... 31
3.3 E-beam lithography..................................................................................................... 34
3.4 Wiring and ESD protection......................................................................................... 39
3.5 Contact resistance and 4-probe measurement............................................................. 45
3.6 Ionic liquid gating....................................................................................................... 50
Chapter 4 Calcium ruthenate thin flake study .................................................................. 62
4.1 Bulk material properties of Ca3Ru2O7 ........................................................................ 62
4.2 MIT in thin flakes ....................................................................................................... 66
4.3 Magnetoresistance and antiferromagnetic transition .................................................. 71
4.4 Lateral dimension effect?............................................................................................ 78
4.5 MIT in Ca2RuO4 flakes............................................................................................... 81
4.6 Summary..................................................................................................................... 84
Chapter 5 Properties of Fe1+yTe thin flakes ...................................................................... 86
5.1 Introduction:................................................................................................................ 86
5.2 Fe1+yTe nanoflake devices .......................................................................................... 90
5.3 Superconductor-like intermediate thin flakes ............................................................. 95
5.4 VRH model in ultra-thin flakes .................................................................................. 98
5.5 Exfoliated FeSe0.5Te0.5 flakes and ionic liquid gating .............................................. 103
5.6 Summary................................................................................................................... 107
Chapter 6 Summary and Future Vision .......................................................................... 109
REFERENCES……………………………………………………………………..…...………113

iv

LIST OF TABLES

Table 1.1: Comparison of measurement techniques available to our group ……………...…..….10
Table 2.1: Fitting parameters of refractive index of Ca3Ru2O7…………………………...……...13

v

LIST OF FIGURES

Figure 1.1: Metal-insulator phase diagram based on the Hubbard model....................................... 6
Figure 2.1: The real and imaginary part of refractive index of Ca3Ru2O7..................................... 12
Figure 2.2: Averaged refractive index of Ca3Ru2O7...................................................................... 13
Figure 2.3: Idea of the reflection contrast...................................................................................... 14
Figure 2.4: Simulated reflectivity of Ca3Ru2O7 few layers on SiO2/Si substrate. ......................... 16
Figure 2.5: Simulated reflectivity contrast for 10 layers of Ca3Ru2O7 on top of SiO2/Si substrates.
....................................................................................................................................................... 17
Figure 2.6: AFM tapping mode. .................................................................................................... 18
Figure 2.7: Several examples of common probems in AFM scanning.......................................... 20
Figure 2.8: Flattening the AFM image by data processing. .......................................................... 21
Figure 2.9: The Stokes/anti-Stokes process................................................................................... 25
Figure 2.10: The experimental setup of Raman spectroscopy....................................................... 26
Figure 3.1: The substrate design.................................................................................................... 30
Figure 3.2: Setup to crack the crystal along the direction of ab-plane. ......................................... 32
Figure 3.3: Collecting flakes from broken crystal......................................................................... 32
Figure 3.4: Examples of exfoliated flakes. .................................................................................... 33
Figure 3.5: The workflow of e-beam lithography. ........................................................................ 34
Figure 3.6: Alignment marks for e-beam lithography................................................................... 36
Figure 3.7: Matching designed patterns with optical images. ....................................................... 37
Figure 3.8: Fabrication of electrodes on flakes. ............................................................................ 38
Figure 3.9: Plasma cleaning of the residue.................................................................................... 39
Figure 3.10: ESD hazard. .............................................................................................................. 40
Figure 3.11: The SEM image of a burned device.......................................................................... 40
Figure 3.12: Grounding aluminum case covering the stereo microscope region. ......................... 42

vi

Figure 3.13:The grounding base of PPMS sample puck. .............................................................. 43
Figure 3.14: Other anti-ESD apparatus. ........................................................................................ 43
Figure 3.15: Baked silver epoxy on the chip. ................................................................................ 44
Figure 3.16: Flow diagram of wiring procedures. ......................................................................... 45
Figure 3.17: The I-V curve of a large FeTe flake from 300K to 10K. .......................................... 47
Figure 3.18: The evolution of Vt as a function of temperature for two pairs of electrodes. ......... 48
Figure 3.19: Contact resistance of a 54 nm Fe1+yTe flake device fabricated with in-situ ion-milling
procedure. ...................................................................................................................................... 49
Figure 3.20: Resistor model of 4-probe measurement. ................................................................. 50
Figure 3.21: The chemical composition of DEME+ (top) and TFSI-............................................. 51
Figure 3.22: The idea of ionic liquid gating (cross-section of a IL-gating device). ...................... 52
Figure 3.23: Microscope setup. ..................................................................................................... 53
Figure 3.24: The toothpick attached on the optical microstage used in dipping ionic liquid on the
chip. ............................................................................................................................................... 53
Figure 3.25: An ionic liquid drop on Ca2RuO4 device. ................................................................. 54
Figure 3.26: SiOx ring surrounded device with ionic liquid dropped............................................ 55
Figure 3.27: Leakage current as a function of gate voltage at room temperature. ........................ 56
Figure 3.28: Gate voltage influence on 4-probe resistance measurement. .................................... 56
Figure 3.29: Breakdown of the ionic liquid at room temperature. Both of the devices are Ca2RuO4
thin flakes. ..................................................................................................................................... 57
Figure 3.30: Etched surface by ionic liquid after the measurement. ............................................. 57
Figure 3.31: Graphene resistance as a function of gate voltage. ................................................... 59
Figure 3.32: The image of graphene device and the R-V curve.................................................... 60
Figure 4.1: Schematics of RuO6 octahedral distortion in Ca3Ru2O7. ............................................ 65
Figure 4.2: The metal-insulator transition and AFM transition of bulk Ca3Ru2O7. ...................... 66
Figure 4.3: AFM images of exfoliated Ca3Ru2O7 flakes. .............................................................. 68
Figure 4.4: R-T curves of Ca3Ru2O7 exfoliated flakes. ................................................................. 69
Figure 4.5: The MR comparison as a function of angle of in-plane magnetic field...................... 73

vii

Figure 4.6: MR of 13/8 nm flake as a function of field strength at various temperatures............. 76
Figure 4.7: The magnetoresistance of 9nm Ca3Ru2O7 flake.......................................................... 77
Figure 4.8: The R-T curves of the 7nm flake. ............................................................................... 78
Figure 4.9: The magnetoresistance of the 7nm flake..................................................................... 78
Figure 4.10: The comparison of Ca3Ru2O7 flakes with minor difference in thickness but dramatic
difference in low-temperature R-T behavior difference................................................................ 79
Figure 4.11: Ca2RuO4 crystal structure. ........................................................................................ 82
Figure 4.12: Ca2RuO4 images of devices. ..................................................................................... 83
Figure 4.13: The comparison of R-T curve of the bulk and 60nm flake around the transition
temperature.................................................................................................................................... 84
Figure 5.1. The structure of Fe1+yTe. ............................................................................................. 87
Figure 5.2: Phase diagram of Fe(Te1-xSex). ................................................................................... 88
Figure 5.3: The bulk properties of Fe1+yTe crystal. ....................................................................... 88
Figure 5.4: Images of an 11 nm exfoliated Fe1+yTe flake.............................................................. 91
Figure 5.5: The images of 15nm Fe1+yTe flake. ............................................................................ 92
Figure 5.6: Raman modes of Fe1+yTe crystal................................................................................. 93
Figure 5.7: Raman spectra of a 12nm Fe1+yTe flake...................................................................... 94
Figure 5.8: The I-V curve of the 15nm flake measured at 10K..................................................... 95
Figure 5.9: R-T curve of thick and intermediate flakes................................................................. 96
Figure 5.10: The increase of resistance over several baking process. ........................................... 97
Figure 5.11: R-T curves of thin insulating flakes.......................................................................... 98
Figure 5.12: Fitting of VRH model ............................................................................................. 100
Figure 5.13: Comparison on MR................................................................................................. 101
Figure 5.14: Magnetoresistance of 16nm and 11nm Fe1+yTe flakes............................................ 101
Figure 5.15: The evolution of superconductivity in exfoliated FeSe0.5Te0.5 flakes. .................... 104
Figure 5.16: The ionic liquid gating effect study on FeSe exfoliated flakes. .............................. 104
Figure 5.17: The fabricated FeSe0.5Te0.5 device with flake thickness 47nm. .............................. 105

viii

Figure 5.18: The R-T curve of superconducting FeSe0.5Te0.5 47nm flake................................... 106
Figure 5.19: The device with the ionic liquid gating................................................................... 107

ix

Chapter 1
Introduction and Motivation
This chapter focuses mainly on the motivation of this thesis study. General
motivation about why we want to study the nanostructured correlated material is
described in Section 1.1. A brief introduction on Mott-insulator and motivation of
calcium ruthenate nanoflake study is given in Section 1.2. The interesting properties and
the possibility of new findings in exfoliation of Fe1+yTe nanoflakes are introduced in 1.3.
Finally, a table of comparison on measurement methods and the reason why we study
these materials mainly by transport method are explained in Section 1.4.

1.1 Motivation
Solid state materials in nature usually fall into three categories: conductors,
insulators and semiconductors based on the separation of energy band gaps. Most of
transport properties can be well predicted by the electronic band theory without
considering electron-electron interaction despite the fact that the Coulomb force between
electrons is actually very strong when they are closed to each other. The reason given by
conventional solid-state-physics textbooks is: electron clouds shield the Coulomb
interaction and the strong repulsion from many electrons cancels each other so the overall
behavior is similar as if there was no interaction (with some correction rules based on
interaction like Hund’s rule). This counter-intuitive picture is amazingly applicable for
many materials but, of course, not always true. Magnetic materials showing macroscopic

1

magnetic moments induced by strongly coupled d-orbital electrons stand as a great
example while conventional superconductors discovered by Kamerlingh Onnes in 1911
with non-dissipating supercurrents (explained by BCS theory, in which electron-phonon
interaction split the energy levels) serves as another one. These materials in which exotic
properties are intrigued by electron-electron-interaction are called (strongly) correlated
electron systems. Many of transition metal oxides are considered correlated electron
systems as de Boer and Verwey first discussed the discrepancy of band theory prediction
on NiO (later called Mott-insulator and features metal-insulator transition), in 19371.
High-Tc superconductor2 and iron-based superconductor3,4 are also members of this
group. The mysterious physical phenomena in correlated systems are interesting not only
because they are not completely understood but also being extremely sensitive to
perturbations. The well-known Tc dependence on minor change of oxygen deficiency in
YBCO superconductor manifests the importance of stoichiometry. In transitional metal
oxides, metal-insulator transition can be manipulated by perturbation parameters like
doping, pressure and fields: VO25-7, manganites8,9, ruthenates10-13 and many other
examples.
Nowadays, nanostructured-material has been a new rising research field. Thanks
to the progress in fabrication technique and thin film technology in decades, preparing
and probing materials in few-atom size in given dimensions become possible and the
interactions on the reduced dimension e. g. the surface of the thin film or the edge of the
constrictions become prominent. Also the dimension reaching the coherence length of the
carriers may induce quantum mechanical effects. For example the conductance in low-

2

dimensional systems can be quantized such as carbon nanotubes14 and two-dimensional
electron gas on GaAs/AlGaAs heterostructures15.
The main motivation of the dissertation work is to combine the nanostructures and
correlated systems. Because correlated systems are extremely sensitive to perturbations
as mentioned above, tuning the systems by the other parameter, the dimension, is an
exciting direction. By constriction of the length scales on correlated systems in
nanoscales, probing single domain behavior, controlling material properties using
electrostatic gating, investigating surface effect and unraveling quantum mechanical
phenomena emerging at nanoscale all become possible. The route has already attracted
intense interests in the field of strongly correlated materials. For instance, researchers
have already tried etching YBCO thin films with focused ion beam (FIB) in narrow
constrictions and found the quantum of resistance R = h/4e2 being the deterministic factor
to suppress high-Tc superconductivity16. Single domain and the role of strain in metalinsulator transition have been observed in VO2 nanobeams17,18 as well. And more
recently the application of ionic-liquid technique is introduced on thin-film transition
metal oxide to tune the transition temperature19 or the magnetic properties20.
To realize the nanostructures of correlated systems, there are several existing
ways. The reduction of the first dimension can be realized by the fabrication of thin films
with the thickness in nanoscale. Conventional thin film deposition technique includes
both chemical vapor deposition (CVD) and physical vapor deposition (PVD). High
quality thin films through CVD usually require epitaxial growth, which is the deposition
on specific substrates with similar lattice structure to the desired material laying atoms
one after another for building the exact crystal structure with atomic precision. Single

3

crystal thin films can be grown using this method. But the cost of the instruments and
limitation of the substrate hindering many research groups from using this technique.
Also, a large substrate-film mismatch strain is inevitable such that the intrinsic property
of the thin film without strain is not reachable through this method. PVD takes a
vaporization process of the parent compound, by thermal energy or laser excitation, to
deposit the thin film much faster than CVD, and usually does not require a specific
substrate. Unfortunately the quality is usually worse than epitaxial growth so growing a
single crystal thin film through this method is almost impossible. The correlated systems
are sensitive to stoichiometry so any growth method without comparable quality to single
crystal undermines further research on fundamental properties.
There is another route invented recently. Since the discovery of peculiar
properties of graphene21, mechanical exfoliation using the “scotch-tape” method has been
widely applied to prepare thin flakes from layered-structured materials which can
preserve the crystallinity and stoichiometry from bulk without the strain effect compared
to epitaxial grown thin films. Graphene, MoS222, topological insulator Bi2Te323 and more
2D materials have been found, broadening the horizon of nanostructured materials.
The dissertation work takes the route in fabrication and characterization of
devices of exfoliated quasi-two-dimensional correlated systems, including calcium
ruthenate Ca3Ru2O7 and iron chalcogenide FeTe. These two categories of materials are
chosen because of their layered structures and rich phases in bulk which are anticipated to
be altered significantly in the exfoliated nanoflake form with reduced dimensionality
along the c-axis. We will discuss on the basic ideas of the interesting phenomena briefly
and the motivation to study these materials in this chapter.

4

1.2 Formation of Mott-insulators
From the Aufbau principle, electrons fill lower energy bands to higher energy
bands. With the Fermi energy lies inside a band, one would expect the system to be
metallic from the classical band theory since an incomplete band can accommodate
excited carriers with little energies. The case is broken in certain materials where the
electron-electron interaction takes part in and prevents the hopping of the carriers. As a
result the band is split, leading to the formation of Mott-insulators. The earliest discussed
Mott-insulator in 1937 is the nickel oxide1 in which Mott argued that most of the
electrons at low temperature are localized and the repulsion between electrons prevent
them from hopping in near positions24. For that reason it takes too much energy for
electrons to hop from one site to the neighboring sites, thus the system becomes insulator.
While the interaction strength (U/t) serves as an essential parameter in the formation of
Mott-insulator, the filling of carrier density which usually leads to the metallicity of the
system serves the other. With the basic idea of Hubbard model going to be introduced
below, the phase diagram of Mott-insulators can be plotted with simply these two
important factors25:

5

Figure 1.1: Metal-insulator phase diagram based on the Hubbard model.
The shaded area is the Mott-insulator, which can be achieved by both
bandwidth-control metal-insulator-transition (BC-MIT) or filling control
(FC-MIT) (From Imada et al., (1998)25)
The origin of theoretical approach to the MIT starts from the well-known
Hubbard model. Based on the tight-binding approximation, the Hubbard model
Hamiltonian contains both the hopping term and the on-site interaction. It is a great
simplification considering only electrons in the single band keeping the essentials of the
electron correlations. The Hamiltonian H takes the form:

H = Ht  H U   N

H t  t  (ci† c j  H.c.)
ij

1
1
H U  U  (n i  )(n i  )
2
2
i
The ciσ is the annihilation operator at site i and spin σ, t is the hopping integral
and the U is the on-site interaction energy. n is the number operator and N   ni .
i

6

From this simple model the single band split into lower and upper Hubbard band that
could explain the behavior of many Mott-insulators called Mott-Hubbard compound.
While in realistic cases, for example in the transitional metal oxides, the hybridization of
adjacent atomic orbitals like oxygen 2p and transition metal d-orbital may induce the
charge transfer as the other important mechanism of Mott-insulators. Most of the known
Mott-insulators are in these two categories.

1.3 Calcium ruthenate Ca3Ru2O7
Among vast choices of strongly correlated materials, we chose the calcium
ruthenate bilayer Ca3Ru2O7 as our main subject of study. Not only because of its layered
structure, there are many interesting aspects arising from the electronic and atomic
structure of this specific material. Two phase transitions at TN = 56K manifesting a
paramagnetic-antiferromagnetic transition and TMIT = 48K corresponding to a metalinsulator-like transition accompanied with the change of in-plane magnetic moment
direction. The magnetic anisotropy is intimately related to the structural anisotropy with
different degrees of RuO6 octahedral tilting angles in the a-axis and b-axis.
We anticipate a radical change of physical properties in exfoliated ruthenate
flakes for a couple of reasons: Firstly, the interbilayer coupling is playing a crucial role in
the antiferromagnetic transitions, thus the symmetry breaking at the surface can alter the
transition dramatically and the effect can be observed with enhanced surface-to-volume
ratio. Secondly, the distortion of RuO6 octahedron, which is a significant factor in the
magnetic anisotropy mentioned in the previous paragraph, may be largely altered in the
nanosheets and resulting a strong modification of electronic and magnetic properties. As

7

discussed in 1.1, the epitaxial growth of the material is extremely difficult especially on
the order of few-nanometer giving us a motivation to develop the exfoliation method.
Our work on the calcium ruthenate focuses on the thickness dependence of metalinsulator transition and magnetotransport behavior, as well as the lateral dimension effect.
Detailed information on device fabrication, characterization and the main result of
measurements will be presented in Chapter 4.

1.4 Antiferromagnetic ground state and transitions in Fe1+yTe
The other subjects of interest is Fe1+yTe, which is an non-superconducting end
member of the 11-type Fe-based superconductor Fe1+y(Sex,Te1-x). The material is
antiferromagnetic at temperature below 65~75K (depending on the concentration of
excess iron).26,27 The transition comes simultaneously with structural tetragonalorthorhombic transition28,29, and a semiconductor-metal transition with a sharp
discontinuous jump of the resistivity. Even though the bulk crystal is nonsuperconductive, superconductivity can be triggered by incorporating oxygen30 , or
introducing a tensile strain31 in the thin film form with Tc around 10K (oxygen) and
9.1~13K (strain) which is comparable to bulk FeSe. Perturbing the Fe1+yTe system and
searching for the superconductivity will be an interesting subject to explore. As we tune
the structure by removing the constraint of adjacent layers, all of the properties and phase
transitions listed above are possibly to change.
Our goal of studying Fe1+yTe exfoliated flakes is to examine the effects of
reduced dimensionality on the multiple transitions and to search the possible
superconductivity without applying pressure/strain and oxidation. Although there are
already a few existing works on FeSe and FeSexTe1-x using the mechanical exfoliation

8

method32,33, Fe1+yTe nanoflakes remain unexplored and calls for detailed investigation
considering the bulk being the parent compound of the superconducting FeSe and
FeSexTe1-x. In addition, the recent SP(spin-polarized)-STM study on Fe1+yTe discovering
the re-orientation of magnetic moments of the surface atoms34 leads to a great inspiration
on the possibility of surface effect in the exfoliated flakes. The main result of the Fe1+yTe
project will be discussed in the Chapter 5 of this thesis.

1.5 Why electrical transport measurements?
The study of the proposed system mostly relies on the electrical transport
measurement of exfoliated flakes. Though the selection of methodology strongly depends
on the existing facility of the lab, transport measurement is the most direct way to
confirm the metal-insulator transition and can probe some magnetic transition indirectly
through magnetoresistance measurements, although the measurements require long-hours
of device fabrication process in order to lay electrodes on the micrometer-sized flakes
and suffer from bad contact resistance, electrostatic discharges, and many other problems.
Note that magnetic moment measurements using SQUID or VSM are not realistic for
small flakes, and the neutron scattering which can determine the microscopic magnetic
structure also requires large bulky samples. There are also Raman spectroscopy and other
optical measurements available for thin flake, but the temperature dependent
measurement is not available for reachable commercial instruments. As a consequence,
we use transport measurement as our principle research method, while on early stage of
study we also use commercial Raman system in Prof. Swain’s lab, and an ellipsometer in
Prof. Bruening’ lab, both in Dept. of Chemistry MSU, in order to confirm the exfoliated
flakes are real ones coming from the bulk material.

9

Table 1.1: Comparison of measurement techniques available to our group

10

Chapter 2
Characterization methods
This chapter discusses the major characterization method utilized in this
dissertation. In Section 2.1 the visibility of Ca3Ru2O7 flakes is analyzed through transfer
matrix method based on the refractive index data we measured on bulk crystal. Section
2.2 contains a brief introduction on atomic force microscope (AFM) and how data is
taken and analyzed, including examples showing typical symptoms of scanning and the
treatment. In Section 2.3 I will discuss briefly the working principle of SEM and EDX.
Section 2.4 will be the basic physical principles of Raman spectroscopy and an example
of experimental setup.

2.1 Visibility of exfoliated flakes on substrates:
Since the monolayers of 2D materials has been obtained by exfoliation methods
including graphene21, MoS222, NbSe235 and many others, a natural question arises: How
can it be possible to see the few-nm flakes with naked eyes? The answer depends on the
optical contrast of the thin flakes on the substrate as proposed by Blake et al, 36. From the
experience human being usually can catch things with ~5% of contrast at the most
sensitive wavelength region to naked eyes (500~550nm). To make sure exfoliated thin
flakes on substrates can be identified, I have measured the refractive index of Ca3Ru2O7
crystal with an ellipsometer in Prof. Merlin Bruening’s group (Dept. of Chemistry, MSU)
and then simulate the reflection contrast of few-layer Ca3Ru2O7 on SiO2 /Si substrate
with the transfer matrix method to see if it is possible to see few-layer flakes on SiO2

11

90nm and 300nm substrates. The refractive index has been measured 3 times and
averaged. According to Prof. Bruening, the values from 400~450 nm and 700~750 nm
are not accurate. The raw data are shown in Figure 2.1.

Figure 2.1: The real and imaginary part of refractive index of Ca3Ru2O7.
Crystal measured at 3 different points (labeled r1, r2 and r3) with an
ellipsometer. Data in 400~450 nm and 700~750 nm should not be trusted.
Ideally, the real part and imaginary part should follow the Kramers-Kronig (KK)
relations in the whole range of wavelength, but because we are only interested in the
visible light region, we use empirical equations which won’t satisfy KK relations to fit
the data. One is the Cauchy’s equation for the real part and the imaginary part is fit with a
linear function. The purpose of fitting is to attain a smooth approximation of refractive
index for the reflective contrast simulation. I averaged the 3 measurements only in the
reliable region (450~700 nm) for fitting. The real data is fitted with MATLAB function
“nlinfit” (nonlinear regression) and imaginary data is fitted with Origin “Fit linear”
function (least-square method)

12

Figure 2.2: Averaged refractive index of Ca3Ru2O7. Left: Real index.
Right: Imaginary index. Black curve: averaged data. Red curve: Fitted line.
Fitted Parameter:
Real: Cauchy’s equation n( ) 

A



2



B



4



C

6

 D , wavelength in micrometers

Imaginary: k ( )  m   b
Parameter
Value

A

B

C

D

m

b

-0.0034

0.0554

-0.1911

2.3141

0.5775

0.4539

Table 2.1: Fitting parameters of refractive index of Ca3Ru2O7

13

Transfer matrix method:

Figure 2.3: Idea of the reflection contrast. Left: The contrast is the
difference between reflectivity on SiO2 (substrate) and Ca3Ru2O7
(material). Here k0 is the wave vector of the incident light and the k0’ is
the reflected light. Right: The reflectivity is calculated on every interface
of different materials.
The calculation of reflectivity from transfer matrix method is based on the
continuous condition of E-M fields at the interface of material n0 and n1. The equation
can be written as:
E0  E0'  E1  E1'
H 0  H 0'  H1  H1'
n0 E0  n0 E0'  n1 E1  n1 E1'

The first equation is from the continuity of the electric field, and the second is the
same from magnetic field. By applying the condition of Fresnel equations H = n*E/c on
the second equation we get the third equation. When the light travels through next layer it

14

picks up a phase ki di 

2 ni di



, assuming the next layer to n1 is semi-infinite denoted as

“T”, the boundary condition is:
E1eikd1  E1' e ikd1  ET
n0 E0 eikd1  n1 E1'e  ikd1  nT ET

By solving the equations with the matrix method, we get the r and t as the ratio of
reflected and transmitted electric field to the incident field:
E0'
n
E
1
 (cos(kd1 )  i T sin(kd1 )) T
E0
n1
E0
E0'
n
n0  n0
 (i T sin(kd1 )  nT cos(kd1 ))
E0
n1
i sin( k1d1 ) / n1   1  ET
 1   1  E0'  cos( k1d1 )
 n     n  E   i sin( k d ) / n
cos (k1d1 )   nT  E0
1 1
1
 0  0 0 
i sin(k1d1 ) / n1   1 
1  1 
 cos(k1d1 )

r

t
 n   n 
 i sin( k d ) / n
cos ( k1d1 )   nT 

1 1
1
 0  0

We define the transfer matrix M here since the same equation applies to the next
layer, M will be a product of a series of matrixes.

 cos(ki di )
Mi  
 i sin(ki di ) / n i
M   Mi

i sin(ki di ) / n i 
cos (ki di ) 

The final reflectivity = |r|2 can be solved by the matrix element of M:
1  1 
1
A B  1 
 n     n  r  M  n  t  C D   n  t

 T
 T
 0  0
An0  BnT n0  C  DnT
r
An0  BnT n0  C  DnT
R | r |2
C

Rmaterial  Rsubstrate
Rmaterial  Rsubstrate

15

Here C is the Michelson’s contrast.
Applying the refractive index of silicon and silicon dioxide, and the Ca3Ru2O7
measured previously, we get the R and C as a function of layers and wavelength. The
result is presented in Figure 2.4 and Figure 2.5.

Figure 2.4: Simulated reflectivity of Ca3Ru2O7 few layers on SiO2/Si
substrate.

Left: SiO2 90nm. Right: SiO2 300nm. The single bi-layer

Ca3Ru2O7 thickness is assumed to be 0.977 nm that is half of the lattice
parameter c because each unit cell contains 2 layers.

16

Figure 2.5: Simulated reflectivity contrast for 10 layers of Ca3Ru2O7 on
top of SiO2/Si substrates. Left: SiO2 90nm Right: SiO2 300nm
C

Rmaterial  Rsubstrate
Rmaterial  Rsubstrate

From the results, we obtain the contrast C larger than 5% (0.05) on both substrates
with few-layer flakes around 550nm indicating it is possible to see them with eyes. And
90 nm SiO2 substrates have better contrast ratio over 300 nm SiO2. The flake will appear
darker on the substrate in the visible wavelength.

2.2 Atomic force microscopy
The thickness of flakes is measured with the atomic force microscopes (AFM)
because it is almost the only way to determine the thickness of micrometer-sized flakes
with nm-precision with reasonable efficiency. There are several different operating
options of AFM, and we will just briefly introduce the tapping mode AFM which we use
for our measurement. The basic working principle, as depicted in Figure 2.6, is sensing
the tiny height differences on the surface with an oscillating tip. While the tip is
oscillating at its resonant frequency, the frequency and the amplitude of the tapping tip
would be sensitive to the weak attraction force between the tip and the surface, and the

17

force would be stronger in the higher area (closer to the tip). By shining laser on the back
of the tip and detecting the reflected signal with optical sensors, it would be possible to
determine the height difference to atomic level (~0.1 nm) at the optimal condition. There
is a piezo stage drags the tip in one direction for a linear scan, and multiple linear scans
compose a complete area scan of the surface. We use both a digital instrument DI-3100
AFM and an Asylum-Cypher AFM.

Figure 2.6: AFM tapping mode. The tapping tip on the flapping cantilever
measures the height profile of a sample (red) on a substrate (blue)
The most important (and the most consumed) part of the scanning is the tip.
Choosing an appropriate tip from multiple options may take some basic knowledge. If
there is no specific need (magnetic force, conductive or insulator) usually people choose
Si tip because it is cheap and available from most of the suppliers. The Si tip is etched to
reach the sharpness about 10nm that is enough for most of the application. The force
constant (of bending the cantilever) is particularly important for contact mode scan of
tiny features which requires the cantilever to be as soft as ~0.1 N/m, so usually the
cantilever is very long. For tapping mode the force constant is much larger, around
several tens of N/m and the cantilever is much shorter.

18

The most problems we face when scanning AFM is usually the artifacts. In Figure
2.7 three kinds of artifacts are shown with real examples. (a,b) is the double-image
problem created by two peaks on the tip because both of them can sense the surface and
the outcome is the superposition of the two images. The distance between the two images
is the distance between the two peaks on the tip. Usually this effect happens with a
damaged tip after repeatedly scanning of rough surface or steep high edges. In (c) the
interference pattern shows up on the graph as a false signal since it is not related to any
height information. Because we use laser beam to probe the oscillation of the tip, and it is
quite closed to the sample surface, thus the ring or stripe patterns generated by
interference between incident and reflected laser light takes place frequently. The
solution to this problem is re-adjusting the tip and laser position so that the tip can block
the laser spot better, preventing the reflection of laser from sample surface. (d) The effect
of dull tips which makes all small dots on the graph look the same weird pattern like the
“c” in the example. Sometimes it can be a triangular pattern or even more irregular
shapes. The origin of this artifact is the tip being worn out and unable to sense features
smaller than the tip size, so the feature will look all the same as the shape of the tip.
Overall, being cautious about the tip status and watch out for the artifact of the scanned
image is the key to make successful AFM scans.

19

Figure 2.7: Several examples of common probems in AFM scanning. (a,b)
The “shifted” image indicates there are two peaks on a single tip both
scanning the feature and generate duplicated images. The images are
Ca3Ru2O7 exfoliated flakes on SiO2/Si substrates. (c) Interference pattern
in AFM scanning. The blue-black stripes on the lower-left of the image is
a false signal induced by the interference pattern on the sample. The
scanned item is a FeTe exfoliated flake. And many “scar” features are also
presented on the upper-middle part. (d) The effect of dull-tip making all
small dot looks like a “c”.
Once the data is taken, analyzing the AFM graph is also an important part. Since
when the AFM tip hits a bad spot, it ruins the rest part of a single line scan leaving a scar-

20

like feature on the graph. And the height reference point may differ between different line
scans and the whole surface may be inclined to a direction. In this case, we first remove
the scars by replacing the ruined part with the two adjacent lines, and then flatten the
whole image by applying a linear correction to every line scan so that the height
references are the same on our assigned line/surface which we assume is flat (usually the
plain SiO2 surface without flakes/glue spots). The work mention above is done by DI
analyze software or an open-source software Gwyddion.

Figure 2.8: Flattening the AFM image by data processing. The raw scan
data on the left is a cross-bar marker fabricated on the SiO2/Si substrate
with some small flakes around. The height reference jumped off between
some sections. After the flattening and removal of scar the essential
information of the graph becomes clear.

2.3 Scanning electron microscopy
There are two applications of SEM in our projects. One is to investigate the
details of broken devices since SEM scanning is much faster than AFM when we don’t
need thickness information on the surface and the other is to utilize e-beam lithography
which will be introduced in the next chapter. Here we just briefly discuss the working

21

principle of SEM. A SEM consists of several parts: beam emitter, condenser lens,
deflection lens and detectors. Electron beams are generated from the beam emitter and
accelerated up to 15~30 keV in the high vacuum, passing through the condenser lens and
condensed onto a very narrow spot, typically ~10 nm in the lateral dimension or less.
And the deflection lens control the direction of the focused electron beam to scan the
surface. Finally the scattered electrons are picked up by the detector for imaging.
There are mainly two kinds of beam emitter: thermal filament and field emission
(FE). The thermal filament is heated up by a large electrical current so that some
electrons can overcome the energy barrier to escape the surface by absorbing thermal
energy. A FE type electron source is usually a sharp tip for creating a large local field on
the tip so that some electrons can be pulled out directly by the field. Thermal filament
features a more stable emission current so it is usually used for e-beam lithography which
requires a precise beam current control and FE is much faster in turning on/off so it is
better for imaging. But currently FE can be controlled quite stable as well so there are
also FE type SEMs used for e-beam lithography. Since the electron carries charge, so the
movement can be controlled with electromagnetic fields. Condenser lens is composed of
a series of coils and plates generating EM fields for focusing the electron beams, and
deflection lens apply usually just electric field to shift the beam spot on the surface.
When the accelerated electrons with the kinetic energy on the order of 10 keV hit
the surface, all of the inner orbital electrons of atoms on the surface will be knocked out
because the binding energies are just on the order of 10 eV. Low energy x-rays will be
produced with the recombination of these electrons and collecting the x-ray spectra helps
analyzing the elemental composition of the sample (EDX). Most of high energy electrons

22

are bounced back without much loss of energy, usually called backscattered electrons
(BSE), while some of them penetrate deeper, keeping knocking out more electrons from
the material called secondary electrons (SE). Typical SEMs have both kinds of detectors
for BSE and SE, while the SE detectors are commonly used for imaging because SE
reveals more details in the scanning.
Aside from imaging the surface, SEM can control the position and on/off of the
focused e-beam to “write” a pre-programed pattern on the sample. This is the basic
concept of the e-beam lithography and the details will be introduced in the device
fabrication process.
We have used a Hitachi SU-4700 SEM (FE) for imaging and EDX analysis, a
JEOL 840A (Thermal filament) and a Hitachi SU-5000 SEM (FE) for e-beam lithography.

2.4 Raman spectroscopy
Probing the structural change of tiny exfoliated flakes with Raman spectroscopy
has been a common method since the graphene is found to have significant peak shift
below 5 layers37. Because the method is fast and does not require extra process or
treatment on the flakes. In our case, though we did not see significant shift from the thin
flakes of Ca3Ru2O7 ~10 nm and the signal from the thinner flakes are too weak to detect,
it still gives us some indication that the flakes keep the similar structure at this thickness.
For Fe1+yTe flakes it is too difficult to have an uncontaminated spectrum since they
oxidize quickly both in air and under strong irradiation.
To explain the Raman effect in one sentence, it is the inelastic scattering of light
by lattices creating/annihilating normal mode phonons. Classically, it can be written in
the form of an oscillating electric field driving the polarization of the material. Suppose

23

the frequency of the field is  0 , the normal mode frequency is  m , for a non-zero
  
polarizability change 
 we will have the emitted light from the oscillation of
 q 0

polarization at frequency  0  m and 0  m as shown in the follow:
E  E0 cos(2 0t )
P   E   E0 cos(2 0t )
q  q0 cos(2 mt )
  
 q
 q 0

  0  

  
P   0 E0 cos(2 0t )  
 E0 q0 cos(2 mt ) cos(2 0t )
 q 0
  
  0 E0 cos(2 0t )  
 E0 q0  cos(2 (( 0  m )t )  cos(2 ( 0  m )t ) 
 q 0

We call the “-” peak “Stokes” and the “+” as the “anti-Stokes”. The light with
original frequency is called Rayleigh-scattered light.
Classical model gives the right frequency of the Raman scattered light, but fails to
describe the correct ratio of intensity between Stokes and anti-Stokes. This is because of
the quantum nature of the light (photon) and the oscillation (phonon). Consider n
phonons in a quadratic potential well V=

1
m 2 2
x in which E =  (n  ) , the excitation
2
2

of light is HE =E0 x since the wavelength of light is way longer than the lattice constant.
From the transition matrix element calculated in basic quantum mechanics we know this
Hamiltonian can only increase/decrease n by one. The result corresponds to Stokes/antiStokes processes mentioned above. The intensity of Stokes and anti-Stokes peaks can be
estimated with the energy (frequency) difference and temperature because phonons are
bosons:

24

E

3

I AS  AS
 3 e k BT *
IS
S

We plot the energy level diagram of phonons and the a real spectrum containing
both Stokes and anti-Stokes peaks in Figure 2.9.

Figure 2.9: The Stokes/anti-Stokes process. Left: The energy level
diagram of the Raman scattering. Right: the actual spectrum of Raman
spectroscopy and the central peak is the laser excitation.
In practical, we use reflective micro-Raman spectroscopy for detecting signals
from the exfoliated flakes on SiO2/Si substrates. The idea is illustrated in the following
picture. Flakes are found under an optical microscope using a 100X objective and laser is
sent through the objective to excite the flakes. We select a certain polarization of analyzer
from the reflected light mixing Raman signals and Rayleigh-scattered light and then
block the latter by a notch filter designed to extinguish the light within a very narrow
wavelength region since the signal of Raman signals is much weaker (typically 10-6 in
ratio) and cannot be seen without reducing the strength of Rayleigh-scattered light. In

25

order to suppress exactly at the laser wavelength, a notch filter is put right in front of the
detector. A diagram of typical Raman spectroscopy setup is plotted in Figure 2.10.

Figure 2.10: The experimental setup of Raman spectroscopy. Usually
100X objective is used for small samples. The polarizer and analyzer
select the polarization of exciting laser and scattered Raman signal. Red
light (deeper) is the incident laser, reflected by a small mirror (or a beam
splitter) and focus on the sample through a microscope objective.
In commercial Raman instruments sometimes there is no analyzer available so we
have plenty of Raman spectra without selection of polarization. The laser source is
532nm Nd-YAG laser producing power on the order of mW. Heating effect is observed
usually with the power over 10 mW (depends on thickness, size and thermal conductivity
of the flake) so at room temperature the power is limited as small as possible around
1~3mW when collecting the data.

26

Chapter 3
Preparation of devices
As a member of layered structured material, Ca3Ru2O7 in principle could be
mechanically exfoliated into thin flakes containing only few atomic layers as well, but
due to the CaO structure formed between RuO6 octahedral and Ca ions, the layer-layer
ionic binding force inside the material is much stronger than graphite (that has van der
waals force) so that exfoliation cannot be carried out easily. In the early studies
researchers found the typical exfoliation method (scotch tape) generates “flakes” with the
thickness on the order of micrometers and the minimum thickness they can find is about
330 nm38, which is apparently not thin enough for any possible transition triggered by
reduction of dimensionality.
In the beginning of this project, our group had tried several methods to exfoliate
the crystal. We tried scribing SiO2/Si substrates with crystal edges hoping the thin flakes
to fall off on the substrates. But it only generates traces which can hardly be distinguished
as the material because they look just like powder and lack characteristic Raman peaks,
also it cannot be used for device fabrication because the size is too small and the surface
is not flat. We also tried adding Ca3Ru2O7 powder into IPA, sonicate it for several hours
and centrifuge the solution, hoping the sonication knocks off some floating flakes in the
solution and we can collect them by dipping a drop of solution on SiO2/Si substrates and
blow-dry. It turns out that there are rarely some tiny pieces with thickness around few
nanometers which can be found by AFM scanning, and they are too small to be observed

27

under optical microscopes. Further characterization to verify if they are really Ca 3Ru2O7
is not possible either because of the size of the flakes.
To deal with this problem, I invented a method which turns the mechanical
exfoliation process of few-nanometer-thick flakes possible with Ca3Ru2O7 and it is also
applicable for other “hard” layer-structured materials. The basic idea of this method is as
follows. First, fix the ab-plane of a crystal on a flat surface by strong glue, and then glue
the other parallel side of the crystal on a top lid. After the glue is totally cured, we crack
the crystal along ab-plane by levitating the top lid. The newly generated surface of the
crystal contains a lot of loosely-connected flakes that can be easily picked up by scotch
tape and transferred onto SiO2/Si substrates. The method can be applied repeatedly on the
newly cracked crystals until they are too thin to be cracked uniformly.
The exfoliated flakes in large sizes can be characterized in the ways mentioned in
previous chapters. Later on they can be made into devices with gold electrodes by e-beam
lithography for electrical transport measurement in the Quantum Design Physical
Property Measurement Systems (PPMS). During the fabrication process, I have
encountered many problems, such as bad contact resistance, the static charge burns the
device instantly when it connected with the probes, the ionic liquid dispersed away and
etched the device, etc. Details of exfoliation, device fabrication and resistance
measurement, including ionic liquid gating will be explained in this chapter. Section 3.1
and 3.2 will cover the preparation of substrates and the method of exfoliation. E-beam
lithography procedures are described in Section 3.3. Problems with contact resistance are
explained in Section 3.5. And finally the idea and the test results with ionic liquid gating
are presented in Section 3.6.

28

3.1 Preparation of substrates
Before we exfoliate flakes onto the substrates, by learning the wisdom of
development of 2D material research we know that without proper label or coordinate it
will be extremely hard to find the same micrometer-wide flakes on a centimeter-wide
substrate repeatedly for characterization and device fabrication purposes. To define the
position of flakes on substrates, we need to pre-pattern the substrates and clean them
before use. Besides the coordinates, we also want to make some large contact pads to
save e-beam writing time and use them for potential ionic-liquid gating purposes.
The substrates used for exfoliation are prepared from SiO2 90 or 300 nm /Si wafer
(100) purchased from Graphene Supermarket. There are plenty of suppliers of this kind
of wafers and we pick up this just for convenience reason. The pre-made coordinates
together with large contact pads (200x100 Âľm2) are made by photolithography and
thermal evaporation. First the SiO2/Si wafer is spin-coated with S1813 photoresist at
3000 rpm for 40 seconds and baked at 95°C in oven for 30 minutes. After that the wafer
is exposed to UV under a designed chromium-quartz mask, fabricated by Photo Science
Inc., for the desired pattern and developed/rinsed with MF-319 developer/DI water
30sec/30sec. The patterned wafer would be then transferred into the thermal evaporator
chamber. Ti/Au thin film with thickness 3/60 nm in high vacuum (~10-7 Torr) is
deposited and lifted off in warm acetone with slight sonication. Fabricated wafers finally
are covered by S1813 and baked again for protection and sliced into 8x8mm pieces by a
digitalized dicing saw. The design and fabricated chip outlook are presented in Figure 3.1.

29

Figure 3.1: The substrate design. Left: 5X optical image of substrate (with
some exfoliated flakes) yellow rectangle blocks and coordinates are
patterned and deposited with Ti/Au thin films. Right: Design of
coordinates and contact pads. The crosses with blocks denotes different
four areas on the substrate avoiding confusing recurring coordinates.
There are two cleaning procedures for the patterned substrates. One is called
“solution clean” by soaking the substrates first in warm acetone for 5 minutes to remove
the photoresist protection layer and rinsed in IPA (isopropanol), followed by blow-dry
IPA with nitrogen gun. And the other is the RCA clean which takes much longer time (in
deed it is the second step of a complete RCA clean of Si wafer without piranha step). The
procedure starts from soaking the chips in the 1:1:5 H2O2 (~25%):NH4OH(~30%):H2O
(ratio in volume) and heat up to 70°C for 30 minutes, rinse in DI water, blow-dry. The
H2O2 should be a fresh bottle or it will not be effective after opened and kept in
refrigerator 24 hours or at room temperature 30 minutes. Solution clean + RCA clean are
recommended to improve flake-substrate attachment and increase the exfoliation yield,
and they usually works better for Fe1+yTe, Fe(Se,Te) and graphene exfoliation. For

30

exfoliation of Ca3Ru2O7 flakes solution clean is good enough. The final step is to put the
substrate in the plasma chamber, pump down below 60mTorr, and clean the substrates
with oxygen plasma at pressure around 500 mTorr and with power = 300W for 2 minutes.
After cleaning, the substrates should be used in exfoliation within a day.to avoid
contamination by exposing them in the air for too long.

3.2 Exfoliation (Gorilla glue method)
The basic idea of this method is to fix the bulk crystal on both sides of ab-plane
with two pieces of microscope slides, and then cleave the crystal by pulling the two slides
apart. To fix a few-millimeter piece of Ca3Ru2O7 crystal we use Gorilla glue which cures
with moisture and glue the ab-plane side of the crystal on a microscope slide. The process
takes 2-3 hours. Until the glue is fully dried, we use another microscope slide with
Gorilla glue to cap the other side of the crystal. To ensure the flat contact between the
crystal and the microscope slide, a small C-clamp is applied over the sandwich and
folded tissue is filled on both sides of the microscope slides avoiding tilted contact. The
setup is drawn in Figure 3.2 with a picture of a real example. After another 2-3 hours we
can break the crystal by levitating the slide so that the crystal breaks through the ab-plane.
Crystal flakes can be then exfoliated by pressing scotch tape over the broken surface and
then transferred onto pre-cleaned SiO2/Si substrates. The schematics of the process are
shown in Figure 3.3. After exfoliation, the substrates are put into oxygen plasma chamber
to clean up the tape residue since Ca3Ru2O7 does not oxidized so it can survive the
oxygen plasma and the substrates are examined under a high magnification optical
microscope for flakes with micrometer size.

31

Figure 3.2: Setup to crack the crystal along the direction of ab-plane. Left:
the procedures of fixing the Ca3Ru2O7 crystal on microslides. (from top to
bottom) Right: a picture of actual setup, tools used in procedures are
labeled with their names.

Figure 3.3: Collecting flakes from broken crystal. After the glue is cured,
the crystal is broken through the layer direction and flakes can be collected
using Scotch tape method on the broken plane of crystal.
The example of typical substrate surfaces are in Figure 3.4. Tanned colored big
flakes are usually much thicker than 100 nm, and we are focusing on blue or dark colored
small flakes. Usually few flakes on the order 20-100 nm could be found in a good-split
sample. But the method depends heavily on the cleaved surface quality. If the broken

32

surface is flat and shiny which usually indicates a nice ab-plane cleavage, we can get
many useful flakes, and the cracked crystals can be glued and split again. If this is not the
case, nearly no useful flakes (under 100nm) can be found in the exfoliated samples and
crushing away the residue glue/bad surface with razor blades gives little hopes on making
another good split.

Figure 3.4: Examples of exfoliated flakes. Left: Typical exfoliation of
large blocks of Ca3Ru2O7, thickness ranging from several hundred nm to
few micrometers. Right: Few flakes can be sometimes found nearby large
blocks, the thickness of the smaller one is about 30 nm for example.

33

3.3 E-beam lithography

Figure 3.5: The workflow of e-beam lithography. The process follows the
number of the diagram and the last one is an optical image of fabricated
device based on a 20 nm Ca3Ru2O7 flake. The orange layers stand for ebeam resist mentioned in the main text and the black layer stands for metal.
E-beam lithography is a necessary step in order to lay sub-micrometer electrodes
on the exfoliated flakes. The work flow is explained in Figure 3.5. First the substrates are
spin-coated with bi-layer e-beam resist MMA/MAA EL9 and PMMA C2, both with
rotation speed 4000rpm and baking condition 180°C on a hot plate for 90 seconds. Bilayer coating is chosen because the advantage of easy liftoff which make less harm to the
flakes. Then the substrates are sent into a Joel-840 (thermal filament) or Hitachi SU-5000
(Field-emission) SEM. Electron beams are accelerated and focused on the desired region
to “write” designed patterns by modifying chemical structure of the e-beam resist so that

34

it becomes soluble in the developer MIBK. Then the sample will be deposited with
material by evaporation or sputtering, and lift-off in warm acetone for 10~15 minutes
following by IPA rinse and blow-dry. Acetone flushing and light sonication will be used
if the film is not lifted-off automatically.
The devices fabricated in the dissertation consist of four electrodes on a flake and
a top gate nearby the flake but not in contact with the flake and other electrodes. The
fabrication process includes putting alignment marks, designing the electrodes, writing
and development, deposition and lift-off. On some devices we have SiO rings over the
device.
Alignment marks are written nearby the sample by referencing the optical images
and the pre-made coordinates on the substrates. Usually two crosses defining an area
about 80x80Âľm2 are written near a flake. In some cases alignment marks will be put a bit
away from the flake to get avoid of pre-made coordinates and contact pads.
Ti(2nm)/Au(60nm) thin film is deposited by thermal evaporation which is also used for
coordinates on the substrates. See Figure 3.6 for the developed align mark pattern and the
finished ones.

35

Figure 3.6: Alignment marks for e-beam lithography. Left: Developed ebeam resist mask for deposition. Right: finished alignment marks. Both
images are taken under a 100X microscope objective.
Electrodes are designed to connect the large pre-made contact pads to the flakes;
the actual process is carried out with the software Design CAD and Microsoft
PowerPoint. Patterns drawn with Design CAD can be pasted onto the optical image in
PowerPoint with appropriate sizing and alignment with the marks to see if the electrodes
will be connecting the flake and the contact pads (See Figure 3.7). In real writing
procedures the patterns are aligned by matching the alignment marks in design and the
actual marks fabricated on substrates from the previous step. Two-step writing at 500X
and 100X covers up a region ~800 Âľm x800 Âľm without losing resolution for fine
patterns. For 100X writing, the aperture is opened up for larger writing current yielding
less writing time when using JEOL SEM.

36

Figure 3.7: Matching designed patterns with optical images. Blue dashed
lines are pasted onto the image from Design CAD. Left: 100X Right: 20X
optical images.
Ni(10nm)/Au(100nm) deposition by sputtering is the best for the device since the
adhesion is better and also the contact resistance is reasonable (few kΊ).
Ti(3nm)/Au(100nm) has been tried a lot before on Ca3Ru2O7 but the contact resistance is
usually too big for measurement and sometimes random (several 10k to MΊ). Au(100nm)
only gives reasonable contact resistance but bad adhesion that sometimes won’t survive
lift-off and get broken with ionic liquid easily. An example of the developed pattern to
finished device is in Figure 3.8.

37

Figure 3.8: Fabrication of electrodes on flakes. Left: Developed e-beam
resist mask on the flake. The center part is blurry because the distance
between electrodes is below diffraction limit of the visible light. Right:
Finished device with deposition of Ni(10nm)/Au(100nm) through
sputtering.
The writing of alignment mark is usually done with a dosage of 300 ÂľC / cm2 well
as electrodes far from the flakes. The dosage of patterns right on the flakes need to be set
separately to 430~480 ÂľC / cm2 depending on the width and separation of electrodes. The
number should be carefully set because the width of each wire is only a few hundred nm
and there are 4 wires running parallel to each other. Under/overdosing may cause a failed
device because gold will be lifted-off in under-dosing case since there is still e-beam
resist in the written pattern or the electrodes will be shorted in the case of overdosing
because the width of patterns expands so the electrodes touch each other. Generally when
electrodes are closer to each other, less dosage should be used and the other way else.
Oxygen plasma can be used to remove under-dosing e-beam resist residue with power =
100W and time = 30s ~ 60s but the patterns will suffer some distortion so it is not

38

recommended for very fine electrodes. Figure 3.9 provides an example of plasma
cleaning improves the as-developed pattern.

Figure 3.9: Plasma cleaning of the residue. Left: Developed e-beam resist
mask with residue in the pattern. Right: Cleaned pattern after oxygen
plasma power = 100W and time = 30 sec. The angles of the rectangular
gate electrode become sharper and the residue in the left electrode is
cleaned up.

3.4 Wiring and ESD protection
The electrostatic discharge (ESD) is a very challenging problem in the early stage
of this project. Once we tried to connect a fabricated device to the probe, the flake and
the electrodes always burned. The reason is because some static charges which have very
high potential conduct a sudden large current through the narrow flake and generate a lot
of heat and destroy the device. Depending on how much current is passed, sometimes we
see the whole device are destroyed including the gold electrode deposited on the top of
the flake, as shown in Figure 3.10. Sometimes only small fractions of the flake are burned
which is difficult to check under the optical microscope so we have to take SEM images
on them as presented in Figure 3.11. The problem is difficult to solve because static

39

charges exist everywhere in the environment and during the wiring process the device
may make electrical contact with many other things. So the process must be taken care
thoroughly from wiring up the device to transferring it into the PPMS chamber to prevent
the risk of burning devices.

Figure 3.10: ESD hazard. The sample burned after being connected to the
probe. Not just the flake but the local gold electrodes are all burned and
vaporized. The device is a flake of exfoliated Ca3Ru2O7.

Figure 3.11: The SEM image of a burned device. The wide bright stripes
are the gold electrodes and the tiny wire across the stripes is the
Ca3Ru2O7 nanoflake. The flake is burned between every two adjacent
electrodes. The distance between electrodes is about 300nm.
The typical route is: A chip with fabricated devices would be placed on a clean
microscope slide under a stereo microscope with a grounded aluminum case (Figure 3.12)

40

in order to eliminate possible charged dusts falling on the device. Anti-ESD lab coat,
grounding wristband and charge dissipating gloves are equipped to prevent the discharge.
Before handling the wires and touching the chips it is also recommended to turn on the
ion fan (Figure 3.14(a)) and blow away charges and particles attached on cloths and skin.
Mixed silver epoxy (Epoxy kit A and B 1:1) is applied on the contact pads connected
with 4-probes or a gate electrode and extended away from the device by a toothpick with
a thin gold wire tip. 25Âľm dia. gold wires are placed at the ends of silver epoxy extension.
To cure the silver epoxy, the chip could be loaded into an oven set at 250°C for 5 minutes
or directly into the PPMS sample chamber with all electrodes grounded and baked at
380K for two hours. The latter method is better because it eliminates the possibility of
discharges completely since the resistance of epoxy gradually goes down and removes
any potential differences between probes slowly. In addition, baking in the vacuumed
chamber prevents oxidation for sensitive materials. The representative device with baked
epoxy is shown in Figure 3.15. A PPMS sample puck is prepared by connecting it with a
home-made socket (Figure 3.13) and grounding all contacts together including the metal
part of the puck such that it can be freely moved between the stereo microscope and the
ionic liquid dipping stage for convenience. The chip is mounted on the PPMS sample
puck with a dip of N grease on the back side for better thermal contact and all extended
gold wires are pressed onto indium contacts. The corresponding position of 4-probes and
the gate electrode will be recorded on the lab book before loading. On the probe side, it is
better to have the instrument (PPMS internal circuit or the Keithly 2400) turned on and
running a very small current (usually 10-8 A) through all grounded 4-terminals before
loading the device in. The reason is because when turning on the instrument there is

41

usually a sudden large pulse current and it may burn the device if it is connected. Two
switch boxes are made to control the probes (Figure 14 (b) (c)): One can turn on/off
grounding each channel and the other can add or the 1MΊ buffer resistance of the current
probe. The purposes of these two boxes is to ground unused probes and also keep device
grounded when loading in, and also prevent a large pulse current when the current go
through. The workflow of the mounting of devices is plotted in Figure 3.16.

Figure 3.12: Grounding aluminum case covering the stereo microscope
region. The case and the blue anti-ESD table mat are connected to the
metal gasline which is connected to the ground.

42

Figure 3.13:The grounding base of PPMS sample puck. The base is made
with metal pins glued on a piece of aluminum. The pins are soldered
together with a long wire connected to the ground.

Figure 3.14: Other anti-ESD apparatus. (a) Ion fan (b,c) grounding switch
boxes for turning on/off grounding on each probe/channel. The buffer
resistors used in the box are 1MΊ for each channel.

43

Figure 3.15: Baked silver epoxy on the chip. Black stripes are silver epoxy
applied by hand with a toothpick. The epoxy stripes connects the
electrodes to gold wires (not shown) and further to the push indium
contacts of the PPMS sample puck. Epoxy looks dark under the
microscope because the surface is rough and does not reflect light
normally back into the objective.

44

Figure 3.16: Flow diagram of wiring procedures. Baking procedures are
labeled as red blocks. The epoxy need to be baked just once so there are
two possible routes.

3.5 Contact resistance and 4-probe measurement
Typically there will be a large resistance generated at the junction of two different
materials called “contact resistance”, especially in the case of metal and semiconductor
because of the so-called Schotkky barrier. It hinders the measurement of intrinsic
resistance in materials because the magnitude of contact resistance could be much larger
than the intrinsic resistance and we could not separate the two through direct two-probe
measurement. So there are two directions to take for measuring the intrinsic resistance:
lower the contact resistance and minimize its effect by 4-probe measurement.
In order to lower the contact resistance, we could not neglect the fact that the
origin of the contact resistance is usually quite complicated and it differs in different
systems. Here we list some of the common sources of the contact resistance: 1. Bad
adhesion of the junction usually caused by rough/dirty surface. 2. Mismatch of work

45

function. 3. Oxidized/deteriorated surface. The methods of reducing contact resistance
with individual materials will be discussed in the following.
First, the contaminant between the metal electrodes and the flake surface can be
from several sources. The un-developed e-beam resist is the most common one, and the
glue from the exfoliation process is the other. To remove the excess e-beam resist, the
light oxygen plasma mentioned in the e-beam lithography chapter is the best way, but
unfortunately only applicable to materials like Ca3Ru2O7 which is invulnerable to
oxidization. In other cases the tuning of e-beam lithography parameters such as dosage
and development time to reduce the possibility of un-developed resist is a crucial issue.
The same problem happens with glue residue since some residue cannot be removed by
acetone or IPA. Even though there are some studies39 about using diluted roomtemperature RCA clean solution mentioned before (original H2O2:NH4OH:H2O 1:1:5,
diluted 1:1:10 or 1:1:20) and soaking graphene for overnight to remove the photoresist
residue, that is not 100% effective to glue and still have potential oxidation concern
(graphene is relatively chemical-inert so it is free of this problem).
The work function mismatch between the flake and the adhesion layer of metal
electrodes has been briefly mentioned in the lithography section. Especially in metalsemiconductor junction, the large difference of work function makes the junction a
Schottky barrier. Ti (work function = 4.33eV) thin film leads to large contact resistance
up to ~MΊ for typical contacts on Ca3Ru2O7 and has been replaced by Ni (5.04eV) which
is closed to gold (5.1eV) and the contact resistance is minimized.
At last we discuss the degraded surface of flakes which mainly occur in Fe1+yTe
or Te-rich Fe(Se,Te). At the early stage of the FeTe project, the tricks of lowering contact

46

resistance we developed for Ca3Ru2O7 does not work for FeTe at all. We tried optimizing
the lithography process to get very clean development, and make many devices with
different metals for electrode deposition but it always shows insulating behavior at low
temperature. The I-V curves of every two electrodes are very non-linear as shown in the
Figure 3.17.

Figure 3.17: The I-V curve of a large FeTe flake from 300K to 10K. The
low-temperature I-V has a turn-on voltage Vt taken from the crossing of
extension of high-voltage part and the zero current voltage axis.

47

Figure 3.18: The evolution of Vt as a function of temperature for two pairs
of electrodes. The rapid raising of Vt starts from around 100K and
becomes saturated at low temperature.
We first suspect the behavior originates from the Schottky barrier, but the
discrepancy between the thermionic emission model and the observed turn-on voltage
cannot be explained since the Vt saturated at low temperature (Figure 3.18). From the
graph it can be understood as an insulating barrier with finite length formed around 100K,
so the voltage needed to tunnel the barrier will not diverge since the length is fixed. By
fitting the I-V curve we found the length at 10K is about 1.2 nm, which coincide with the
observation with TEM of the surface layer degradation of Fe1+yTe40. To solve this
problem, we use in-situ ion-milling before the metal deposition to remove 2 nm of
surface layer of flakes, and the I-V curve of newly fabricated devices becomes linear at
all temperature range with a contact resistance on the order of 100Ί at room temperature
estimated by the difference of 4-probe and 2-probe measurement in the example in Figure

48

3.19. Other flakes like the 15nm one which is insulating shows the similar result (in
Chapter 5).

Figure 3.19: Contact resistance of a 54 nm Fe1+yTe flake device fabricated
with in-situ ion-milling procedure. Inset: The 4-probe measurement retains
the bulk-like behavior. The R-T curve shows nice metallic behavior
confirming the ohmic contact properties in all temperature range.
Finally we introduce briefly the four-probe measurement. The idea of four-probe
measurement is shown in the Figure 3.20. To measure the intrinsic resistance of the
sample, e.g. RS2, we cannot avoid a contact resistance RC when we use a metal electrode
deposited on the surface of the sample. The solution is applying a larger current IS
through the sample with two outer electrodes and measuring the voltage difference of two
inner electrodes with a smaller current IP. Since the voltage drop will be

V  I P  RC 2  RC 3   ( I P  I S ) RS 2
IP
If I S

RS
RC , V  I R .
S S2

49

(RC, RS here is the order of magnitude of contact and intrinsic resistance.)
Usually the probe current is about the order of pA (e. g. Keithley 2400 we used),
given that our IS is on the order of ÂľA, the 6 orders of difference will be able to eliminate
the effect of contact resistances.

Figure 3.20: Resistor model of 4-probe measurement. RC# stands for the
contact resistance of #th contact and RS# for the intrinsic resistance of the
flake. Blue ellipsoid denotes the flake.

3.6 Ionic liquid gating
Ionic liquid is defined as a kind of organic salt with melting point below room
temperature. Because of the ions flowing freely inside the liquid it usually has pretty high
dielectric constant compared to common gating oxide material such as SiO2 or Al2O3.41
The liquid forms electric double layer on the interface when a potential difference is
applied on both sides and generates very large carrier density on the electrodes in contact
with the liquid. The liquid we used in our experiment is DEME-TFSI (N,N-diethyl-N-(2methoxyethyl)-N-methylammonium

bis(trifluoromethylsulphonyl-imide)),

which

is

commonly used in many EDL (electric double layer) devices. It breaks into DEME + and

50

TFSI- when the voltage is applied on two electrodes in contact of the liquid, and it freezes
below 180K and the induced charge on surface will be frozen as well. The chemical
composition is shown below in Figure 3.21.

Figure 3.21: The chemical composition of DEME+ (top) and TFSI-.
The working cycle depends on the type of the ionic liquid. Take DEME-TFSI as
an example, the liquid freezes at 180K so the electric field has to be applied above the
temperature so the molecules could still move around to form electric double layer on the
gated surface. And usually gate voltage is scanned within positive to negative 6V at this
temperature to see the resistance voltage change as a first observation of gate effect. Then
the gate voltage will be fixed and the device will be cooled down to low temperature to
see if low-temperature properties can be change by this gating voltage.
To apply a reasonable amount of ionic liquid on the device, I built up a setup for
precise dipping purpose by using existing material in the lab. First I attached a flash light
on one of the eye pieces of a biological microscope which has only bottom light source so
that it can serve as a one-eyed confocal microscope. To record the images I used cell-

51

phone camera and took the image through the eyepiece. Second, I sharpened a toothpick
and taped it on a wedge of microscope slides combined with an optical micro-stage which
can move in three dimensions with micrometer precision. The tip of the toothpick could
pick up very little ionic liquid and move under the microscope so that the liquid can be
dropped at desired location by lowering the tip carefully toward the surface. When the tip
gets very closed I would pull it up and see if there is some liquid left on the surface, this
usually takes a couple times of practice to get a desired size of droplet and precise
location.

Figure 3.22: The idea of ionic liquid gating (cross-section of a IL-gating
device). The yellow pad is the gold pad deposited on the substrate as the
electrode for the device and the large gate pad near the sample. When the
voltage is turned on, there will be high charge density induced on the
surface of the gate electrode and the device.

52

Figure 3.23: Microscope setup. Left: The microscope with flashlight
attached to it. Right: A image seen from the other eyepiece, taken by cell
phone camera. The little droplets are dips of ionic liquid dropped on
surface for practice.

Figure 3.24: The toothpick attached on the optical microstage used in
dipping ionic liquid on the chip.

53

Figure 3.25: An ionic liquid drop on Ca2RuO4 device. This is a nice
example because we want the area ratio covered by ionic liquid of device
to gate electrode to be very small, and the drop position and size is just
desired.
SiOx rings are put onto the device as the optional step for ionic liquid application.
The purpose of this ring is to block the contact between large gold pad nearby that shares
most of the voltage drop. A ring pattern with inner diameter ~100 Âľm and outer diameter
~ 200 Âľm is written on a device rounding the flake and an ionic liquid gate inside. When
the devices are made they are visible in the SEM so the last step does not require an
alignment. The SiOx is deposited by thermal evaporation in another chamber. The
finished device with ionic liquid added is in Figure 3.26. Later we found it is not efficient
to put such a ring compares to designing the electrodes not using large gold pads nearby
except the gate electrode. Thus it was not used in devices fabricated later on. To avoid the
contact pad, we just pull the electrode far away in the lithography process.

54

Figure 3.26: SiOx ring surrounded device with ionic liquid dropped. The
SiOx ring is transparent, covering the nearby wires and circuit. The ionic
liquid drop changes the optical length so the center of device and the flake
are blurred, not focused simutaneously with other parts of the image.
The key problem in the ionic liquid gating is the leakage current. Ideally ionic
liquid should be insulating and work as a dielectric media, but just like other dielectric,
the current will leak through with high electric field. The problem is more serious at high
temperature. At room temperature, only 3 volts of gating voltage can create a significant
leakage current as seen in the Figure 3.27. The current induces a voltage drop between
the electrodes of the device which interferes with the 4-probe measurement of the
intrinsic resistance. The Figure 3.28 shows a measurement on the same sample using
different excitation current with the leakage current induced voltage drop, confirming the
drop is not an intrinsic effect of the flake resistance since it shrinks with larger excitation
current.

55

Figure 3.27: Leakage current as a function of gate voltage at room
temperature.

Figure 3.28: Gate voltage influence on 4-probe resistance measurement.
The limitation on both voltage and current throught the drain-source
electrodes is applied. The device is a 40nm Ca2RuO4 thin flake.
If a large voltage is applied on the ionic liquid, eventually the liquid will break
down. In Figure 3.29 we saw two examples of the breakdown with positive and negative
voltage. As mentioned above with only 3 Volts there is already significant leakage
current. If 6 Volts is applied at room temperature the liquid will break down for sure and

56

cause a large voltage drop as observed in the resistance drop in the Figure. The difference
in resistance drop from the two graphs is again from the difference of excitation current.
After the breakdown, the device is usually destroyed by the electro-etching effect and
both flakes and electrodes are etched away (Figure 3.30).

Figure 3.29: Breakdown of the ionic liquid at room temperature. Both of
the devices are Ca2RuO4 thin flakes.

Figure 3.30: Etched surface by ionic liquid after the measurement. The
liquid dispersed all over the surface.

57

Based on the experience, we prepared a batch of test devices using exfoliated
graphene to examine the gate effect. Graphene as a well-known material is good for the
test, since the exfoliation and device fabrication is relatively easier than all of materials
studied in our projects, and also the thickness is below 1nm which makes carrier doping
very effective. We successfully retained the typical gate effect of graphene by scanning
the voltage at 200K right above the freezing point of the liquid. The results are shown in
Figure 3.31 and Figure 3.32. The R-V curves exhibit a peak in resistance because
graphene has a Dirac-cone band structure and allows electron or hole-doping depending
on gate voltage. The resistance peak is the charge neutral point (CNP) where ideally there
is no free electron or hole so the resistance reaches the maximum. The peak position is
shifted after each scan probably because of the gating voltage moved the local charged
impurities and changed the CNP. We have also tried several scan speed and as expected,
slower scan speed leads to smaller peak shift and the peak shape is also sharper. After
several rounds we froze the liquid by cooling down to 160K and warmed up to 200K
again, tested with scan speed = 8.8mV/s again and see similar results. We have confirmed
the similar result on other graphene devices as well.

58

Figure 3.31: Graphene resistance as a function of gate voltage. The curve
is scanned at various scanning speeds scanned at 200K. Scanning route is
always 0V, 3V, -3V, 0V. The resistance at high gate voltage reaches about
25% of the peak value.

59

Figure 3.32: The image of graphene device and the R-V curve. The device
shows sharper charge neutral point peak but it is finally destroyed at low
temperature because the frozen ionic liquid broke.
After the gate voltage scanning, we tried to cool the device down to lower
temperature. But the frozen ionic liquid always broke and destroyed the device. Slow
cooling rate (<1K/min) is used but the frozen liquid still breaks. Somehow on FeSe0.5Te0.5

and some Ca2RuO4 flakes we don’t have this problem. Currently we still cannot

control this problem at low temperature and it may be the most frustrating problem in
using ionic liquid. I have a suspect that is correlated to the substrate surface properties
such as hydrophilic or hydrophobic since I have treated the graphene surface with diluted
RCA clean solution and it will make the surface hydrophilic, or the moisture absorbed on
the surface but has not tried it out. Ionic liquid is super sensitive to the water moisture
and that could be the reason. For the future plan it may be necessary to examine this

60

problem before re-starting projects related to ionic liquid. More information about the
future plan can be found in Chapter 6.

61

Chapter 4
Calcium ruthenate thin flake study
From the method established in previous chapters, we exfoliated calcium ruthenate flakes
for the first time and fabricated 4-probe devices with various thickness using mechanical
exfoliation method. In this chapter we mainly focus on the study of nanostructured
Ca3Ru2O7. Through the (magneto)transport measurement on exfoliated Ca3Ru2O7
nanoflakes, we find that thin flakes below 10 nm become more insulating and the
magnetoresistance behavior is drastically different from that of thicker samples, implying
a change in both electronic state and magnetic order. Such a change may be associated
with the enhanced RuO6 octahedral distortion and weakening of ferromagnetic interaction
due to the loss of oxygen at the cleaved surfaces, whose contribution to the physical
properties becomes more significant as the surface-to-volume ratio increases in the
thinner flakes. In section 4.5, we briefly discuss our preliminary study on Ca2RuO4
nanoflakes and the potential electric gating effect using ionic liquid. These studies not
only bring another dimensionality to control the physical properties of ruthenates, but
also in general establish the marriage between correlated electron systems and two
dimensional materials.

4.1 Bulk material properties of Ca3Ru2O7
The strong electron-electron Coulomb interaction and the intimate interplay
between spin, charge, lattice, and orbital degrees of freedom in transition-metal oxides
(TMOs) give rise to a rich variety of interesting electronic and magnetic phenomena 42 43.
In general, the strong magnetic and electronic instabilities in these systems render their
properties very susceptible to subtle non-thermal stimuli, such as chemical doping,

62

magnetic field, pressure, etc., which can readily tip the balance between competing
interactions and generate new quantum phases and complex phase diagrams.
Recently, there have been increasing efforts in studying nanostructured correlated
TMOs which enable ones to probe the exciting collective phenomena arising from
quantum confinement, to tune materials properties via electrostatic gating effect, as well
as to probe the non-equilibrium physics of correlated TMOs stemming from the
competition of various interactions44. Thus far, most of studies on nanostructured TMOs
have been focused on oxide nanowires or epitaxial oxide thin films. However, for TMOs
with layer structure, earlier attempts on thin film growth often led to intergrowth of the
materials which change the materials properties in unintentional ways45. In contrast, the
mechanical exfoliation approach not only can circumvent the intergrowth problem, but
also offers the feasibility to deposit the exfoliated oxide nanosheets on a variety of
templates for potential device applications. Although research along this line is still in its
infancy, early attempts have shown some success. For instance, recent Raman
spectroscopy on the nanosheet structure of high-Tc superconductor, Bi2Sr2CaDyxCa1xCu2O8+δ,

reveals a significant change in the strength of magnetic exchange interaction

compared with the bulk material46; and transport measurements on a monolayer of
superconducting Bi2Sr2CaCu2Ox show an insulating behavior22. Thus, the cleaved oxide
nanosheets provide a unique platform to investigate physical properties of the correlated
oxides at nanoscale.
Ruthenates with Ruddlesden-Popper (RP) layer structure, (Sr,Ca)n+1RunO3n+1,
represent an exciting class of 4d TMOs, which display a broad spectrum of perplexing
ground state properties47

48 49

. With reduced n (e.g., n = 1, or 2), although the associated

63

RP materials can generally be regarded as quasi-two dimensional systems, the correlation
between layers along the c-axis still plays a non-negligible role in determining physical
properties of bulk materials. For example, the observation of c-axis coherent transport in
bulk Ca3Ru2O7, Sr2RuO4 and Sr3Ru2O7 at low temperatures implies that the charge
carriers are not fully confined to the layer planes50 51 52. Besides, the interbilayer coupling
in Ca3Ru2O7, albeit weak, is crucial in leading to the antiferromagnetic transition
resulting from the coupling between bilayers53

54 55 56

. Thus, provided the nanosheet

structures of RP ruthenates with the reduced dimensionality along the c-axis, one may
anticipate the physical properties of ruthenates to change dramatically. In addition, due to
the symmetry breaking and the absence of constraint at the surface and interface, the
structural distortion of RuO6 octahedron may be markedly altered considering the
enhanced surface-to-volume ratio in the nanosheet structure, which is also expected to
strongly modify magnetic and electronic properties in nanosheets as a result of the strong
correlation between lattice and other degrees of freedom in RP ruthenates.
The material of interest in this work is the bilayer ruthenate (n = 2),
Ca3Ru2O7.(Figure 4.1) Ca3Ru2O7 exhibits two phase transitions: one at TN = 56 K
corresponding to the paramagnetic-antiferromagnetic phase transition, and the other at
TMIT = 48 K referring to a metal-insulator-like phase transition that is accompanied by the
change of the magnetic moment direction from the a-axis at high temperature to the baxis at low temperature53

57 55

(Figure 4.2). Such magnetic anisotropy is known to arise

from spin-orbit coupling via lattice, which is intimately related to the structural
anisotropy with different degree of RuO6 octahedral titling along the a-axis and b-axis (as
seen in Figure 4.1)58. And it was proposed that the change of spin anisotropy below TMIT

64

is essential to the change of the electronic band structure59. The metallicity of Ca3Ru2O7
reappears below 30 K50, originating from a small Fermi surface pocket surviving through
the MIT60. And this material possesses strong magnetic and electronic instabilities, as
evidenced by their susceptibility to chemical substitution, pressure, and magnetic field
reported previously61 56 62 55 63 64.

Figure 4.1: Schematics of RuO6 octahedral distortion in Ca3Ru2O7. ϕ and θ
are defined as rotation and tilting angles.(From J. Peng Phys. Rev. B
(2010))

65

Figure 4.2: The metal-insulator transition and AFM transition of bulk
Ca3Ru2O7. Left: the resistivity along a and c axis and the magnetic
susceptbility gives the TMIT = 48K and TN = 56K. Right: The magnetic
structure of. Ca3Ru2O7 (from Yoshida, Y. et al.,(2004) Phys. Rev. B and
Wei Bao et al., Phys. Rev. Lett. (2008)).

4.2 MIT in thin flakes
Ca3Ru2O7 single crystals were grown using floating zone method. By gluing both
sides (ab plane) of the crystal on two pieces of microslides and plying open the crystal,
flakes are generated on the newly broken surface which are then transferred to the prepatterned SiO2 (285 nm)/Si substrates using Scotch tape method. Oxygen plasma
cleaning is performed to remove excess glue from the tape. Flakes with various thickness
ranging from ~3 nm up to ~100 nm were located under an optical microscope and then
the topology was measured using the atomic force microscope. Nanoflake devices with

66

four-terminal electrode pattern were fabricated using the standard e-beam lithography
procedure (inset of Figure 4.3). Ohmic contacts were formed by depositing 10 nm Ni
(used as the adhesive layer) and 100 nm Au electrodes by DC sputtering. Electronic
transport measurements were carried out using Quantum Design PPMS cryostat. Note
that an earlier study of exfoliated Ca3Ru2O7 can only achieve flakes as thin as 300 nm38.
Details of the procedures are described in Chapter 2 and 3.
Figure 4.3 presents the atomic force microscopy images of several representative
flakes of different thickness. All the flakes show atomically flat surface, as illustrated by
the line profile scan done on a 9 nm flake sample shown in the inset. It is worth noting
that the thinnest flake with sizeable lateral dimension that we achieved is ~ 3 nm depicted,
which corresponds to 1.5 unit cell thick (note the lattice parameter c = 1.956 nm).

67

Figure 4.3: AFM images of exfoliated Ca3Ru2O7 flakes. Representative
atomic force microscopy images of the exfoliated Ca3Ru2O7 flakes with
different thickness. Insets show the scanning electron microscope image
with false color of a fabricated flake device with 4 gold electrodes and the
cross section line profile of the 9 nm flake
Figure 4.4 shows the resistivity (ρ) as a function of temperature (T) measured on
devices with various flake thickness. For the 90 nm flake, the ρ(T) curve behaves
identical to what is observed in the bulk counterpart50: a large drop in resistivity occurs at
TN = 56 K followed by a sudden (first-order type) increase at TMIT = 48 K, and eventually
the metallic resistivity behavior reappears below 26 K. With decreasing the flake
thickness (25 nm and 13 nm), while the overall MIT and the reentrance of metallicity
behaviors are preserved, the first-order type MIT is smeared out. As the flake thickness

68

further decreases, TMIT observed in the device of 9 nm and 8 nm flakes clearly shifts to 33
K and 30 K respectively without the re-entrance of metallicity at lower temperature.
Interestingly, for the thinnest measurable flake (~ 7 nm), no transition is observed and the
device shows an insulating behavior all the way to the lowest temperature, with the
resistivity two orders of magnitude higher than that for thicker flakes at T = 2 K. And
flakes thinner than 7 nm are very resistive and not measurable.

Figure 4.4: R-T curves of Ca3Ru2O7 exfoliated flakes. Thickness is labeled
in each graph and the dashed line indicates the TMIT in bulk. Three thicker
flakes on the left follows the TMIT but the other three thinner flakes on the
right the TMIT deviate significantly.
There are a couple of scenarios that may happen to the structure as the flake
becomes thinner and thinner. On the one hand, it is known that compared to Sr-based
counterparts, Ca-based ruthenates tend to have much larger structural distortion due to
the smaller ionic radius of Ca2+ than Sr2+

65

. Thus, naively, one may anticipate that the

69

RuO6 octahedral distortion might be (partially) relaxed due to the absence of constraint
on the surface and interface. Provided this was the case, however, the flakes should
become more metallic as the thickness decreases, which is opposite to the experimental
observation. On the other hand, earlier scanning probe microscopy and low-energy
electron diffraction studies on Sr2RuO4, which has no structural distortion in bulk, found
that the RuO6 octahedra on the cleaved surface layers are distorted with an octahedral
rotation around the c-axis by about 9 degree66. Similarly, larger RuO6 octahedral
distortion was also observed on cleaved surface of Ca1.9Sr0.1RuO4, which gives rise to
lower TMIT on the surface compared to the bulk67. Therefore, as the exfoliated nanoflake
gets thinner, its RuO6 octahedral distortion might become more severe. A natural
question then arises: What drives the larger structural distortion on the surface of
ruthenates? One plausible mechanism is the loss of oxygen on the cleaved surfaces. It is
highly likely that the apical oxygen atoms in the RuO6 octahedra are (partially) lost on
the cleaved surface, which leads to oxygen deficient on the surface layer. Recently, it has
been found that oxygen deficiency in Sr3Ru2O7-δ and Ca2RuO4-δ give rises to a decrease
in the c-lattice parameter but an increase in the in-plane lattice parameters, thus resulting
in larger RuO6 distortion compared to the stoichiometric compounds68 69.
Accordingly, the enhanced structural distortion may account for the decrease of
TMIT in the thinner Ca3Ru2O7 nanoflakes shown in Figure 4.4, a feature similar to the
reduced TMIT on the cleaved surface of Ca1.9Sr0.1RuO4 compared to the bulk counterpart67.
Furthermore, the enhanced structural distortion in thin nanoflakes is anticipated to reduce
the electronic bandwidth as well, leading to the tendency of opening the bandgap. Thus,
as the nanoflake gets thinner and thinner such that the surface-to-volume ratio become

70

larger, the electronic state arising from the cleaved surface and interface layers becomes
dominant, resulting in the observed more insulating transport behavior for flakes thinner
than 8 nm. Additionally, the reduced bandwidth, i.e., the decrease of the carrier hopping
energy (t) tends to soften the antiferromagnetic phase transition (TN ~ t2/U, U: Coulomb
energy). For instance, TN is ~ 42 K in the 8 nm flake compared to 56 K in the bulk
sample. It is worth noting that the partial reduction of Ru4+ to Ru3+ due to the loss of
oxygen atom can weaken magnetic exchange interactions and contribute to the decrease
of TN too.

4.3 Magnetoresistance and antiferromagnetic transition
In addition to the insulating behavior at low temperature, thinner flakes also
exhibit markedly different magnetic anisotropy and magnetoresistance (MR, which is
defined as

𝜌(𝐻)−𝜌(0)
𝜌(0)

) behavior from those of thicker films or bulk. Figure 4.5 shows the

comparison of the in-plane angle dependence of MR for two flakes of 90 nm and 8 nm at
various temperatures. The measurement field is 9 T, and the angle (θ) is relative to the
chip edge. For the 90 nm flakes, the MR(θ) exhibits two-fold symmetry with troughs
positioning around 220° and 40° at T = 60 K but with troughs around 130° and 310° at T
= 2 K, and the MR(θ) curves at these two temperatures can be roughly fitted using a sinesquared function. Such an angular dependent MR behavior mainly stems from spinscattering associated with the field-dependent spin structure, and is consistent with the
magnetic anisotropy determined from neutron diffraction studies on bulk Ca3Ru2O7 with
the easy-axis along the a-axis and b-axis (see the labels in Fig. 3) above and below TMIT,
respectively55. Note that the MR behavior at T = 20 K and 40 K cannot be fitted using a
sine-squared function, a feature similar to the out-of-plane MR behavior observed at this

71

temperature regime70, which was attributed to the angle-tuned spin-valve effect that is
affected by the change of magnetic anisotropy when T approaches TMIT. For the 8 nm
flake sample, besides the smaller TN (~ 42 K) and TMIT (~ 30 K) as discussed above,
interestingly, the angular dependence of MR(θ) curves display no much variation below
and above TMIT. Similar feature is also observed in the 9 nm flake sample. As discussed
previously, the magnetic anisotropy in this compound arises from spin-orbit coupling.
This implies that, unlike thick or bulk samples where spin-orbit coupling changes below
TMIT at which the c-axis compresses and b-axis expands71, there is no drastic change in
spin-orbit coupling as a function of temperature in thinner flakes, potentially associated
with the absence of first-order like structural change as inferred from the transport
measurements shown in Figure 4.4. Furthermore, owing to the strong spin-charge
coupling in ruthenates, the distinction in MR(θ) behavior in the 90 nm and 8 nm flakes
suggests that their ground states at low temperature possess different magnetic structures.
By comparing the MR(θ) curves measured at T = 60 K depicted in Figure 4.5(d,h), we
tentatively denote a and b-axes for the 8nm flake as labeled in Figure 4.5(h).

72

Figure 4.5: The MR comparison as a function of angle of in-plane
magnetic field. The left graphs are taken from the 90nm thick flake and
the right figure from the 8 nm thin flake. The MR is defined as the
percentage change compared with the resistance measured with zero
magnetic field. Inset: geometry of the chip (crystal plane) and the rotation
of in-plane magnetic field.
The alteration in the magnetic structures as the flakes becomes thinner is
supported by the MR(H) measurements. Figure 4.6 presents the MR(H) data measured at
various temperatures with the magnetic field applied along both a and b-axes for 13 nm
and 8 nm flakes. Note that a and b-axes are determined via MR(θ) measurements
discussed above. For the 13 nm flake, the overall MR behavior is similar to that observed
in the bulk sample50. As shown in Fig. 4(a), with the magnetic field up to 8 T applied

73

along the a-axis, which corresponds to the magnetic hard-axis, the flake exhibits positive
MR at T = 2 K and negative MR at higher temperatures. And with the field applied along
the b-axis, i.e., the magnetic easy-axis, a field-induced phase transition is clearly
observed, as inferred from the hysteresis loop in MR around 6 T. The similarity in
MR(H) data of 13 nm flake to those of the bulk55

63

implies that the magnetic structures

observed in the bulk highly likely persists in the nanoflakes down 13 nm thick. In sharp
contrast, the 8 nm flake exhibits remarkably different MR(H) behavior. For H // a (Figure
4.6(c)), overall the sample displays negative MR at all temperatures measured except that
there show a small positive MR around 1 T at T = 30 K which is close to the TMIT. The
MR magnitude reaches about 90% at 2 K and 9 T, which then decreases monotonically
with increasing temperature, which is distinct from the observation on the 13 nm flake
shown in Figure 4.6(a). For H // b (Figure 4.6(d)), the MR behavior is also different from
that observed in the 13 nm flake. At T = 2 K, the MR shows a non-monotonic
dependence on the magnetic field, with a dip around 3 T. With increasing temperature,
the MR value gradually changes the sign and the dip feature in MR disappears at T = 30
K with the MR value at 9 T reaching the maximum followed by a decrease at higher
temperature. Again, the drastic distinction in MR between 13 nm and 8 nm flakes reflects
the change in magnetic structure as the flake thickness decreases. Very interestingly, it is
worth noting that the MR phenomena shown in Figure 4.6(c,d) are quite similar to the
MR observed in the bulk Fe-doped Ca3Ru2O772, which possesses a coexistence of
commensurate and incommensurate spin structure at low temperature in the absence of
magnetic field62 and a field-induced incommensurate-to-commensurate (canted
antiferromagnet) magnetic phase transition73.

74

Finally, we briefly comment the mechanism of the change in the magnetic
structure of thinner flakes. Firstly, as discussed previously, the RuO6 octahedron is
compressed along the c-axis in the presence of oxygen loss at the cleaved surfaces, which
is expected to change the magnetic structure due to the strong spin-lattice coupling in
rutehnates. However, the octahedral compression tends to result in enhanced
antiferromagnetic

superexchange

interactions

that

can

give

rise

to

G-type

antiferromagnetic structure as reported in both Ca2RuO474 and Ti-doped Ca3Ru2O775.
Secondly, the partial reduction of Ru4+ to Ru3+ due to the loss of oxygen atom can
weaken the original ferromagnetic interaction and locally induce the antiferromagnetic
interaction between Ru4+ and Ru3+, which may consequently lead to an incommensurate
magnetic structure. Note that incommensurate magnetic structure has been observed in
Fe-doped Ca3Ru2O7 due to the competing interactions in combination with magnetic
anisotropy and the Dzyaloshinskii-Moriya interaction62. Although challenging, future
studies using microprobe resonant x-ray diffraction are very desirable to directly probe
the nature of the magnetic structure of the nanoflakes as a function of thickness.

75

Figure 4.6: MR of 13/8 nm flake as a function of field strength at various
temperatures. Left: 13nm flake with the field on b-axis. Bottom left: 13nm
flake, field on a-axis. Right: 8 nm flake, field on “b-axis”. Bottom right: 8
nm flake, field on “a-axis”.
It is worth mentioning that the MR behavior in the 9 nm flake (Figure 4.7) is very
similar to that of 8 nm flake but with smaller amplitude. The result reinforces the
reproducibility of the feature and also the thickness-related origin of the magnetic
structure.

76

Figure 4.7: The magnetoresistance of 9nm Ca3Ru2O7 flake. Left: B
parallel to b-axis. Right: B parallel to a-axis.
We also studied in detail on the only measurable insulating flake with thickness =
7 nm, although thus far we could not completely understand our observation on this
sample. Figure 4.8 plotted with resistivity in log scale with respect to T or 1/T, which
does not follow a simple linear trend in any temperature region. This indicates that the
insulating behavior cannot be simply ascribed to thermal activated semiconducting gap
(   exp(

(   exp(


) ) or Mott’s variable range hopping regime in any dimensions
k BT

 d11
) ) leaving the detailed mechanism of conductance in thinner flakes to be
k BT

explored more in the future.
Additionally, the MR of the 7 nm flake shown in Figure 4.9 also behaves
differently from that of 8 nm and 9 nm flakes. The incommensurate and commensurate
magnetic order we proposed in the previous two flakes might be modified further with
larger distortion by increased surface-bulk ratio. But at this point there are not enough
clues to give a definite picture of the origin of the behavior.

77

Figure 4.8: The R-T curves of the 7nm flake. Left: The metal-insulator
transition disappeared completely and the low-temperature behavior is
very insulating. Right: The resistance in log-scale is not linear versus 1/T.

Figure 4.9: The magnetoresistance of the 7nm flake. Left: The hard-axis
positive MR is enhanced at low-temperature. Right: The easy-axis
negative MR is also larger at lower temperature.

4.4 Lateral dimension effect?
In some of the thicker flakes (>13nm) we observe the re-entrance of the metallic
state at low temperature (<30K) similar to the bulk, however, strangely some other flakes
(with comparable thickness but narrower) show insulating behavior below TMIT and never

78

return to metallic again. Thus such an insulating in the narrower flakes should not be
ascribed the thickness effect for the very thin flakes below 10nm as we discussed in
previous sections. In Figure 4.10 we plotted the R-T curve comparison of two set of
flakes for clearance. Two flakes with thickness near 50nm and two near 30nm confirm
that the insulating behavior happens in narrow flakes regardless of the thickness. This is
interesting because the width of the “narrow” flakes is still several hundreds of
nanometers, which is still pretty wide compared to the nanowires people study about the
quantized conductance and 1D Peierls transition. We will discuss several possible reasons
as follows.

Figure 4.10: The comparison of Ca3Ru2O7 flakes with minor difference in
thickness but dramatic difference in low-temperature R-T behavior
difference. Left: the comparison of 54nm (wide) and 51nm (narrow) flakes.
Inset: the AFM images of the two flakes. Right: The same comparison
between 25nm and 30nm flakes.
An intuitive guess would be the size of the nanowire that quantizes the
conductance of the electrons and prevents the re-entrance of the metallic state. But the

79

calculation in detail will show that is not possible. Consider that the conductance in near
1D constriction is given:
G=

g s e2
h



 Wk F

  
m 1


 m



Here gs is the degeneracy of spin = 2, θ is the step function, W is the width of
constriction and kF is the Fermi wavenumber. From the literature the kF of Ca3Ru2O7 at
low temperature is estimated to be ~3.2•108/m, which means the unit conductance
corresponds to W ~ 10nm. Take a 30nm flake for example, the width of a typical wireshaped flake is about 200 nm, the conductance of the constriction will be 2•20•e2/h =
0.001545, equivalent to 647Ί in resistance. But the value is calculated for a single layer,
in Ca3Ru2O7 conduction is through quasi-2D bilayers at the low temperature and a 30 nm
flake roughly have 30 bilayers. Consider 30 bilayers connected in parallel, the lowest
possible resistance value is ~21Ί which is not going to limit the behavior of the flake
since the lowest resistance in the actual 30 nm flake at the TMIT exceeds 500Ί.
Even though the quantized conductance seems unlikely to be the reason of the
absence of low-temperature metallic state, the calculation does provide some ideas about
the effect of width in the constriction. In the literature the origin of the metallic state in
bulk Ca3Ru2O7 is attributed to the incomplete gapping of Fermi surface arising from a
small difference in hopping distances on a and b axes. We are not sure if the anisotropy
will be removed in the lateral constriction, but the dimension seems to be too long
compared to the thickness to talk about the surface effect. Another possibility left is the
phase coherence length of electrons in the system that could probably exceed the width,
making the interference in the width direction significant. The interaction might further

80

gap the Fermi surface and prevent the return of metallic state, but for the present we have
too little clue to propose a practical model and examine this observation.

4.5 MIT in Ca2RuO4 flakes
The motivation on the study of Ca2RuO4 arises from the structural similarity to
Ca3Ru2O7. The Ca2RuO4 has a single RuO2 layer spaced by Ca2+ ions (see Figure 4.11)
instead of bilayers in Ca3Ru2O7. Thus an interesting question would be: Will the
thickness effect be different in Ca2RuO4 and Ca3Ru2O7 because of the bilayer/single layer?
Indeed we would expect even stronger thickness effect in Ca2RuO4 because the intrabilayer bonding is absent so it could possibly be easier for the surface monolayer to relax
the distortion imposed by Jahn-Teller effect. Also the effect can be observed by transport
measurement because it is also a Mott-insulator with high TMIT (357K)76 above the room
temperature with an order of magnitude of resistivity change. The transition accompanies
with the change of lattice constant77 so it occasionally break the crystal. As an possible
effect of substrate mismatch strain or decrease in crystallinity, previous studies of the
epitaxial thin films often observe broader transition at lower temperature78 or the
transition completely disappeared79. After we established the exfoliation method on the
Ca3Ru2O7 crystal, we tried to apply this method to the Ca2RuO4 crystals as well. And the
transport measurement on the 60 nm flake indicates that the MIT transition is preserved.

81

Figure 4.11: Ca2RuO4 crystal structure. The unit cell (left), b-axis view
(top right) and c-axis view (bottom right) are presented. The crystal has
single layer RuO6 structure instead of bi-layer. The octahedra are still
rotated and tilted because of the Jahn-Teller effect.
It turns out that it is much more challenging to obtain Ca2RuO4 thin flakes
compared to Ca3Ru2O7 because of the very small Ca2RuO4 crystals that we have. We
made the 60 nm flake device (Figure 4.12 (b)) and measured the transport (Figure 4.13)
at high temperature. The transition is very similar like the bulk crystal confirming the
quality of exfoliated flake is comparable with the bulk single crystal without the strain
effect in the epitaxial thin films even though the flake is not quite flat. It is worth noting
that the resistivity of the flake is much smaller than the bulk. The literature76 reports that
the resistivity of the crystal is on the order of 100 Ω•cm at room temperature which is not
very different from our bulk measurement. But the resistivity of the flake is on the order
of 0.1 Ω•cm which is 3 orders smaller that cannot be explained by geometrical effect or

82

any mistake in estimation of the shape. We observed low resistivity on most of flakes but
they are not very consistent with each other in the R-T. A more systematic study of this
material is needed for a conclusion.

Figure 4.12: Ca2RuO4 images of devices. (a) The AFM image and crosssection of the white line of the 40nm flake showing the flake is a flat
narrow wire. Inset: the image of fabricated device. (b) The AFM image
and cross-section of the 60nm showing the flake is slightly tilted.

83

Figure 4.13: The comparison of R-T curve of the bulk and 60nm flake
around the transition temperature. The red curve corresponds to the flake
and the resistance value in kiloohms is on the left, and the black bulk
curve is on the right.

4.6 Summary
In summary, we report the electronic transport properties of mechanically
exfoliated Ca3Ru2O7 nanoflakes as a function of thickness. We show that thin flakes
below 10 nm become more insulating and the magnetoresistance behavior is drastically
different from that of the thicker samples, implying a change in both electronic state and
magnetic order. Such a change may be associated with the enhancement in RuO6
octahedral distortion and weakening of ferromagnetic interaction due to the loss of
oxygen at the cleaved surfaces, whose contribution to the physical properties becomes

84

more significant as the surface-to-volume ratio increases in the thinner flakes. This study
not only brings another dimensionality to control the physical properties of ruthenates,
but also in general establishes the marriage between correlated electron systems and twodimensional materials.
There are two side projects with great potential of further discoveries. One is the
lateral dimension constriction effect on the low-temperature metallic state in Ca3Ru2O7
nanoflakes which cannot be explained simply by the quantization of conductance based
on the Fermi wavenumber estimated. The other one is the study on exfoliated Ca2RuO4
nanoflakes which needs more samples for examination. This research has proven the
possibility of exfoliation of layer-structured material with tight interlayer binding
showing dramatic changes of the physical properties of thin flakes. The result may bring
more understanding of the intrinsic properties related to the dimensionality in correlated
systems, motivating further studies of the combination of correlated systems and 2D
materials.

85

Chapter 5
Properties of Fe1+yTe thin flakes
In this chapter we are going to discuss the man result from exfoliated Fe1+yTe
flakes. The material properties and a brief review of Fe(Se,Te) system will be given in
Section 5.1. Characterization of Fe1+yTe flakes and devices are in Section 5.2. In Section
5.3 we will discuss the broadened transition and superconductor-like feature observed in
flakes with intermediate thickness. In very thin flakes, the insulating behavior and
enhanced magnetoresistance are described within variable range hopping (VRH) regime
in Section 5.4. And prior to the summary, we discuss a mini side project in Section 5.5
about the superconducting FeSe0.5Te0.5 flake and the application of ionic liquid gating.

5.1 Introduction:
Iron-based superconductors have been intensely researched since the discovery of
superconductivity in iron pnictides LaOFeP 80 4, mainly stimulated by the close
correlation between the superconductivity and the magnetic correlation. Among the
various types of iron pnictide and ion chalcogenides81 82 83, including 1111 (RFeAsO, R =
La, Ce …), 122 (AFe2As2, A = Ba, Sr, Ca and K) , 111 (AFeAs, A = Li, Na), and 11
(FeTe1-xSex) families, the FeTe1-xSex series possesses the simplest crystal structure
composed of edge-sharing Fe(Te,Se)4 tetrahedra layers (Figure 5.1)84 85 27 86 and attracts a
lot of attention87. At ambient pressure, one of the end members, FeTe, exhibits
antiferromagnetic metallic ground state with the Neel temperature TN = 65 K ~75 K
(which occurs concurrently with the tetragonal-orthogonal structure phase transition)27,85;

86

the other end member, FeSe, becomes superconducting below the transition temperature
Tc ~8 K without the presence of preceding magnetic phases27,84-86, which arises from the
increase of Fe-Se-Fe bond angle of FeSe compared to that of FeTe. Various approaches,
such as chemical substitution (Figure 5.2)88,89,90,91 and pressure 92, have been applied to
bulk FeTe1-xSex in order to enhance the superconducting phase transition temperature as
well as to better understand the mechanism of the unconventional superconductivity in
the iron based systems.

Figure 5.1. The structure of Fe1+yTe. (a) unit cell of Fe1+yTe. The Fe atoms
sit in the center of the layers, surrounded by Te atoms forming FeTe4
tetragons. The empty sphere with a colored section is the site of exces iron.
(b) Fe1+yTe layer structures. (c) top view of the crystal. Fe atoms show
checkboard patterns. (d) a-axis view of the crystal. The bonding angle of
Fe-Te-Fe which will be a key parameter in thin flakes is labeled θ here.

87

Figure 5.2: Phase diagram of Fe(Te1-xSex). Small number of Se doping
breaks the AFM state and leads to superconductivity. (From T. J. Liu et al.,
Nat. Mater. 9, 716 (2010))

Figure 5.3: The bulk properties of Fe1+yTe crystal. (a) The inverse χ curves
indicate a clear AFM transition and the Curie-Weiss linear asymptotic at
high temperature. (From I. A. Zaliznyak et al., Phys. Rev. B (2012)) (b)
The resistance-temperature plots of bulk Fe1+yTe crystal show the
semiconductor-metal transition and a sharp drop of resistance.
Recently, significant efforts have been invested on FeSe in the epitaxial thin film
form as well as in the mechanically-exfoliated thin flake form, which have not only
unraveled novel phenomena that are distinct from the physical behavior observed in the

88

bulk counterpart, but also provided unprecedented insight into the nature of the
superconductivity in bulk. For instance, it was discovered that Tc of FeSe monolayer
grown on various substrates can be dramatically enhanced93,94,95,96,97,98,99, e.g., Tc ~ 109 K
for FeSe monolayer deposited on SrTiO3 substrates99. Interestingly, beyond the
monolayer of FeSe, the superconductivity gets suppressed in FeSe thin films 100 and even
exhibits an insulator-like behavior for relatively thin films while thicker FeSe films
behave like the bulk sample101. Various mechanisms93-97,101,102,103,104, such as electronphonon coupling and charge transfer from substrates, phonon softening or structural
modification induced by the lattice mismatch between the monolayer and the substrate,
have been put forward to account for the change in Tc. Alternative to thin film growth,
when being exfoliated down to nanometer thick, it is found that exfoliated FeSe thin
flakes (~ 10 nm) still show metallic and superconducting behavior32, in contrast to the
insulating behavior observed in FeSe thin films101. Furthermore, Tc of thin FeSe flakes is
significantly enhanced upon ionic liquid gating, which has been ascribed to the change of
the dominant carriers from holes to electrons32.
In contrast, relatively fewer studies of nanostructured Fe1+yTe have been reported.
It was found that tensile strain of epitaxial Fe1+yTe films softens the first-order magnetic
and structural phase transition observed in bulk (Figure 5.3), followed by an emergence
of superconductivity upon further decreasing temperature due to the strain-induced
increase of Fe-Te-Fe bond angle31. However, Nie et al reported that superconductivity
was only induced in Fe1+yTe films by incorporation of oxygen via low temperature
thermal annealing30. More recently, scanning tunneling microscopy/spectroscopy study
has revealed interfacial superconductivity in Fe1+yTe monolayer grown on a topological

89

insulator Bi2Te3 which coexists with the same antiferromagnetic order as the one
observed in bulk Fe1+yTe105. These studies raise a fundamental question of what happens
to the electronic and magnetic properties of nanostructured Fe1+yTe, i.e., exfoliated
Fe1+yTe nanoflakes, in the absence of epitaxial strain or interfacial charge transfer.

5.2 Fe1+yTe nanoflake devices
Fe1+yTe (y = 0.15) bulk single crystals were grown using a flux method89.
Nanoflakes of this material were prepared by mechanically exfoliating bulk crystals using
Scotch tape method and then transferred to the pre-patterned SiO2 (285 nm)/Si substrates.
Flakes with various thickness ranging from ~10 nm up to ~100 nm were located under an
optical microscope right after exfoliation, and then the topology was measured using the
atomic force microscope (AFM). To reduce air exposure, samples were covered by spincoated photoresist after AFM measurements such that the typical exposure time did not
exceed 1 hour after exfoliation. Nanoflake devices with four-terminal electrode pattern
were fabricated using the standard e-beam lithography procedure described in Chapter 2.
Ohmic contacts were formed by depositing 10 nm Ni (used as the adhesive layer) and
100 nm Au electrodes by DC sputtering, preceded by an in-situ ion-milling to remove ~ 2
nm of Fe1+yTe over the contact area which may be oxidized due to the air exposure.
Electronic transport measurements were carried out using Quantum Design PPMS
cryostat.

90

Figure 5.4: Images of an 11 nm exfoliated Fe1+yTe flake. (a, b) 100X
Optical images before and after metal electrode deposition. (c, d) Atomic
force image and the cross-section profile of the 11nm flake. The thickness
can be determined within Âą1 nm.

91

Figure 5.5: The images of 15nm Fe1+yTe flake. (a,b) optical images before
and after device fabrication. There are actually four electrodes but the
other current lead is out of the view. (c,d) the AFM images and profile of
cross-section.
Figure 5.4, 5.5 (a-b) display the optical images of 11 and 15 nm exfoliated flakes
with and without patterned electrodes respectively, and Figure 5.4, 5.5(c-d) present the
AFM images and the line scan profiles of the same flake, showing a flat surface topology
with ~ 1nm roughness.
Although it would be nice to have some lattice information through nondestructive optical measurement, but unfortunately Raman spectroscopy in air is not a
good way to probe these Fe1+yTe flakes. We have tried to take the Raman spectra of

92

Fe1+yTe flakes right after the exfoliation, but the peaks of the intrinsic phonons are
always overwhelmed by signals from degraded layers, as the same mentioned in [106]
(Figure 5.6 Right). Figure 5.7 shows the typical Raman spectrum we have for Fe1+yTe
flakes.

Figure 5.6: Raman modes of Fe1+yTe crystal. Left: (a,d) Fe1+yTe Raman
spectra with various geometry where A1g and B1g should both present in
parallel geometry and only B1g will present in Z(X’Y’)-Z geometry. Right:
The spectra of aged or long irradiated crystal shows strong peak at
~150cm-1 and Raman modes are not distinguishable. (From Xia, T. L. et al.
Phys. Rev. B - Condens. Matter Mater. Phys. 79, 1–4 (2009).)

93

Figure 5.7: Raman spectra of a 12nm Fe1+yTe flake. The black curve is
taken with parallel Z(XX)-Z geometry and the red curve is with
orthogonal Z(XY)-Z geometry. The Raman peaks are not observable in
the strong background peak similar to degraded crystal. The peak at
520cm-1 is from Si background.
After device fabrication, I-V measurements between any two probes of the
devices show linear behavior and the resistance values measured between any two
electrodes are scalable with their separation (Figure 5.8), affirming ohmic contacts with
negligible contact resistance. It is worth mentioning that two-probe measurements on
devices without the ion-milling step prior to electrode deposition tend to show non-linear
I-V character below 100 K, which suggests the importance of the in-situ ion-milling
process (details described in Chapter 3).

94

Figure 5.8: The I-V curve of the 15nm flake measured at 10K (without
magnetic field.) The linear curves determine the resistance between each
two probes as depicted on the right. And the resistance between two
current leads is 55.69 kΊ, meaning the contact resistance is on the order of
~100Ί which is negligible to the 4-probe resistance value we measured.

5.3 Superconductor-like intermediate thin flakes
Figure 5.9 depict the measured resistivity (ρ) as a function of temperature (T) of
devices with typical flake thickness. One can see that the thicker flakes (e.g., 84 nm)
shows a sharp semiconducting-metallic feature around 62 K, a feature which is also
observed in bulk Fe1+yTe and corresponds to the paramagnetic-antiferromagnetic phase
transition that occurs simultaneously with tetragonal-orthorhombic structural phase
transition89,107. For the intermediate thickness range (e.g., 25 nm flake), such a
semiconducting-metallic transition becomes broadened, as shown in Figure 5.3(b). In
addition, there is a sharp drop in resistivity occurring below 10K, as seen in the inset.
Such a feature is observed, although weaker, on devices with 27 nm and 30 nm flakes too,
and resembles the onset of superconducting transition found in the FeTe thin films31,30,40

95

or even bulk crystals108. Note that the kink point around 10 K shifts slightly to lower
temperature upon applying 9 T magnetic field, affirming the superconducting transition.

Figure 5.9: R-T curve of thick and intermediate flakes. (a) The R-T curve
of the bulk-like flake (84nm) shows the transition similar to the bulk
material. (b) The R-T curve of intermediate-thickness flake (25nm) shows
broadened transition and a dip at 10K signatures the onset of
superconductivity.
As discussed above, two different scenarios have been proposed to account for
emergence of superconductivity in FeTe: one is due to the change of Fe-Te-Fe bond
angle arising from the tensile strain in thin films31, and the other is due to the oxidization
which changes the valence state of Fe 3d electrons and may suppress the local magnetic
moment and enhance the carrier itinerancy30,108. However, it is likely that the
superconducting-like feature observed in the flakes with intermediate thickness is not
ascribable to aging or oxidation of the samples, as it is present only for devices with
intermediate thickness (e.g. see Figure 5.9 (a)) but absent for the devices with thicker

96

flakes. To further affirm this, we purposely baked the 25nm flake several times including
in-situ and in air at 100°C for an hour and found that the overall features of the ρ(T)
curves did not change after all, except an increase in resistivity (Figure 5.10).

Figure 5.10: The increase of resistance over several baking process. After
the first measurement without baking (blue-dashed line), the sample went
through (1) in-situ 100°C baking for 2 hours (2) in air 100°C baking for 1
hour (3) in-air 200°C for 5 min. and R-T curves were measured after each
process.
Instead, the appearance of the superconducting-like feature in the devices with
flakes of intermediate thickness may presumably associate with the change of crystal
structure. For exfoliated flakes with thin enough thickness, the absence of constraints on
both surface and interface likely relaxes (partially) the structural distortion towards the
ideal FeTe4 tetrahedral structure by enlarging the in-plane lattice parameters. This results
in an enhancement of Fe-Te-Fe bond angle, similar to that observed in the thin films with

97

tensile epitaxial strain31, leading to the broadening/smearing of the structural transition
observed around 60 K and the onset of superconducting-like behavior31,40.

5.4 VRH model in ultra-thin flakes
Interestingly, with further decreasing the thickness, both structural phase
transition and the superconducting-like features are absent and the flakes becomes more
insulating, as shown in Figure 5.11(a) for the 11 nm flake. The evolution of R(T) of
flakes with various thickness is presented in Figure 5.11 (b). The absence of
superconducting-like feature in the thinner flakes originates from the predominant
charge-carrier localization as the dimensionality reduces, which will be discussed next.
The rapid increases in resistivity observed in thinner Fe1+yTe flakes can be
associated with charge-carrier localization due to the excess of Fe, and the transport
behavior can be well captured by the variable range hopping (VRH) model109.

Figure 5.11: R-T curves of thin insulating flakes. (a) The R-T curve of
11nm flake showing the insulating behavior and the absence of both AFM
transition and superconductor-like dip at 10K. (b) The evolution of R-T
behavior in various thickness of Fe1+yTe flakes. R is normalized to
R(300K) and plotted in log-scale to contain the changes in orders.

98

Within this model, the charge-carriers in a highly disordered system hop between
impurity sites with a variable hopping distance instead of fixed distance in normal
crystals, and the electronic conductance can be fitted using the equation
T
T

1

 (T)   0  A exp(( 0 )1 d ) , where T0 is the hopping parameter that is related to the
localization length  of charge-carriers and d is the dimensionality of the sample. To
validate if the VRH mechanism is applicable to Fe1+yTe thin flakes, we fit the
conductance-temperature (T) curve of the 15 nm flake using the aforementioned
equation, as shown in Figure 5.12(a). One can see that the VRH model fits the (T) curve
below 50 K very well, with the fitting parameter d = 2 reflecting the two-dimensional
1

nature of the nanoflakes and T0 = 147 K. Considering that  = 13.8 / kB DT0  2 110,111,
where D is density of states, one can calculate the localization length . Taking the
𝑚∗

effective mass m* = 3me obtained from the ARPES experiments112, 𝐷 = 𝜋ħ2 (neglecting
the valley degeneracy) is calculated to be 1.251015 /eVcm2, which gives  = 9.34 nm.
Indeed there are three different mass enhancement on different carrier pockets, and 3me is
the one estimated with the hole pocket. In bulk FeTe the two types of carriers coexist in
the system, while below the transition electron-type carriers get largely enhanced and
drive the system metallic40, but at low temperature we did not observe this transition in
very thin flakes and the behavior is quite insulating. Though the exact carrier composition
in highly disordered flakes cannot be examined in our devices, since the systems show
non-metallic properties it should be reasonable to assume those electron-type carriers are

99

largely suppressed and the conduction of current is through hole-type carriers. It is known
that in Te-rich Fe(Te,Se) the presence of excess Fe leads to enhanced incoherent
magnetic scattering and results in charge-carrier localization89 33. Thus, while the chargecarrier localization due to the excess Fe yields to the metallic behavior in bulk/thick
Fe1+yTe at low temperature, it is anticipated that the effects of charge-carrier localization
becomes more prominent in thin flakes with reduced dimensionality when the
localization length is comparable to the flake thickness, i.e, being essentially a disordered
two-dimensional system which exhibits localized states that can be well described using
the VRH model.

Figure 5.12: Fitting of VRH model (a) The zero field conductance σ
(normalized to the value at 300 K) vs T and (b) the magnetoresistance MR
data at T = 2 K for the devices on the 15 nm flake. The scatter points are
experimental data and the red solid curve is the fit using the variable range
hopping model described in the main text.

100

Figure 5.13: Comparison on MR. (a) The magnetoresistance of 15nm
Fe1+yTe flake at various temperature. In comparison of the largely
enhanced MR the 27nm data at 2K is also plotted in the same graph. (b)
The MR at 9 Tesla as a function of temperature of different samples. MR
decreases quite fast with increasing temperature and thickness.

Figure 5.14: Magnetoresistance of 16nm and 11nm Fe1+yTe flakes. The
magnitude of MR is similar to 15nm flake. The 11nm flake curve is more
noisy than the other two since the contact is less stable due to very
insulating state at low temperature.
Figure 5.13 presents the magnetoresistance (MR), which is defined as [R(B)R(0)]/R(0), for insulating thin flakes (t = 15 nm). There are two features worth pointing

101

out. First, compared to the MR value of thicker and metallic films (e.g., t = 27 nm), the
MR of thinner and insulating flakes are largely enhanced. For instance, MR measured at
B = 9 T and T = 2 K exceeds 2% for the 15 nm flake while MR is ~ 0.5% only for the 27
nm flake. And the MR decreases with increasing temperature, as shown in both Fig.
5.13(a) and Fig. 5.13(b). Second, interestingly, the MR value is positive and scales
quadratically on magnetic field in the low field region, and such MR behavior can be
attributed to the shrinkage of the overlap of the tails of wave functions of localized states
in the VRH regime113. The similar behavior is observed also in 16nm and 11nm flakes
(Figure 5.14). Within this model, MR depends on B2,
𝛼ℏ2

Δ𝑅

=
𝑅(0)

𝑅(𝐵)−𝑅(0)
𝑅(0)

𝐵2

= 𝐵2 , with
0

𝑇

𝐵02 = (𝑒 2 𝜉4 )(𝑇 )3/2 where 𝛼 ≈ 6.6 × 102 is a numerical constant. Based on the obtained
0

localization length Ξ (= 9.34 nm) and the hopping parameter T0 (=147 K) discussed above,
B0 is calculated to be B0cal ~ 39 T at T = 2 K. By fitting the MR data (0 < B < 3 T) as
shown in Fig. 5.12(b), we get the quadratic dependence B0exp = 48 Tesla, which is
comparable to the calculated B0cal and thus affirms the validity of the VRH model in
describing both (T) and MR data in our system.
Finally, we would like to briefly comment on the thickness dependent transport
behavior in Fe1+yTe nanoflakes and raise a few interesting questions. Is there an
antiferromagnetic order in the thinner flakes (thinner than 21 nm)? If so, would the
magnetic order occur with potential structural phase transition, if the latter does exist as
well? If there was no magnetic and structural phase transition in the thinner flakes, what
would be the driving force for their absence? Future in-situ Raman scattering study is
desirable to address these questions. Furthermore, it would be interesting to explore ionic

102

liquid gating on Fe1+yTe nanoflakes to examine potential carrier doping-induced
superconductivity as what has been recently reported on nanostructured FeSe101 32.

5.5 Exfoliated FeSe0.5Te0.5 flakes and ionic liquid gating
The motivation on studying the properties of exfoliated superconducting flakes of
Fe(Se,Te) comes from two studies shown in Figure 5.15 and Figure 5.16. The first one is
the evolution of the transport properties in exfoliated FeSe0.5Te0.5 as a function of
thickness, which indicates that the thin flakes start showing semiconducting behavior as
the thickness decreases and eventually the superconductivity disappears for flakes with
thickness 12nm and below33. The researchers argued that the inhomogeneity of Se and Te
atoms in thin flakes caused the inhomogeneous superconducting regions in the flake and
the regions cannot form continuous path when the thickness is below a certain length. On
the other hand, the ionic liquid gating effect study of the FeSe thin flakes (Figure 5.16)
with thickness ~10nm shows the application of positive gate voltage (which should draw
negative charges on the sample surface) dramatically increases the Tc and at the same
time reversing the Hall number (nH)32. The study implies the close relationship between
the superconductivity in FeSe and the heavy doping of electron-type carriers. These
studies raise a question: Can we also increase Tc in FeSe0.5Te0.5 flakes by heavy electron
doping using ionic liquid gating?

103

Figure 5.15: The evolution of superconductivity in exfoliated FeSe0.5Te0.5
flakes. The superconductivity is completely suppressed in flakes with
thickness = 12nm or lower and the R-T curve shows insulating behavior.
(From Yue, C. et al. ACS Nano 10, 429–435 (2016) )

Figure 5.16: The ionic liquid gating effect study on FeSe exfoliated flakes.
(a) The optical image of the device with thickness ~10nm and the R-Vg
curve. (b) Tc and Hall number as a function of Vg. As the nH is suppressed
by positive Vg, Tc is enhanced up to 45K. (From Lei, B. et al., Physical
Review Letters (2016))

104

Figure 5.17 shows both optical and AFM images of a fabricated device. With the
large size shown in optical image (Figure 5.17(a)) and a nice flat surface and uniform
thickness of 47nm all around shown in Figure 5.17(b-c) it ensures good quality contacts
for ionic-liquid gating purpose since the application of gate voltage is often harmful to
devices. The R-T measurement without ionic liquid found the Tc round 10K shown in
Figure 5.18. We examined the magnetic field dependence of Tc with the field applied
perpendicularly, as shown in the inset figure. The result confirms that the Tc is
suppressed with magnetic field as a typical effect, setting up a good test base of the
gating effect.

Figure 5.17: The fabricated FeSe0.5Te0.5 device with flake thickness 47nm.
(a) optical image of the device (b) AFM image (c) profile of the white line
in (b) shows the thickness 47 nm.

105

Figure 5.18: The R-T curve of superconducting FeSe0.5Te0.5 47nm flake.
Inset: Tc is suppressed by the presence of magnetic field. The sequence
started with no field applied (0T) and scanned with 5 Tesla and 9 Tesla of
perpendicular field. Finally 0T was scanned again for consistency.
After the initial test we applied the ionic liquid on the device and cooled it down
again (details described in Chapter 3). Right above the freezing point of the IL (around
180K), we scanned the gate voltage at 200K and found no significant effect to the
resistance value. Then the sample was cooled down with 6V gating voltage. Interestingly,
the change in Tc for this device is much smaller than what has been observed in FeSe
nanoflakes.
There could be two reasons regarding the smaller electrostatic gating effect
observed in our devices. The most possible one is the flake (47nm) is too thick so that the
number of doped carrier on the first or two layers on the surface is not large enough to
change the overall behavior, though the resistivity reduced ~10% at low temperature right

106

above Tc implies an increase of carrier number via gating. The other possibility comes
from the insulating surface of the exfoliated flakes which we removed before depositing
ionic liquid on the top of the device. The surface of FeTe or FeSeTe flakes is quite
insulating possibly because of oxidation or degradation leading to a deformed structure
(observed with TEM on FeTe surface.) The deformed surface could shield the underneath
layers so the carrier doping becomes not quite efficient.

Figure 5.19: The device with the ionic liquid gating. (a) The resistance
change as a function of gate voltage scanned from 0 to 6V, 6V to -6V and
-6V to 6V at 200K (right above the freezing point of the ionic liquid). The
change is actually negligible because the small non-reversable drift of
resistance value is common when sweeping gate voltage since the liquid
could be flowing. (b) Scanning of Tc with the ionic liquid and G is the gate
voltage. The G = 6V data has the resistivity slightly lower than the G = 0V
but Tc does not change much.

5.6 Summary
In summary, we report the study of electronic transport of exfoliated FeTe1-x
nanoflakes with various thickness. We find that while thick flakes behave similar to the

107

bulk counterpart and exhibit sharp first-order magnetic and structural phase transition
simultaneously, flakes with intermediate thickness show a broadening of the structural
phase transition followed by an onset of superconducting-like transition, which might be
associated with the (partial) relaxation of structural distortion of FeTe4 tetrahedra due to
the absence of the constraint on the surface and interface of exfoliated thin flakes. With
further decreasing thickness, the nanoflakes behave as a highly disordered twodimensional system exhibiting charge-carrier localized states with largely enhanced
magnetoresistance at low temperature, both of which can be well captured using the
variable range hopping model with a localization length of ~ 10 nm due to the excess of
Fe.

108

Chapter 6
Summary and Future Vision
In this dissertation, we have discussed the motivation in Chapter 1, several
characterization methods in Chapter 2, detailed procedures and a lot of problem-solving
in Chapter 3. The results from the studies of Ca3Ru2O7, Ca2RuO4, FeTe0.85 and
FeSe0.5Te0.5 exfoliated flakes are presented and discussed in Chapter 4 and 5. At the end
of this dissertation, I would like to summarize the major progress we made and discuss
the possibility of future experiment and expected discoveries.
On the part of preparation of samples, the invention of the “Gorilla glue method”
used to exfoliate calcium ruthenate is certainly the greatest breakthrough. The cleavage of
large single crystals in-situ is pretty common in surface studies, but such a method
combined with mechanical exfoliation and turning the seemingly impossible exfoliation
to sub-10 nm flakes possible is unprecedented. The contact resistance issue solved by
changing adhesion layer material and in-situ ion milling, as well as the working
procedure set up to prevent ESD hazard also play important roles and pave the way to the
real study of atomically thin strongly correlated materials.
The decrease of transition temperature TMIT in Ca3Ru2O7 flakes below 10 nm
indicates the existence of pure thickness effect in the system without substrate-induced
strain. The effect may be related to the increased distortion of the surface layer by oxygen
loss. The dramatic change in magnetoresistance (MR) as a function of field strength or
in-plane rotation angle implies a very different magnetic structure in thinner flakes,

109

which may be closed to the incommensurate magnetic order observed in Fe-doped
Ca3Ru2O7. Even though the transport measurement can only provide indirect evidences, it
is still quite exciting to see such a nanoscale size effect on magnetic structure. The lateral
width effect on the low-temperature metallic state is still an unresolved one. Quantized
conductance is calculated but it does not account for the experimental observation.
In the chapter of Fe1+yTe flakes, the AFM transition and the transport properties
are found to change drastically in sub-100 nm regime. The transition becomes broad
around 30 nm and disappeared completely when it approaches 20 nm. Thinner flakes
show insulating behavior in all temperature range accompanied enhanced MR. The
behaviors can be described using the VRH model. Fitting both the R-T curve and the
MR-H curves yields self-consistent result indicating the localization length around 10 nm.
A 47 nm FeSe0.5Te0.5 flake show superconductor transition around 12K the same
as the bulk. We tried to apply the ionic liquid gating but did not see any significant effect
on Tc, the reason could be either the flake being too thick or the surface layer blocking
the carrier doping.
Besides of unresolved projects discussed in Chapter 4&5, there is still plenty of
possibility in our attempt to combine the correlated system to nanostructures. In the short
term the characterization of Ca3Ru2O7 flakes with micro-XRD is definitely important as a
direct evidence for surface relaxation of distortion. And magnetic force microscopy
scanning of the flake before and after applied field may provide possible evidence of
magnetic structure change in thinner flakes.
In the middle term, the solution to the problems encountered in ionic liquid gating
experiments is desired. In my opinion there are some clues on the possible ways to avoid

110

these problems. One is the gate should never be connected to the source meter before it is
cooled down to 200K or below. Another one is the test of low temperature ionic liquid
attachment on hydrophilic and hydrophobic surfaces. Glass vessels could be used to
avoid the dispersing of ionic liquid and platinum wires serves as the gate. Even though
this method use much more ionic liquid on gating one device, it still worth a try on
important devices because it solves the problem and promises the liquid is always in
place when the gate voltage is applied. Again as mentioned in Chapter 3, initial tests
could be done on graphene devices since the preparation is way much easier, parent
material HOPG is cheap and available from vast choices of providers and the tolerance of
ESD is huge compared to our ruthenate devices. For training purposes, I strongly suggest
every new student following this project starts with preparing and measuring a gate-effect
device of graphene with both bottom gate and ionic-liquid gate. This route contains
almost all necessary skills needed in the project and will significantly reduce the
frustration of beginners and the waste of precious crystals. If the ionic liquid problem can
be finally solved, the carrier doping effect will be explored on thinner flakes of FeTe,
Fe(Se,Te) and calcium ruthenates and I expect the shift of TMIT or induced
superconductivity can happen in these materials.
In the long term, the exfoliation method can be applied on more layered materials
with strong binding forces, especially those materials which cannot be made single
crystal thin film through epitaxial growth. At the beginning, it would be reasonable to
explore more layered structured compounds in Ruddlesden-Popper series beyond
ruthenate. Manganites are the most possible choices, particularly La2-2xSr1+2xMn2O7 with
the intra-bilayer and inter-bilayer magnetic coupling changing dramatically with different

111

La/Sr ratio. Summarized by T. Kimura and Y. Tokura114 from a list of literature115-119,
with x ~ 0.3 it has perpendicular intra-bilayer FM coupling and inter-bilayer AFM
coupling but when x increases up to 0.32 the inter-bilayer coupling turns FM. Further
increasing the x up to 0.4 makes all bi-layers coupling FM at in-plane direction. For x >
0.4 canted AFM is formed within the bilayer and the cant angle increases with increasing
x. Finally A-type AFM is achieved at x = 0.5. Even though there has been the epitaxial
MBE thin film growth of the material already120, the grown thin film (50 cycles which
should yield 50 unit cells) had the Curie temperature TC = 75K, which is much lower than
bulk (~120K), indicating there might be still influence from the substrate or the
crystallinity might be lower in the grown thin film compared to bulk (the authors consider
small amount of 1-1-3 phases in grown thin films). The problems in the epitaxially grown
thin film actually inspire new hopes of applying the exfoliation method in these decadeold materials. For x = 0.3 and x = 0.32 compound, my expectation is the effect of
increased distortion may shift the configuration to different x ratio when it is exfoliated
into sub-10 nm thin flakes. In x > 0.4 compounds, it would be interesting to examine if
the canted-AFM or the A-type AFM (x=0.5) order will vanish in thin flakes as in the
Ca3Ru2O7 flakes.
Like the old proverb says, “The work will teach you how to do it”. Many of the
solutions to the problems of this project did not even exist before one takes a look on
them seriously. Fortunately most of hurdles are removed and interesting physics are dug
out. I hope more surprising phenomena can be discovered in nanoscale from follow-up
people in this thesis project.

112

REFERENCES

113

REFERENCES
1

Boer, J. H. d. & Verwey, E. J. W. Semi-conductors with partially and with completely
filled 3d -lattice bands. Proceedings of the Physical Society 49, 59-71, (1937).

2

Wu, M. K. et al. Superconductivity at 93 K in a new mixed-phase Yb-Ba-Cu-O
compound system at ambient pressure. Physical Review Letters 58, 908-910, (1987).

3

Kamihara, Y. et al. Iron-Based Layered Superconductor: LaOFeP. Journal of the
American Chemical Society 128, 10012-10013, (2006).

4

Kamihara, Y., Watanabe, T., Hirano, M. & Hosono, H. Iron-Based Layered
Superconductor La[O1- xFx]FeAs ( x = 0.05−0.12) with Tc = 26 K. Journal of the
American Chemical Society 130, 3296-3297, (2008).

5

Berglund, C. N. & Jayaraman, A. Hydrostatic-Pressure Deyendence of the Electronic
Properties of VO2, Near the Semiconductor-Metal Transition Temperature. Physical
Review 185, 1034-1039, (1969).

6

Neuman, C. H., Lawson, A. W. & Brown, R. F. Pressure dependence of the resistance of
VO2. The Journal of Chemical Physics 41, 1591-1595, (1964).

7

Stefanovich, G., Pergament, A. & Stefanovich, D. Electrical switching and Mott
transition in VO2. Journal of Physics: Condensed Matter 12, 8837-8845, (2000).

8

Kubota, M. et al. Neutron scattering studies on magnetic structure of the double-layered
manganite La2-2xSr1+2xMn2O7 (0.30 ≤ x ≤ 0.50). Journal of Physics and Chemistry of
Solids 60, 1161-1164, (1999).

9

Tomioka, Y., Asamitsu, A., Kuwahara, H., Moritomo, Y. & Tokura, Y. Magnetic-fieldinduced metal-insulator phenomena in Pr1-xCaxMnO3 with controlled charge-ordering
instability. Physical Review B 53, R1689-R1692, (1996).

10

Nakatsuji, S. & Maeno, Y. Switching of magnetic coupling by a structural symmetry
change near the Mott transition in Ca2-xSrxRuO4. Physical Review B - Condensed Matter
and Materials Physics 62, 6458-6466, (2000).

11

Nakamura, F. Pressure-induced Mott transition and related novel quantum phenomena in
Ca2RuO4. Journal of the Physical Society of Japan 76, 96-99, (2007).

12

Snow, C. S. et al. Pressure-tuned collapse of the Mott-like state in Can+1RunO3n+1 (n=1,2):
Raman spectroscopic studies. Physical review letters 89, 226401-226401, (2002).

13

Bohnenbuck, B. et al. Magnetic structure and orbital state of Ca3Ru2O7 investigated by
resonant x-ray diffraction. Physical Review B 77, 224412-224412, (2008).

14

Saito, R., Fujita, M., Dresselhaus, G. & Dresselhaus, M. S. Electronic structure of chiral
graphene tubules. Applied Physics Letters 60, 2204-2206, (1992).

114

15

Van Wees, B. J. et al. Quantized conductance of point contacts in a two-dimensional
electron gas. Physical Review Letters 60, 848-850, (1988).

16

Mikheenko, P. et al. Phase slips in submicrometerYBaCu3O7− bridges. Physical Review
B 72, 174506-174506, (2005).

17

Wei, J., Wang, Z., Chen, W. & Cobden, D. H. New aspects of the metal-insulator
transition in single-domain vanadium dioxide nanobeams. Nature nanotechnology 4, 420424, (2009).

18

Cao, J. et al. Strain engineering and one-dimensional organization of metal-insulator
domains in single-crystal vanadium dioxide beams. Nature nanotechnology 4, 732-737,
(2009).

19

Garcia-Barriocanal, J. et al. Electronically driven superconductor-insulator transition in
electrostatically doped La2CuO4+ thin films. Physical Review B - Condensed Matter and
Materials Physics 87, 1-11, (2013).

20

Hong, X., Posadas, A., Lin, A. & Ahn, C. H. Ferroelectric-field-induced tuning of
magnetism in the colossal magnetoresistive oxide La1-xSrxMnO3. Physical Review B 68,
134415-134415, (2003).

21

Novoselov, K. S. et al. Electric field effect in atomically thin carbon films. Science (New
York, N.Y.) 306, 666-669, (2004).

22

Novoselov, K. S. et al. Two-dimensional atomic crystals. Proceedings of the National
Academy of Sciences of the United States of America 102, 10451-10453, (2005).

23

Teweldebrhan, D., Goyal, V. & Balandin, A. A. Exfoliation and characterization of
bismuth telluride atomic quintuples and quasi-two-dimensional crystals. Nano Letters 10,
1209-1218, (2010).

24

Mott, N. F. & Peierls, R. Discussion of the paper by de Boer and Verwey. Proceedings of
the Physical Society 49, 72-73, (1937).

25

Imada, M., Fujimori, A. & Tokura, Y. Metal-insulator transitions. Reviews of Modern
Physics 70, 1039-1263, (1998).

26

Yeh, K.-W. et al. Tellurium substitution effect on superconductivity of the alpha-phase
Iron Selenide. Epl 37002, 14-14, (2008).

27

Fang, M. H. et al. Superconductivity close to magnetic instability in Fe(Se1-xTex)0.82.
Physical Review B 78, 224503, (2008).

28

Hu, R., Bozin, E. S., Warren, J. B. & Petrovic, C. Superconductivity, magnetism, and
stoichiometry of single crystals ofFe1+y(Te1−xSx)z. Physical Review B 80, 214514-214514,
(2009).

29

Li, S. et al. First-order magnetic and structural phase transitions in Fe1+ySexTe1-x. Physical
Review B - Condensed Matter and Materials Physics 79, 1-7, (2009).

115

30

Nie, Y. F., Telesca, D., Budnick, J. I., Sinkovic, B. & Wells, B. O. Superconductivity
induced in iron telluride films by low-temperature oxygen incorporation. Physical Review
B - Condensed Matter and Materials Physics 82, 2-5, (2010).

31

Han, Y. et al. Superconductivity in Iron telluride thin films under tensile stress. Physical
Review Letters 104, 8-11, (2010).

32

Lei, B. et al. Evolution of High-Temperature Superconductivity from a Low-Tc Phase
Tuned by Carrier Concentration in FeSe Thin Flakes. Physical Review Letters 116,
077002, (2016).

33

Yue, C. et al. Nanoscale Inhomogeneous Superconductivity in Fe(Te1–xSex) Probed by
Nanostructure Transport. ACS Nano 10, 429-435, (2016).

34

Hänke, T. et al. Reorientation of the diagonal double-stripe spin structure at Fe1+yTe bulk
and thin-film surfaces. Nature Communications 8, 13939-13939, (2017).

35

Xi, X. et al. Strongly enhanced charge-density-wave order in monolayer NbSe2. Nature
Nanotechnology 10, 1-6, (2015).

36

Blake, P. et al. Making graphene visible. Applied Physics Letters 91, (2007).

37

Ferrari, A. C. et al. Raman Spectrum of Graphene and Graphene Layers. Physical Review
Letters 97, 187401-187401, (2006).

38

Puls, C. P. et al. Structural and metal-insulator transitions in ionic liquid-gated Ca3Ru2O7
surface. Applied Physics Letters 104, 253503, (2014).

39

Liang, X. et al. Toward clean and crackless transfer of graphene. ACS Nano 5, 9144-9153,
(2011).

40

Tsukada, I. et al. Mobility analysis of FeTe thin films. Journal of the Physical Society of
Japan 80, 1-4, (2011).

41

Goldman, a. M. Electrostatic Gating of Ultrathin Films. Annual Review of Materials
Research 44, 45-63, (2014).

42

Dagotto, E. Complexity in Strongly Correlated Electronic Systems. Science 309, 257,
(2005).

43

Dagotto, E. & Tokura, Y. Strongly Correlated Electronic Materials: Present and Future.
MRS Bulletin 33, 1037-1045, (2011).

44

Wei, J. & Natelson, D. Nanostructure studies of strongly correlated materials. Nanoscale
3, 3509-3521, (2011).

45

Tian, W., Pan, X. Q., Haeni, J. H. & Schlom, D. G. Transmission electron microscopy
study of n= 1–5 Srn+1TinO3n+1 epitaxial thin films. Journal of Materials Research 16,
2013-2026, (2011).

116

46

Sandilands, L. J. et al. Stability of exfoliated Bi2Sr2DyxCa1-xCu2O8+ studied by Raman
microscopy. Physical Review B 82, 064503, (2010).

47

Mackenzie, A. P. & Grigera, S. A. Electronic Properties of the Layered Perovskite
Ruthenates: Correlated Electron Physics Approaching the Low-Disorder Limit. Journal
of Low Temperature Physics 135, 39-50, (2004).

48

Andrew, P. M., Yoshiteru, M. & Stephen, R. J. Novel quantum order in the ruthenates.
Physics World 15, 33, (2002).

49

Cao, G., Alexander, C. S., McCall, S., Crow, J. E. & Guertin, R. P. From
antiferromagnetic insulator to ferromagnetic metal: a brief review of the layered
ruthenates. Materials Science and Engineering: B 63, 76-82, (1999).

50

Yoshida, Y. et al. Quasi-two-dimensional metallic ground state of Ca3Ru2O7. Physical
Review B 69, 220411, (2004).

51

Mackenzie, A. P. et al. Quantum Oscillations in the Layered Perovskite Superconductor
Sr2RuO4. Physical Review Letters 76, 3786-3789, (1996).

52

Liu, Y. et al. Electrical transport properties of single-crystal Sr3Ru2O7 The possible
existence of an antiferromagnetic instability at low temperatures. Physical Review B 63,
174435, (2001).

53

Cao, G., McCall, S., Crow, J. E. & Guertin, R. P. Observation of a Metallic
Antiferromagnetic Phase and Metal to Nonmetal Transition in Ca3Ru2O7. Physical
Review Letters 78, 1751-1754, (1997).

54

Yoshida, Y. et al. Crystal and magnetic structure of Ca3Ru2O7. Physical Review B 72,
054412, (2005).

55

Bao, W., Mao, Z. Q., Qu, Z. & Lynn, J. W. Spin Valve Effect and Magnetoresistivity in
Single Crystalline Ca3Ru2O7. Physical Review Letters 100, 247203, (2008).

56

Ke, X. et al. Spin-wave excitation in the antiferromagnetic bilayer ruthenate Ca3Ru2O7.
Physical Review B 84, 014422, (2011).

57

Yoshida, Y. et al. Crystal and magnetic structure of Ca3Ru2O7. Physical Review B 72,
054412-054412, (2005).

58

Cao, G., Balicas, L., Xin, Y., Crow, J. E. & Nelson, C. S. Quantum oscillations, colossal
magnetoresistance, and the magnetoelastic interaction in bilayered Ca3Ru2O7. Physical
Review B 67, 184405, (2003).

59

Liu, G.-Q. Mott transition and magnetic anisotropy in Ca3Ru2O7. Physical Review B 84,
235137, (2011).

60

Baumberger, F. et al. Nested Fermi Surface and Electronic Instability in Ca3Ru2O7.
Physical Review Letters 96, 107601, (2006).

117

61

Qu, Z. et al. Unusual heavy-mass nearly ferromagnetic state with a surprisingly large
Wilson ratio in the double layered ruthenates (Sr1-xCax)3Ru2O7. Physical Review B 78,
180407, (2008).

62

Ke, X. et al. Commensurate-incommensurate magnetic phase transition in the Fe-doped
bilayer ruthenate Ca3Ru2O7. Physical Review B 89, 220407, (2014).

63

Lin, X. N., Zhou, Z. X., Durairaj, V., Schlottmann, P. & Cao, G. Colossal
Magnetoresistance by Avoiding a Ferromagnetic State in the Mott System Ca 3Ru2O7.
Physical Review Letters 95, 017203, (2005).

64

Snow, C. S. et al. Pressure-Tuned Collapse of the Mott-Like State in Can+1RunO3n+1
(n=1,2): Raman Spectroscopic Studies. Physical Review Letters 89, 226401, (2002).

65

Peng, J. et al. Interplay between the lattice and spin degrees of freedom in (Sr 1xCax)3Ru2O7. Physical Review B 82, 024417, (2010).

66

Matzdorf, R. et al. Ferromagnetism Stabilized by Lattice Distortion at the Surface of the
p-Wave Superconductor Sr2RuO4. Science 289, 746, (2000).

67

Moore, R. G. et al. A Surface-Tailored, Purely Electronic, Mott Metal-to-Insulator
Transition. Science 318, 615-619, (2007).

68

MartĂ­nez-Anaya, O., GarcĂ­a-ValdĂŠs, J., de la Mora, P. & TavizĂłn, G. Synthesis and
Crystal Structure of Oxygen-deficient Bilayer Ruthenate Sr3Ru2O7–δ. Australian Journal
of Chemistry 67, 777-783, (2014).

69

Takenaka, K., Okamoto, Y., Shinoda, T., Katayama, N. & Sakai, Y. Colossal negative
thermal expansion in reduced layered ruthenate. Nature Communications 8, 14102,
(2017).

70

Fobes, D., Peng, J., Qu, Z., Liu, T. J. & Mao, Z. Q. Magnetic phase transitions and bulk
spin-valve effect tuned by in-plane field orientation in Ca3Ru2O7. Physical Review B 84,
014406, (2011).

71

Cao, G., Balicas, L., Xin, Y., Crow, J. & Nelson, C. Quantum oscillations, colossal
magnetoresistance, and the magnetoelastic interaction in bilayered Ca3Ru2O7. Physical
Review B 67, 184405-184405, (2003).

72

Zhu, M. & al, e. unpublished, (2017).

73

Zhu, M. Phys. Rev. B in press, (2017).

74

Braden, M., AndrĂŠ, G., Nakatsuji, S. & Maeno, Y. Crystal and magnetic structure of
Ca2RuO4 Magnetoelastic coupling and the metal-insulator transition. Physical Review B
58, 847-861, (1998).

75

Ke, X. et al. Emergent electronic and magnetic state in Ca3Ru2O7 induced by Ti doping.
Phys. Rev. B 84, 201102-201102, (2011).

118

76

Alexander, C. et al. Destruction of the Mott insulating ground state of Ca2RuO4 by a
structural transition. Physical Review B 60, R8422-R8425, (1999).

77

Friedt, O. et al. Structural and magnetic aspects of the metal-insulator transition in Ca2xSrxRuO4. Physical Review B 63, 174432-174432, (2001).

78

Miao, L. et al. Itinerant ferromagnetism and geometrically suppressed metal-insulator
transition in epitaxial thin films of Ca2RuO4. Applied Physics Letters 100, 052401052401, (2012).

79

Wang, X., Xin, Y., Stampe, P. a., Kennedy, R. J. & Zheng, J. P. Epitaxial thin film
growth of Ca2RuO4+δ by pulsed laser deposition. Applied Physics Letters 85, 6146-6146,
(2004).

80

Kamihara, Y. et al. Iron-Based Layered Superconductor:  LaOFeP. Journal of the
American Chemical Society 128, 10012-10013, (2006).

81

Johnston, D. C. The puzzle of high temperature superconductivity in layered iron
pnictides and chalcogenides. Advances in Physics 59, 803-1061, (2010).

82

Paglione, J. & Greene, R. L. High-temperature superconductivity in iron-based materials.
Nature Physics 6, 645-658, (2010).

83

Stewart, G. R. Superconductivity in iron compounds. Reviews of Modern Physics 83,
(2011).

84

Hsu, F.-C. et al. Superconductivity in the PbO-type structure Îą-FeSe. Proceedings of the
National Academy of Sciences 105, 14262-14264, (2008).

85

Kuo-Wei, Y. et al. Tellurium substitution effect on superconductivity of the Îą-phase iron
selenide. EPL (Europhysics Letters) 84, 37002, (2008).

86

McQueen, T. M. et al. Extreme sensitivity of superconductivity to stoichiometry in
Fe1+Se. Physical Review B 79, 014522, (2009).

87

Mizuguchi, Y. & Takano, Y. Review of Fe Chalcogenides as the Simplest Fe-Based
Superconductor. Journal of the Physical Society of Japan 79, 102001, (2010).

88

Mizuguchi, Y., Tomioka, F., Tsuda, S., Yamaguchi, T. & Takano, Y. Substitution effects
on FeSe superconductor. Journal of the Physical Society of Japan 78, 1-5, (2009).

89

Liu, T. J. et al. Charge-carrier localization induced by excess Fe in the superconductor
Fe1+yTe1-xSex. Physical Review B 80, 174509, (2009).

90

Liu, T. J. et al. From (,0) magnetic order to superconductivity with (,) magnetic
resonance in Fe1.02Te1−xSex. Nat Mater 9, 718-720, (2010).

91

Zhang, A. M., Xia, T. L., Kong, L. R., Xiao, J. H. & Zhang, Q. M. Effects on
superconductivity of transition-metal doping in FeSe0.5Te0.5. Journal of Physics:
Condensed Matter 22, 245701, (2010).

119

92

Medvedev, S. et al. Electronic and magnetic phase diagram of -Fe1.01Se with
superconductivity at 36.7K under pressure. Nat Mater 8, 630-633, (2009).

93

Wang, Q.-Y. et al. Interface-Induced High-Temperature Superconductivity in Single
Unit-Cell FeSe Films on SrTiO3. Chinese Physics Letters 29, 037402-037402, (2012).

94

Liu, D. et al. Electronic origin of high-temperature superconductivity in single-layer FeSe
superconductor. Nature Communications 3, 931, (2012).

95

He, S. et al. Phase diagram and electronic indication of high-temperature
superconductivity at 65 K in single-layer FeSe films. Nat Mater 12, 605-610, (2013).

96

Tan, S. et al. Interface-induced superconductivity and strain-dependent spin density
waves in FeSe/SrTiO3 thin films. Nature Materials 12, 634-634, (2013).

97

Lee, J. J. et al. Interfacial mode coupling as the origin of the enhancement of T c in FeSe
films on SrTiO3. Nature 515, 245-248, (2014).

98

Shiogai, J., Ito, Y., Mitsuhashi, T., Nojima, T. & Tsukazaki, A. Electric-field-induced
superconductivity in electrochemically etched ultrathin FeSe films on SrTiO3 and MgO.
Nat Phys 12, 42-46, (2016).

99

Ge, J.-F. et al. Superconductivity above 100 K in single-layer FeSe films on doped
SrTiO3. Nat Mater 14, 285-289, (2015).

100

Qingyan, W. et al. Thickness dependence of superconductivity and superconductor–
insulator transition in ultrathin FeSe films on SrTiO3 (001) substrate. 2D Materials 2,
044012, (2015).

101

Hanzawa, K., Sato, H., Hiramatsu, H., Kamiya, T. & Hosono, H. Electric field-induced
superconducting transition of insulating FeSe thin film at 35 K. Proceedings of the
National Academy of Sciences 113, 3986-3990, (2016).

102

He, J. et al. Electronic evidence of an insulator–superconductor crossover in single-layer
FeSe/SrTiO3 films. Proceedings of the National Academy of Sciences 111, 18501-18506,
(2014).

103

Zhang, W. et al. Interface charge doping effects on superconductivity of single-unit-cell
FeSe films on SrTiO3 substrates. Physical Review B 89, 060506, (2014).

104

Yang, S. et al. Thickness-Dependent Coherent Phonon Frequency in Ultrathin
FeSe/SrTiO3 Films. Nano Letters 15, 4150-4154, (2015).

105

Manna, S. et al. Interfacial superconductivity in a bi-collinear antiferromagnetically
ordered FeTe monolayer on a topological insulator. Nature Communications 8, 14074,
(2017).

106

Xia, T. L. et al. Raman phonons of Îą-FeTe and Fe1.03Se0.3Te0.7 single crystals. Physical
Review B - Condensed Matter and Materials Physics 79, 1-4, (2009).

120

107

Bao, W. et al. Tunable (, )-Type Antiferromagnetic Order in -Fe(Te,Se)
Superconductors. Physical Review Letters 102, 247001, (2009).

108

Zhang, Z. T. et al. Superconductivity in Fe1.05Te:Ox single crystals. Physical Review B 88,
214511, (2013).

109

Mott, N. F. Conduction in non-crystalline materials. Philosophical Magazine 19, 835-852,
(1969).

110

Van Keuls, F. W., Hu, X. L., Jiang, H. W. & Dahm, A. J. Screening of the Coulomb
interaction in two-dimensional variable-range hopping. Physical Review B 56, 1161-1169,
(1997).

111

Ovchinnikov, D., Allain, A., Huang, Y.-S., Dumcenco, D. & Kis, A. Electrical Transport
Properties of Single-Layer WS2. ACS Nano 8, 8174-8181, (2014).

112

Tamai, A. et al. Strong Electron Correlations in the Normal State of the Iron-Based
FeSe0.42Te0.58 Superconductor Observed by Angle-Resolved Photoemission Spectroscopy.
Physical Review Letters 104, 097002, (2010).

113

Schoepe, W. Variable-range hopping conduction in doped germanium at very low
temperatures and high magnetic fields. Zeitschrift fĂźr Physik B Condensed Matter 71,
455-463, (1988).

114

Kimura, T. & Tokura, Y. Layered magnetic manganites. Annual review of materials
science 30, 451-474, (2000).

115

Kubota, M. et al. Neutron scattering studies on magnetic structure of the double-layered
manganite La2−2xSr1+2xMn2O7 (0.30≤x≤0.50). Journal of Physics and Chemistry of Solids
60, 1161-1164, (1999).

116

Hirota, K. et al. Neutron-Diffraction Studies on the Magnetic Ordering Process in
theLayeredMn Perovskite La2-2xSr1+2xMn2O7 (x=0.40, 0.45 and 0.48). Journal of the
Physical Society of Japan 67, 3380-3383, (1998).

117

Perring, T. G., Aeppli, G., Kimura, T., Tokura, Y. & Adams, M. A. Ordered stack of spin
valves in a layered magnetoresistive perovskite. Physical Review B 58, R14693-R14696,
(1998).

118

Argyriou, D. N. et al. Unconventional magnetostriction in layered La1.2Sr1.8Mn2O7:
Evidence for spin-lattice coupling above TC. Physical Review B 55, R11965-R11968,
(1997).

119

Battle, P. D. et al. Antiferromagnetism, Ferromagnetism, and Phase Separation in the
GMR System Sr2-xLa1+xMn2O7. Chemistry of Materials 9, 1042-1049, (1997).

120

Tanaka, H. & Kawai, T. Artificial construction of SrO/(La, Sr)MnO3 layered perovskite
superlattice by laser molecular-beam epitaxy. Applied Physics Letters 76, 3618-3620,
(2000).

121

122