university of groningen quantum transport in molybdenum ... · the phd research proposal was...
TRANSCRIPT
University of Groningen
Quantum transport in molybdenum disulfide and germanane transistorsChen, Qihong
IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite fromit. Please check the document version below.
Document VersionPublisher's PDF, also known as Version of record
Publication date:2017
Link to publication in University of Groningen/UMCG research database
Citation for published version (APA):Chen, Q. (2017). Quantum transport in molybdenum disulfide and germanane transistors. RijksuniversiteitGroningen.
CopyrightOther than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of theauthor(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).
Take-down policyIf you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediatelyand investigate your claim.
Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons thenumber of authors shown on this cover page is limited to 10 maximum.
Download date: 21-04-2021
Quantum transport in
molybdenum disulfide and
germanane transistors
Chen Qihong
Quantum transport in molybdenum disulfide and germanane transistors Chen Qihong
PhD thesis
University of Groningen
The Hong Kong University of Science and Technology
The work described in this thesis was performed in the research group “Device
Physics of Complex Materials” of the Zernike Institute for Advanced Materials at the
University of Groningen, the Netherlands, and the research group “Nanostructured
Materials” of the Physics department at the Hong Kong University of Science and
Technology, Hong Kong, China.
The PhD research proposal was awarded an Ubbo Emmius Scholarship in 2014,
funded by the Zernike Institute for Advanced Materials at the University of Groningen.
The instruments and lab facilities were supported by the European Research Council
(consolidator grant no. 648855 Ig-QPD).
Zernike Institute PhD thesis series: 2017-7
ISSN: 1570-1530
ISBN (Printed version): 978-90-367-9559-3
ISBN (Electronic version): 978-90-367-9560-9
Cover design: Lei Liang
Printed by: Ipskamp Printing
Quantum transport in molybdenum disulfide and germanane transistors
Proefschrift
ter verkrijging van de graad van doctor aan de Rijksuniversiteit Groningen
op gezag van de rector magnificus prof. dr. E. Sterken
en volgens besluit van het College voor Promoties.
De openbare verdediging zal plaatsvinden op
vrijdag 27 januari 2017 om 09.00 uur
door
Qihong Chen
geboren op 4 augustus 1987 te Chongqing, China
Promotores
Prof. dr. J. Ye
Prof. dr. P. Sheng
Beoordelingscommissie
Prof. dr. B. Noheda Pinuaga
Prof. dr. ir. H. Hilgenkamp
Prof. dr. D. Daghero
Prof. dr. R. Lortz
Contents
Chapter 1. Introduction .............................................................................................................. 1
1.1 Two-dimensional (2D) Layered Materials .................................................................................................. 2
1.2 Molybdenum disulfide (MoS2) .......................................................................................................................... 4
1.2.1 Band structure of MoS2 .............................................................................................................................. 5
1.2.2 Optical properties ......................................................................................................................................... 7
1.2.3 Field effect transistors (FETs) of MoS2 ............................................................................................ 10
1.2.4 Ionic gating and superconductivity in MoS2 ................................................................................. 12
1.2.5 Shubnikov–de Haas oscillation in high mobility MoS2 transistors .................................... 15
1.3 Germanane (GeH) ................................................................................................................................................ 17
1.4 Motivation and outline of this thesis .......................................................................................................... 21
References ....................................................................................................................................................................... 25
Chapter 2. Two channels transport in MoS2 .................................................................... 33
2.1 Introduction ............................................................................................................................................................ 34
2.2 Device fabrication ................................................................................................................................................ 36
2.3 Results and discussions .................................................................................................................................... 37
2.3.1 MoS2 encapsulated between top and bottom h-BN ................................................................... 37
2.3.2 MoS2-EDLT on h-BN substrate without top cover ..................................................................... 41
2.4 Conclusion ............................................................................................................................................................... 52
References ....................................................................................................................................................................... 54
Chapter 3. Dual-gate MoS2 superconducting transistor .............................................. 57
3.1 Introduction ............................................................................................................................................................ 58
3.2 Device fabrication ................................................................................................................................................ 59
3.3 Results and discussion ...................................................................................................................................... 61
3.3.1 Sample State A (close to QCP) .............................................................................................................. 62
3.3.2 Sample State B (deep into the superconducting dome) .......................................................... 68
3.4 Conclusion ............................................................................................................................................................... 76
References ....................................................................................................................................................................... 78
Chapter 4. Superconductor-normal metal junction in gated MoS2 ......................... 81
4.1 Introduction ............................................................................................................................................................ 82
4.2 Device fabrication ................................................................................................................................................ 84
4.3 Field effect transistor characteristics ......................................................................................................... 85
4.4 Superconductivity in the exposed MoS2 channel (contacts 4 and 5) .......................................... 88
4.5 Magnetoresistance in the h-BN covered MoS2 channel (contacts 2 and 3) ............................. 88
4.5.1 SdH quantum oscillations ...................................................................................................................... 89
4.5.2 Anomalous magnetoresistance ........................................................................................................... 93
4.6 Conclusion ............................................................................................................................................................... 97
References ....................................................................................................................................................................... 98
Chapter 5. Transport properties of multi-layer germanane ................................... 101
5.1 Introduction .......................................................................................................................................................... 102
5.2. Structure and characterization .................................................................................................................. 104
5.3 Electrical transport ........................................................................................................................................... 106
5.3.1 Low annealing temperature, ~ 170 °C .......................................................................................... 107
5.3.2 Intermediate annealing temperature, ~190 °C ......................................................................... 109
5.3.3 High annealing temperature ~ 210 °C ........................................................................................... 110
5.3.4 Discussion .................................................................................................................................................... 111
5.4 Strong spin-orbit interaction in annealed GeH .................................................................................... 113
5.5 Conclusion ............................................................................................................................................................. 115
References ..................................................................................................................................................................... 117
Appendix.Ultra-small carbon nanotubes in zeolite template ................................. 119
1 Introduction ............................................................................................................................................................. 120
2 Experimental ............................................................................................................................................................ 122
3 Results and discussion ........................................................................................................................................ 123
3.1 Raman spectra ............................................................................................................................................... 123
3.2 Simulation of the RBM ............................................................................................................................... 130
3.3 Thermal-gravimetric analysis (TGA) .................................................................................................. 130
4 Conclusion ................................................................................................................................................................. 132
References ..................................................................................................................................................................... 134
Summary ..................................................................................................................................... 137
Samenvatting ............................................................................................................................. 143
Acknowledgements ................................................................................................................. 149
1
Chapter 1. Introduction
Abstract
This chapter presents a brief introduction to the two 2D materials studied in
this thesis: semiconducting transition metal dichalcogenide—molybdenum
disulfide (MoS2), and graphane analogue—germanane (GeH). The crystal
structures, sample preparation methods and basic physical properties are
discussed. A brief overview of this thesis is presented at the end of this chapter.
Chapter 1
2
1.1 Two-dimensional (2D) layered materials
In 2010, the Nobel Prize in physics was awarded to Andre Geim and Konstantin
Novoselov for their groundbreaking experiments in isolating two-dimensional (2D)
carbon: graphene [1]. Figure 1.1(a) shows the lattice structure of graphene, which is
composed of a flat monolayer of carbon atoms arranged in a honeycomb lattice.
Graphene shows many extraordinary properties: it is the thinnest and strongest
material known—about 100 times stronger than steel [2]; it is almost transparent
yet impermeable to gases [3]; it exhibits very high thermal conductivity and
stiffness [4]; furthermore, graphene shows extremely high conductivity and can
withstand current densities six orders of magnitude higher than copper [5].
Figure 1.1 (a) Crystal structure of graphene, carbon atoms are arranged in a hexagonal lattice.
The dashed rectangle indicates the unit cell, which is composed of inequivalent A and B carbon
atoms. The A and B carbon atoms form two sub-lattices. (b) The Brillouin zone in reciprocal
space. (c) Left: energy dispersion relations of graphene. Right: a blowup at the corner of
Brillouin zone, showing the linear relation close to the Fermi surface with low energy excitation.
The hallmark feature of graphene’s electronic properties is the linear dispersion
relation 𝐸 = ℏ𝑣𝐹𝑘. Close to the Dirac point, the density of states reaches zero, as
shown in Figure 1.1 (b) and (c). Compared to the quadratic bands in normal
semiconductors, the linear band implies vanishing effective mass; therefore the
Fermi velocity of the Dirac quasi-particle can reach 106 m/s. In practice, very high
mobility up to 106 cm2V-1s-1 has been observed at low temperatures [6]. Despite the
rich physics in this Dirac material, the application of graphene is very limited
because of its electronic band structure: as a semimetal with zero band gap, field
effect transistors (FETs) made with graphene cannot be switched off. Chemical
modification or nanofabrication can create energy gaps in graphene, but its
distinctive properties, specifically its high mobility, will also be greatly reduced,
making the complicated fabrication process inferior to the mature technology-
based conventional semiconductors.
Introduction
3
On the other hand, the discovery of graphene and the methodology of isolating
2D materials have stimulated research into a variety of other 2D materials that are
attractive for applications in electronic and photonic devices. Transition metal
dichalcogenides (TMDs) are one of the most widely studied graphene counterparts.
TMDs refer to a family of compounds with the formula MX2, where M is a
transition metal atom (Mo, W, Ta, Nb, etc.) and X is a chalcogen atom (S, Se or Te).
Similar to the layered structure of graphite, bulk TMDs are stacked by monolayers
with weak van der Waals force, and hence can also be easily exfoliated into
atomically thin layers. In contrast to graphene, many monolayer TMDs are
intrinsically semiconducting with a moderate band gap. TMDs have shown a wide
range of electronic properties for developing transistors [7–11], optical properties
suitable for optoelectronics [12–14] and mechanical properties for potential
applications in flexible and transparent electronic devices [15–18].
TMDs have different electronic properties, which are primarily determined by
the different transition metal atoms and the polytypes with different crystal
structures. The electronic properties of the most commonly studied TMDs are
summarized in Table 1.1 [10]. In general, TMDs composed of Mo or W are
semiconducting and those made with Ta or Nb are metallic. For metallic TMDs,
exceptional properties such as superconductivity and charge density waves (CDW)
are well known in NbSe2 [19] and TaS2 [20], among others. On the other hand,
chemical doping using alkali metals can also create a metallic phase in
semiconducting MoS2, which also shows superconductivity if the correct dopants
are chosen [21,22]. The richness of TMDs’ physical properties and chemical
flexibilities provide an ideal platform for studying 2D electronic phases.
Table 1.1 Summary of the electronic properties of TMD materials. TMDs with Nb or Ta
metal atoms are metallic, showing superconducting and charge density wave (CDW) properties.
Those with Mo or W metal atoms are semiconducting. Adapted from Ref. 10.
Chapter 1
4
For semiconducting TMDs, recent studies indicate that the electronic properties
are dependent on the number of layers [10,23]. The energy gap becomes larger
with a decreasing number of layers, and the indirect band gap in bulk changes to a
direct band gap in monolayers, which is indicated by the great enhancement of
photoluminescence (PL) in monolayer compared with bulk. More importantly,
monolayer TMDs show valley-dependent spin textures due to the breaking of
inversion symmetry and large spin-orbit interactions in transition metal
atoms [24–27]. Most of the semiconducting TMDs have a band gap size of 1.1 – 2
eV, which are considered to be attractive candidates for replacing silicon in 2D
electronic transistors.
In terms of being applied to optoelectronics, semiconducting TMDs can be used
to generate, detect, control or interact with light [28–30]. Modern electronics such
as transparent displays, wearable electronics and solar cells demand
semiconductors that are flexible and transparent. TMDs can be made one atom
thick, which is highly flexible and transparent. There have been many prototypes
showing attractive potential applications in optoelectronic devices such as
photodetectors [31–34], light emitters [35–38], photovoltaics [39–41] and
valleytronic devices [42,43]. Furthermore, direct band gap and valley-dependent
spin texture can also support the design of optoelectronic devices sensitive to the
circular polarization of light [12,42].
1.2 Molybdenum disulfide (MoS2)
Molybdenum disulfide (MoS2) is one of the most intensively studied materials
in the TMDs family. In daily applications, MoS2 is widely used as a lubricant and
catalyst. As an archetypal TMD, bulk MoS2 is stacked by individual layers with
weak van der Waals coupling; it can therefore be easily thinned down to uniform
few-layer or monolayer flakes by mechanical exfoliation, using the simple scotch
tape method that is widely used in graphene research [1]. Figure 1.2(a) illustrates
the crystal structure of monolayer MoS2. The side view (left panel) shows that a
MoS2 monolayer is composed of three atom planes, a plane of Mo atoms
sandwiched between two S atom planes. From the top view (right panel) it is a
hexagonal lattice similar to graphene, except that the C atoms in A and B
sublattices are respectively replaced by one Mo and two S atoms. Correspondingly,
MoS2 has a hexagonal Brillouin zone with the corners denoted as K (K’) points.
Depending on the stacking order of each layer and the metal atom coordination,
MoS2 has three different polytypes: 1T, 2H and 3R as shown in Figure 1.2(b). The
1T phase is composed of octahedral units with AA stacking, where inversion
symmetry is preserved in the monolayer. The 2H and 3R phases are composed of
trigonal prismatic units, with AB and ABC stacking, respectively. For the 2H
Introduction
5
polytype, the inversion symmetry is broken in the monolayer, while for the 3R type,
the inversion symmetry is broken both in monolayer and in bulk phase [44]. The
lattice constants of MoS2 are a = 3.15 Å and c = 12.3 Å; the interlayer spacing is
around 6.5 Å. The different compositions of the A and B sublattices make the
electronic properties of MoS2 very different from graphene. The most important
feature is the presence of a substantial energy gap in the band structure, which will
be further elaborated below.
Figure 1.2 (a) Side (left panel) and top (right panel) views of monolayer MoS2. One Mo atom
plane is sandwiched between two S planes. The structure is a honeycomb lattice with A and B
sublattices occupied by Mo and S atoms, respectively. (b) Schematics of the structural
polytypes: 1T (tetragonal symmetry, one layer per repeat unit, octahedral coordination), 2H
(hexagonal symmetry, two layers per repeat unit, trigonal prismatic coordination) and 3R
(rhombohedral symmetry, three layers per repeat unit, trigonal prismatic coordination).
1.2.1 Band structure of MoS2
It is known that bulk MoS2 is a semiconductor with an indirect band gap of ~1.3
eV, which is similar to that of silicon. For the free-standing monolayer, however,
Chapter 1
6
the indirect band gap becomes a direct band gap of ~1.9 eV [45]. Theoretical
calculations of the band structure for bulk MoS2 and ultrathin MoS2 layers with
different thicknesses are summarized in Figure 1.3. For bulk MoS2, the valence and
conduction band extrema are away from the K point indicated by the black arrows.
As the thickness decreases from bulk to monolayer, the direct transition at the K
point stays almost unchanged. However, the indirect transition from the Γ point to
the minimum of the conduction band (denoted as the Q point) increases by more
than 0.6 eV as the thickness decreases. In particular for monolayer MoS2, the
indirect transition energy exceeds the direct transition energy at the K point of the
Brillouin zone, so monolayer MoS2 changes to a direct band gap 2D semiconductor.
Figure 1.3 Band structures of MoS2 calculated by the first-principle density functional theory
(DFT) from bulk to monolayer. The horizontal red dashed lines denote the Fermi levels. The
black arrows show the fundamental band gaps (direct or indirect) for each system. The top of
the valence bands and bottom of the conduction band are highlighted by blue and green solid
lines, respectively. Adapted from Ref. 47.
The evolution of the band structure from bulk to monolayer can be qualitatively
understood by quantum confinement in reduced dimensions, which modifies the
hybridization between pz orbitals of the S atoms and d orbitals of the Mo
atoms [45,46]. Theoretical calculations show that electronic states of different
Introduction
7
wave vectors are associated with electron orbitals with different spatial
distributions: conduction band states at the K point are composed primarily of the
localized d orbitals of the Mo atoms [46]. Since the Mo atoms are sandwiched
between two S atom planes, states at the K point are barely affected by interlayer
coupling. While the states at the Γ point are combinations of antibonding pz
orbitals of the S atoms and d orbitals of the Mo atoms, the influence from
interlayer coupling is much stronger. The energy gap at the Γ point is thus closely
related to the number of layers.
1.2.2 Optical properties
The evolution of the electronic band structure of MoS2 with different numbers
of layers can be monitored by absorption spectra, photoluminescence (PL) spectra
and photoconductivity [46,47].
Figure 1.4 (a) PL spectra for mono- and bilayer MoS2 samples. A sharp increase in the PL
intensity is observed from bilayer to monolayer MoS2. Inset: PL quantum yield of thin layers of
MoS2 for number of layers N = 1-6 in log scale. (b) Band-gap energies of thin layers of MoS2,
extracted from the energy of the PL features. The dashed line indicates the (indirect) band-gap
energy of bulk MoS2. Adapted from Ref. 45.
Bulk MoS2 has negligible PL because of the indirect band gap; the absorption or
emission of a photon must be assisted by a phonon to compensate for the
momentum difference, which means the efficiency is low. This limitation does not
exist in the case of monolayer MoS2 due to the nature of the direct band gap:
photons with energies that are higher than the band gap can be absorbed or
emitted efficiently. A dramatic increase in the PL is observed from double-layer to
monolayer MoS2, as shown in Figure 1.4(a). The lowest PL peak energies for
different numbers of layers are shown in Figure 1.4(b), which matches well with
the calculated band gaps shown in Figure 1.3.
Chapter 1
8
The inset of Figure 1.4(a) shows that with the change from the indirect band
gap of bulk to the direct band gap of the monolayer, the PL quantum yield
increases by more than four orders of magnitude. In spite of this enhancement, the
quantum yield of MoS2 is much lower than the expected value for a direct band gap
semiconductor. It has been reported that suspended [45] or boron nitride-
supported [48] MoS2 monolayers can reach higher quantum yields, but still below
the theoretical value. This is because both the unintentional doping during
fabrication and the unavoidable crystal defects suppress the intensity of PL
significantly. However, it was recently reported that the quantum yield of MoS2
treated by solution-based organic superacid can reach 95% [49,50]. The
underlying mechanism is still under exploration in order to develop more efficient
light emitters and other optoelectronic devices.
Figure 1.5 (a) Simplified band structure of bulk MoS2 showing the lowest conduction band c1
and the highest split valence bands v1 and v2. A and B are the direct band gap transitions and I
is the indirect gap transition. Eg’ is the indirect gap for bulk and Eg is the direct band gap for
monolayer. (b) Optical (left) and photoluminescence (right) images of a MoS2 flake on SiO2
substrate with etched holes. The large holes in the left panel are 1.5 μm in diameter. The
photoluminescence emission is enhanced in suspended regions over the holes, but is not
detected in the multilayer region. (c) Absorption spectra (left axis, normalized by layer number)
and the corresponding PL intensity (right axis, normalized by the intensity of the peak A).
Adapted from Ref. 10.
The absorption spectra of MoS2 are dominated by two main features,
corresponding to the direct transition between the split valence band maxima (v1
and v2) and the conduction band minima (c1) at the K point of the Brillouin
zone [51], as shown in Figure 1.5(a). The band splitting is a result of inversion
symmetry breaking and strong spin-orbit coupling. The spin splitting at the
valence band is about 0.16 eV for monolayer MoS2 [13,52]. The so-called A and B
Introduction
9
excitons can be distinguished in the absorption and photoluminescence spectra of
MoS2 [45], as shown in Figure 1.5(c). The PL peak A matches well with the lower
absorption peak, which can be attributed to the transition of the direct band gap,
and peak I is attributed to the indirect transition from the Γ point of the valence
band to the minima of the conduction band.
Present electronic devices are primarily operated based on controlling charges,
where the number of charge carriers bears information. As an alternative, it has
been proposed that devices operated by controlling the spin and valley degrees of
freedom will lead to new microelectronic devices that can increase computing
power as well as reduce energy consumption [48,53–55]. Excitation and
manipulation of the valley degree of freedom have attracted much interest for both
the underlying physics and new classes of electronic devices.
Figure 1.6 Schematic illustration of proposed valley-dependent selection rules at K and K’
points in the crystal momentum space for monolayer MoS2. Due to angular momentum
conservation and time reversal symmetry, left-/right- handed circularly polarized light σ+/σ-
only couples to the band-edge transition at the K/K’ point. Adapted from Ref. 55.
For most crystals, the valley degree of freedom is degenerated due to the
symmetric crystal structure. It is difficult to selectively pump a specific valley to
produce valley polarization. However, monolayer MoS2 with two inequivalent
valleys K and K’ at the corners of the Brillouin zone, has recently been proven to be
effective for the control of valley population using optical pumping with circularly
polarized light [48,55–57]. Because of the broken inversion symmetry in
monolayer, strong spin-orbit interactions split the valence bands by ~0.16 eV. The
splitting is opposite at the K and K’ valleys because of the time reversal symmetry,
so the spin and valley degrees of freedom are inherently coupled [13,27]. When the
optical transition happens between the upper valence band and the conduction
band using circularly polarized light pumping, excitation only occurs at either the
K or K’ valley depending on the helicity of the light [58]. It has been observed that
the valley polarization remains largely preserved during the lifetime of the
excitons [59]. Due to the coupling between the valley and spin degrees of freedom,
Chapter 1
10
the excitation process also gives rise to photo-generated particles of a well-defined
spin state [48,55,56,60].
The coupling between the valley and spin largely relies on the spin splitting at
the K/K’ points of the valence bands, which have been well characterized by a wide
range of optical experiments. Electronic properties at the conduction band are
much less explored. For the purpose of studying the electronic properties at the
conduction band, electrical transport becomes very useful. This is crucial especially
for the multilayer MoS2, because the optical measurement becomes much less
efficient due to the indirect band gap nature. By the electrical field tuning, the
Fermi level can be easily shifted to access the conduction band since MoS2 shows
n-type semiconductor characteristics. In practice, however, this is not easy to
accomplish because the quality requirement for electrical measurement is high. In
addition, the ohmic contact is difficult to make on the semiconducting materials
due to the Schottky barrier.
1.2.3 Field effect transistors (FETs) of MoS2
Field effect transistors (FETs) are the fundamental building blocks of integrated
circuits. A conventional transistor is composed of source and drain electrodes
connected by the channel material made of semiconducting materials. The current
that flows through this channel is controlled by the electrical field applied through
the gate [61]. In industrial applications, typical transistors must provide digital
logics with high on/off ratios for effective switching, high carrier mobility for fast
operation, and low off state conductivity for low power consumption. Silicon is
currently the most widely used channel material in all kinds of electronic devices.
In 1975, Gordon Moore predicted that the number of transistors in a densely
integrated circuit would double approximately every two years, which is known as
the famous Moore’s Law. This prediction has proved valid for decades. However,
silicon-based transistors are reaching their limit in decreasing size as well as
improving performance. In order to keep Moore’s Law, the possibility of using
nano-structures, such as carbon nanotubes [62,63] and semiconductor
nanowires [64], as transistor channel materials has been extensively explored.
Since the discovery of graphene, much effort has been devoted to studying its
electronic applications because graphene is strictly two-dimensional and has very
high carrier mobility and its conductivity can be effectively tuned by gate
voltage [62,65–68]. As mentioned previously, the biggest disadvantage of
graphene is its lack of an intrinsic band gap. Hence, for graphene-based transistors,
the on/off ratio is very low at room temperature and the off-state conductivity is
relatively high, which limits its application in digital logic transistors.
In contrast to graphene, transistors made of MoS2 are expected to show current
on/off ratios in the order of 1010 because of the large band gap. Simulations of the
Introduction
11
device characteristics show that for a monolayer MoS2 transistor with 15 nm gate
length, 2.8 nm-thick HfO2 top-gate dielectric and 0.5 V source-drain voltage, the
sub-threshold swing can reach as low as 60 mV/dec. Furthermore, the off-state
current is very low, making MoS2 an attractive candidate for low-power
applications [69,70].
Figure 1.7 (a) Schematic illustration of top-gated monolayer MoS2 FET device. Light blue color
represents the high-κ HfO2 dielectrics, which improves the carrier mobility due to the charge
screening effect. (b) Room-temperature transfer characteristics for the FET with applied bias
voltage VDS = 10 mV. Back gate voltage Vbg is applied to the substrate while the top gate is
disconnected. The inset shows the IDS–VDS curves acquired for VBG values of 0, 1 and 5 V from
bottom to top. Adapted from Ref. 74.
About a decade ago, a thin-layer MoS2-based transistor was successfully
realized, with mobility in the range of 0.1-10 cm2V-1s-1 [71,72]. In 2011, Kis and co-
workers reported a monolayer MoS2 transistor with a HfO2 top gate, the structure
of which is shown in Figure 1.7 [31]. The use of high-κ HfO2 as the gate dielectric in
the transistor configuration enables the channel to be effectively switched on and
off with small gate voltage. The device can be completely turned off by changing
the top gate voltage from -2 to -4 V. For VDS = 10 mV, the Ion/Ioff ratio is in the
order of 106. For VDS = 500 mV, the Ion/Ioff ratio is higher than 108 in the measured
range, while the sub-threshold swing S = 74 mV/dec. From the FET fitting, the
mobility is higher than 200 cm2V-1s-1 at room temperature. The high-κ dielectric
nature of HfO2 helps in screening the Coulomb scatterings from the charged
impurities, which greatly enhances electron mobility. In addition to the
mechanically exfoliated sample, scalable sample preparation methods such as
liquid-exfoliated thin MoS2 flakes have shown similar FET performance [73],
which is an important step for industrial production.
In addition to single transistors, integrated logic circuits based on monolayer
MoS2 have also been extensively studied [75]. The logic circuit consisting of two
Chapter 1
12
connected transistors achieves the logic operation of converting logic “1” to “0” or
logic “0” to “1”, as shown in Figure 1.8. The NOT logic operation shows room
temperature voltage gains higher than 1.
To operate the MoS2 logic circuit, the input voltage Vin is applied to the local
gate of the lower transistor, and the VDD = 2 V is applied to the drain electrode of
the upper transistor. When the input voltage is negative corresponding the logic
“0”, the resistance of the lower transistor is much higher than that of the upper one,
and hence the output voltage is close to the supply voltage of VDD = 2 V. When the
input voltage becomes positive corresponding to the logic “1”, the lower transistor
is switched on and becomes more conductive than the upper one, and hence the
output voltage is close to zero. The NOT logic operation is achieved. In addition,
more complicated multistage logic circuits based on few-layer MoS2 were also
realized, with functions of an inverter, a NAND gate, a static random access
memory and a five-stage ring oscillator [76].
Figure 1.8 (a) Single-layer MoS2 is deposited on top of a silicon chip covered with 270 nm-thick
SiO2. Two n-type transistors are fabricated on the same monolayer MoS2 flake. (b) The output
voltage and transfer characteristic of the inverter as a function of the input voltage Vin. Inset:
Schematic drawing of the electronic circuit and the truth table for the NOT logic operation. The
middle electrode is connected to the local gate of the upper transistor, serving as output voltage
Vout. Adapted from Ref. 75.
1.2.4 Ionic gating and superconductivity in MoS2
As mentioned above, field effect transistors work with a source and a drain
electrode connected by a semiconducting channel material, where the current that
flows through the channel is controlled by gate voltage. SiO2 formed by directly
oxidizing the Si substrate is widely used for the gate dielectric between gate
electrode and channel. Other oxides with higher dielectric constants, such as HfO2
and ZrO2, can reduce the gate voltage required for accumulating the same amount
of carriers. The high dielectric constant nature might help to improve carrier
Introduction
13
mobility as well. For these oxides, the carrier accumulation capabilities are
determined by the largest electric field that can be applied across the oxides
without destroying them, often measured in units of MV/cm. Typically,
conventional solid oxide dielectrics such as SiO2, Al2O3 and HfO2 cannot withstand
too high electric fields (usually in the order of ~10 MV/cm), which limits the ability
to tune the carrier density.
Figure 1.9 Schematic illustration of charge carrier induction for an EDL device. For illustration,
a positive gate voltage is applied between the gate electrode and channel material; cations are
driven very close to the channel, which induces electrons at the surface. The carriers are mainly
confined to the top surface due to the strong Thomas-Fermi screening effect.
Alternatively, electrical double layer (EDL) has been proven to be an efficient
way to improve device performance, especially the ability to accumulate large
amounts of carriers [77]. In this method, especially in its recent form, ionic liquid
replaces the traditional oxide dielectrics between the gate electrode and channel
material. The working principle of ionic gating is shown schematically in Figure 1.9.
Because cations/anions can freely move in the ionic liquid, when a gate voltage is
applied between the ionic liquid and channel material, cations/anions are driven
very close (~1 nm) to the channel surface and induce a layer of carriers with
opposite polarity right beneath the surface of the channel material. The narrowly
separated space charge between the ions and channel material can be regarded as a
nano-size capacitor [78], and the capacitance is huge because the distance between
the ions and channel surface is in the order of ~1 nm. As a result, the electric
double layer induces very large carrier density, usually more than 1-2 orders of
magnitude higher than that of conventional oxide dielectrics.
Since EDL can accumulate carrier density that is beyond the limit of solid gates,
novel electronic states [80,81] and new device functionalities [82–85] that appear
only in the high carrier density regime can be accessed. Electric double layer
transistors (EDLTs) fabricated with TMDs have shown various attractive electronic
properties, including ambipolar transport [79,86], circularly polarized
electroluminescence [42] and electric field-controlled spin polarization [87]. As
Chapter 1
14
shown in Figure 1.10, EDLTs made of thin MoS2 flakes show ambipolar transistor
operation, with an on/off ratio greater than 102 for both hole and electron
transport [79].
Figure 1.10 (a) Schematic of electric double-layer transistor (EDLT, a FET gated by ionic
liquid). VDS is the source-drain voltage and VG is the gate voltage. (b) Conductivity as a function
of top gate voltage for both bulk and thin-flake MoS2 EDLT devices. Adapted from Ref. 79.
Figure 1.11 (a) Temperature dependence of the channel sheet resistance Rs at different liquid
gate biases ranging from 0 to 6 V (indicated on the right). (b) Dome-shaped phase diagram
showing the evolution of different electronic phases as a function of carrier density n2D. Solid
dots represent data from ionic gated MoS2 transistor. Empty circles are the data from alkali
metal intercalated MoS2. The quantum critical point (QCP) resides at 6 × 1013 cm-2,
corresponding to the onset of superconductivity. Adapted from Ref. 89.
Introduction
15
Superconductivity, as a special property of certain materials at low
temperatures, has been observed in various metals (Al, Pb, Nb, etc.) and
compounds (MgB2, NbN, YBa2Cu3O4 etc.). It is well known that superconducting
properties can be controlled by modifying carrier density. In compounds with wide
chemical flexibility, e.g. cuprates, doping is the most effective way to identify the
electronic phase of materials. In cuprates, superconductivity can be induced with
either electron or hole doping by chemically introducing impurities, forming a
dome-like superconducting phase diagram as a function of the chemical doping
concentration. On the other hand, chemical doping can also be achieved using the
strong field effect. Recently, superconductivity was observed in ionic gated
SrTiO3 [82], ZrNCl [83], KTaO3 [84], WS2 [88] and MoS2 [89] etc. Particularly,
with ionic liquid gating MoS2 becomes superconducting when the carrier density
reaches the level of 1014 cm-2 [89]. A dome-shaped phase diagram is established as
shown in Figure 1.11(b). At absolute zero temperature, the critical carrier density
for superconductivity is ~6×1013 cm-2, corresponding to the quantum critical point
(QCP, the onset of superconductivity). At finite temperature, the phase boundary
between superconducting and non-superconducting states is specified by the left
edge of the shaded area. Within this range, the transition temperature Tc increases
rapidly with the increase of carrier density. For a real device, this means that Tc
increases with the increase of gate voltage. Recently, it is found that
superconductivity in MoS2-EDLT shows Ising pairing mechanism [90]. The spins
of the Cooper pairs are strongly pinned to the out-of-plane direction by the
effective Zeeman field at the K/K’ points of the Brillouin zone. Therefore, the
superconductivity is protected from very strong external magnetic fields that are
parallel to the ab-plane of MoS2. The effective Zeeman field is from the strong
spin-orbit interaction with the breaking of inversion symmetry.
1.2.5 Shubnikov–de Haas oscillation in high mobility MoS2
transistors
For semiconductors, high carrier mobility is a very important factor for creating
high-speed electronic devices. Various approaches have been studied to improve
the performance of MoS2 FET devices. While mobility can be greatly enhanced by
choosing high-κ gate dielectrics such as HfO2 [74] or Al2O3 [91], or by suspending
MoS2 which shows 2-10 times of mobility and on/off ratio improvement compared
to conventional substrate-supported devices [92]. However, the absolute mobility
is still far from the expectation for high-speed electronics. In addition, suspended
MoS2 is not easy to prepare due to the complicated fabrication process.
For MoS2 on the traditional oxide substrate, the carrier mobility is largely
limited by the roughness, charged surface states and surface optical phonons of the
substrates. Hexagonal boron nitride (h-BN) is a very attractive substrate for
Chapter 1
16
improving the carrier mobility of MoS2 transistors [94–99]. The structure of h-BN
is similar to graphene, with the A and B sublattices occupied by boron and nitrogen
atoms, respectively. Because of the strong in-plane covalent bonding, h-BN is
expected to exhibit minimized dangling bonds and surface charge traps [94]. The
surface of h-BN is atomically flat. Therefore by using h-BN as the substrate for
MoS2 devices, the fluctuations in the potential and roughness of MoS2 can be
significantly reduced compared to MoS2 on SiO2. Furthermore, the surface optical
phonon modes of h-BN have energies much larger than similar modes in SiO2, the
scattering from the substrate is considerably reduced. Due to these reasons, the
mobility of MoS2 can be significantly improved when h-BN is used as the substrate.
The van der Waals heterostructure devices created by encapsulating MoS2 between
two h-BN flakes showed exceptional high mobility [93,100]. The contact resistance
could be significantly reduced especially when graphene was used as contact
electrodes. The Hall mobility reached 34,000 cm2V-1s-1 for a 6-layer flake at low
temperatures, where pronounced Shubnikov-de Haas (SdH) oscillations were
observed, as shown in Figure 1.12. The SdH oscillations are the periodic oscillation
of the conductivity at low temperatures in the presence of a strong external
magnetic field, due to the quantization of energy levels. The SdH oscillation is a
macroscopic quantum phenomenon, which serves as direct evidence for the high
quality and homogeneity of a device.
Figure 1.12 Longitudinal magnetoresistance Rxx (red line) and Hall resistance Rxy (blue line) as
a function of the B field at T = 0.3 K for monolayer MoS2 encapsulated between two h-BN flakes.
The contacts are made of graphene. Inset: oscillation amplitude (black curve) as a function of
1/B after subtracting the magnetoresistance background. Red dashed line is the best fitting by
Dingle plot. The extracted quantum scattering time is 176 fs, corresponding to a mobility of 619
cm2V-1s-1. Adapted from Ref. 93.
Introduction
17
1.3 Germanane (GeH)
In addition to TMDs, novel layered materials composed of other group IV
elements such as silicon (Si) and germanium (Ge) have recently been discovered as
silicene [101,102] and germanene [103]. Although they have the same honeycomb
lattice as graphene, the 2D planes of silicene and germanene are periodically
buckled rather than being flat. Their electrical transport is dominated by the
unbonded pz orbitals. For silicene, both theoretical calculations [104–106] and
experimental observations [101,102,107,108] reveal that the band structure
exhibits a crossing at the K/K’ points of the Brillouin zone with a linear energy
dispersion near the crossing, thus the carriers behave like massless Dirac fermions,
making silicene almost equivalent to graphene. A field effect transistor was created
based on single-layer silicene [109], showing ambipolar transport operation with
an on/off ratio of ~10. For germanene, although theory has predicted similar
electronic band structure [105], no relevant transport study has been reported.
Overall, properties of silicene and germanene are very similar to graphene. The
lack of intrinsic band gap largely limits their applications. It should be noted that
silicene and germanene are not stable when exposed to air.
It has been shown that a band gap can be created in graphene by hydrogen
termination, the product is known as graphane [110]. Each C atom is covalently
bonded to three other C atoms in the ab-plane and one H atom in the c-direction.
Graphane is predicted to have a band gap of 3.5 eV by LDA calculations [111],
while a larger band gap of 5.4 eV is obtained by GW quasiparticle calculations [112].
In general, the gap calculated using the GW quasiparticle method is considered to
be more accurate, therefore hydrogenation transforms graphene into a good
insulator.
Similar to graphane, germanane (GeH) is obtained by adding covalently bonded
hydrogen atoms to both sides of germanene. The covalent bonding with hydrogen
atoms involves the pz orbitals of the Ge atoms; hence the electron transport from
the contribution of pz orbitals is significantly suppressed. As a result, electron
transport is dominated by the conduction band derived from s orbitals and the
valence band derived from the px and py orbitals near the Γ point [113]. A band gap
opens in germanane with a small effective mass at the K point of the Brillouin zone.
Theoretical calculations using the LDA method predict that germanane has a direct
band gap of about 1.5 eV at the K/K’ points [114,115]. Considering that the LDA
method normally underestimates the gap size, it is expected that the real band gap
for germanane is close to 3.0 eV [115,116]. Theoretically, the electron mobility is
predicted to be 18000 cm2V-1s-1, which is five times higher than bulk germanium
and ten times higher than silicon. Furthermore, the σ bond (composed of px and py
orbitals) dominates the electron transport in germanane, instead of the original π
Chapter 1
18
bond (composed of pz orbitals). Since the σ bond has a stronger spin-orbit coupling
(SOC) than the π bond, a non-trivial spin-orbit gap of ~0.2 eV is expected at the
valence band. Combined with the large direct band gap, the traditional spin-
selective optical process will also be applicable in germanane, providing attractive
potential for optoelectronic applications in the blue/violet spectral range.
To make this promising 2D material, a novel method has recently been
developed to synthesize large-scale multilayer germanane by topochemical
deintercalation from the CaGe2 precursor [117]. CaGe2 is prepared by reaction
between high-purity germanium (Ge) and calcium (Ca) at high temperatures in
vacuum. Then, CaGe2 is put into a concentrated HCl solution and stirred for a few
days at low temperatures. Finally, GeH is obtained by substituting Ca with H atoms
in CaGe2 following an ion exchange process that can be described by the following
chemical equation:
CaGe2 + HCl → GeH + CaCl2
In this way, millimeter-sized crystallites and gram-scale multilayer GeH can be
synthesized. It has been proved that the germanane grown in this manner is
resistant to oxygen and thermally stable in ambient conditions. Single- and few-
layer GeH can subsequently be obtained by mechanical exfoliation on the SiO2
substrate. The cleaved thin flakes have a structure similar to hydrogen-terminated
graphene. With hydrogen-termination, GeH becomes stable in an ambient
environment.
Figure 1.13 (a) Side view of the lattice structure of multi-layer germanane. Blue spheres
represent Ge atoms and black spheres represent H atoms. Each Ge atom is covalently bonded to
one H atom in the c-direction. (b) Structural diagram of crystalline germanium. Adapted from
Ref. 118.
Figure 1.13 (a) and (b) show the crystal structure of GeH and a bulk germanium
crystal, respectively. For the bulk germanium crystal, each Ge atom is bonded to
Introduction
19
four Ge atoms with sp3 hybridization, while for GeH, each Ge atom is bonded to
three Ge atoms in the ab plane and one H atom in the c-direction. In monolayer
GeH, Ge atoms form a buckled honeycomb lattice similar to graphene. Lattice
constants are a = 3.880 Å and c = 11.04 Å, with an interlayer distance of 5.5 Å.
Figure 1.14 (a) Picture of the synthesized bulk GeH crystals, they are black flakes with a typical
size of ~ 1 mm. (b) Optical image of a thin GeH flake on a 100 nm SiO2/Si wafer. (c) AFM image
(top) and height profile (bottom) of a monolayer flake, with thickness around 0.6 nm, which is
in line with the expected value of 0.55 nm. Adapted from Ref. 117.
Figure 1.14(a) shows a picture of the crystals grown using the method outlined
above. The dimensions of the flakes are typically one millimeter in diameter and a
few hundreds of micrometers in thickness. Similar to other layered materials, bulk
GeH is held together by weak van der Waals force, which means it can be
mechanically exfoliated into thin flakes. Optical microscopy, atomic force
microscopy (AFM) and Raman spectroscopy are used to characterize the GeH
crystal. Figure 1.14 shows the optical (b) and AFM (c) images of a 2 μm × 2 μm
monolayer GeH on a 100 nm SiO2/Si substrate. From the AFM height profile in the
bottom panel of Figure 1.14(c), we see that the thickness of the monolayer is ~0.6
nm. For thin flakes exfoliated on SiO2 substrate, the measured height is normally
Chapter 1
20
larger than the real thickness of sample due to the absorption of water molecules
and the different attractive potentials between sample flake, substrate and AFM tip.
Considering this effect, the measured height matches well with the theoretical
value of 0.55 nm.
Figure 1.15 Experimental Raman spectra of GeH measured at ambient conditions. Inset:
magnification of the featured peak of GeH around 300 cm-1, corresponding to the E2 vibration
mode of Ge-Ge. Adapted from Ref. 117.
Raman characterization has been a very useful tool to study the structures of
many 2D materials [119,120]. Figure 1.15 (blue curve) shows the typical Raman
spectra of GeH. The main peak appears at ~300 cm-1, which corresponds to the E2
mode of the Ge-Ge vibration in the in-plane direction [117]. Compared to the
crystalline germanium, this main vibration is slightly blue shifted but very close to
the 297 cm-1 Ge-Ge stretching mode of crystalline germanium, as plotted in the
inset of Figure 1.15(red curve). This suggests that the light H atom has a nearly
negligible influence on the vibration mode of GeH. The ab initio calculations of the
Γ-point phonon modes in GeH predict the emergence of Ge-based A1 and E2
Raman modes (assuming a C6v point group) at 223 and 289 cm-1. Experiment
observations indicate an additional vibrational mode at 228 cm-1, which matches
well with the theoretical prediction. The symmetries of the A1 and E2 vibrational
modes are shown schematically in the inset of Figure 1.15. It has been reported that
with the application of high pressure, the E2 stretching mode of the Ge-Ge shifts to
a higher frequency, suggesting that the peak position is strongly influenced by the
strain [121].
Introduction
21
1.4 Motivation and outline of this thesis
With aforementioned introduction, the general motivations of my thesis are as
follows:
(1) Recent studies showed that superconductivity can be induced in few-layer
MoS2 flakes by ionic liquid gating. The induced superconducting state shows
strong 2D nature: while it is easily suppressed under magnetic fields that are
perpendicular to the ab-plane of MoS2, it survives in very large magnetic fields that
are parallel to the ab-plane of MoS2. Theoretical calculations show that the
superconductivity is protected by the effective Zeeman field at the K valleys of the
Brillouin zone. This may suggest that the induced carriers are mainly confined to
the topmost layer hence it is well isolated electronically and acts like a free-
standing monolayer; superconductivity exists in this layer. The bottom layers of
the MoS2 flake remain intrinsically undoped. However, how to verify this
configuration is still a challenge—no direct experimental evidence has been
reported to explicitly prove this argument.
(2) Optical characterization has been widely used to study the electronic
properties of monolayer MoS2. However, optical measurement loses its power
when dealing with few-layer MoS2, because the optical transition efficiency is low
due to the indirect band gap characteristic. The study of the few-layer MoS2 is also
very important because it usually shows higher mobility compared to the
monolayer. Nevertheless, the electronic properties at the Q valleys of the
conduction band of few-layer MoS2 are much less studied. For this purpose,
electrical transport measurement becomes a very useful tool. Usually, MoS2 shows
n-type semiconductor characteristics due to the defects and unintentional doping
during device fabrication etc., so that the Fermi level can be easily shifted to access
the conduction band by the field effect. However, electrical transport measurement
still remains a challenge, because (a) the quality of MoS2 is usually degraded
during device fabrication, how to make high-quality MoS2 devices with high-
mobility is a long pursuing goal for the research community of this field; (b) the
ohmic contact required for the electrical transport measurement is not easy to
accomplish, due to the Schottky barrier between the semiconducting MoS2 and the
metal electrodes.
(3) Realizing continuous tuning between superconducting and non-
superconducting states has attracted both scientific and technological interest. In
some superconducting systems, this can be achieved by adjusting static parameters
such as film thickness, disorder concentration and chemical doping; but they are
not practically efficient. From the perspective of real applications, the best way is
to use field effect tuning. Nevertheless, previous attempts of field effect tuning of
Chapter 1
22
superconductivity either require very high gate voltages or ultralow temperatures.
Regarding this concern, superconductivity induced in a MoS2-ELDT device shows
very promising potential because of the well-defined crystal structure, device
fabrication flexibility and relatively high transition temperature (helium-4
temperature range). Yet, how to achieve this goal remains a challenge.
(4) The boom of graphene initiated the studies of other novel low-dimensional
materials. Germanene—the germanium version of graphene—is very attractive
because it is highly compatible with the current silicon- or germanium-based
technology, showing great potential for making 2D electronic devices. However,
germanene is a semimetal with zero band gap and it is not stable in air, hence it is
difficult to be applied in practical devices. These drawbacks can be overcome by
adding hydrogen termination on both sides of germanene, which is termed as
germanane (chemical formula: GeH). Germanane is stable under ambient
conditions, and more importantly, a gap opens in the band structure. The
structural, thermal and optical properties have been studied, but very limited
reports are related to its electrical transport properties. Fundamental questions
include: whether germanane is an insulator or a conductor, whether it is a good
candidate for field effect transistors and how strongly the spin-orbit coupling
affects the electrical conductivity.
(5) Besides the widely explored 2D materials, 1D structures represented by
carbon nanotubes (CNTs) are regarded as another basic unit for the future
electronic industry. CNTs have been intensively studied since their discovery in
1991, owing to their unique physical properties and the 1D character. Much effort
has been devoted to the synthesis of single-walled carbon nanotubes (SWCNTs)
with desired chirality, at a given location and with well-controlled direction and
length. It was reported that ultra-small SWCNTs can be produced inside the linear
pores of aluminophosphate, AlPO4-5 (AFI) zeolite single crystals, and the
diameters of the SWCNTs were determined to be 0.4 nm by high-resolution
transmission electron microscopy. However, the overall quality of the
SWCNTs@AFI is poor. How to improve the quality still needs to be explored.
Furthermore, with such a small diameter and extremely large curvature, these
well-aligned and mono-sized SWCNT arrays showed interesting physical
properties, e.g. evidence for superconductivity has been observed in this system. It
is well-known that for CNTs, there are Van Hove singularities in the band structure,
where the density of states (DOS) is very high. It is interesting to explore the
transport properties of these ultra-small SWCNTs; especially when combined with
the ionic liquid gating technique, the Fermi level can be effectively shifted to access
the Van Hove singularities where fascinating phenomena could appear.
Introduction
23
To tackle all the questions and challenges above, in the main text of this thesis I
study the quantum transport properties of MoS2 and GeH transistors, focusing on
controlling the electronic properties of these 2D materials using field effect gating.
This part of work was mainly accomplished at the University of Groningen (RUG).
The Appendix discusses my earlier works about the fabrication of 0.4 nm-SWCNTs
in the linear pores of the AFI crystals, which was carried out at the Hong Kong
University of Science and Technology (HKUST). The outline of the thesis is as
follows:
Chapter 2: It has been argued that in a few-layer MoS2-EDLT device, the ionic
liquid gating confines carriers mostly to the topmost layer, while the bottom layers
stay almost unaffected. To verify this model, we fabricate dual-gate MoS2
transistors, combining the ionic liquid top gate and solid state back gate. In this
configuration, ionic liquid gate induces a large carrier density in the topmost layer,
which becomes superconducting. The carrier density of the bottom layers can be
well controlled by the back gate. Because the carrier density induced by back gate
is much smaller, the bottom layers cannot reach superconducting states according
to the established phase diagram in the previous study. As a result, two parallel
conducting channels form in a single few-layer MoS2 crystal: the top
superconducting channel and the bottom normal channel. The electrical
conductivity contributed by the top channel can be characterized by the
superconducting transition at low temperatures. In order to evaluate the electrical
contribution by the bottom conducting channel, we use the h-BN on a SiO2/Si
wafer as the substrate for the dual-gate MoS2 transistor. In this way, the carrier
mobility in the bottom channel is greatly improved, which leads to the observation
of the Shubnikov-de Haas (SdH) oscillations. The behavior of the
superconductivity and SdH oscillations are carefully characterized. Electronic
properties such as the effective mass of charge carriers, Landau level degeneracy
and spin splitting texture at the conduction band are discussed.
Chapter 3: To tune the superconductivity in MoS2-EDLT by field effect, another
device is fabricated on the HfO2 substrate. We use HfO2 substrate because the
dielectric constant of HfO2 is high which means the carrier density can be tuned in
a large amount. Ionic liquid gate induces superconductivity in the topmost layer
and we use the back gate to tune the superconductivity at low temperatures. At
different back gate voltages, the superconducting properties including transition
temperatures, critical fields and critical currents are characterized. By applying
different ionic gate voltages, the superconductivity in the topmost layer is prepared
in two different states labeled as I and II. Specifically, state I with a lower carrier
density, which is very close to the quantum critical point of the dome-shaped phase
diagram (onset of superconductivity); and state II with a higher carrier density,
Chapter 1
24
which is deep into the superconducting dome. For both states, responses with
respect to different back gate voltages and the physical mechanisms of tuning
superconductivity are discussed.
Chapter 4: This chapter deals with the intrinsic properties of few-layer MoS2
that is not doped by ionic gating. For the device fabrication, a few-layer MoS2-
EDLT is prepared on the h-BN substrate. Compared to the device configuration in
chapter 2, the difference is that a top h-BN flake is used to partially separate the
ionic liquid from the MoS2 surface. As a result, the h-BN covered MoS2 surface is
not doped by ionic gating and remains intrinsically semiconducting, while the
exposed MoS2 surface becomes highly doped and superconducting. The
conductivity of h-BN covered MoS2 surface can be tuned by the back gate. SdH
oscillations are observed in this channel and the electronic properties of the
conduction band of few-layer MoS2 are discussed. At the interface between the
exposed and the h-BN covered MoS2 surface, a superconductor-normal metal (SN)
junction forms. The transport properties at the SN interface and the probabilities
of Andreev reflections are studied.
Chapter 5: To study the electronic transport properties of 2D hydrogen-
terminated germanene (GeH) as well as search for new device applications, we
fabricate field effect transistors based on thin GeH flakes and measure their
electrical transport properties at low temperatures and under magnetic fields. The
samples include pristine and thermally annealed flakes, whose FET characteristics
with respect to both solid and ionic liquid gating are studied. At last, the spin-orbit
interaction of GeH at low temperatures is discussed. The quantum phase
coherence length is obtained from the weak antilocalization behavior.
Appendix: In the first half of my PhD study, I focused on the synthesis of ultra-
small SWCNTs in the linear channels of AlPO4-5 (AFI) zeolite crystals
(SWCNTs@AFI). Different from the old method which uses the organic template
inside the linear pores of AFI crystals as the carbon source for growing SWCNTs, a
new CVD process is developed. Various parameters are tested for growing the
SWCNTs@AFI including heating temperatures, pressures and different feedstock
gases such as ethylene and carbon monoxide. The products are characterized by
the optical microscopy, Raman spectroscopy and Thermo-gravimetry analyses,
based on which the quality and quantity of the SWCNTs inside the AFI channels
are discussed. Simulations are carried out to study the vibrational modes of
different chirality of the ultra-small SWCNTs. The possibility of developing a
chirality-selective process for growing the SWCNTs@AFI is discussed.
Introduction
25
References
[1] K.S. Novoselov, A.K. Geim, S.V. Morozov, D. Jiang, Y. Zhang, S.V. Dubonos, I.V.
Grigorieva, A.A. Firsov, Electric Field Effect in Atomically Thin Carbon Films, Science.
306 (2004) 666–669.
[2] G. Tsoukleri, J. Parthenios, K. Papagelis, R. Jalil, A.C. Ferrari, A.K. Geim, K.S. Novoselov,
C. Galiotis, Subjecting a Graphene Monolayer to Tension and Compression, Small. 5 (2009)
2397–2402.
[3] J.S. Bunch, S.S. Verbridge, J.S. Alden, A.M. van der Zande, J.M. Parpia, H.G. Craighead,
P.L. McEuen, Impermeable Atomic Membranes from Graphene Sheets, Nano Lett. 8 (2008)
2458–2462.
[4] E. Pop, V. Varshney, A.K. Roy, Thermal properties of graphene: Fundamentals and
applications, MRS Bull. 37 (2012) 1273–1281.
[5] A.K. Geim, Graphene: Status and Prospects, Science. 324 (2009) 1530–1534.
[6] D.C. Elias, R.V. Gorbachev, A.S. Mayorov, S.V. Morozov, A.A. Zhukov, P. Blake, L.A.
Ponomarenko, I.V. Grigorieva, K.S. Novoselov, F. Guinea, A.K. Geim, Dirac cones
reshaped by interaction effects in suspended graphene, Nat. Phys. 7 (2011) 701–704.
[7] L.F. Mattheiss, Band Structures of Transition-Metal-Dichalcogenide Layer Compounds,
Phys. Rev. B. 8 (1973) 3719–3740.
[8] A.D. Yoffe, Layer Compounds, Annu. Rev. Mater. Sci. 3 (1973) 147–170.
[9] A.D. Yoffe, Low-dimensional systems: quantum size effects and electronic properties of
semiconductor microcrystallites (zero-dimensional systems) and some quasi-two-
dimensional systems, Adv. Phys. 42 (1993) 173–262.
[10] Q.H. Wang, K. Kalantar-Zadeh, A. Kis, J.N. Coleman, M.S. Strano, Electronics and
optoelectronics of two-dimensional transition metal dichalcogenides, Nat. Nanotechnol. 7
(2012) 699–712.
[11] J.A. Wilson, A.D. Yoffe, The transition metal dichalcogenides discussion and interpretation
of the observed optical, electrical and structural properties, Adv. Phys. 18 (1969) 193–335.
[12] K.F. Mak, J. Shan, Photonics and optoelectronics of 2D semiconductor transition metal
dichalcogenides, Nat. Photonics. 10 (2016) 216–226.
[13] A. Pospischil, T. Mueller, Optoelectronic Devices Based on Atomically Thin Transition
Metal Dichalcogenides, Appl. Sci. 6 (2016) 78.
[14] H. Tian, M.L. Chin, S. Najmaei, Q. Guo, F. Xia, H. Wang, M. Dubey, Optoelectronic
devices based on two-dimensional transition metal dichalcogenides, Nano Res. 9 (2016)
1543–1560.
[15] S. Bertolazzi, J. Brivio, A. Kis, Stretching and Breaking of Ultrathin MoS2, ACS Nano. 5
(2011) 9703–9709.
[16] C. Ataca, H. Şahin, S. Ciraci, Stable, Single-Layer MX2 Transition-Metal Oxides and
Dichalcogenides in a Honeycomb-Like Structure, J. Phys. Chem. C. 116 (2012) 8983–8999.
[17] J. Pu, Y. Yomogida, K.-K. Liu, L.-J. Li, Y. Iwasa, T. Takenobu, Highly Flexible MoS2 Thin-
Film Transistors with Ion Gel Dielectrics, Nano Lett. 12 (2012) 4013–4017.
[18] M.A. Khan, A. Bhattarai, J.N. Kuznia, D.T. Olson, High electron mobility transistor based
on a GaN‐AlxGa1−xN heterojunction, Appl. Phys. Lett. 63 (1993) 1214–1215.
Chapter 1
26
[19] X. Xi, Z. Wang, W. Zhao, J.-H. Park, K.T. Law, H. Berger, L. Forró, J. Shan, K.F. Mak,
Ising pairing in superconducting NbSe2 atomic layers, Nat. Phys. 12 (2016) 139–143.
[20] J.A. Wilson, F.J. Di Salvo, S. Mahajan, Charge-Density Waves in Metallic, Layered,
Transition-Metal Dichalcogenides, Phys. Rev. Lett. 32 (1974) 882–885.
[21] R.B. Somoano, V. Hadek, A. Rembaum, S. Samson, J.A. Woollam, The alkaline earth
intercalates of molybdenum disulfide, J. Chem. Phys. 62 (1975) 1068–1073.
[22] J.A. Woollam, R.B. Somoano, Physics and chemistry of MoS2 intercalation compounds,
Mater. Sci. Eng. 31 (1977) 289–295.
[23] T. Heine, Transition Metal Chalcogenides: Ultrathin Inorganic Materials with Tunable
Electronic Properties, Acc. Chem. Res. 48 (2015) 65–72.
[24] Z.Y. Zhu, Y.C. Cheng, U. Schwingenschlögl, Giant spin-orbit-induced spin splitting in two-
dimensional transition-metal dichalcogenide semiconductors, Phys. Rev. B. 84 (2011)
153402.
[25] X. Xu, W. Yao, D. Xiao, T.F. Heinz, Spin and pseudospins in layered transition metal
dichalcogenides, Nat. Phys. 10 (2014) 343–350.
[26] A. Kormányos, V. Zólyomi, N.D. Drummond, G. Burkard, Spin-Orbit Coupling, Quantum
Dots, and Qubits in Monolayer Transition Metal Dichalcogenides, Phys. Rev. X. 4 (2014)
11034.
[27] D. Xiao, G.-B. Liu, W. Feng, X. Xu, W. Yao, Coupled Spin and Valley Physics in
Monolayers of MoS2 and Other Group-VI Dichalcogenides, Phys. Rev. Lett. 108 (2012)
196802.
[28] G.D. Scholes, G. Rumbles, Excitons in nanoscale systems, Nat. Mater. 5 (2006) 683–696.
[29] M. Law, J. Goldberger, P. Yang, Semiconductor Nanowires and Nanotubes, Annu. Rev.
Mater. Res. 34 (2004) 83–122.
[30] P.V. Kamat, Quantum Dot Solar Cells. Semiconductor Nanocrystals as Light Harvesters, J.
Phys. Chem. C. 112 (2008) 18737–18753.
[31] Z. Yin, H. Li, H. Li, L. Jiang, Y. Shi, Y. Sun, G. Lu, Q. Zhang, X. Chen, H. Zhang, Single-
Layer MoS2 Phototransistors, ACS Nano. 6 (2012) 74–80.
[32] H.S. Lee, S.-W. Min, Y.-G. Chang, M.K. Park, T. Nam, H. Kim, J.H. Kim, S. Ryu, S. Im,
MoS2 Nanosheet Phototransistors with Thickness-Modulated Optical Energy Gap, Nano Lett.
12 (2012) 3695–3700.
[33] O. Lopez-Sanchez, D. Lembke, M. Kayci, A. Radenovic, A. Kis, Ultrasensitive
photodetectors based on monolayer MoS2, Nat. Nanotechnol. 8 (2013) 497–501.
[34] D. Kufer, I. Nikitskiy, T. Lasanta, G. Navickaite, F.H.L. Koppens, G. Konstantatos, Hybrid
2D–0D MoS2–PbS Quantum Dot Photodetectors, Adv. Mater. 27 (2015) 176–180.
[35] A. Pospischil, M.M. Furchi, T. Mueller, Solar-energy conversion and light emission in an
atomic monolayer p-n diode, Nat. Nanotechnol. 9 (2014) 257–261.
[36] F. Withers, O. Del Pozo-Zamudio, A. Mishchenko, A.P. Rooney, A. Gholinia, K. Watanabe,
T. Taniguchi, S.J. Haigh, A.K. Geim, A.I. Tartakovskii, K.S. Novoselov, Light-emitting
diodes by band-structure engineering in van der Waals heterostructures, Nat. Mater. 14
(2015) 301–306.
[37] S. Wu, S. Buckley, J.R. Schaibley, L. Feng, J. Yan, D.G. Mandrus, F. Hatami, W. Yao, J.
Vučković, A. Majumdar, X. Xu, Monolayer semiconductor nanocavity lasers with ultralow
thresholds, Nature. 520 (2015) 69–72.
Introduction
27
[38] Y. Ye, Z.J. Wong, X. Lu, X. Ni, H. Zhu, X. Chen, Y. Wang, X. Zhang, Monolayer excitonic
laser, Nat. Photonics. 9 (2015) 733–737.
[39] B.W.H. Baugher, H.O.H. Churchill, Y. Yang, P. Jarillo-Herrero, Optoelectronic devices
based on electrically tunable p-n diodes in a monolayer dichalcogenide, Nat. Nanotechnol. 9
(2014) 262–267.
[40] M.M. Furchi, A. Pospischil, F. Libisch, J. Burgdörfer, T. Mueller, Photovoltaic Effect in an
Electrically Tunable van der Waals Heterojunction, Nano Lett. 14 (2014) 4785–4791.
[41] C.-H. Lee, G.-H. Lee, A.M. van der Zande, W. Chen, Y. Li, M. Han, X. Cui, G. Arefe, C.
Nuckolls, T.F. Heinz, J. Guo, J. Hone, P. Kim, Atomically thin p–n junctions with van der
Waals heterointerfaces, Nat. Nanotechnol. 9 (2014) 676–681.
[42] Y.J. Zhang, T. Oka, R. Suzuki, J.T. Ye, Y. Iwasa, Electrically Switchable Chiral Light-
Emitting Transistor, Science. 344 (2014) 725–728.
[43] K.F. Mak, K.L. McGill, J. Park, P.L. McEuen, The valley Hall effect in MoS2 transistors,
Science. 344 (2014) 1489–1492.
[44] R. Suzuki, M. Sakano, Y.J. Zhang, R. Akashi, D. Morikawa, A. Harasawa, K. Yaji, K.
Kuroda, K. Miyamoto, T. Okuda, K. Ishizaka, R. Arita, Y. Iwasa, Valley-dependent spin
polarization in bulk MoS2 with broken inversion symmetry, Nat. Nanotechnol. 9 (2014)
611–617.
[45] K.F. Mak, C. Lee, J. Hone, J. Shan, T.F. Heinz, Atomically Thin MoS2: A New Direct-Gap
Semiconductor, Phys. Rev. Lett. 105 (2010) 136805.
[46] A. Kuc, N. Zibouche, T. Heine, Influence of quantum confinement on the electronic
structure of the transition metal sulfide TS2, Phys. Rev. B. 83 (2011) 245213.
[47] A. Splendiani, L. Sun, Y. Zhang, T. Li, J. Kim, C.-Y. Chim, G. Galli, F. Wang, Emerging
Photoluminescence in Monolayer MoS2, Nano Lett. 10 (2010) 1271–1275.
[48] K.F. Mak, K. He, J. Shan, T.F. Heinz, Control of valley polarization in monolayer MoS2 by
optical helicity, Nat. Nanotechnol. 7 (2012) 494–498.
[49] M. Amani, D.-H. Lien, D. Kiriya, J. Xiao, A. Azcatl, J. Noh, S.R. Madhvapathy, R. Addou,
S. Kc, M. Dubey, K. Cho, R.M. Wallace, S.-C. Lee, J.-H. He, J.W. Ager, X. Zhang, E.
Yablonovitch, A. Javey, Near-unity photoluminescence quantum yield in MoS2, Science.
350 (2015) 1065–1068.
[50] H.-V. Han, A.-Y. Lu, L.-S. Lu, J.-K. Huang, H. Li, C.-L. Hsu, Y.-C. Lin, M.-H. Chiu, K.
Suenaga, C.-W. Chu, H.-C. Kuo, W.-H. Chang, L.-J. Li, Y. Shi, Photoluminescence
Enhancement and Structure Repairing of Monolayer MoSe2 by Hydrohalic Acid Treatment,
ACS Nano. 10 (2016) 1454–1461.
[51] A. Kormányos, G. Burkard, M. Gmitra, J. Fabian, V. Zólyomi, N.D. Drummond, Vladimir
Fal’ko, k · p theory for two-dimensional transition metal dichalcogenide semiconductors, 2D
Mater. 2 (2015) 22001.
[52] A. Chernikov, T.C. Berkelbach, H.M. Hill, A. Rigosi, Y. Li, O.B. Aslan, D.R. Reichman,
M.S. Hybertsen, T.F. Heinz, Exciton Binding Energy and Nonhydrogenic Rydberg Series in
Monolayer WS2, Phys. Rev. Lett. 113 (2014) 76802.
[53] S.A. Wolf, D.D. Awschalom, R.A. Buhrman, J.M. Daughton, S. von Molnár, M.L. Roukes,
A.Y. Chtchelkanova, D.M. Treger, Spintronics: A Spin-Based Electronics Vision for the
Future, Science. 294 (2001) 1488–1495.
[54] F. Pulizzi, Spintronics, Nat. Mater. 11 (2012) 367–367.
Chapter 1
28
[55] H. Zeng, J. Dai, W. Yao, D. Xiao, X. Cui, Valley polarization in MoS2 monolayers by
optical pumping, Nat. Nanotechnol. 7 (2012) 490–493.
[56] T. Cao, G. Wang, W. Han, H. Ye, C. Zhu, J. Shi, Q. Niu, P. Tan, E. Wang, B. Liu, J. Feng,
Valley-selective circular dichroism of monolayer molybdenum disulphide, Nat. Commun. 3
(2012) 887.
[57] A.R. Botello-Méndez, F. López-Urías, M. Terrones, H. Terrones, Metallic and
ferromagnetic edges in molybdenum disulfide nanoribbons, Nanotechnology. 20 (2009)
325703.
[58] S.Z. Butler, S.M. Hollen, L. Cao, Y. Cui, J.A. Gupta, H.R. Gutiérrez, T.F. Heinz, S.S. Hong,
J. Huang, A.F. Ismach, E. Johnston-Halperin, M. Kuno, V.V. Plashnitsa, R.D. Robinson, R.S.
Ruoff, S. Salahuddin, J. Shan, L. Shi, M.G. Spencer, M. Terrones, W. Windl, J.E.
Goldberger, Progress, Challenges, and Opportunities in Two-Dimensional Materials Beyond
Graphene, ACS Nano. 7 (2013) 2898–2926.
[59] A.M. Jones, H. Yu, N.J. Ghimire, S. Wu, G. Aivazian, J.S. Ross, B. Zhao, J. Yan, D.G.
Mandrus, D. Xiao, W. Yao, X. Xu, Optical generation of excitonic valley coherence in
monolayer WSe2, Nat. Nanotechnol. 8 (2013) 634–638.
[60] G. Sallen, L. Bouet, X. Marie, G. Wang, C.R. Zhu, W.P. Han, Y. Lu, P.H. Tan, T. Amand,
B.L. Liu, B. Urbaszek, Robust optical emission polarization in MoS2 monolayers through
selective valley excitation, Phys. Rev. B. 86 (2012) 81301.
[61] Sze, S. M., Ng, K. K., Physics of semiconductor devices, Wiley, 2007.
[62] P. Avouris, Z. Chen, V. Perebeinos, Carbon-based electronics, Nat. Nanotechnol. 2 (2007)
605–615.
[63] P. Avouris, M. Freitag, V. Perebeinos, Carbon-nanotube photonics and optoelectronics, Nat.
Photonics. 2 (2008) 341–350.
[64] W. Lu, C.M. Lieber, Nanoelectronics from the bottom up, Nat. Mater. 6 (2007) 841–850.
[65] Y. Wu, K.A. Jenkins, A. Valdes-Garcia, D.B. Farmer, Y. Zhu, A.A. Bol, C.
Dimitrakopoulos, W. Zhu, F. Xia, P. Avouris, Y.-M. Lin, State-of-the-Art Graphene High-
Frequency Electronics, Nano Lett. 12 (2012) 3062–3067.
[66] Y.-M. Lin, A. Valdes-Garcia, S.-J. Han, D.B. Farmer, I. Meric, Y. Sun, Y. Wu, C.
Dimitrakopoulos, A. Grill, P. Avouris, K.A. Jenkins, Wafer-Scale Graphene Integrated
Circuit, Science. 332 (2011) 1294–1297.
[67] F. Schwierz, Graphene transistors, Nat. Nanotechnol. 5 (2010) 487–496.
[68] Y. Wu, Y. Lin, A.A. Bol, K.A. Jenkins, F. Xia, D.B. Farmer, Y. Zhu, P. Avouris, High-
frequency, scaled graphene transistors on diamond-like carbon, Nature. 472 (2011) 74–78.
[69] L. Liu, S.B. Kumar, Y. Ouyang, J. Guo, Performance Limits of Monolayer Transition Metal
Dichalcogenide Transistors, IEEE Trans. Electron Devices. 58 (2011) 3042–3047.
[70] Y. Yoon, K. Ganapathi, S. Salahuddin, How Good Can Monolayer MoS2 Transistors Be?
Nano Lett. 11 (2011) 3768–3773.
[71] K.S. Novoselov, D. Jiang, F. Schedin, T.J. Booth, V.V. Khotkevich, S.V. Morozov, A.K.
Geim, Two-dimensional atomic crystals, Proc. Natl. Acad. Sci. U. S. A. 102 (2005) 10451–
10453.
[72] A. Ayari, E. Cobas, O. Ogundadegbe, M.S. Fuhrer, Realization and electrical
characterization of ultrathin crystals of layered transition-metal dichalcogenides, J. Appl.
Phys. 101 (2007) 14507.
Introduction
29
[73] K. Lee, H.-Y. Kim, M. Lotya, J.N. Coleman, G.-T. Kim, G.S. Duesberg, Electrical
Characteristics of Molybdenum Disulfide Flakes Produced by Liquid Exfoliation, Adv.
Mater. 23 (2011) 4178–4182.
[74] B. Radisavljevic, A. Radenovic, J. Brivio, V. Giacometti, A. Kis, Single-layer MoS2
transistors, Nat. Nanotechnol. 6 (2011) 147–150.
[75] B. Radisavljevic, M.B. Whitwick, A. Kis, Integrated Circuits and Logic Operations Based
on Single-Layer MoS2, ACS Nano. 5 (2011) 9934–9938.
[76] H. Wang, L. Yu, Y.-H. Lee, Y. Shi, A. Hsu, M.L. Chin, L.-J. Li, M. Dubey, J. Kong, T.
Palacios, Integrated Circuits Based on Bilayer MoS2 Transistors, Nano Lett. 12 (2012)
4674–4680.
[77] M.A.V. Devanathan, B.V.K.S.R.A. Tilak, The Structure of the Electrical Double Layer at
the Metal-Solution Interface, Chem. Rev. 65 (1965) 635–684.
[78] H. Shimotani, H. Asanuma, A. Tsukazaki, A. Ohtomo, M. Kawasaki, Y. Iwasa, Insulator-to-
metal transition in ZnO by electric double layer gating, Appl. Phys. Lett. 91 (2007) 82106.
[79] Y. Zhang, J. Ye, Y. Matsuhashi, Y. Iwasa, Ambipolar MoS2 Thin Flake Transistors, Nano
Lett. 12 (2012) 1136–1140.
[80] S. Ono, N. Minder, Z. Chen, A. Facchetti, A.F. Morpurgo, High-performance n-type organic
field-effect transistors with ionic liquid gates, Appl. Phys. Lett. 97 (2010) 143307.
[81] K. Hong, S.H. Kim, K.H. Lee, C.D. Frisbie, Printed, sub-2V ZnO Electrolyte Gated
Transistors and Inverters on Plastic, Adv. Mater. 25 (2013) 3413–3418.
[82] K. Ueno, S. Nakamura, H. Shimotani, A. Ohtomo, N. Kimura, T. Nojima, H. Aoki, Y. Iwasa,
M. Kawasaki, Electric-field-induced superconductivity in an insulator, Nat. Mater. 7 (2008)
855–858.
[83] J.T. Ye, S. Inoue, K. Kobayashi, Y. Kasahara, H.T. Yuan, H. Shimotani, Y. Iwasa, Liquid-
gated interface superconductivity on an atomically flat film, Nat. Mater. 9 (2010) 125–128.
[84] K. Ueno, S. Nakamura, H. Shimotani, H.T. Yuan, N. Kimura, T. Nojima, H. Aoki, Y. Iwasa,
M. Kawasaki, Discovery of superconductivity in KTaO3 by electrostatic carrier doping, Nat.
Nanotechnol. 6 (2011) 408–412.
[85] J. Ye, M.F. Craciun, M. Koshino, S. Russo, S. Inoue, H. Yuan, H. Shimotani, A.F.
Morpurgo, Y. Iwasa, Accessing the transport properties of graphene and its multilayers at
high carrier density, Proc. Natl. Acad. Sci. 108 (2011) 13002–13006.
[86] V. Podzorov, M.E. Gershenson, C. Kloc, R. Zeis, E. Bucher, High-mobility field-effect
transistors based on transition metal dichalcogenides, Appl. Phys. Lett. 84 (2004) 3301–3303.
[87] H. Yuan, M.S. Bahramy, K. Morimoto, S. Wu, K. Nomura, B.-J. Yang, H. Shimotani, R.
Suzuki, M. Toh, C. Kloc, X. Xu, R. Arita, N. Nagaosa, Y. Iwasa, Zeeman-type spin splitting
controlled by an electric field, Nat. Phys. 9 (2013) 563–569.
[88] S. Jo, D. Costanzo, H. Berger, A.F. Morpurgo, Electrostatically Induced Superconductivity
at the Surface of WS2, Nano Lett. 15 (2015) 1197–1202.
[89] J.T. Ye, Y.J. Zhang, R. Akashi, M.S. Bahramy, R. Arita, Y. Iwasa, Superconducting Dome
in a Gate-Tuned Band Insulator, Science. 338 (2012) 1193–1196.
[90] J.M. Lu, O. Zheliuk, I. Leermakers, N.F.Q. Yuan, U. Zeitler, K.T. Law, J.T. Ye, Evidence
for two-dimensional Ising superconductivity in gated MoS2, Science. (2015) aab2277.
[91] S. Kim, A. Konar, W.-S. Hwang, J.H. Lee, J. Lee, J. Yang, C. Jung, H. Kim, J.-B. Yoo, J.-Y.
Choi, Y.W. Jin, S.Y. Lee, D. Jena, W. Choi, K. Kim, High-mobility and low-power thin-film
transistors based on multilayer MoS2 crystals, Nat. Commun. 3 (2012) 1011.
Chapter 1
30
[92] T. Jin, J. Kang, E.S. Kim, S. Lee, C. Lee, Suspended single-layer MoS2 devices, J. Appl.
Phys. 114 (2013) 164509.
[93] X. Cui, G.-H. Lee, Y.D. Kim, G. Arefe, P.Y. Huang, C.-H. Lee, D.A. Chenet, X. Zhang, L.
Wang, F. Ye, F. Pizzocchero, B.S. Jessen, K. Watanabe, T. Taniguchi, D.A. Muller, T. Low,
P. Kim, J. Hone, Multi-terminal transport measurements of MoS2 using a van der Waals
heterostructure device platform, Nat. Nanotechnol. 10 (2015) 534–540.
[94] C.R. Dean, A.F. Young, I. Meric, C. Lee, L. Wang, S. Sorgenfrei, K. Watanabe, T.
Taniguchi, P. Kim, K.L. Shepard, J. Hone, Boron nitride substrates for high-quality
graphene electronics, Nat. Nanotechnol. 5 (2010) 722–726.
[95] L. Wang, I. Meric, P.Y. Huang, Q. Gao, Y. Gao, H. Tran, T. Taniguchi, K. Watanabe, L.M.
Campos, D.A. Muller, J. Guo, P. Kim, J. Hone, K.L. Shepard, C.R. Dean, One-Dimensional
Electrical Contact to a Two-Dimensional Material, Science. 342 (2013) 614–617.
[96] A. Castellanos-Gomez, M. Buscema, R. Molenaar, V. Singh, L. Janssen, H.S.J. van der Zant,
G.A. Steele, Deterministic transfer of two-dimensional materials by all-dry viscoelastic
stamping, 2D Mater. 1 (2014) 11002.
[97] J. Chang, Q. Zhou, A. Zettl, Facile electron-beam lithography technique for irregular and
fragile substrates, Appl. Phys. Lett. 105 (2014) 173109.
[98] P.J. Zomer, M.H.D. Guimarães, J.C. Brant, N. Tombros, B.J. van Wees, Fast pick up
technique for high quality heterostructures of bilayer graphene and hexagonal boron nitride,
Appl. Phys. Lett. 105 (2014) 13101.
[99] Z. Wu, Y. Han, J. Lin, W. Zhu, M. He, S. Xu, X. Chen, H. Lu, W. Ye, T. Han, Y. Wu, G.
Long, J. Shen, R. Huang, L. Wang, Y. He, Y. Cai, R. Lortz, D. Su, N. Wang, Detection of
interlayer interaction in few-layer graphene, Phys. Rev. B. 92 (2015) 75408.
[100] Z. Wu, S. Xu, H. Lu, G.-B. Liu, A. Khamoshi, T. Han, Y. Wu, J. Lin, G. Long, Y. He, Y.
Cai, F. Zhang, N. Wang, Observation of Valley Zeeman and Quantum Hall Effects at Q
Valley of Few-Layer Transition Metal Disulfides, ArXiv151100077 Cond-Mat. (2015).
[101] B. Aufray, A. Kara, S. Vizzini, H. Oughaddou, C. Léandri, B. Ealet, G.L. Lay, Graphene-
like silicon nanoribbons on Ag(110): A possible formation of silicene, Appl. Phys. Lett. 96
(2010) 183102.
[102] B. Lalmi, H. Oughaddou, H. Enriquez, A. Kara, S. Vizzini, B. Ealet, B. Aufray, Epitaxial
growth of a silicene sheet, Appl. Phys. Lett. 97 (2010) 223109.
[103] M.E. Dávila, L. Xian, S. Cahangirov, A. Rubio, G.L. Lay, Germanene: a novel two-
dimensional germanium allotrope akin to graphene and silicene, New J. Phys. 16 (2014)
95002.
[104] G.G. Guzmán-Verri, L.C. Lew Yan Voon, Electronic structure of silicon-based
nanostructures, Phys. Rev. B. 76 (2007) 75131.
[105] S. Lebègue, O. Eriksson, Electronic structure of two-dimensional crystals from ab initio
theory, Phys. Rev. B. 79 (2009) 115409.
[106] Z. Ni, Q. Liu, K. Tang, J. Zheng, J. Zhou, R. Qin, Z. Gao, D. Yu, J. Lu, Tunable Bandgap in
Silicene and Germanene, Nano Lett. 12 (2012) 113–118.
[107] L. Chen, C.-C. Liu, B. Feng, X. He, P. Cheng, Z. Ding, S. Meng, Y. Yao, K. Wu, Evidence
for Dirac Fermions in a Honeycomb Lattice Based on Silicon, Phys. Rev. Lett. 109 (2012)
56804.
[108] P. Vogt, P. De Padova, C. Quaresima, J. Avila, E. Frantzeskakis, M.C. Asensio, A. Resta, B.
Ealet, G. Le Lay, Silicene: Compelling Experimental Evidence for Graphenelike Two-
Dimensional Silicon, Phys. Rev. Lett. 108 (2012) 155501.
Introduction
31
[109] L. Tao, E. Cinquanta, D. Chiappe, C. Grazianetti, M. Fanciulli, M. Dubey, A. Molle, D.
Akinwande, Silicene field-effect transistors operating at room temperature, Nat.
Nanotechnol. 10 (2015) 227–231.
[110] D.C. Elias, R.R. Nair, T.M.G. Mohiuddin, S.V. Morozov, P. Blake, M.P. Halsall, A.C.
Ferrari, D.W. Boukhvalov, M.I. Katsnelson, A.K. Geim, K.S. Novoselov, Control of
Graphene’s Properties by Reversible Hydrogenation: Evidence for Graphane, Science. 323
(2009) 610–613.
[111] J.O. Sofo, A.S. Chaudhari, G.D. Barber, Graphane: A two-dimensional hydrocarbon, Phys.
Rev. B. 75 (2007) 153401.
[112] S. Lebègue, M. Klintenberg, O. Eriksson, M.I. Katsnelson, Accurate electronic band gap of
pure and functionalized graphane from GW calculations, Phys. Rev. B. 79 (2009) 245117.
[113] W. Amamou, P.M. Odenthal, E.J. Bushong, D.J. O’Hara, Y.K. Luo, J. van Baren, I. Pinchuk,
Y. Wu, A.S. Ahmed, J. Katoch, M.W. Bockrath, H.W.K. Tom, J.E. Goldberger, R.K.
Kawakami, Large area epitaxial germanane for electronic devices, 2D Mater. 2 (2015)
35012.
[114] X. Luo, E. Zurek, Crystal Structures and Electronic Properties of Single-Layer, Few-Layer,
and Multilayer GeH, J. Phys. Chem. C. 120 (2016) 793–800.
[115] L.C.L.Y. Voon, E. Sandberg, R.S. Aga, A.A. Farajian, Hydrogen compounds of group-IV
nanosheets, Appl. Phys. Lett. 97 (2010) 163114.
[116] M. Houssa, E. Scalise, K. Sankaran, G. Pourtois, V.V. Afanas’ev, A. Stesmans, Electronic
properties of hydrogenated silicene and germanene, Appl. Phys. Lett. 98 (2011) 223107.
[117] E. Bianco, S. Butler, S. Jiang, O.D. Restrepo, W. Windl, J.E. Goldberger, Stability and
Exfoliation of Germanane: A Germanium Graphane Analogue, ACS Nano. 7 (2013) 4414–
4421.
[118] S. Jiang, E. Bianco, J.E. Goldberger, The structure and amorphization of germanane, J.
Mater. Chem. C. 2 (2014) 3185–3188.
[119] A.C. Ferrari, J.C. Meyer, V. Scardaci, C. Casiraghi, M. Lazzeri, F. Mauri, S. Piscanec, D.
Jiang, K.S. Novoselov, S. Roth, A.K. Geim, Raman Spectrum of Graphene and Graphene
Layers, Phys. Rev. Lett. 97 (2006) 187401.
[120] A.C. Ferrari, D.M. Basko, Raman spectroscopy as a versatile tool for studying the properties
of graphene, Nat. Nanotechnol. 8 (2013) 235–246.
[121] Y. Ma, Y. Chen, Y. Ma, S. Jiang, J. Goldberger, T. Vogt, Y. Lee, Inter- and Intralayer
Compression of Germanane, J. Phys. Chem. C. 118 (2014) 28196–28201.
Chapter 1
32
33
Chapter 2. Two channels transport in
MoS2
Abstract
Two-dimensional MoS2 has shown promising characteristics for application in
electronic devices. In this chapter, we study the quantum transport properties of
MoS2 field effect transistors on h-BN substrates. The results consist of two parts.
First, the electron mobility of MoS2 is found to be significantly improved by
encapsulation between top and bottom h-BN flakes. With greatly reduced contact
resistance due to ionic liquid gating, the insulator to metal transition is observed.
Second, for the MoS2/h-BN structure without the top h-BN cap, superconductivity
is induced at the top layer of the MoS2 by ionic liquid gating as expected;
surprisingly, the metallic bottom layers controlled by the back gate show
Shubnikov-de Haas oscillation under perpendicular magnetic fields after the
suppression of superconductivity. Hence, artificial electronic structures could be
developed by controlling the interaction between the hetero-states at the top and
bottom surfaces of a single few-layer MoS2 flake. Variable electronic functionalities
could be developed through the interaction between the different electronic states,
similar to conventional van der Waals heterostructures that are prepared by
stacking different two-dimensional (2D) materials, where novel electronic
properties emerge from the realignment of dissimilar bands or tunneling through
them.
Submitted as
“Electrically tunable coupling between hetero surface states in few-layer MoS2”
Q.H. Chen, J.M. Lu, L. Liang, O. Zheliuk, A. Ali, P. Sheng and J.T. Ye
Chapter 2
34
2.1 Introduction
In the modern electronic industry, silicon-based transistors are reaching their
scaling limit below 5 nm in size. Exploration of novel device structures and new
channel materials has therefore attracted growing interest. As a potential
alternative to Si, MoS2 exhibits larger band gaps and lower dielectric constants [1].
Because of its layered structure, MoS2 can be easily thinned down to uniformly thin
layers. Prototypes of few-layer and monolayer MoS2 transistors have been reported,
showing promising device operation characteristics for applications [2–4].
In order to meet the requirements for high-speed electronic devices, it is
desirable to improve the electron mobility of MoS2-based transistors. Suspended
MoS2 shows 2-10 times greater mobility enhancement compared to that supported
by conventional oxide substrates [5], but is still lower than the expectation for
high-speed electronic devices. In addition, the fabrication process for suspended
MoS2 is very complicated, which means it is not practically efficient. Alternatively,
using hexagonal boron nitride (h-BN) as the substrates is an appealing approach to
improve the performance of MoS2 transistors. h-BN is thermally and chemically
inert and electrically insulating. More importantly, the surface of h-BN is
atomically flat and free of dangling bonds, and it has a lattice constant similar to
that of MoS2. As a result, when a thin MoS2 flake is encapsulated between two h-
BN flakes, the performance can be significantly improved; a Hall mobility of
34000 cm2V-1s-1 has been reported [6,7].
On the other hand, improving the functional properties of gate control on the
channel material is beneficial for reaching a high on/off ratio as well as low power
consumption. Recently, ionic liquid has been increasingly used as gate media for
replacing the traditional oxide gate dielectrics. This is known as the electrical
double layer (EDL) technique. The biggest advantage of EDL is its ability to induce
extremely large carrier density at the sample surface. New electronic states are
observed in this regime that is not accessible with conventional solid gates. In two-
dimensional (2D) transition metal dichalcogenides (TMDs), the field effect doping
by ionic liquid induces high conductivity states for both electron and hole
transport [8,9]. Remarkably, superconductivity is observed at the liquid/TMD
interface in MoS2 [10,11], MoTe2 [12] and WS2 [13], providing a playground for
studying and controlling 2D superconductivity. Interesting superconducting
properties in this system have been discovered, such as Ising pairing due to large
effective Zeeman fields, which results from the broken inversion symmetry and
strong spin-orbit coupling. As a result, the superconductivity is protected from
very strong magnetic fields parallel to the ab-plane of MoS2 [14–16].
Two channels transport in MoS2
35
It is predicted that for multilayer MoS2 electrical double layer transistors
(EDLTs), ionic gating confines the charge carriers mainly to the top-most layer,
hence the superconductivity at the liquid/MoS2 interface shows strictly 2D
characteristics [14]. The top layer is well isolated electronically, acting like a free-
standing monolayer, whereas the bottom layers (two layers away from the top)
remain almost unaffected by ionic gating. It is expected that the top
superconducting layer and bottom layers are electronically separated, forming two
parallel conducting channels: a superconducting top channel and a normal bottom
channel. The different electronic states at the top and bottom surfaces mimic the
artificial van der Waals heterostructures stacked by different 2D materials, and
new functionalities could be developed through their interaction. Nevertheless, no
direct observation has been reported to corroborate this model of two conducting
channels.
In this chapter, we first study the quantum transport properties of a MoS2
transistor encapsulated between two h-BN flakes, with a partially exposed surface
designed for electrical contacts. With the application of ionic gating, the exposed
MoS2 surface is turned into a metallic state by inducing a large carrier density,
greatly improving the contact between the MoS2 and electrodes. As a result,
transport properties can be precisely measured at low temperatures, where the
device shows improved carrier mobility. In addition, an insulator to metal
transition controlled by the back gate is observed, suggesting that h-BN
encapsulation suppresses the disordered potential so that the raised Fermi levels
by the back gate can easily exceed the mobility edge.
In this device configuration, however, superconductivity cannot be induced
because the ionic liquid and MoS2 channel are separated by h-BN and the gating
effect on the channel is very weak, ionic gating mainly helps with improving the
contact. To overcome this difficulty and verify the two conducting channels model,
we fabricated a MoS2-EDLT on the h-BN substrate without a top h-BN capping
flake. In this way, superconductivity can be easily induced at the liquid/MoS2
interface, while high carrier mobility is maintained because of the atomically flat h-
BN substrate. At low temperatures a sharp resistance drop from finite to zero
resistance is observed, corresponding to the superconducting transition. After
suppressing superconductivity in the out-of-plane magnetic field B, the parallel
metallic state at the bottom surface of the MoS2 is revealed by the observation of
Shubnikov-de Haas (SdH) oscillation, which serves as direct evidence for the
existence of a high mobility and low carrier density channel. The carrier density
extracted from the oscillation of high-mobility electrons is found to be more than
one order of magnitude smaller than that derived from the Hall effect, which
includes carrier contributions from the whole flake, thus unambiguously
confirming the formation of two spatially separated conducting channels.
Chapter 2
36
2.2 Device fabrication
MoS2 flakes were prepared by mechanical exfoliation of the 2H polytype bulk
crystals (SPI supplies), using the well-known scotch tape method [17]. Silicon
wafers coated with 300 nm-thick SiO2 were used as substrates. Optical microscopy
and atomic force microscopy (AFM) were used to identify the number of layers of
MoS2. h-BN flakes were prepared on different silicon wafers in a similar way.
Uniformly thin MoS2 flakes in the range of 1-5 layers and h-BN flakes with
thicknesses of 10-30 nm were selected for device fabrication.
After preparing MoS2 and h-BN flakes on separate Si/SiO2 substrates, we use
the so-called dry transfer technique [18,19] to fabricate the van der Waals
heterostructure. The transfer process is illustrated schematically in Figure 2.1,
which shows an example of making the h-BN/MoS2/h-BN heterostructure. A thin
film of polycarbonate (PC) is prepared by first dissolving the PC pellets in
chloroform and putting a few droplets of the solution onto a cover glass. Another
cover glass is put on top to spread out the solution, and then the two cover glasses
are separated by sliding them over each other. After a few minutes, the chloroform
solvent evaporates and a thin uniform PC film is left on the glass. The PC film is
then carefully removed from the glass substrate and fixed on top of a piece of
polydimethylsiloxane (PDMS) on a thin glass slide with dimensions of ~3×3×1 mm.
The cover glass with the PDMS and PC film used as a stamp to pick up MoS2 and h-
BN flakes; it is inverted and fixed on a micro-manipulator with xyz-movement. The
stamp is gradually brought into contact with the SiO2/Si substrate until the PC film
almost touches the target h-BN flake. Then the SiO2/Si substrate with h-BN is
heated to about 90 °C, allowing the real contact of PC film and h-BN due to
thermal expansion. After that, the heating is stopped and the stamp gradually
retracts from the substrate. Because the adhesion between the h-BN and PC film is
stronger than that between the h-BN and substrate, the h-BN flake is picked up by
the PC film. The whole structure (h-BN + PC + PDMS) is used as a stamp to pick
up the target MoS2 flake in a similar way, and this process can be repeated to pick
up other crystals to form the desired stacking of different 2D materials.
After all the target flakes are picked up, the whole stamp with the stacked
heterostructure is put in contact with the target substrate and heated up to 170 °C
to melt the PC film. The stamp is carefully retracted, leaving the PC film as well as
the stacked structure on the substrate. The PC film on top of the heterostructure
can be removed by dissolving it in chloroform. The flakes are not exposed to any
solvent until the whole heterostructure is completed; the interface remains
relatively clean, which is important for achieving high carrier mobility. Then
reactive ion etching (RIE) is used to shape the sample into Hall bar geometry, and
metal contacts are made by the standard electron beam lithography procedure.
Two channels transport in MoS2
37
Figure 2.1 Dry transfer process for preparing the h-BN/MoS2/h-BN heterostructure. The top h-
BN flake is carefully chosen so that it covers the MoS2 surface only partially; the exposed area is
used for contact electrodes as well as ionic gating. Sandwiched MoS2 is etched into Hall-bar
geometry with RIE etching, and Ti/Au electrodes are patterned using standard electron beam
lithography. The process can be modified to fabricate any other desired device structures, such
as MoS2/h-BN or MoS2 transistors on an HfO2 substrate.
2.3 Results and discussions
In the following, we focus on the measurement results of two device
configurations. The first is a MoS2 transistor encapsulated between top and bottom
h-BN flakes. For ionic gating, the MoS2 conducting channel and liquid is separated
by h-BN; only the exposed area for contact electrodes is heavily doped by ionic
gating. The second is a MoS2-EDLT device on an h-BN substrate, without any h-BN
capping on top. The whole MoS2 surface can be gated by ionic liquid and becomes
superconducting at low temperatures. The bottom layers stay semiconducting
which can be tuned by the back gate as a conventional transistor.
2.3.1 MoS2 encapsulated between top and bottom h-BN
Figure 2.2(a) shows the optical image of an h-BN/MoS2/h-BN heterostructure
on SiO2/Si substrate. A few-layer MoS2 flake is sandwiched between two h-BN
flakes, with a partially exposed surface designed for metal contacts. Figure 2.2(b)
shows the transfer curve at different temperatures before applying ionic gating.
The device can be switched on and off by the back gate, showing an n-type
Chapter 2
38
transistor operation with a current on/off ratio higher than 102. However, a clear
decrease of the source-drain current is observed as the temperature decreases,
which is most likely associated with the increase in contact resistance. It is well
known that because of the work function difference between metal electrodes and
semiconducting MoS2, there is a potential barrier for electrons, namely the
Schottky barrier. For a fixed source-drain voltage, the temperature dependence of
the current is given by 𝐼 = 𝐴∗𝑇2𝑒−𝑞𝛷𝑏𝑘𝑇 (𝑒
𝑞𝑉
𝑘𝑇 − 1), where 𝐴∗ is a constant and q and k
are the electron charge and Boltzmann constant, respectively. V is the source-drain
voltage and 𝛷𝑏 is the height of the barrier. From this relation we see that the
source-drain current decreases when the temperature decreases, in line with our
observation.
Figure 2.2 (a) Optical image of a typical h-BN/MoS2/h-BN device. Top h-BN is highlighted by
the blue dashed line in the center. The green dashed line depicts the edge of the MoS2, and the
large dark area surrounded by the blue dashed line is the bottom h-BN flake. The top h-BN is
smaller than the MoS2 flake, metal contacts are made in the exposed area. Electrodes are
composed of 5 nm Ti / 65 nm Au. (b) Transfer curves of transistor operations by the SiO2 back
gate at different temperatures, with fixed source-drain voltage VDS = 0.1 V. An n-type transistor
operation is observed with source-drain current on/off ratio ~ 102.
As mentioned above, the contact resistance can be greatly reduced by applying
ionic liquid gating. Figure 2.3(a) shows a schematic of the gating configuration.
When a positive voltage is applied between the gate pad and the MoS2 flake,
cations are driven very close to the surface of the exposed MoS2, forming a
nanogap capacitor with very large capacitance. The induced carrier density is in the
order of ~1014 cm-2; hence the Fermi level can be significantly altered by ionic
gating. According to the phase diagram [10,20], the insulator-metal transition
occurs at 8×1012 cm-2 for multilayer MoS2. As a result, the exposed MoS2 surface
Two channels transport in MoS2
39
behaves like a normal metal, and the contact resistance is significantly reduced
locally at the exposed area.
Figure 2.3 (a) Schematics of the MoS2-EDLT device and measurement configuration. With
positive liquid gate voltage, cations are driven very close to the exposed surface of the MoS2
and induce a lot of electrons in these local areas. (b) Transfer curves of transistor operations by
the liquid gate at T = 220 K, for the same device as shown in Figure 2.2, showing a remarkable
enhancement of gating effect compared with the SiO2 solid gate.
The transfer curves by liquid gating are shown in Figure 2.3(b), where the
arrows indicate the gate voltage sweeping direction. To retain the high mobility of
cations/anions as well as protect the sample from the electrochemical reaction, the
ionic gating procedure is conducted at T = 220 K, as has been well established in
previous studies [9–14,21,22]. The device can also be switched on and off with
ionic gating, and the required voltage is much lower compared to that of the SiO2
back gate, suggesting that the efficiency of the liquid gate is much higher than that
of the solid gate. The source-drain current is nearly two orders of magnitude
higher than SiO2 back gating at T = 200 K, showing a great improvement in the
FET performance. We also see that the first cycle (red) and the second cycle (blue)
almost coincide with each other, indicating that no chemical reaction happens
during the gating procedure, which means the doping is electrostatic. However, a
large hysteresis is observed due to the low ion mobility at relatively low
temperatures. This is the disadvantage of gating at low temperatures, and the
uniformity of the gating will also be affected.
We cooled the device down to liquid nitrogen temperature to measure the
temperature dependence of the MoS2 channel resistance. Normally the gate voltage
is retained before it is cooled down to freeze the ion movement. Once the
temperature is below the freezing point of the ionic liquid, ions cannot move, and
the gate voltage can be retracted without losing the gating effect. As shown in
Figure 2.4(a), the two black curves represent the temperature dependence of the
Chapter 2
40
channel resistance before applying ionic gating. If the back gate voltage is zero,
which means that the MoS2 channel is not gated, it behaves as an intrinsic
insulator with very high resistance (top black curve with label VBG = 0 V, VLG = 0 V).
With applied back gate voltage VBG = 60 V, the MoS2 channel shows a metallic
behavior at T > 120 K, where the resistance decrease as the temperature goes down.
With further cooling down to below 120 K, sample resistance starts to increase. We
did not perform the Hall measurement on this device because it was measured in a
setup that is not equipped with magnet. Therefore the carrier density cannot be
determined accurately. Alternatively, because the carriers are induced by back gate,
the corresponding carrier density can be estimated from the geometric capacitance
of back gate through 𝑛2D = 𝐶g(𝑉BG − 𝑉th)/𝑒, where VBG is the applied back gate
voltage; e is the charge of electron; 𝐶g = 10.5 nF/cm2, which is the capacitance per
unit area considering the back gate of 300 nm SiO2 and 30 nm h-BN used in this
experiment; and Vth = 40 V, which is the threshold voltage that can be extracted
from the transfer curve, as shown in Figure 2.4(b) (black curve with VLG = 0 V).
Therefore, the induced 2D carrier density is n2D = 1.3×1012 cm-2. This relatively low
carrier density with metallic behavior proves that the quality of the sample device
is good.
Figure 2.4 (a) Comparison of the temperature dependence of the channel resistance before
(black curves) and after (red curves) applying ionic gating. With the application of ionic gating
VLG = 5 V, channel resistance undergoes a transition from insulating behavior (VBG = 0 V) to
metallic behavior (VBG = 20 V). (b) Transfer curve with SiO2 back gate at low temperatures (T =
80 K). Mobility extracted from the gate dependence of conductivity is 2700 cm2V-1s-1 before and
12000 cm2V-1s-1 after applying ionic gating, showing great enhancement.
The red curves in Figure 2.4(a) show the temperature dependence of the MoS2
channel resistance under different back gate voltages, with the ionic gate voltage
Two channels transport in MoS2
41
fixed at VLG = 5 V. A different behavior is observed comparing to before applying
ionic gating. For VLG = 5 V and VBG = 0 V, the RT curve is very similar to that of VLG
= 0 V and VBG = 60 V. This result implies that even though there is a top h-BN flake
covering the MoS2 conducting channel, ions can still accumulate at the top surface
of the h-BN and induce carriers in the MoS2 channel. With a 20 nm-thick h-BN
flake in this device, the induced carrier is similar to that of the 60 V back gate. For
VLG = 5 V and VBG = 20 V, the sample shows metallic behavior (dR/dT > 0) until
the liquid nitrogen temperature, and becomes more conducting at the higher back
gate VBG = 60 V.
Figure 2.4 (b) shows the conductivity as a function of back gate voltage, before
(black curve) and after (red curve) applying the liquid gate. The threshold voltage
shifts to the left side due to the weak doping effect by the ionic gating through the
h-BN, and for VLG = 5 V the conductivity increases much faster than VLG = 0 V,
corresponding to a higher mobility. The FET mobility can be estimated by the back
gate dependence of conductivity through 𝜇𝐹𝐸𝑇 =1
𝐶g
d𝜎
d𝑉BG , where 𝜎 is the
conductivity and Cg is the gate capacitance per unit area. Here, for 300 nm SiO2
and 30 nm h-BN 𝐶g = 10.5 nF/cm2. We can estimate the mobility from the linear
fitting in the vicinity of threshold voltage. Before and after applying the ionic
gating, the extracted mobilities are 2700 and 12000 cm2V-1s-1, respectively. The
enhancement of mobility may be related to the fact that there is ion accumulation
at the surface of the top h-BN, these ions form a charge screening layer which helps
to reduce the scattering from the charged impurities in the MoS2 channel. Hence
the mobility increases. It should be noted that the off-state current for VLG = 5 V is
higher than that of VLG = 0 V. While this may be due to the gating effect of the
MoS2 channel edge, further investigation is needed to study this behavior.
2.3.2 MoS2-EDLT on h-BN substrate without top cover
With MoS2 covered by a top h-BN, ionic gating is able to induce carriers in the
MoS2 channel. However, the carrier density is not high enough to reach
superconducting state. To prove the two parallel conducting channels that form
due to the dual-gate configuration, we modified the device structure by removing
the top capping h-BN flake. With the application of ionic gating in this
configuration, the top surface of the MoS2 becomes highly doped, whereas the
bottom layers remain almost unaffected. The high electron mobility nature of the
bottom layers is preserved due to the atomically flat bottom h-BN substrate. Novel
electronic functionalities emerge through the interplay of the superconducting top
surface and high mobility bottom surface. The device structure is similar to that of
high electron mobility transistors (HEMTs), which are conventionally fabricated
Chapter 2
42
with semiconducting planar junctions with different band structures, such as GaAs
and AlGaAs, using a molecular beam evaporation technique [23].
Figure 2.5 Dual-gate device configuration of MoS2/h-BN transistor on SiO2/Si++ substrate.
Without top h-BN cover, the MoS2 surface is in contact with the ionic liquid; thus the whole
surface becomes highly doped by ionic gating.
Figure 2.6 Illustration of the doping profile by the ionic liquid gate and solid-state back gate.
The layer number is labeled as “1” for the topmost layer and “5” for the bottommost layer. Due
to the strong Thomas-Fermi screening effect, the doping effect from ionic gating decays
exponentially for individual layers from top to bottom. The back gating effect also decays from
bottom to top, but it is much slower because the carrier density induced by the back gate is
much smaller.
The schematic device configuration of the MoS2-EDLT device on the h-BN
substrate is shown in Figure 2.5. Since the top surface of the MoS2 is not covered
by h-BN, superconductivity can be easily induced by ionic gating. Theoretical
calculations show that with 2D carrier density n2D of the topmost surface induced
Two channels transport in MoS2
43
by ionic gating in the order of ~1014 cm-2, the number of accumulated carriers for
individual layers decays exponentially from the top to bottom due to the strong
Thomas-Fermi screening effect [24]. The n2D of the second layer is reduced by
almost 90% compared to that of the first one. Therefore, the superconducting top-
most layer is well isolated electronically from the rest of the layers below and acts
like a freestanding monolayer [25]. The bottom layers remain almost undoped. On
the other hand, the back gate mainly induces carriers in the bottom layers. The
electron mobility in the bottom layers on h-BN is very high, suitable for making
HEMTs. The rough carrier density profile is shown in Figure 2.6, where the layer
number “1” denotes the topmost layer and “5” denotes the bottommost layer. The
screening effect is inversely scaled with the carrier density; the ionic gating induces
large carrier density but decays abruptly from top to bottom, while the back gating
induces smaller carrier density but tends to influence more layers.
Figure 2.7 (a) Top panel: optical image of a MoS2-EDLT device on an h-BN substrate, with MoS2
highlighted by the dashed line. Bottom panel: AFM height profile of the MoS2 flake, showing
that the thickness of the MoS2 in this study is 3.5 nm, corresponding to a layer number of five.
(b) Temperature dependence of the channel resistance with fixed ionic liquid gate voltage VLG =
6 V.
For ionic gating, all the conditions are kept the same as detailed in the previous
section. Figure 2.7(a) shows the optical image (top panel) and the AFM height
profile (bottom panel) of the measured device. For a layered material with layer
number N, the corresponding thickness is 𝑑 = 𝑎 + (𝑁 − 1) × 𝑏 , where a is the
height of the bottom layer (including the space between the bottom layer and the
Chapter 2
44
substrate) and b is the layer thickness. In MoS2, the typical value of a ranges from
0.8 to 1 nm depending on the adhesion between the flake and the underlying
substrates, while b is around 0.62 nm [26,27]. The thickness of the studied MoS2
flake is 3.5 nm characterized by AFM (Figure 2.7(a), bottom panel), and
consequently the number of layers is five. The ionic gate voltage VLG was applied at
T = 220 K, and then the sample was cooled down to measure the temperature
dependence of channel resistance. The channel resistance decreases as the
temperature decreases, showing metallic behavior. The RT curve becomes flat
below T = 30 K, which is likely due to the impurity scatterings and doping
inhomogeneity. A sharp resistance drop is observed below T = 10 K, corresponding
to the superconducting transition with Tc = 7 K. The transition temperature Tc is
defined as 50% of the normal state resistance. Zero resistance is observed with
further cooling down.
The RT behavior is different with different ionic gate voltages, as shown in
Figure 2.8. For VLG = 3 V, the sample shows metallic behavior, but no
superconductivity is observed until T = 2 K, this state is very close to the quantum
critical point (QCP). With increasing liquid gate voltage, superconductivity starts
to appear. Tc shifts to higher temperatures for higher gate voltages with a tendency
to saturate at VLG > 4.6 V.
Figure 2.8 (a) Temperature dependence of MoS2 channel resistance for different ionic gate
voltages. (b) Low temperature region of the same data set as in Figure 4-7(a).
At low temperatures, we measured the temperature dependence of resistance
under different magnetic fields, which were applied perpendicularly to the surface
of the MoS2. Figure 2.9 shows the results for liquid gate voltage VLG = 4.6 V. The
superconducting transition is dramatically suppressed by the magnetic field.
Two channels transport in MoS2
45
Figure 2.9 Sheet resistance of the MoS2-EDLT device as a function of temperature for VLG = 4.6
V under perpendicular magnetic fields, 0, 0.05, 0.1T, 0.2 to 2.4 T in 0.2 T steps, 2.7, 3, 3.5, 4, 5
and 6 T. Superconductivity is gradually suppressed by the magnetic field.
Without applying back gate voltage, the bottom layers remain insulating and
make no contribution to the conductivity of the MoS2 channel. Superconductivity
originates solely from the top layer, which has a high carrier density (~1014 cm-2).
With positive back gate voltage applied, electrons start to accumulate at the bottom
layers. As a result, the conductivity of the bottom layers increases and two parallel
conducting channels form in a single few-layer MoS2 flake.
Figure 2.10 (a) Longitudinal resistance as a function of the magnetic field measured at T = 2 K
and VLG = 6 V, for back gate voltages from 90 to 110 V in 5 V steps. (b) Magnified view of the
region between 6 and 12 T.
Figure 2.10 shows the longitudinal magnetoresistance (MR) for different VBG
values at T = 2 K and VLG = 6 V. Without the magnetic field, the MoS2 is in a
Chapter 2
46
superconducting state, and resistance is zero. When the applied magnetic field
increases, the resistance also increases as expected due to the suppression of
superconductivity. Surprisingly, pronounced Shubnikov-de Haas (SdH)
oscillations are observed after the superconductivity is suppressed by the magnetic
field. As the back gate voltage increases from 90 to 110 V, the amplitude of the
oscillation becomes notably larger, due to the enhanced conductivity of the bottom
layers. The peak and valley positions also change with different back gate voltages.
In contrast, all the curves almost coincide with each other at the low field region,
which suggests that the superconducting properties are not affected by the back
gate. This can also be verified from the doping profile in Figure 2.6; the influence
of the back gate on the superconducting layer is negligible due to the screening
effect.
Figure 2.11 Determination of the magnetoresistance background for back gate voltage VBG =
110 V. The green and red dashed lines denote the top and bottom envelop which are
determined by the peaks and valleys of the oscillation, respectively. The average of the top and
bottom envelope (blue dashed line) is taken as the magnetoresistance background for
extracting the oscillation component.
According to theory, SdH quantum oscillations are a periodic function of the
inverse of the magnetic field 1/B. To carefully examine the oscillation behavior, we
first subtract the MR background. As shown in Figure 2.11, the top (green dashed
line) and bottom (red dashed line) envelope lines are determined by the peaks and
valleys of the oscillations, respectively. The blue dashed line is the average of the
top and bottom envelope lines and is taken as the MR background for extracting
the oscillation component.
Two channels transport in MoS2
47
Figure 2.12 Oscillation amplitudes as a function of 1/B after subtracting the MR background
for different back gate voltages corresponding to the same data set as shown in Figure 2.10.
Curves are vertically shifted for clarity.
After subtracting the MR background and plotting as a function of the inverse
of the magnetic field 1/B, we see a clearly periodic oscillation and the oscillation
period decreases with the increasing back gate voltage, as shown in Figure 2.12.
According to theory, the envelope function of the oscillation in longitudinal
resistance can be described by the Ando formula [28,29]:
D
04 / sinh( )TR R e T T
Where 𝑅0 is the classical resistance in zero applied field, = 2𝜋2𝑘B/ℏ𝜔c , 𝜔c =
𝑒𝐵/𝑚∗ is the cyclotron frequency and 𝑚∗ is the electron effective mass. ℏ is the
reduced Planck constant, 𝑘B is the Boltzmann constant. 𝑇D = ℏ/2𝜋𝑘B𝜏 is the Dingle
temperature and 𝜏 is the scattering time. For 2D electron gas, the carrier density
𝑛2D is related to the period of the quantum oscillations through 𝑛2D = 𝑔𝑒𝐵F/ℎ,
where g is the Landau level degeneracy and BF is the oscillation frequency in 1/B
which can be obtained from Figure 2.12. On the other hand, it is clear that the
oscillation behavior is well controlled by back gate, hence 2Dn can be obtained
from the gate capacitance 𝑛2D = 𝐶g(𝑉BG − 𝑉th)/𝑒. Here Vth is the threshold voltage
Chapter 2
48
and 𝐶g = 10.5 nF/cm2 is the capacitance per unit area for the back gate of 300 nm
SiO2 and 30 nm h-BN used in this experiment. Thus, we have
𝐵𝐹 =ℎ𝐶g
𝑔𝑒2(𝑉BG − 𝑉th)
Hence Landau level degeneracy g can be obtained from the linear fitting of the
back gate dependence of BF, as shown in Figure 2.13. The result shows that g = 3.16,
consistent with that of the previous report [7].
Figure 2.13 (a) Oscillation period BF as a function of back gate voltage obtained from the
period of oscillation. The linear fitting yields a Landau level degeneracy of 3.16. (b) The density
of carriers that accounts for the quantum oscillations at different back gate voltages (solid dots
connected by black line), extracted from the oscillation period. Shaded area indicates the
carrier densities where MoS2 is superconducting, the bottom of this area represents the
quantum critical point (QCP) of the superconducting phase diagram.
With the obtained Landau level degeneracy, we are able to estimate the density
of carriers that accounts for the SdH oscillations through 𝑛2D = 𝑔𝑒𝐵F/ℎ . The
results are shown in Figure 2.13(b). For all VBG, the carrier densities that
accumulate at the bottom surface are more than one order of magnitude lower
than the lowest value for reaching superconductivity (~6×1013 cm-2). This
unambiguously confirms the existence of two conducting channels with high and
low carrier densities responsible for superconductivity and SdH oscillation,
respectively. This is exactly what was expected: ionic gating confines carriers to the
topmost layer, which becomes superconducting, while bottom layers remain
intrinsically semiconducting. With a positive back gate, the bottom channel
becomes electron doped and conducting. The electron mobility in the bottom
channel is very high due to the atomically flat h-BN substrate, and SdH oscillations
are contributed by this channel. SdH oscillations in MoS2 have been observed in
Two channels transport in MoS2
49
previous reports, either when graphene was used as contacts [6] or when MoS2 was
encapsulated between the top and bottom h-BN [7]. In our case, the MoS2 flake
was not covered with a top h-BN; the top superconducting layer of MoS2 itself
becomes a piece of normal metal after the superconductivity is suppressed by the
magnetic field, acting as a screening layer that functions the same as or even better
than a top protecting h-BN. Furthermore, the induced superconducting top
channel (metallic at the high magnetic fields) serves as a metallic contact for the
encapsulated bottom layers that ensures the observation of SdH oscillations. These
advantages make the fabrication process very simple.
Figure 2.14 (a) ∆𝑅 as a function of 1/B measured at different temperatures from 2 to 6 K, for
VBG = 110 V. The oscillation is gradually suppressed with increased temperature and almost
disappears at T = 6 K, owing to the enhanced electron-phonon scattering. (b) Solid dots show
the temperature dependence of the oscillation amplitude at B = 9.9 T. By using the Ando
formula ∆𝑅 ∝ 𝛼𝑇/sinh (𝛼𝑇), the best fitting (red curve) gives an electron effective mass of 0.59
me. Fittings at other magnetic fields are not shown here but give similar and consistent results.
Figure 2.14(a) shows the temperature dependence of the SdH oscillation
amplitudes at VBG = 110 V. With increased temperatures, the quantum oscillations
are gradually suppressed due to the enhanced electron-phonon scattering. The
effective mass of the conducting electron can be deduced by fitting the
temperature-dependent oscillation amplitudes with the Ando formula at a fixed
magnetic field. With the amplitude proportional to / sinh( )R T T , as
shown in Figure 2.14 for B = 9.9 T, the best fitting yields an effective mass of m* =
0.59 me, where me is the bare electron mass. In theoretical calculations of the MoS2
band structure there are two bands near the Fermi level (K and Q), while the
lowest energy is at the K valley (m* = 0.5 me) for monolayer MoS2, the Q valley (m*
= 0.6 me) becomes the lowest for layer number greater than 2L [6,30]. Since the
Chapter 2
50
MoS2 flake in this study is five layers, the result is in excellent agreement with the
theoretical value.
Figure 2.15 Conduction band edge corresponding to the Brillouin zone of the top and bottom
channels, where blue and red represent the spin-up and spin-down bands, respectively. For the
top superconducting channel, the minima of the conduction band reside at the K/K’ valleys;
whereas for the bottom high mobility channel, the conduction band minima are at the Q/Q’
valleys which sit in between the Γ and K/K’ points. The large red and blue arrows indicate the
effective Zeeman fields that cause the spin splitting, which is opposite for the K(Q) and K’(Q’)
points.
To summarize, the electronic states of two channels are shown schematically in
Figure 2.15. The top-most layer is highly doped and well isolated
electronically [14,15], which behaves like a monolayer, with the lowest energy of
the conduction band located at the K/K’ valleys, whereas for the rest of the flake
the lowest energy of the conduction band is at the Q/Q’ valleys. Spin degeneracies
in each Q and Q’ valleys are lifted by the intrinsic Zeeman field Beff, which
accordingly aligns spin in opposite directions at the Q and Q’ valleys of the
Brillouin zone. The degeneracy 6 composed of 3Q and 3Q’ valleys is further
removed by the valley Zeeman effect with an applied external magnetic field [7,31].
Thus the Landau level degeneracy g equals 3, which is consistent with the g = 3.16
observed experimentally.
Two channels transport in MoS2
51
An unusual behavior is observed in the transition temperature Tc of the top
superconducting channel, as shown in Figure 2.16. As the back gate voltage
increases, Tc first increases and then decreases. According to the phase diagram,
the positive back gate accumulates electrons in the MoS2 channel, which in
principle leads to the increase in Tc. The unusual decrease in Tc suggests that due
to the growing conductivity of the bottom channel, enhanced interaction with the
top surface plays an important role in determining superconducting properties.
This behavior will be discussed in more detail in the next chapter.
Figure 2.16 Superconducting transition temperature Tc as a function of applied back gate
voltage, for VLG = 6 V. As back gate voltage increases, Tc first increases and then decreases,
which is not consistent with the phase diagram.
As a summary, Figure 2.17 plots the magnetoresistance of the MoS2 device for
different applied ionic gate voltages at the lowest temperature T = 2 K, with fixed
back gate voltage VBG = 100 V. The oscillations become more evident as the ionic
gate voltage increases, because higher ionic voltage induces higher carrier density
at the top surface, which leads to a stronger screening effect, so the bottom layers
are better protected from the environment. The peak and valley positions of the
oscillation remain almost unchanged, indicating that the band structure of the
bottom layers is not affected by the top ionic gating, consistent with the band edge
of the two parallel channels as shown in Figure 2.16.
Chapter 2
52
Figure 2.17 Magnetoresistance of MoS2 for different ionic gate voltages, with fixed back gate
VBG = 100 V, at T = 2 K. The peak and valley positions of the oscillations remain almost the same.
2.4 Conclusion
This chapter studied the quantum transport of two types of MoS2 transistor
devices:
(1) MoS2 encapsulated between two h-BN flakes. With protection from top and
bottom h-BN flakes, high electron mobility reveals the high quality and
homogeneity of our MoS2 transistors. The insulator to metal transition is observed
at relatively low carrier density, providing additional evidence for the good quality
of the heterostructure. Ionic liquid gating is found to improve device performance
by greatly reducing the contact resistance. Surprisingly, although ionic liquid and
MoS2 surface are separated by a 20 nm-thick h-BN flake, ions can still accumulate
at the surface of the h-BN. These ions act as a screening layer which further
improves the electron mobility of the MoS2 channel.
(2) MoS2 FET without a top h-BN flake. The top surface of the MoS2 channel
can be directly gated by ionic liquid and becomes superconducting at low
temperatures. Clear SdH quantum oscillations are observed in the
magnetoresistance after the superconductivity is suppressed by the magnetic field.
The carrier density is more than one order of magnitude lower than the value
derived from the Hall measurements, which unambiguously confirms the existence
of two parallel conducting channels in few-layer MoS2-EDLT devices.
Two channels transport in MoS2
53
Superconductivity is induced at the top surface of the MoS2 by ionic gating; the
induced carrier is in the order of ~1014 cm-2. Due to the strong screening effect of
the large carrier density, the bottom layers remain unaffected by the ionic gating,
and the conductivity of the bottom layers can be continuously controlled by the
back gate as a conventional field effect transistor. Therefore, two parallel
conducting channels form at the top and bottom surfaces of the MoS2, accounting
for the superconductivity and SdH quantum oscillation, respectively. The
superconducting channel acts as a screening layer as well as metal contact for
probing SdH oscillations under a high magnetic field when superconductivity is
suppressed. Furthermore, the effective mass and Landau level degeneracy
extracted from the oscillation amplitude are in excellent agreement with
theoretical calculations. This is the first experimental observation that firmly
confirms the coexistence of two parallel conducting channels in a single MoS2 flake.
Chapter 2
54
References
[1] S.B. Desai, S.R. Madhvapathy, A.B. Sachid, J.P. Llinas, Q. Wang, G.H. Ahn, G. Pitner, M.J.
Kim, J. Bokor, C. Hu, H.-S.P. Wong, A. Javey, MoS2 transistors with 1-nanometer gate
lengths, Science. 354 (2016) 99–102.
[2] K.S. Novoselov, D. Jiang, F. Schedin, T.J. Booth, V.V. Khotkevich, S.V. Morozov, A.K.
Geim, Two-dimensional atomic crystals, Proc. Natl. Acad. Sci. U. S. A. 102 (2005) 10451–
10453.
[3] A. Ayari, E. Cobas, O. Ogundadegbe, M.S. Fuhrer, Realization and electrical characterization
of ultrathin crystals of layered transition-metal dichalcogenides, J. Appl. Phys. 101 (2007)
14507.
[4] B. Radisavljevic, A. Radenovic, J. Brivio, V. Giacometti, A. Kis, Single-layer MoS2
transistors, Nat. Nanotechnol. 6 (2011) 147–150.
[5] T. Jin, J. Kang, E.S. Kim, S. Lee, C. Lee, Suspended single-layer MoS2 devices, J. Appl. Phys.
114 (2013) 164509.
[6] X. Cui, G.-H. Lee, Y.D. Kim, G. Arefe, P.Y. Huang, C.-H. Lee, D.A. Chenet, X. Zhang, L.
Wang, F. Ye, F. Pizzocchero, B.S. Jessen, K. Watanabe, T. Taniguchi, D.A. Muller, T. Low, P.
Kim, J. Hone, Multi-terminal transport measurements of MoS2 using a van der Waals
heterostructure device platform, Nat. Nanotechnol. 10 (2015) 534–540.
[7] Z. Wu, S. Xu, H. Lu, G.-B. Liu, A. Khamoshi, T. Han, Y. Wu, J. Lin, G. Long, Y. He, Y. Cai,
F. Zhang, N. Wang, Observation of Valley Zeeman and Quantum Hall Effects at Q Valley of
Few-Layer Transition Metal Disulfides, ArXiv151100077 Cond-Mat. (2015).
[8] Y. Zhang, J. Ye, Y. Matsuhashi, Y. Iwasa, Ambipolar MoS2 Thin Flake Transistors, Nano Lett.
12 (2012) 1136–1140.
[9] Y.J. Zhang, J.T. Ye, Y. Yomogida, T. Takenobu, Y. Iwasa, Formation of a Stable p–n
Junction in a Liquid-Gated MoS2 Ambipolar Transistor, Nano Lett. 13 (2013) 3023–3028.
[10] J.T. Ye, Y.J. Zhang, R. Akashi, M.S. Bahramy, R. Arita, Y. Iwasa, Superconducting Dome in
a Gate-Tuned Band Insulator, Science. 338 (2012) 1193–1196.
[11] D. Costanzo, S. Jo, H. Berger, A.F. Morpurgo, Gate-induced Superconductivity in atomically
thin MoS2 crystals, ArXiv151203222 Cond-Mat. (2015).
[12] W. Shi, J. Ye, Y. Zhang, R. Suzuki, M. Yoshida, J. Miyazaki, N. Inoue, Y. Saito, Y. Iwasa,
Superconductivity Series in Transition Metal Dichalcogenides by Ionic Gating, Sci. Rep. 5
(2015) 12534.
[13] S. Jo, D. Costanzo, H. Berger, A.F. Morpurgo, Electrostatically Induced Superconductivity at
the Surface of WS2, Nano Lett. 15 (2015) 1197–1202.
[14] J.M. Lu, O. Zheliuk, I. Leermakers, N.F.Q. Yuan, U. Zeitler, K.T. Law, J.T. Ye, Evidence for
two-dimensional Ising superconductivity in gated MoS2, Science. (2015) aab2277.
[15] Y. Saito, Y. Nakamura, M.S. Bahramy, Y. Kohama, J. Ye, Y. Kasahara, Y. Nakagawa, M.
Onga, M. Tokunaga, T. Nojima, Y. Yanase, Y. Iwasa, Superconductivity protected by spin-
valley locking in ion-gated MoS2, Nat. Phys. 12 (2016) 144–149.
[16] X. Xi, Z. Wang, W. Zhao, J.-H. Park, K.T. Law, H. Berger, L. Forró, J. Shan, K.F. Mak, Ising
pairing in superconducting NbSe2 atomic layers, Nat. Phys. 12 (2016) 139–143.
[17] K.S. Novoselov, A.K. Geim, S.V. Morozov, D. Jiang, Y. Zhang, S.V. Dubonos, I.V.
Grigorieva, A.A. Firsov, Electric Field Effect in Atomically Thin Carbon Films, Science. 306
(2004) 666–669.
Two channels transport in MoS2
55
[18] L. Wang, I. Meric, P.Y. Huang, Q. Gao, Y. Gao, H. Tran, T. Taniguchi, K. Watanabe, L.M.
Campos, D.A. Muller, J. Guo, P. Kim, J. Hone, K.L. Shepard, C.R. Dean, One-Dimensional
Electrical Contact to a Two-Dimensional Material, Science. 342 (2013) 614–617.
[19] P.J. Zomer, M.H.D. Guimarães, J.C. Brant, N. Tombros, B.J. van Wees, Fast pick up
technique for high quality heterostructures of bilayer graphene and hexagonal boron nitride,
Appl. Phys. Lett. 105 (2014) 13101.
[20] X. Chen, Z. Wu, S. Xu, L. Wang, R. Huang, Y. Han, W. Ye, W. Xiong, T. Han, G. Long, Y.
Wang, Y. He, Y. Cai, P. Sheng, N. Wang, Probing the electron states and metal-insulator
transition mechanisms in molybdenum disulphide vertical heterostructures, Nat. Commun. 6
(2015) 6088.
[21] K. Ueno, S. Nakamura, H. Shimotani, A. Ohtomo, N. Kimura, T. Nojima, H. Aoki, Y. Iwasa,
M. Kawasaki, Electric-field-induced superconductivity in an insulator, Nat. Mater. 7 (2008)
855–858.
[22] J.T. Ye, S. Inoue, K. Kobayashi, Y. Kasahara, H.T. Yuan, H. Shimotani, Y. Iwasa, Liquid-
gated interface superconductivity on an atomically flat film, Nat. Mater. 9 (2010) 125–128.
[23] M.A. Khan, A. Bhattarai, J.N. Kuznia, D.T. Olson, High electron mobility transistor based on
a GaN‐AlxGa1−xN heterojunction, Appl. Phys. Lett. 63 (1993) 1214–1215.
[24] R. Roldán, E. Cappelluti, F. Guinea, Interactions and superconductivity in heavily doped
MoS2, Phys. Rev. B. 88 (2013) 54515.
[25] T. Eknapakul, P.D.C. King, M. Asakawa, P. Buaphet, R.-H. He, S.-K. Mo, H. Takagi, K.M.
Shen, F. Baumberger, T. Sasagawa, S. Jungthawan, W. Meevasana, Electronic Structure of a
Quasi-Freestanding MoS2 Monolayer, Nano Lett. 14 (2014) 1312–1316.
[26] C. Lee, H. Yan, L.E. Brus, T.F. Heinz, J. Hone, S. Ryu, Anomalous Lattice Vibrations of
Single- and Few-Layer MoS2, ACS Nano. 4 (2010) 2695–2700.
[27] J. Chang, L.F. Register, S.K. Banerjee, Ballistic performance comparison of monolayer
transition metal dichalcogenide MX2 (M = Mo, W; X = S, Se, Te) metal-oxide-semiconductor
field effect transistors, J. Appl. Phys. 115 (2014) 84506.
[28] T. Ando, A.B. Fowler, F. Stern, Electronic properties of two-dimensional systems, Rev. Mod.
Phys. 54 (1982) 437–672.
[29] Y.Z. Chen, N. Bovet, F. Trier, D.V. Christensen, F.M. Qu, N.H. Andersen, T. Kasama, W.
Zhang, R. Giraud, J. Dufouleur, T.S. Jespersen, J.R. Sun, A. Smith, J. Nygård, L. Lu, B.
Büchner, B.G. Shen, S. Linderoth, N. Pryds, A high-mobility two-dimensional electron gas at
the spinel/perovskite interface of γ-Al2O3/SrTiO3, Nat. Commun. 4 (2013) 1371.
[30] G.-B. Liu, W.-Y. Shan, Y. Yao, W. Yao, D. Xiao, Three-band tight-binding model for
monolayers of group-VIB transition metal dichalcogenides, Phys. Rev. B. 88 (2013) 85433.
[31] D. MacNeill, C. Heikes, K.F. Mak, Z. Anderson, A. Kormányos, V. Zólyomi, J. Park, D.C.
Ralph, Breaking of Valley Degeneracy by Magnetic Field in Monolayer MoSe2, Phys. Rev.
Lett. 114 (2015) 37401.
Chapter 2
56
57
Chapter 3. Dual-gate MoS2
superconducting transistor
Abstract
In this chapter, we study a dual-gated MoS2 transistor on HfO2 substrate. Two
parallel conducting channels (top superconducting layer and bottom normal layers)
form due to ionic liquid and solid back gating. Superconductivity in the top
channel can be effectively tuned through interaction with the bottom channel.
When the bottom channel is insulating, superconductivity can be switched on and
off continuously by field effect tuning of the carrier density of the top channel. For
the metallic bottom channel, the superconducting state is suppressed by the anti-
proximity effect between these two channels. A “bipolar”-like superconducting
transistor is realized based on this configuration.
Submitted as
“Continuous electrical tuning of superconductivity in two-dimensional MoS2”
Q.H. Chen, L. Liang, J.M. Lu, O. Zheliuk, A. Ali, P. Sheng and J.T. Ye
Chapter 3
58
3.1 Introduction
Switching between super/non-superconducting states as controllable quantum
phase transitions (QPT) has been widely studied in low-dimensional
superconducting systems by adjusting static parameters such as film thickness [1],
disorder [2], and chemical doping [3]. Alternative tunable parameters, best
exemplified by field effect gating, have been increasingly used as continuous
switching knobs to control superconductivity in systems with given chemical
stoichiometry and disorder morphology. Continuous control of superconductivity
through field effects is limited by the existing device technologies. General solid
state gating is too weak to achieve this goal because switching superconductivity on
and off usually requires tuning large amounts of carriers. Most previous efforts
were focused on maximizing the field effect by using special gate dielectrics such as
those with ferroelectric [4] or quantum paraelectric [5] properties. Solid state
gating has been effective only in examples where tuning superconductivity requires
changing a small amount of carriers such as realizing superconductor-insulator
transition at the LaAlO3/SrTiO3 interface [5–7]. In previous reports, switching the
super/non-superconducting states achieved by field effect tuning usually requires
high gate voltages or very low temperatures, or is not continuous at the
superconducting transition temperature.
Recently, field effect tuning using ionic dielectric media has been demonstrated
to be an effective method for inducing and manipulating superconductivity in
many material systems including oxides [8–10] and layered compounds [11–15].
While ionic gating is very capable of handling large amount of carriers, it functions
only when the ions are mobile above the glass transition temperature of the liquid.
For this reason, ionic gating is not continuous at temperatures where most
superconducting transitions happen. Hence, continuous tuning of
superconductivity in systems with neither low carrier density nor exceptional
dielectric properties remains a challenge that requires not only new ways of
utilizing existing device technologies, but also alternative device operation
concepts that go beyond the common wisdom of maximizing the carrier tunability
through optimizing dielectrics.
In 2D transition metal dichalcogenides (TMDs), field effect doping by ionic
liquid induces superconductivity in MoS2 [11,14], MoTe2 [12] and WS2 [15] at the
liquid/TMD interface. Interesting superconducting properties, such as Ising
pairing induced by broken inversion symmetry and strong spin-orbit coupling,
have recently been discovered [16–18]. In ionic gated MoS2 transistors on h-BN
substrates, two parallel conducting channels have been firmly confirmed by the
observed coexistence of superconductivity and SdH quantum oscillation in the
Dual-gate MoS2 superconducting transistor
59
same few-layer MoS2 flake. The corresponding Cooper pairs and high mobility
electrons are contributed by the top and bottom surface, respectively.
In this chapter, we fabricated a dual-gate MoS2 transistor combining the
advantages of ionic and solid gating. While ionic liquid gating induces
superconductivity at the top surface, the solid back gate can continuously tune the
electronic state of the bottom channel. Superconductivity can be effectively
controlled by different working principles according to the electronic state of the
bottom channel. When the bottom channel is insulating, it serves as an additional
dielectric layer for the field effect tuning of the top surface by the back gate; when
the bottom channel becomes metallic, strongly enhanced interaction between the
top superconducting channel and the bottom normal channel leads to the anti-
proximity effect that influences the superconducting properties of the top channel.
By precisely positioning the system close to the quantum critical point (QCP,
the onset of superconductivity), the field effect tuning of carrier density by the back
gate can easily change the state across the phase boundaries. When ionic gating
induces states deep into the superconducting dome [11], the carrier density is
beyond the tunable range of the back gate, and alternative control of
superconductivity is achieved by inducing anti-proximity effect between the
spatially separated top superconducting channel and bottom normal channel. The
anti-proximity effect can be continuously controlled in a way that weakens and
eventually suppresses the superconductivity of the top channel.
3.2 Device fabrication
Thin MoS2 flakes were prepared on substrates by micro-mechanically
exfoliating bulk MoS2 crystals (2H polytype, SPI supplies), following the standard
scotch tape method [19]. Optical microscopy and atomic force microscopy (AFM)
were used to select thin and uniform samples. High-quality flakes with thicknesses
of 2-5 nm were chosen for device fabrication. To improve the carrier tunability of
the back gate, we chose high-κ dielectric HfO2 (50 nm, dielectric constant of ~24),
which was deposited by atomic layer deposition (ALD) onto a Si substrate.
Through the HfO2 back gate, the carrier density can be tuned in the order of 1013
cm-2 with a moderate gate voltage and is effective at low temperatures. Electrodes
composed of Ti/Au (5/65 nm) were deposited in a high vacuum evaporator, after
patterning by standard e-beam lithography.
Figure 3.1(a) shows the typical device configuration of a dual-gate few-layer
MoS2 transistor on HfO2 substrates. The voltage bias (VLG) through ionic liquid
induces superconductivity at the topmost layer by accumulating carriers of high
density up to 1014 cm-2. Back gate bias (VBG) through high-κ HfO2 (50 nm) can be
Chapter 3
60
applied up to ±20 V (with leak current less than 1 nA), reaching a continuous
carrier density tuning of ±3×1013 cm-2. The left panel of Figure 3.1(b) displays the
optical image of the studied device, and the right panel is the corresponding AFM
height profile of this sample. As mentioned in the previous chapter, for a layered
material with layer number N, the corresponding thickness is 𝑑 = 𝑎 + (𝑁 − 1) × 𝑏,
where a is the height of the bottom layer (including the space between the bottom
layer and the substrate) and b is layer thickness. In MoS2, the typical value of a
ranges from 0.8 to 1 nm depending on the adhesion between the flake and
underlying substrates, while b is around 0.62 nm [20,21]. AFM characterizes the
thickness of the studied MoS2 flake as 2.85 nm (Figure 3.1(b) right panel).
Consequently, the layer number is four.
Figure 3.1 (a) Schematics illustration of the dual-gate MoS2 transistor on HfO2 substrates. With
applied ionic gate voltage, ions move very close to the surface of the MoS2 and induce a high
density of charge carriers (~1014 cm-2). (b) The optical image (left panel, dashed line depicts
the edge of the MoS2 flake) and AFM height profile (right panel) of the MoS2 flake used in this
study. The thickness of the MoS2 in this study is 2.85 nm, corresponding to a layer number of
four.
For ionic liquid gating, we used the well-known ionic liquid: N,N-diethy-N-(2-
methoxyethyl)-N-methylammonium bis-(trifluoromethylsulfonyl)-imide (DEME-
TFSI). To retain the high mobility of cations/anions as well as prevent
electrochemical reactions, all the ionic gating procedures were conducted at T =
220 K, which has been well established in previous studies [11,15–17,13,22,14].
When a gate voltage is applied between the ionic liquid and the sample, ions in the
liquid are driven very close (~1 nm) to the sample surface, which induces carriers
with the opposite polarity. The dynamic gating is transformed into static doping
when the movements of ions is frozen, which is done by retaining the voltage bias
while cooling the device to below the glass transition temperature of the ionic
liquid. At low temperatures, the induced carriers are maintained even if the gate
bias is grounded. After measurement, the gating effect can be thermally released by
warming the sample up to T = 220 K, where ion movement resumes.
Dual-gate MoS2 superconducting transistor
61
For transport measurement, liquid gate and back gate voltages were set by a DC
source meter (Keithley 2450). Transport measurements at low temperatures were
done by measuring voltage drops across the sample with a constant AC current,
using standard lock-in amplifiers (Stanford Research SR830). I-V curves were
measured with Keithley 2450.
3.3 Results and discussion
Figure 3.2 (a) Transfer curves measured at T = 220 K using liquid top gate (VLG) and solid back
gate (VBG, HfO2), with fixed source-drain voltage VDS = 0.1 V. (b) Side view of a few-layer MoS2
flake (left panel) and rough estimation of the carrier distribution (right panel) induced by
liquid top gate (red) and back gate (green), respectively.
At T = 220 K, the device can be switched on with either liquid gate (VLG) or back
gate (VBG) as shown in the respective transfer curves (Figure 3.2(a)). To reach the
same conductivity, VLG is typically ~1/10 of VBG, suggesting that the efficiency of
the ionic gating is almost one order of magnitude higher than the back gating with
50 nm HfO2 gate dielectric. Theoretical calculations showed that with 2D carrier
density (n2D) of the topmost layer induced by ionic gating in the order of ~1014 cm-2,
the amount of accumulated carriers decays exponentially for individual layers from
top to bottom [23]. The n2D of the second layer is reduced by almost 90%
compared to that of the first one. Therefore, the superconducting top most layer is
well isolated electronically from the rest of the layers below and acts like a
freestanding monolayer [24]. On the other hand, VBG alone in limited bias VBG <
Vth (where Vth is the threshold voltage indicated by the black arrow in Figure 3.2(a))
is not able to induce carriers as shown by the transfer curve (red) in Figure 3.2(a),
which leaves the bottom layers of the flake simply as additional dielectric layers for
Chapter 3
62
VBG to tune the superconductivity of the topmost layer. When VBG > Vth, carriers
start to accumulate at the bottom layers. Although the bottom-most layer helps to
screen out the applied VBG when it becomes more metallic, the field effect of the
back gate can extend longer to more layers than the ionic liquid gate, because the
screening effect scales inversely with induced carrier density [23]. The doping
profile is shown schematically in Figure 3.2(b). Interaction appears when the
electron wave functions of the top and bottom states overlap. Overall, doping by
VLG exclusively affects the topmost layer, whereas doping by VBG tends to influence
the whole flake, which tunes the superconductivity effectively. When VBG controls
the bottom layers to be insulating or metallic, correspondingly, the operation
mechanisms of controlling superconductivity could be switched between field
effect and anti-proximity effect, respectively.
By carefully adjusting VLG, we prepared the device in two different
representative states: state A with n2D ~6×1013 cm-2 close to the QCP, and state B
with n2D ~1×1014 cm-2, deep into the superconducting dome. In this device, some of
the electrodes were not working so we were not able to determine the carrier
density through the Hall measurement. The values shown here are estimated by
mapping the transition temperature Tc with the established phase diagram [11].
The results for each sample state are presented below.
3.3.1 Sample State A (close to QCP)
In Figure 3.3 we summarize the measured data from sample state A. The
temperature dependence of the sheet resistance Rs was measured at different VBG
between -20 and 20 V (Figure 3.3(a)). At the most negative VBG hence the lowest
n2D by depletion, 𝑑𝑅𝑠 𝑑𝑇 < 0⁄ indicates insulating behavior [25]. Increasing VBG
gradually switches on the superconductivity. Transition temperature Tc shifts
remarkably with higher VBG, where Tc is defined as 50% of the normal state
resistance RN (Rs at T = 13 K). The evolution of the superconducting state is plotted
as a quasi-continuous 2D map of the logarithm of Rs, as a function of temperature
and VBG (Figure 3.3(b)), where the super/non-superconducting phase boundary
(border line between the yellow and green areas) can be continuously accessed by
varying VBG. It should be noted that with negative VBG the n2D spanned from 3×1013
to 6×1013 cm-2; in this range, a metallic state should be observed referring to the
phase diagram [11,26]. The observed insulating behavior can be attributed to
reduced localization length and stronger inhomogeneity during the depletion of the
superconducting top layer, which leads to faster crossover from weak to strong
localization compared with pure ionic gating. In the normal state, RN decreases
rapidly with increasing VBG (Figure 3.3(c), red triangle) from -10 to 10 V. A clear
saturation of RN appears at VBG ~10 V accompanying the saturation of the Tc.
Dual-gate MoS2 superconducting transistor
63
Figure 3.3 (a) Field effect switching of superconductivity near QCP for sample state A. (b) A
quasi-2D color map of the logarithm of Rs versus T and VBG, showing the evolution between
superconducting (green) and insulating (orange) phases. (c) The left axis shows changes of TBKT
as a function of VBG. The green solid line describes the scaling relation 𝑇BKT ∝ (𝑉 − 𝑉c)𝑧𝜈, where
zν = 2/3. Red triangles show the VBG dependence of normal state Rs measured at T = 13 K.
The Mermin-Wagner theorem [27] has proven that no normal second-order
phase transition occurs in the frame of the 2D XY model, which is a vector spin
model of statistical mechanics in two dimensions. Nevertheless, there is a kind of
transition that exists in two dimensions called the BKT transition [28,29], named
after Vadim Berezinskii, John M. Kosterlitz and David J. Thouless. In the 2D XY
model, vortices are topologically stable configurations. Thermal generation
produces vortices with opposite sign, i.e. vortex and antivortex pairs. At low
temperatures, the vortex and antivortex are bounded to pairs so that the free
energy is lower, forming a quasi-long range order phase where the correlation
function decays with distance like a power law. At high temperatures, the vortex-
antivortex pairs break into unpaired vortices, forming an unordered phase where
Chapter 3
64
the correlation function decays exponentially. The BKT transition describes the
transition from the disordered phase with bound vortex-antivortex pairs to
unpaired vortices and antivortices that happens at some critical temperature TBKT.
The BKT transition can be found in several systems, such as Josephson junction
arrays and 2D superconducting systems [5,30]. Experimentally, the BKT transition
can be observed by measuring the current and voltage V-I relation. Above a certain
temperature, the voltage is a linear function of current 𝑉 ∝ 𝐼, while it becomes
𝑉 ∝ 𝐼3 right below the critical temperature. The jump from linear to 𝐼3 relation is
an indication of the BKT transition and the temperature where it happens is
defined as TBKT.
For the sheet resistance Rs, the temperature dependence of Rs above the critical
temperature TBKT can be described by 𝑅s(𝑇) ∝ exp (−𝑏
√𝑇−𝑇BKT), where b is a constant
related to the vortex-antivortex interaction strength. The TBKT extracted by fitting
Rs (T) with the above relation at different VBG is shown in Figure 3.3(c) (dark blue
circles). Access to the QCP is controlled by VBG, where TBKT = 0 is found at a critical
back gate of VGC = -1 V. The related variations of carrier concentration ∆𝑛2D = 𝑛 −
𝑛c serves as a direct control parameter for the quantum phase transition, where nc
denotes the critical carrier density at QCP. Following the scaling theory of
continuous quantum phase transition [31], TBKT is expected to scale near the QCP
as 𝑇𝐵𝐾𝑇 ∝ (𝛿𝑛2𝐷)𝑧𝜈, where 𝑧𝜈 is the scaling factor and 𝛿𝑛2𝐷 is the variation of 𝑛2𝐷
which is proportional to δV. Figure 3.3(c) shows that the TBKT at different VBG can
be well described by 𝑇𝐵𝐾𝑇 ∝ (𝛿𝑉)𝑧𝜈, with 𝑧𝜈 = 2/3 being consistent with the 2D XY
model at non-zero temperature [32], which was also observed in amorphous
bismuth film [25] and at the LaAlO3/SrTiO3 interface [5].
By crossing the super/non-superconducting phase boundary, other physical
parameters associated with the superconducting phase transition can also be
continuously tuned by VBG. Figure 3.4(a) shows the temperature dependence of the
critical field Bc. For each magnetic field, we measured the corresponding transition
temperature and plot it as one data point in Figure 3.4(a). The transition
temperatures are determined by the RT scans, using the 50% of normal state
resistance criterion. The temperature dependence of the upper critical field Bc
(𝐵 ∥ 𝑐 ) at different VBG shows a close association with the change in Tc. With nearly
constant Tc when VBG > 10 V, Bc(T) almost coincides for VBG = 13 V (stars) and 20 V
(pentagons). With Tc reduced by decreasing VBG, the Bc(T) curve shifts as parallel
lines. It should be noted that a clear decrease of the slope of the Bc(T) curve is
observed close to the QCP, which might be related to the progressive decrease of
uniformity by depletion close to the QCP. Additionally, the VBG changes the I–V
characteristics by controlling the superconducting critical current Ic, as shown in
Figure 3.4(b). Corresponding to the continuous suppression of Ic towards the QCP,
Dual-gate MoS2 superconducting transistor
65
the switching is characterized by the change in the line shape from the abrupt
transition in the superconducting state to the linear behavior in the normal state.
Figure 3.4 (a) Temperature dependence of Bc (perpendicular) for different VBG. (b) The I–V
characteristics at T = 2.5 K for different VBG between -20 and 20 V (using the same color coding
as data shown in Figure 3.3(a)).
We also measured the temperature dependence of sheet resistance under
different magnetic fields. Consistent with the device configuration that the strong
field effect induces a highly two-dimensional electron gas [16], the present
superconducting state is highly anisotropic as shown in the temperature
dependence of Bc under perpendicular and parallel magnetic fields.
Superconductivity is progressively suppressed by an increasing out-of-plane
magnetic field before being completely quenched at 12 T (Figure 3.5(a)). In
contrast, under an in-plane magnetic field, only a small shift of Tc is observed up to
12 T (Figure 3.5(b)), suggesting a very strong 2D nature. The field-induced
superconductivity can be described by the Tinkham model for 2D
superconductors [33]:
𝐵c⊥(𝑡) =
Φ0
2𝜋𝜉GL(0)2(1 − 𝑡)
𝐵c∥(𝑡) =
Φ0
√12
2𝜋𝜉GL(0)𝑑Tinkham
(1 − 𝑡)1/2
where 𝑡 = 𝑇/𝑇c denotes the reduced temperature, Φ0
= ℎ/2𝑒 = 2.07 × 1015Wb is
the flux quantum and 𝜉GL(0) is the Ginzburg-Landau coherence length at zero
temperature. By fitting experimental data, as shown in Figure 3.5(c), we obtained
𝐵c⊥(𝑡) = 9.6 T and 𝐵c
∥(𝑡) = 60 T for VBG = 20 V. Estimation using the Ginzburg-
Chapter 3
66
Landau theory gives coherence length 𝜉GL(0) = 5.9 nm and the thickness of
superconducting layer 𝑑Tinkham = 3.2 nm. Here, 𝜉GL is larger than 𝑑Tinkham, which is
consistent with the 2D nature of superconductivity. However, it should be noted
that dTinkham tends to overestimate the thickness of the superconducting layer, since
the present estimation is even larger than the thickness of the whole MoS2 flake.
This is because the Tinkham model does not consider spin-orbit interaction or
Pauli paramagnetism, both of which are essential in our devices. Hence, the
extracted dTinkham is an upper limit of the thickness of the superconducting layer,
which is consistent with previous reports [34,35].
Figure 3.5 Temperature dependence of Bc for state A at VBG = 20 V. Temperature dependence
of normalized Rs under two magnetic field configurations. (a) Magnetic fields perpendicular to
the ab plane of MoS2. (b) Magnetic fields parallel to the ab plane. Contrary to the effective
suppression of superconductivity by the perpendicular field, the parallel fields have very little
effect on the superconducting state. (c) Temperature dependence of Bc at VBG = 20 V in
perpendicular (black open square) and parallel (red open circle) magnetic fields. Solid lines are
the fittings by the 2D Ginzburg-Landau theory, which give 𝐵c⊥(𝑡) = 9.6 T and 𝐵c
∥(𝑡) = 60 T. (d)
Temperature dependence of in-plane Bc for different back gate voltages. Shaded areas in panels
(c) and (d) denote the corresponding Pauli limit.
Dual-gate MoS2 superconducting transistor
67
Figure 3.5(d) shows the temperature dependence of in-plane Bc at different
back gate voltages, with the Pauli paramagnetic limit 𝐵p ≈ 1.86 𝑇c(0) denoted by
the shaded areas. We consistently observed that in-plane Bc can easily exceed Bp
for all back gate voltages, in agreement with 2D Ising superconductivity protected
by large effective Zeeman field [16]. It is worth noting that despite the significant
change of Tc at different back gate voltages (Figure 3.5(d)), Bc remains almost a
vertical line without any sign of saturation in the whole accessible range of the
magnetic field, indicating strong Ising protection throughout the whole
superconducting dome.
Figure 3.6 Evolution of current-voltage characteristics from superconductor to normal metal
as temperature increases from 2.5 (red) to 13 K (blue). The exponent 𝛼 = 3 derived from
𝑉 ∝ 𝐼𝛼 at TBKT = 6.6 K is denoted by the black dashed line.
Figure 3.6 plots the V-I curves at different temperatures for state A at VBG = 20
V. The V-I behavior changes from the abrupt transition to the linear relation with
increasing temperature, indicating the transition from superconducting to normal
metallic state. In 2D superconductors, this transition is consistent with the BKT
theory. At the critical transition temperature TBKT, the V-I curve follows 𝑉 ∝ 𝐼𝛼
with 𝛼 = 3 because the density of free vortices, which causes energy dissipation,
scales with the square of the applied current. Figure 3.6 shows that this
relationship is best fitted with TBKT = 6.6 K. At temperatures lower than TBKT, free
vortices are bounded as pairs. The system enters into a quasi-long range ordered
phase where the correlation function decays like a power law. Resistance becomes
zero below TBKT.
Chapter 3
68
3.3.2 Sample State B (deep into the superconducting dome)
Figure 3.7 shows the measurement results from sample state B, where
superconductivity was induced deep into the superconducting dome (n2D ~1×1014
cm-2). As shown in Figure 3.7(a), the Tc could be effectively tuned by VBG within the
same bias range (as in Figure 3.3), but depletion by negative VBG is insufficient to
reach the super/non-superconducting phase boundary. As a function of VBG from -
20 to 20 V, Tc first increases rapidly and reaches the maximum at VBG = 3 V. With
further increases in VBG, Tc surprisingly goes down. From the phase diagram of
MoS2, starting from QCP Tc rises with increasing n2D until dome peak, at 1.2×1014
cm-2. Therefore, starting from n2D = 1×1014 cm-2 accessed by ionic gating, the
maximum charge carrier density that can be achieved at VBG = 6 V is n2D ~1.1×1014
cm-2, estimated by the geometrical capacitance of the back gate, which still remains
on the left side of the dome peak. Tc is expected to increase monotonically with the
increase of n2D. The decrease of Tc is not expected in this range.
From the transfer curve of the normal state at T = 13 K (Figure 3.7(b) inset),
depletion at negative VBG causes oscillations of Rs, a clear sign of universal
conductance fluctuation due to the disordered potential. For positive VBG, the Rs
decreases anomalously by over 40%. This amount of change is beyond the limit of
the back gate, which is not capable of doping 40% more carriers to the MoS2
channel (assuming that the mobility remains constant for the metallic state).
Considering that pure ionic gating confines carriers and induces superconductivity
mostly in the topmost layer, bottom layers (exclude the topmost layer) remain
almost undoped. In the previous chapter, two conducting channels in dual-gate
MoS2 transistors were confirmed by the coexistence of superconductivity and SdH
oscillations. Positive back gate bias has negligible influence on the carrier density
of top superconducting layer, but can significantly enhance the conductivity of the
bottom channel when VBG > Vth. The significant resistance drop and overall
metallic behavior are mainly contributed by the conductivity enhancement of the
bottom channel.
It is well known that the physical properties of superconducting thin films
depend greatly on the choice of substrate. The Tc of superconducting thin film is
close to that of the bulk material on insulating substrates, but is reduced on
metallic substrates due to suppression by the anti-proximity effect [36–39]. In the
present gating configuration, negative VBG depletes the carriers of the top
superconducting layer continuously, and Tc decreases in accordance with the phase
diagram. Positive VBG induces a metallic state in the bottom layers when VBG > Vth,
similar to a conventional solid-state transistor. The conductivity of the bottom
layers is significantly enhanced, and the structure becomes similar to that of a
superconducting thin film sitting on a metallic substrate. As a result, the
Dual-gate MoS2 superconducting transistor
69
superconductivity in the top layer is effectively weakened by the anti-proximity
effect in this superconductor-normal metal heterostructure. This explains why Tc
starts to decrease when VBG > 3 V, a behavior which is not consistent with the
phase diagram. Here, we use the terminology “anti-proximity effect” to describe
this phenomenon because in general, the proximity effect refers to the effect in
which a normal metal becomes superconducting due to contact with a
superconductor. It should be noted that the meaning of the proximity effect is two-
fold: (1) the pairing potential of the superconductor extends into the normal metal,
so part of the normal metal becomes superconducting, and (2) the
superconducting gap is lowered near the interface, so the transition temperature of
the superconductor is reduced. These two processes always happen at the same
time. To avoid any confusion, we use the “anti-proximity effect” to describe the
influence of the bottom normal layers on the superconducting properties of the
topmost layer.
Figure 3.7 Interacting hetero-states for tuning superconductivity far from the QCP (sample
state B). (a) Change of Tc as a function of VBG, where the solid line is a guideline for eyes. Inset:
expanded temperature dependence of Rs for different VBG close to the superconducting
transition, where Rs is normalized by the normal resistance measured at 13 K. (b) Calculated
conductivity of bottom layers is plotted to the left axis (black solid curve). Red dots connected
by the solid line are the inverse of the calculated inter-surface resistances 𝜌int (in analog of the
inter-surface conductivity) between the top and bottom channels. The inset shows VBG
dependence of Rs at 13 K, from which normal resistance of the top superconducting channel is
estimated as 750 Ω (dashed red line).
For a standard superconductor-normal metal (SN) sandwich structure where a
superconductor is in contact with a normal metal [38–45], the interaction between
them can be described by the Usadel equation [46], where the anti-proximity effect
in the system is introduced by defining the inter-surface resistance 𝜌𝑖𝑛𝑡, which is
small if the coupling between SN layers strong. The Tc of the SN sandwich
Chapter 3
70
structure decreases exponentially with increasing normal metal thickness dN,
approaching a finite value defined by the superconducting layer on the bulk metal
substrate [47–50], as shown in the following equation [36]:
In𝑇cs
𝑇c
=𝜏N
𝜏S + 𝜏N
[𝜓 (1
2+
ℎ(𝜏S + 𝜏N)
2𝜋𝑘B𝑇c𝜏S𝜏N
) − 𝜓 (1
2) − In√1 + (
𝜏S + 𝜏N
𝜏S𝜏N𝜔D
)2] (1)
where 𝜓(𝑥), h and kB are the digamma function, Planck constant and Boltzmann
constants, respectively. Tcs and Tc are the transition temperatures of the bulk
superconductor and SN sandwich structure. We denote 𝜏𝑁 = 2𝜋𝑉N𝑑N
𝑉S2 𝜌int and
𝜏𝑆 = 2𝜋𝑑S
𝑉S𝜌int, where 𝑉N and 𝑉S are the Fermi velocities, dS and dN are the thickness
of the superconducting and normal layers, respectively. 𝜌int defines the normalized
inter-surface resistance between superconducting and metallic layers. The
logarithmic term at the right hand side is only important for a perfect interface
( 𝜌int ⟶ 0), which is omitted in our case because the superconducting and metallic
layers are well separated.
At VBG = 3 V, just before switching on the bottom parallel channel (when the
bottom layers are still insulating), we obtain Tc = Tcs = 8.75 K. At higher VBG, using
equation (1) we can estimate 1/ρint (as an analogy of inter-surface conductivity) as
the coupling parameter between the SN heterostructure (Figure 3.7(b), dots and
red guiding curve). The small inter-surface conductivity [51,52] also corroborates
the omission of the logarithmic term in equation (1). With increasing VBG, 1/ρint
increases rapidly as a result of conductivity enhancement in bottom layers. At
higher VBG, due to the saturation of metallic conduction and reduced screening
length, 1/ρint also saturates. Considering the rough doping profile as shown in
Figure 3.2(b), we extract the contribution of the bottom channel by assuming that
the VBG dependence of Rs (T = 13 K) (Figure 3.7(b) inset) is contributed by parallel
conducting channels as 1
𝑅𝑠=
1
𝑅top+
1
𝑅bottom . Rtop ≈ 750 Ω is obtained at negative VBG
when the bottom channel is in off state. Using this model, the extracted gate
dependence of the conductance of the bottom layers is shown in Figure 3.7(b)
(black curve). The gate dependence of 1/ρint at positive VBG (Figure 3.7(b), right
axis) is clearly correlated with the increasing conductivity of the bottom layers,
which suggests that the increased conductivity in the bottom layers suppresses
superconductivity in the topmost layer, due to the anti-proximity effect.
As shown in Figure 3.7(a), compared with direct carrier depletion (VBG < 0), the
anti-proximity effect (VBG > 3V) changes the Tc in much smaller amounts. The
special carrier configuration might affect more subtly the other physical
parameters associated with varying superconducting states. As shown in Figure
Dual-gate MoS2 superconducting transistor
71
3.8(a), the upper critical field Bc (𝐵 ∥ 𝑐) is compared at two VBG regimes: field effect
depletion (VBG < 0 V) and anti-proximity effect (VBG > 3 V). For VBG < 0 V, the
temperature-dependent Bc curves are almost parallel to each other that is
consistent with sample state A close to the QCP (Figure 3.4(a)), where the decrease
of Bc is proportional to the decrease of Tc with reducing n2D. However, for VBG > 3V,
the temperature dependence of Bc becomes clearly different with the increase of
VBG. Phenomenologically, this is characterized as a significant decrease in the slope
of the Bc(T) curve even though the change in Tc is very small.
Figure 3.8 (a) Temperature dependence of Bc at different VBG. The top and bottom panels
correspond to the field effect regime (VBG < 0 V) and anti-proximity effect regime (VBG > 0 V),
respectively. (b) The calculated temperature dependence of Bc corresponding to the same
regimes shown in panel (a). The trend of the Bc(T) curves agrees with the experimental
observation.
Chapter 3
72
To further investigate this unusual critical field behavior, we simulated Bc(T) at
different VBG (Figure 3.8(b)), as described theoretically by Fominov etc. [36], and
find good agreement between experimental data and calculations. Details of the
simulation are described below. The behavior of Bc can be described by the
following equation:
ln𝑇cs
𝑇= −
𝜏N
𝜏S + 𝜏N
ln√1 + (𝜏S + 𝜏N
𝜏S𝜏N𝜔D
)2
− 𝜓 (1
2)
+1
2[1 +
휀S − 휀N
(휀S − 휀N)2 +4ℎ2
𝜏S𝜏N
] 𝜓 (1
2+
1
4𝜋𝑘B𝑇) [휀S + 휀N + √(휀S − 휀N)2 +
4ℎ2
𝜏S𝜏N
]
+1
2[1 −
휀S − 휀N
(휀S − 휀N)2 +4ℎ2
𝜏S𝜏N
] 𝜓 (1
2+
1
4𝜋𝑘B𝑇) [휀S + 휀N − √(휀S − 휀N)2 +
4ℎ2
𝜏S𝜏N
]
where 휀S = 𝐷S𝑒𝐵c +ℎ
𝜏S and 휀N = 𝐷N𝑒𝐵c +
ℎ
𝜏N depend on the upper critical field Bc.
Here, 𝐷 =1
3𝑉F𝑙 denotes the diffusion constant, where 𝑉F is the Fermi velocity and 𝑙
is the mean free path. The diffusion time constants 𝜏S and 𝜏N of normal carrier and
Cooper pairs are determined by the interlayer resistance 𝜌int as 𝜏N = 2𝜋𝑉N𝑑N
𝑉S2 𝜌int
and 𝜏S = 2𝜋𝑑S
𝑉S𝜌int, respectively. The mean free path 𝑙 can be determined by the
Ginzburg-Landau theory: 𝑙 =2.4𝜋𝐾B𝑇c𝜉(0)2
ℏ𝑉F. Here, 𝜉(0) is the coherence length
derived from 𝐵c(0) =Φ
0
2𝜋𝜉(0)2, where Φ0
= ℎ/2𝑒 = 2.07 × 10−15 Wb is the quantum
flux. Using 𝜌int obtained in the calculation of Tc, Bc(T) can be solved numerically by
ignoring the negligible contribution of the logarithmic term due to the large 𝜌int.
Here, Fermi velocity can be determined by 𝑉F = ℏ𝑘F/𝑚∗ ; 𝑘F = (4𝜋𝑛2D/𝑔s𝑔v)1/2 ,
where 𝑘F , 𝑚∗ , 𝑔s , 𝑔v are Fermi wave number, effective mass, spin degree of
freedom, and valley degree of freedom, respectively. In this device, the
superconducting top channel (one layer) is characterized by 𝑑S = 0.6 nm , 𝑚∗ =
0.5 𝑚𝑒, 𝑔s = 2, and 𝑔v = 2 (lowest band at the K/K’ point). For the bottom normal
channel (three layers) 𝑑N = 1.8 nm, 𝑚∗ = 0.6 𝑚e, 𝑔s = 1, and 𝑔v = 6 (lowest band
at the Q/Q’ points).
As shown in Figure 3.9, the fitting results match well with the experimental
data, except for the growing deviation at low temperatures. In our device, the
mesoscopic universal conductance fluctuation is observed in the transfer
characteristics measured at 13 K (inset of Figure 3.7(b)), indicating strong
interference within the long coherence length. On the other hand, the magnetic
Dual-gate MoS2 superconducting transistor
73
field plays an important role in setting the global phase coherence length. As a
consequence, the interplay between interference and dephasing process might
account for the deviation observed progressively at low temperatures, which
manifests as an upturn in the temperature dependence of Bc. From this point of
view, our result is in line with previous observations of a re-entrant
superconducting phase at low temperatures [53–55].
Figure 3.9 Theoretical fittings of the temperature dependence of Bc at different back gate
voltages. The solid dots and lines are experimental data and calculation results, respectively.
The calculations fit well with the experimental data at relatively high temperatures, but not
very well at low temperatures.
Due to the different charge configurations hence different underlying
mechanism, the theoretical simulation shown in Figure 3.8(b) clearly follows the
experimental characteristics found in Figure 3.8(a), where the upper and lower
panels can be described by the change of Tc and ρint due to the anti-proximity effect,
respectively. It should be noted that although the significant suppression of the
upper critical field at the positive back gate is clearly associated with the enhanced
interaction between the top and bottom channels, the underlying physical
interpretation is not established yet at this moment. One possible explanation is
that according to the WHH theory, d𝐵c
d𝑇|𝑇c
is inversely correlated with the scattering
time 𝜏, the anomalously large suppression of Bc with only small change of Tc might
be explained by the significant enhancement of the screening effect due to the
formation of metallic bottom state. With a more conductive bottom channel at
Chapter 3
74
higher VBG, the scattering time of the top channel increases due to screening, which
leads to the suppression of Bc.
Figure 3.10 Continuous “bipolar” switching characteristics. (a) Back gate dependence of Rs at
different temperatures. (b) Comparison between states A and B and the phase diagram from
Ref. 11. For small or negative VBG, Tc of both states fit well with the phase diagram, while for
both states A and B, the behavior of Tc shows clear deviations from the phase diagram due to
the anti-proximity effect at large VBG.
Based on the previous observations, we realize a “bipolar-like” superconducting
transistor operation (Figure 3.10(a)). Here, “ON” means zero resistance at
superconducting state, “OFF” means finite resistance at non-superconducting state.
The device can be switched off by both positive and negative back gate voltages,
supported by two different modes of working principles. With VBG < 0 V (field
effect regime), superconductivity is tuned by charge carrier depletion; with VBG > 3
V (anti-proximity effect regime), carriers accumulate at the bottom layers of MoS2,
which interact with the top superconducting layer through the anti-proximity
effect that suppresses the superconductivity. It is worth noting that moving deep
into the superconducting dome (close to the dome peak) can enhance the doping
uniformity which is crucial for device applications. On the other hand, larger n2D
also makes back gate tuning more challenging because of the difficulty in handling
the more robust superconducting phase compared with the low doping state close
to the QCP. As shown in Figure 3.10(a), well below the Tc (for T < 5 K) where
pairing is strong, both the carrier depletion and anti-proximity effect are too weak
to affect the superconducting state. The transfer curve shows little VBG dependence.
With weakened pairing at higher temperature, a bipolar response could be
established in which the superconducting state is tuned with an OFF/ON/OFF
pattern.
Dual-gate MoS2 superconducting transistor
75
Figure 3.10(b) summarizes the data from both sample states A and B and
compares them with the superconducting phase diagram of MoS2 (Ref. 11). For
negative or small positive VBG, the data of states A and B can be mapped well to the
phase diagram. For higher positive VBG, however, Tc starts to deviate significantly
from the phase diagram due to the anti-proximity effect between SN channels. It
should be noted that the 2D carrier densities at different back gate voltages are
estimated from the HfO2 back gate capacitance through 𝑛2D = 𝑛0 + 𝐶g ∙ 𝑉BG, where
𝑛0 is the carrier density at zero back gate voltage, induced by ionic liquid gating.
𝐶g = 260 nF/cm2 is the gate capacitance per unit area for 50 nm-HfO2. We noticed
that this estimation is reasonable at negative back gate voltages because the
bottom channel is insulating, it serves as an additional dielectric layer for the field
effect tuning of the top channel. The capacitance change due to this additional
layer is negligible because it is very thin compared with HfO2, the estimation
remains reasonable. However, at positive back gate the bottom layers become
metallic, so the back gating effect decays from the bottom to top, which means the
carrier densities of the top superconducting channel may deviate from the values
calculated by the back gate capacitance. However, even after taking this into
account, the decrease of Tc still disagrees with the phase diagram. The reason is
that a positive back gate cannot reduce the carrier density in the top
superconducting channel hence Tc should not decrease. Combined with the
unusual behavior of critical field, we can conclude that the deviation is caused by
the anti-proximity effect between the SN channels.
Finally, we compare our transistor with other materials and devices reported to
exhibit the field effect tuning of superconductivity (Figure 3.11). From application
perspective, we seek ways of creating superconducting transistors that can operate
at high temperatures with low gate voltage, corresponding to the top left corner of
the diagram. The high-Tc cuprate YBa2Cu3O7-x is the closest to this goal, but the
switching is realized through ionic gating hence superconductivity cannot be tuned
continuously at low temperatures. Device has to be warmed up to above the glass
transition temperature to change the gating status, therefore it is not practically
efficient. This disadvantage also exists in LSCO and ZrNCl with relatively high Tc.
Amorphous bismuth and LaAlO3/SrTiO3 can achieve continuous switching, but
they either work at very low temperatures or require high gate voltages. Proximity-
induced superconductivity in metal-decorated graphene can be tuned at low
temperatures with moderate gate voltages, by controlling the channel
transmittance for Cooper pairs without turning the channel material into
superconducting phase, which means it is not able to withstand a large
supercurrent.
Compared with previous materials, our MoS2 superconducting transistor can
achieve high speed switching operations because it is continuously tuned by back
Chapter 3
76
gate; this is an important factor towards the practical applications of
superconducting transistors. In addition, it requires only a small gate voltage of 10
V with bipolar response to the voltage bias and works in the helium-4 temperature
range. The combination of the ionic liquid gate and HfO2 back gate presented in
this work demonstrates an excellent candidate for balancing basic research and
possible applications.
Figure 3.11 Comparison of the performance of reported material systems that are capable of
switching superconductivity on and off.
3.4 Conclusion
In this chapter, we studied the transport properties of MoS2-EDLT on HfO2
substrate. Superconductivity is induced by ionic gating. Making use of the two
conducting channels model, the superconducting state in the top surface of few-
layer MoS2 can be tuned as a function of the electronic state of the bottom normal
layers. It is found that when the bottom channel is insulating (VBG < Vth), it serves
as an additional dielectric layer for the field effect tuning of superconductivity by
the back gate. When the bottom channel becomes metallic (VBG > Vth), strong anti-
proximity between these two channels could change the superconducting
properties significantly.
By applying different gate voltages, two different superconducting states in the
top channel were studied. For state A with lower carrier density close to the
quantum critical point, a continuous switching of superconductivity by the HfO2
back gate is realized. For state B with higher carrier density close to the dome peak
of the phase diagram, the unusual decrease in the transition temperature Tc with
Dual-gate MoS2 superconducting transistor
77
increasing back gate voltage provides strong evidence that enhanced anti-
proximity can suppress and eventually destroy the superconductivity in the top
layer.
A “bipolar”-like superconducting transistor was realized based on the two
conducting channels model. Different working principles are responsible for each
regime: field effect tuning for VBG < Vth, and anti-proximity effect for VBG > Vth. Our
superconducting transistor works in the helium-4 temperature range with
moderate gate voltages, providing an ideal candidate both for basic research and
device applications.
Chapter 3
78
References
[1] D.B. Haviland, Y. Liu, A.M. Goldman, Onset of superconductivity in the two-dimensional limit, Phys. Rev. Lett. 62 (1989) 2180–2183.
[2] J.M. Graybeal, M.R. Beasley, Localization and interaction effects in ultrathin amorphous
superconducting films, Phys. Rev. B. 29 (1984) 4167–4169.
[3] H. Takagi, T. Ido, S. Ishibashi, M. Uota, S. Uchida, Y. Tokura, Superconductor-to-
nonsuperconductor transition in (La1-xSrx)2CuO4 as investigated by transport and magnetic measurements, Phys. Rev. B. 40 (1989) 2254–2261.
[4] K.S. Takahashi, M. Gabay, D. Jaccard, K. Shibuya, T. Ohnishi, M. Lippmaa, J.-M. Triscone,
Local switching of two-dimensional superconductivity using the ferroelectric field effect,
Nature. 441 (2006) 195–198.
[5] A.D. Caviglia, S. Gariglio, N. Reyren, D. Jaccard, T. Schneider, M. Gabay, S. Thiel, G.
Hammerl, J. Mannhart, J.-M. Triscone, Electric field control of the LaAlO3/SrTiO3 interface ground state, Nature. 456 (2008) 624–627.
[6] J. Biscaras, N. Bergeal, S. Hurand, C. Feuillet-Palma, A. Rastogi, R.C. Budhani, M. Grilli, S.
Caprara, J. Lesueur, Multiple quantum criticality in a two-dimensional superconductor, Nat.
Mater. 12 (2013) 542–548.
[7] P.D. Eerkes, W.G. van der Wiel, H. Hilgenkamp, Modulation of conductance and
superconductivity by top-gating in LaAlO3/SrTiO3 2-dimensional electron systems, Appl. Phys. Lett. 103 (2013) 201603.
[8] A.T. Bollinger, G. Dubuis, J. Yoon, D. Pavuna, J. Misewich, I. Božović, Superconductor-
insulator transition in La2 - xSrxCuO4 at the pair quantum resistance, Nature. 472 (2011) 458–460.
[9] K. Ueno, S. Nakamura, H. Shimotani, A. Ohtomo, N. Kimura, T. Nojima, H. Aoki, Y. Iwasa,
M. Kawasaki, Electric-field-induced superconductivity in an insulator, Nat. Mater. 7 (2008) 855–858.
[10] X. Leng, J. Garcia-Barriocanal, S. Bose, Y. Lee, A.M. Goldman, Electrostatic Control of the
Evolution from a Superconducting Phase to an Insulating Phase in Ultrathin YBa2Cu3O7-x Films, Phys. Rev. Lett. 107 (2011) 27001.
[11] J.T. Ye, Y.J. Zhang, R. Akashi, M.S. Bahramy, R. Arita, Y. Iwasa, Superconducting Dome in
a Gate-Tuned Band Insulator, Science. 338 (2012) 1193–1196.
[12] W. Shi, J. Ye, Y. Zhang, R. Suzuki, M. Yoshida, J. Miyazaki, N. Inoue, Y. Saito, Y. Iwasa,
Superconductivity Series in Transition Metal Dichalcogenides by Ionic Gating, Sci. Rep. 5 (2015) 12534.
[13] J.T. Ye, S. Inoue, K. Kobayashi, Y. Kasahara, H.T. Yuan, H. Shimotani, Y. Iwasa, Liquid-gated interface superconductivity on an atomically flat film, Nat. Mater. 9 (2010) 125–128.
[14] D. Costanzo, S. Jo, H. Berger, A.F. Morpurgo, Gate-induced Superconductivity in atomically
thin MoS2 crystals, ArXiv151203222 Cond-Mat. (2015).
[15] S. Jo, D. Costanzo, H. Berger, A.F. Morpurgo, Electrostatically Induced Superconductivity at the Surface of WS2, Nano Lett. 15 (2015) 1197–1202.
[16] J.M. Lu, O. Zheliuk, I. Leermakers, N.F.Q. Yuan, U. Zeitler, K.T. Law, J.T. Ye, Evidence for two-dimensional Ising superconductivity in gated MoS2, Science. (2015) aab2277.
[17] Y. Saito, Y. Nakamura, M.S. Bahramy, Y. Kohama, J. Ye, Y. Kasahara, Y. Nakagawa, M.
Onga, M. Tokunaga, T. Nojima, Y. Yanase, Y. Iwasa, Superconductivity protected by spin-
valley locking in ion-gated MoS2, Nat. Phys. 12 (2016) 144–149.
Dual-gate MoS2 superconducting transistor
79
[18] X. Xi, Z. Wang, W. Zhao, J.-H. Park, K.T. Law, H. Berger, L. Forró, J. Shan, K.F. Mak, Ising pairing in superconducting NbSe2 atomic layers, Nat. Phys. 12 (2016) 139–143.
[19] K.S. Novoselov, A.K. Geim, S.V. Morozov, D. Jiang, Y. Zhang, S.V. Dubonos, I.V.
Grigorieva, A.A. Firsov, Electric Field Effect in Atomically Thin Carbon Films, Science. 306
(2004) 666–669.
[20] C. Lee, H. Yan, L.E. Brus, T.F. Heinz, J. Hone, S. Ryu, Anomalous Lattice Vibrations of Single- and Few-Layer MoS2, ACS Nano. 4 (2010) 2695–2700.
[21] J. Chang, L.F. Register, S.K. Banerjee, Ballistic performance comparison of monolayer
transition metal dichalcogenide MX2 (M = Mo, W; X = S, Se, Te) metal-oxide-semiconductor field effect transistors, J. Appl. Phys. 115 (2014) 84506.
[22] Y.J. Zhang, J.T. Ye, Y. Yomogida, T. Takenobu, Y. Iwasa, Formation of a Stable p–n
Junction in a Liquid-Gated MoS2 Ambipolar Transistor, Nano Lett. 13 (2013) 3023–3028.
[23] R. Roldán, E. Cappelluti, F. Guinea, Interactions and superconductivity in heavily doped MoS2, Phys. Rev. B. 88 (2013) 54515.
[24] T. Eknapakul, P.D.C. King, M. Asakawa, P. Buaphet, R.-H. He, S.-K. Mo, H. Takagi, K.M.
Shen, F. Baumberger, T. Sasagawa, S. Jungthawan, W. Meevasana, Electronic Structure of a Quasi-Freestanding MoS2 Monolayer, Nano Lett. 14 (2014) 1312–1316.
[25] K.A. Parendo, K.H.S.B. Tan, A. Bhattacharya, M. Eblen-Zayas, N.E. Staley, A.M. Goldman,
Electrostatic Tuning of the Superconductor-Insulator Transition in Two Dimensions, Phys. Rev. Lett. 94 (2005) 197004.
[26] X. Chen, Z. Wu, S. Xu, L. Wang, R. Huang, Y. Han, W. Ye, W. Xiong, T. Han, G. Long, Y.
Wang, Y. He, Y. Cai, P. Sheng, N. Wang, Probing the electron states and metal-insulator
transition mechanisms in molybdenum disulphide vertical heterostructures, Nat. Commun. 6
(2015) 6088.
[27] N.D. Mermin, H. Wagner, Absence of Ferromagnetism or Antiferromagnetism in One- or Two-Dimensional Isotropic Heisenberg Models, Phys. Rev. Lett. 17 (1966) 1133–1136.
[28] V.L. Berezinskiǐ, Destruction of Long-range Order in One-dimensional and Two-dimensional
Systems having a Continuous Symmetry Group I. Classical Systems, Sov. J. Exp. Theor. Phys. 32 (1971) 493.
[29] J.M. Kosterlitz, D.J. Thouless, Ordering, metastability and phase transitions in two-
dimensional systems, J. Phys. C Solid State Phys. 6 (1973) 1181.
[30] N. Reyren, S. Thiel, A.D. Caviglia, L.F. Kourkoutis, G. Hammerl, C. Richter, C.W. Schneider,
T. Kopp, A.-S. Rüetschi, D. Jaccard, M. Gabay, D.A. Muller, J.-M. Triscone, J. Mannhart, Superconducting Interfaces Between Insulating Oxides, Science. 317 (2007) 1196–1199.
[31] T. Schneider, J.M. Singer, Universal Critical Quantum Properties of Cuprate Superconductors,
in: M. Ausloos, A.A. Varlamov (Eds.), Fluct. Phenom. High Temp. Supercond. 32 (1997)
361–368.
[32] S.L. Sondhi, S.M. Girvin, J.P. Carini, D. Shahar, Continuous quantum phase transitions, Rev. Mod. Phys. 69 (1997) 315–333.
[33] M. Tinkham, Effect of Fluxoid Quantization on Transitions of Superconducting Films, Phys. Rev. 129 (1963) 2413–2422.
[34] M. Ben Shalom, M. Sachs, D. Rakhmilevitch, A. Palevski, Y. Dagan, Tuning Spin-Orbit
Coupling and Superconductivity at the SrTiO3/LaAlO3 Interface: A Magnetotransport Study, Phys. Rev. Lett. 104 (2010) 126802.
[35] M. Kim, Y. Kozuka, C. Bell, Y. Hikita, H.Y. Hwang, Intrinsic spin-orbit coupling in
superconducting δ-doped SrTiO3 heterostructures, Phys. Rev. B. 86 (2012) 85121.
Chapter 3
80
[36] Y.V. Fominov, M.V. Feigel’man, Superconductive properties of thin dirty superconductor-normal-metal bilayers, Phys. Rev. B. 63 (2001) 94518.
[37] A. Stępniak, A. Leon Vanegas, M. Caminale, H. Oka, D. Sander, J. Kirschner, Atomic layer superconductivity, Surf. Interface Anal. 46 (2014) 1262–1267.
[38] P.G. DE GENNES, Boundary Effects in Superconductors, Rev. Mod. Phys. 36 (1964) 225–
237.
[39] M. Wolz, C. Debuschewitz, W. Belzig, E. Scheer, Evidence for attractive pair interaction in
diffusive gold films deduced from studies of the superconducting proximity effect with aluminum, Phys. Rev. B. 84 (2011) 104516.
[40] L.N. Cooper, Superconductivity in the Neighborhood of Metallic Contacts, Phys. Rev. Lett. 6 (1961) 689–690.
[41] W.L. McMillan, Tunneling Model of the Superconducting Proximity Effect, Phys. Rev. 175
(1968) 537–542.
[42] A.A. Golubov, Proximity effect in dirty N/S multilayers, Proceedings of the SPIE. 2157 (1994) 353-362.
[43] Y.V. Fominov, N.M. Chtchelkatchev, A.A. Golubov, Nonmonotonic critical temperature in superconductor/ferromagnet bilayers, Phys. Rev. B. 66 (2002) 14507.
[44] J. Kim, Y.-J. Doh, K. Char, H. Doh, H.-Y. Choi, Proximity effect in Nb/Au/CoFe trilayers, Phys. Rev. B. 71 (2005) 214519.
[45] K. Kim, J.H. Kwon, J. Kim, K. Char, H. Doh, H.-Y. Choi, Proximity effect in Nb/Cu/CoFe
trilayers, Phys. Rev. B. 74 (2006) 174503.
[46] K.D. Usadel, Generalized Diffusion Equation for Superconducting Alloys, Phys. Rev. Lett. 25 (1970) 507–509.
[47] N.R. Werthamer, Theory of the Superconducting Transition Temperature and Energy Gap Function of Superposed Metal Films, Phys. Rev. 132 (1963) 2440–2445.
[48] J.J. Hauser, H.C. Theuerer, N.R. Werthamer, Superconductivity in Cu and Pt by Means of Superimposed Films with Lead, Phys. Rev. 136 (1964) A637–A641.
[49] J. Clarke, the Proximity Effect Between Superconducting and Normal Thin Films in Zero
Field, J. Phys. Colloq. 29 (1968) C2-3-C2-16.
[50] G. Brammertz, A.A. Golubov, P. Verhoeve, R. den Hartog, A. Peacock, H. Rogalla, Critical
temperature of superconducting bilayers: Theory and experiment, Appl. Phys. Lett. 80 (2002)
2955–2957.
[51] S. Das, J. Appenzeller, Where Does the Current Flow in Two-Dimensional Layered Systems?, Nano Lett. 13 (2013) 3396–3402.
[52] J. Na, M. Shin, M.-K. Joo, J. Huh, Y.J. Kim, H.J. Choi, J.H. Shim, G.-T. Kim, Separation of
interlayer resistance in multilayer MoS2 field-effect transistors, Appl. Phys. Lett. 104 (2014) 233502.
[53] Z. Han, A. Allain, H. Arjmandi-Tash, K. Tikhonov, M. Feigel’man, B. Sacépé, V. Bouchiat,
Collapse of superconductivity in a hybrid tin-graphene Josephson junction array, Nat. Phys. 10 (2014) 380–386.
[54] B. Spivak, F. Zhou, Mesoscopic Effects in Disordered Superconductors near Hc2, Phys. Rev. Lett. 74 (1995) 2800–2803.
[55] V.M. Galitski, A.I. Larkin, Disorder and Quantum Fluctuations in Superconducting Films in
Strong Magnetic Fields, Phys. Rev. Lett. 87 (2001) 87001.
81
Chapter 4. Superconductor-normal
metal junction in gated MoS2
Abstract
In previous chapters, we studied the transport properties of MoS2-EDLT
devices on h-BN and HfO2 substrates, hetero electronic states can be induced in a
single few-layer MoS2 flake and new device functionalities have be developed by
controlling the interaction between the two states. The hetero-states at the top and
bottom surfaces of few-layer MoS2 form a vertical junction similar to the
conventional van der Waals heterostructure. In this chapter, we focus on the lateral
transport properties of few-layer MoS2 superconducting transistor. By controlling
the doping level in different regions of the MoS2 surface, we study the transport
properties of the superconductor-normal metal junction formed at the surface of
MoS2.
In preparation
“An unusual magnetoresistance in a superconductor-normal junction of MoS2”
Q.H. Chen, J.M. Lu, L. Liang and J.T. Ye
Chapter 4
82
4.1 Introduction
Electrical transport of the superconductor-normal conductor (SN) junction has
shown a variety of interesting mesoscopic phenomena [1–3]. Given a particular
superconductor, the normal conductor part of the junction can be a metal [4,5],
semimetal [6–9] or degenerate semiconductor [10–12]. As shown in Figure 4.1, at
the interface of an SN junction, the current is converted between a dissipative flow
and non-dissipative supercurrent carried by normal electrons and Cooper pairs,
respectively. The physical mechanism of this process was first proposed by
Andreev in 1964 [13]. At the superconducting side of the junction, an energy gap
opens up for single electrons. An electron injected from the normal conductor,
having excitation energy slightly above the Fermi level but lower than the
superconducting energy gap, cannot be accommodated in the superconductor as a
single particle because only Cooper pairs are allowed in the superconductor [14].
Hence an additional electron with excitation energy below the Fermi level is
required from the normal conductor to form a Cooper pair consisting of an up and
down spin electron. This process leaves behind a hole in the normal conductor,
with opposite spin and velocity to the incident electron. The energy state is shown
schematically in Figure 4.2. The reflected hole traces the same path as the electron
in the opposite direction. The reflection procedure is similar for an incident hole
through time-reversal symmetry. This process is known as retro-reflection or
Andreev reflection.
Figure 4.1 Andreev reflection of an electron excitation from a normal metal (N) by a
superconductor (S). A spin-up electron incident from the normal metal is reflected as a spin-
down hole with opposite velocity, at the same time a Cooper pair forms in the superconductor.
Since the reflected hole has spin polarization opposite to that of the electron,
then if only one spin band is occupied by the conduction electrons in the normal
Superconductor-normal metal junction in gated MoS2
83
conductor, Andreev reflection will be suppressed due to the lack of spin to form
Cooper pairs. For example in a ferromagnetic material where spins have preferred
orientation, the probability of the Andreev reflection (and hence the conductance
of the junction) is a function of the spin-polarization in the normal state.
Figure 4.2 Illustration of the energy state for Andreev reflection. Due to the quasiparticle gap
in the superconductor, a potential barrier is formed at the interface between the normal
conductor and superconductor.
In a traditional SN junction, the superconductor and normal conductor are
constructed by different materials. For electrical measurement, contact resistance
between two different materials is unavoidable due to work function difference,
residues from the fabrication process, etc. A perfectly clean interface is impossible,
which means the probability of Andreev reflection is more or less affected by
external factors. In order to study the intrinsic properties of Andreev reflection, we
fabricated superconductor-normal metal (SN) junctions at the surface of few-layer
MoS2 crystal. This idea was implemented by partially covering the surface of a
MoS2-EDLT using an h-BN flake. The exposed surface of the MoS2 flake is turned
into a superconductor by ionic gating, while the h-BN covered surface remains
normal, forming an atomically sharp SN junction. For the normal side,
pronounced Shubnikov-de Haas oscillations reveal the high quality and
homogeneity of the MoS2. The SN interface is perfectly clean because both the
superconducting and normal sides are formed in a single crystal; the transparency
is expected to be high. Yet an intrinsic energy gap forms at the interface, due to
different doping levels at the S and N sides. At low temperatures, a non-monotonic
behavior of the Andreev reflection is observed as the magnetic field increases. This
result is in line with the theoretical calculations of the Andreev reflection
probabilities at the SN interface with intermediate barrier strength, providing an
excellent platform for studying the transport properties of clean SN junctions.
Chapter 4
84
4.2 Device fabrication
To make the SN junction in a single crystal, we fabricated a MoS2-EDLT device
on an h-BN substrate following the same fabrication procedure outlined in Chapter
2. A few-layer MoS2 flake (2H polytype) was first transferred onto a uniformly thin
h-BN flake, which served as the substrate. Another carefully chosen h-BN flake
with a quasi-rectangular shape was put on top of the MoS2 flake to locally separate
ionic liquid from the surface of the MoS2. Electrodes composed of Ti/Au (5/60 nm)
were patterned with the standard e-beam lithography method.
Figure 4.3 (a) Optical image of a typical h-BN/MoS2/h-BN device. The big dark blue flake with
label “Bottom h-BN” is the bottom substrate, the light blue color in the middle is the few-layer
MoS2 flake, a small h-BN flake is laid across the MoS2 channel, indicated by the black arrow with
label “Top h-BN”. Scale bar: 5 μm. (b) Cartoon picture showing the SN junction. (c) AFM height
profile of the MoS2 flake, showing that the thickness of the MoS2 in this study is 3.5 nm,
corresponding to a layer number of five.
The exposed MoS2 surface is in contact with the ionic liquid, so
superconductivity can be easily induced in this area through ionic gating. For the
h-BN covered area, however, the ionic liquid is separated from the sample surface.
In our previous study we found that ions can still accumulate at the surface of the
top h-BN, but the capacitance is much smaller compared with the exposed area,
thus the gating effect is very weak. This area remains almost unaffected by ionic
gating, showing intrinsic semiconducting properties of few-layer MoS2. Instead,
the conductivity of this channel can be well controlled by the back gate through
dielectrics composed of 300 nm SiO2 and the bottom h-BN flake. Figure 4.3(a)
shows the device we measured in detail for the remaining part of this chapter,
which was fabricated according to the aforementioned configuration. Electrodes
Superconductor-normal metal junction in gated MoS2
85
were patterned for both the exposed area and h-BN covered area. The
corresponding AFM height profile of this device is shown in Figure 4.3(b), where
the thickness of the studied MoS2 flake is 3.5 nm. Using the same calculation
procedure as in previous chapters, the thickness was calculated by 𝑑 = 𝑎 +
(𝑁 − 1) × 𝑏, where a is the height of the bottom layer (including the space between
the bottom layer and the substrate) and b is the layer thickness. In MoS2, the
typical value of a ranges from 0.8 to 1 nm, while b is around 0.62 nm [15,16]. As a
result, the number of layers of the studied flake is five.
To avoid any confusion, in the following description “exposed MoS2 channel”
refers to the surface of MoS2 that is exposed to ionic liquid, it becomes
superconducting due to ionic gating; “h-BN covered MoS2 channel” refers to the
surface of the MoS2 that is covered by the top h-BN flake, it remains
semiconducting. The SN junction forms at the interface between the exposed and
h-BN covered channel. For most of the electrical measurement (as shown in Figure
4.3(a)), contact 1 and 6 are used as the source-drain electrodes to send current
through the whole MoS2 channel, contacts 2 and 3 are used to measure the
resistance of the h-BN covered MoS2 channel; contacts 4 and 5 are used to measure
the resistance of the exposed MoS2 channel. The resistance measured by contacts 2
and 3 includes both the resistance from MoS2 channel and the resistance from two
SN interfaces (left and right side of the h-BN covered channel). The properties at
these two interfaces are the same so it can be regarded as one SN junction.
4.3 Field effect transistor characteristics
Figure 4.4 Transfer curve by ionic gating at T = 220 K for the exposed area (contacts 4 and 5)
as shown in Figure 4.3(a), with fixed VDS = 0.1 V. Arrows indicate the scan direction of the gate
voltage. Large hysteresis is observed due to low ion mobility at relatively low temperatures.
The off-state gap ΔVLG = 2 V is consistent with the value from previous reports [17,18].
Chapter 4
86
The transfer curve for the exposed area (contacts 4 and 5) by ionic gating is
shown in Figure 4.4, where the arrows indicate the gate voltage scan direction.
Consistent with previous measurements, ionic gating procedure was conducted at
T = 220 K. The device can be easily switched on and off with small ionic gate
voltages. Both electron and hole conduction are observed, showing a bipolar
transistor operation with a current on/off ratio higher than 105 for the electron side.
Similar to the previous case, large hysteresis was also observed due to the low ion
mobility at relatively low temperatures. With ionic gate voltage fixed, the
temperature is quickly cooled to freeze the ionic liquid. Below T = 180 K, the gate
voltage can be retracted without losing the gating effect, because ions cannot move.
The gating effect can be released by warming up the sample to T = 220 K, where
ion movement is resumed. Different ionic gate voltages were tested by repeatedly
warming up and cooling down.
Figure 4.5 (a) Channel resistance as a function of VBG for different temperatures from 170 to 2
K. (b) FET mobility obtained by fitting the transfer curve with the back gate capacitance
𝜇FET =1
𝐶g
d𝜎
d𝑉BG. The red dashed line shows the power law relation 𝜇(𝑇)~𝑇−𝛾 , with 𝛾 = 1.73
which is smaller than the theoretical prediction but consistent with previous experimental
observations [23,24].
At low temperatures where the ionic liquid is frozen, we measured the
temperature dependence of the transfer curve by the back gate for the h-BN
covered area (contacts 2 and 3). The measured four-probe resistance in the range
of VBG = -50 to 80 V is shown in Figure 4.5(a). For all different temperatures, R
decreases with increasing VBG, as expected for an n-type semiconductor. For VBG >
0 V, sample resistance shows metallic behavior, where resistance decreases
significantly with decreasing temperature. The FET mobility can be extracted from
the gate dependence of conductivity by 𝜇FET =1
𝐶g
d𝜎
d𝑉BG , where 𝜎 is the conductivity
Superconductor-normal metal junction in gated MoS2
87
and Cg is the gate capacitance per unit area. Here, for 300 nm SiO2 and 30 nm h-
BN 𝐶g = 10.5 nF/cm2. The extracted FET mobilities at different temperatures are
shown in Figure 4.5(b). At T = 170 K, 𝜇FET ~ 100 cm2/Vs which increases with
decreasing temperature, with a tendency to saturate below T = 80 K. At T = 2 K,
𝜇FET ~ 1000 cm2/Vs, which shows almost one order of magnitude improvement. As
we know, the measured mobility can be described by a simple formula: 1/𝜇(𝑇) =
1/𝜇imp + 1/𝜇ph(𝑇), where 𝜇imp(𝑇) is the contribution from impurity scatterings
and 𝜇ph(𝑇) is the temperature-dependent contribution from phonon
scatterings [19]. At high temperatures, phonon scatterings are the main source that
limits the mobility of electrons. As the temperature decreases, phonon scatterings
are gradually suppressed, and hence the mobility increases dramatically. The
temperature dependence of 𝜇ph(𝑇) is described by the power law 𝜇ph(𝑇)~𝑇−𝛾. This
behavior is consistent with mobility contributions from optical phonons.
Theoretical calculations showed that the exponent is ~1.69 for monolayer [20] and
~2.5 for bulk [21] MoS2. Here, in our measurement the best fitting gives an
exponent of 𝛾 = 1.73, which is in the middle of previously reported values 1.9-
2.5 [19,22] and 0.55-1.7 [23,24]. At sufficiently low temperatures, the phonon
scatterings are completely suppressed and the main scattering sources are from
long-range Coulomb impurities and short-range atomic defects [25–27]. Since
these scatterings have very weak temperature dependence, mobility saturates at
low temperatures.
Figure 4.6 Chanel resistance as a function of SiO2 back gate voltage at T = 2 K, VLG = 3 V. Black
and red curves represent the MoS2 channel that is covered (contacts 2 and 3) and not covered
(contacts 4 and 5) by h-BN, respectively. For the h-BN covered MoS2 channel, the transfer curve
becomes flat at VBG < 30 V because the resistance exceeds the detecting limit of our setup.
At low temperatures, the FET characteristics for the h-BN covered channel and
exposed channel show distinctive response by the back gate tuning, as shown in
Chapter 4
88
Figure 4.6. Since the exposed area is already highly doped (n2D ~ 1014 cm-2) by ionic
gating, the back gate has almost no effect on the channel resistance. The h-BN
covered MoS2 channel is only slightly doped by ionic gating, and thus can be easily
switched on and off by the back gate. The on/off ratio is about 105.
4.4 Superconductivity in the exposed MoS2 channel
(contacts 4 and 5)
No clear superconductivity is observed for the first cooling down with VLG = 3 V.
When the sample is warmed up to T = 220 K and a higher ionic gate voltage is
applied VLG = 3.8 V, the RT curve of the exposed MoS2 channel shows a clear
superconducting transition, as shown in Figure 4.7(a). A more robust
superconducting state with higher Tc is observed after applying a 100 V back gate
voltage (red curve in Figure 4.7(a)). At low temperatures, the RT curves also show
strong perpendicular magnetic field dependence. The superconducting transition
is suppressed by applying magnetic fields perpendicular to the ab-plane of the
MoS2 (Figure 4.7(b)).
Figure 4.7 (a) Temperature dependence of the channel resistance for exposed MoS2 channel.
Without applying back gate voltage (black curve), the superconducting state is close to the QCP.
(b) Exposed MoS2 channel resistance as a function of temperature at VBG = 100 V, under
different perpendicular magnetic fields: 0, 0.05 T, 0.1 to 1.5 T in 0.1T steps, 1.7, 2, 2.5, 3 and 5 T.
Superconductivity is gradually suppressed by the magnetic field.
4.5 Magnetoresistance in the h-BN covered MoS2 channel
(contacts 2 and 3)
When measuring the magnetoresistance (MR) of the h-BN covered MoS2
channel, an unusual behavior is observed as shown in Figure 4.8(a). With the
Superconductor-normal metal junction in gated MoS2
89
applied perpendicular magnetic field increases from 0 to 12 T, the resistance of the
h-BN covered MoS2 channel first decreases and then increases. This behavior in
the small field region is magnified and shown in Figure 4.8(b). At B > 2.5 T, MR
shows a parabolic B-field dependence, which results from the semi-classical
Lorentz deflection of carriers. As the field further increases to B > 5 T, pronounced
Shubnikov-de Haas (SdH) oscillations appear as a manifestation of the high-
quality MoS2 channel covered by the h-BN flake. Thus, we divide the MR into two
regions: the low field region with anomalous MR behavior, and the high field
region with SdH oscillations.
Figure 4.8 (a) Magnetoresistance of MoS2 covered by h-BN at T = 2 K and VBG = 100 V. (b)
Magnification of the small field region in panel (a).
4.5.1 SdH quantum oscillations
We first focus on the SdH oscillations of the h-BN covered MoS2 channel in the
high field region. MR curves for different back voltages are shown in Figure 4.9(a).
Both the amplitude and frequency of the SdH oscillations change significantly with
the increase of back gate voltage. Similar to the description in previous chapters,
the oscillation components are extracted by subtracting the MR background. The
top and bottom envelope lines are determined by the peak and valley positions of
the oscillations, and the average of the two envelope lines is taken as the MR
background. After subtracting the MR background, the oscillations shown in
Figure 4.9(b) are unique periodic functions of 1/B. The period decreases with
increasing back gate voltage.
In 2D electron gas, the carrier density 2Dn is directly related to the period of
the quantum oscillations through 𝑛2D = 𝑔𝑒𝐵F/ℎ, where 𝑔 = 𝑔s ∙ 𝑔v is the Landau
level degeneracy and BF is the oscillation frequency in 1/B, which can be obtained
Chapter 4
90
from Figure 4.9(b). On the other hand, carriers of the h-BN covered MoS2 channel
are induced by the back gate (in the h-BN covered MoS2 channel, the ionic gating
effect can be neglected because the top h-BN separate the ions from the MoS2
surface), and thus the carrier density can be calculated through the geometric
capacitance and gate voltage 𝑛2D = 𝐶g(𝑉BG − 𝑉th)/𝑒 . Here, Vth is the threshold
voltage and 𝐶g = 10.5 nF/cm2 is the capacitance per unit area for the back gate of
300 nm SiO2 and 30 nm h-BN used in this experiment. Hence 𝐵F =ℎ𝐶g
g𝑒2 (𝑉BG − 𝑉th),
the Landau level degeneracy g can be obtained from the linear fitting of back gate
dependence of BF as shown in Figure 4.9(c). The result shows that g = 3.06, in
excellent agreement with what was obtained in previous chapters, and also with
values from the literature [22].
Figure 4.9 (a) Magnetoresistance at high field region for VBG = 90 (blue), 95 (red) and 100 V
(black), at T = 2 K. (b) After subtracting the MR background, ΔR is a unique period function of
1/B, consistent with the behavior of a standard SdH oscillation. Curves are vertically shifted for
clarity. (c) Oscillation period BF as a function of back gate voltage obtained from panel (b). With
𝐵F =ℎ𝐶g
g𝑒2 (𝑉BG − 𝑉th), the linear fitting yields a Landau level degeneracy of g = 3.06.
According to the relation 𝜇𝑞 = 𝑒𝜏𝑞/𝑚∗ ≈ 1/𝐵𝑞, the quantum scattering time 𝜏𝑞
can be obtained from the magnetic field corresponding to the onset of SdH
oscillations [28]. Here with the onset of the oscillations at 𝐵𝑞 = 5 T, the quantum
mobility is ~ 2000 cm2V-1s-1. A more accurate method of estimating the mobility is
to use the Dingle plot in the Ando formula:
04 exp( )th
c q
R R
Superconductor-normal metal junction in gated MoS2
91
where, 𝛾𝑡ℎ = 𝛼𝑇/sinh (𝛼𝑇) is the Dingle term 𝛼 = 2𝜋2𝑘B/ℏ𝜔c , 𝑅0 is the classical
resistance in the zero applied field, 𝜔c = 𝑒𝐵/𝑚∗ is the cyclotron frequency and 𝑚∗
is the electron effective mass. ℏ is the reduced Planck constant and 𝑘B is the
Boltzmann constant. 𝜏𝑞 is the quantum scattering time, determined by the
scattering event that destroy quantized cyclotron orbits.
In Figure 4.10(a) the red dashed lines are the amplitude fitted by the Ando
formula for the oscillations at VBG = 100 V and T = 2 K. Figure 4.10(b) shows the
ln (∆𝑅/4𝑅0𝛾𝑡ℎ) as a function of the inverse of the magnetic field. The best linear
fitting gives the quantum scattering of 𝜏𝑞 = 801 𝑓𝑠 . From 𝜇 = 𝑒𝜏/𝑚∗ we can
estimate the mobility of the electron to be 2348 cm2V-1s-1. This result is in good
agreement with the value obtained from the onset of quantum oscillation.
Quantum oscillation is the most accurate way to extract the carrier mobility, which
is a well-established method in conventional 2D electron gas. In general, the
mobility obtained from the SdH oscillations is smaller than the value obtained
from FET fitting. In our device, however, it is the opposite, suggesting that the FET
fitting underestimates the carrier mobility. As far as we know, the mobility
extracted from the SdH oscillations in our experiment is the highest compared
with previous reports [19,22]. This result provides strong evidence that the quality
and homogeneity of the MoS2 channel covered by the top h-BN are very good.
Figure 4.10 (a) ∆𝑅 plotted as a function of the inverse of the magnetic field at T = 2 K and VBG =
100 V. The red dashed line is the amplitude fitted by the Ando formula. (b) Dingle plot of ∆𝑅
(solid squares). ln (∆𝑅/4𝑅0𝛾𝑡ℎ) is proportional to the inverse of the magnetic field 1/B. From
the best linear fitting (black solid line), the extracted quantum scattering time is 801 fs. The
corresponding mobility is 2348 cm2V-1s-1.
Chapter 4
92
Figure 4.11 (a) Magnetoresistance of h-BN covered MoS2 channel at temperatures from 2 K to
10 K. Curves are vertically shifted by 30 Ω for comparison. The SdH oscillation disappears
above 6 K due to enhanced electron-phonon scatterings. (b) Oscillation amplitude after
subtracting the MR background, plotted with the inverse of the magnetic field 1/B. (c)
Temperature dependence of oscillation amplitude at B = 9.4 T. The best fitting using the Ando
formula yields an electron effective mass of 0.60 me. (d) Effective mass fitted at different
magnetic fields. All the fittings give similar results of 0.6 ± 0.02 me, in excellent agreement with
the theoretical calculation of 0.6 me for multilayer MoS2.
Figure 4.11(a) and (b) show the temperature dependence of the SdH oscillations
for VBG = 100 V, before (a) and after (b) subtracting MR background. According to
the Ando formula, the oscillation amplitude at the fixed magnetic field is described
by ∆𝑅 ∝ 𝛼𝑇/sinh (𝛼𝑇) . The effective mass of the conduction electrons can be
deduced by fitting the temperature dependent oscillation amplitudes with this
formula. As shown in Figure 4.11(c), the best fitting yields an effective mass of
𝑚∗ = 0.60 𝑚𝑒 at B = 9.4 T, where me is the bare electron mass. Figure 4.11(d) shows
Superconductor-normal metal junction in gated MoS2
93
the fitting results for other magnetic fields, all of which give very similar values.
This result matches well with the theoretical calculation of the MoS2 band
structure, which predicts there are two bands near the Fermi level (K and Q
valleys), with the lowest energy of the conduction band at the K/K’ valleys (𝑚∗ =
0.5 𝑚𝑒) for monolayer MoS2, the Q/Q’ valleys (𝑚∗ = 0.6 𝑚𝑒) become the lowest of
conduction band for multilayer MoS2. Since the MoS2 flake in this study is
composed of five layers, the result is consistent with the theoretical calculation.
4.5.2 Anomalous magnetoresistance
Figure 4.12 (a) Magnetoresistance at different temperatures from 2 to 10 K, with VBG = 100 V.
Each curve is shifted by 50 Ω for clarity. (b) Magnified view of the low field region, showing the
evolution of the unusual MR behavior at the low field region with different temperatures.
As shown in Figure 4.8, for the h-BN covered MoS2 channel, an unusual MR
behavior is observed. We measured the resistance of the h-BN covered MoS2
channel using contacts 2 and 3, this also involves the resistance at the interface
between exposed and h-BN covered channels. More specifically, the resistance
between contacts 2 and 3 can be written as 𝑅23 = 𝑅SN + 𝑅MoS2+ 𝑅SN, where 𝑅SN is
the resistance at the SN interface and 𝑅MoS2 is the channel resistance of MoS2. The
interface resistance behaves similarly at the left and right side of the h-BN covered
channel, so it can be regarded as one SN junction. Figure 4.12(a) shows the whole
field range of MR at different temperatures from 2 to 10 K, and the magnified
small field region is shown in Figure 4.12(b). An obvious negative MR is observed
at low temperatures and low magnetic fields, which becomes less pronounced as
temperature increases. At T > 3 K resistance increases monotonously with the
increasing magnetic field. The positive MR with a dip at B = 0 T is clearly related to
superconductivity, as it disappears above the superconducting transition
temperature at T = 5 K. When the temperature further increases above Tc, the MR
Chapter 4
94
shows a normal parabolic B-field dependence due to the Lorentz deflection of
carriers.
The behavior of the negative MR looks similar to that of weak localization, but it
changes to a positive MR above 3 K, and the whole low field anomalous behavior
disappears at temperatures above 5 K. It is therefore unlikely to be explained by
the weak localization. In Figure 4.13, the different MR features at low fields are
schematically mapped to the RT curve. It clearly shows that the unusual MR
behavior is correlated with the superconducting transition. Above the
superconducting transition, the MR shows a normal parabolic shape. Positive MR
is observed in the middle of the superconducting transition. Negative MR appears
when a more robust superconducting state with zero resistance is developed below
T = 3.5 K.
In previous reports, negative MR has been observed in granular or amorphous
thin superconducting films and wires [29–31]. In these systems, Cooper pairs are
localized in individual islands, and electrical transport is dominated by the
tunneling between coupled superconducting grains. The tunneling effect is
weakened due to the formation of a superconducting gap in the density of states.
With the application of perpendicular magnetic fields, the superconducting gap is
effectively reduced. As a result, the quasi-particle tunneling is enhanced, which
leads to the increase in conductivity. However, this negative effect contributed by
the tunneling effect normally occurs in relatively high fields, which is contrary to
our case. In addition, although doping inhomogeneity may exist in our system, it is
unlikely to be a granular system, as proved by the positive MR of the exposed MoS2
channel. Hence, this model is not applicable to our system.
Figure 4.13 (a) Schematic illustration of the unusual MR behavior of the h-BN covered MoS2
channel corresponding to the RT curve of the exposed area at VBG = 100 V. Light grey region
represent the negative MR, in light pink region positive MR is observed, and in the light blue
region MR shows a normal parabolic behavior. (b) RT curves of the h-BN covered MoS2 channel
under different perpendicular magnetic fields from 0 to 1.2 T.
Superconductor-normal metal junction in gated MoS2
95
Figure 4.14 (a) Energy band diagram showing the NS interface. The exposed MoS2 channel
becomes highly doped by ionic gating and behaves similarly to a metal, while semiconducting
properties are preserved in the h-BN covered MoS2 channel. The interface is analogous to a
metal-semiconductor junction. A potential barrier forms at the interface with intermediate
strength, judged by the intermediate resistance of the junction. (b) A blow-up around the Fermi
level showing the ideal superconductor density of states (DOS). When the temperature is below
the superconducting transition temperature, a gap opens in the energy states. The quasi-
particle tunneling is suppressed due to the lack of energy states to accommodate the incident
electron. Andreev reflection could occur at this interface, but the probability is lowered by the
finite potential barrier. (c) Plots of transmission and reflection coefficients at the NS interface
for an intermediate barrier strength Z = 0.3. Curves A and B are the Andreev and ordinary
reflection probabilities, respectively. Curve C(D) gives the transmission probability
without(with) branch crossing. With intermediate barrier strength, the probability of Andreev
reflection first increases and then decreases as a function of the electron energy. (d) Calculated
differential conductance as a function of electron energy for barrier strength Z = 0.5 at absolute
zero temperature, showing similar non-monotonic behavior with the energy of the electron. (c)
and (d) are adapted from Ref. 32.
Chapter 4
96
In our device configuration, the current flows through a SN junction. Figure
4.13(b) shows the RT curves of the h-BN covered MoS2 channel under different
magnetic fields. The drop in resistance around Tc in the RT curve and the
resistance increase around Bc in the MR curve are clear signs of Andreev reflection
at the SN interface. As a result, the observed negative and positive MR are likely to
be explained by the BTK model [32] and Andreev reflection [13]. Because of the
large inequivalent doping levels between ionic gated MoS2 channel and h-BN
covered MoS2 channel, a Schottky-like barrier forms at the interface. The strength
of the barrier is not very high, as inferred from the intermediate resistance (~1 kΩ,
which is neither very high nor very low). The Andreev reflection process in our
sample is shown schematically in Figure 4.14 (a) and (b). A potential barrier forms
at the interface due to the large Fermi level difference between the exposed and h-
BN covered MoS2 channels.
Blonder and coworkers [32] developed a theory for superconductor-normal
metal micro-constrictions contacts, which describes the crossover from metallic to
tunnel junction behaviors. In this model, a dimensionless δ-function potential
barrier Z is used to describe the strength of the electrical tunneling at the NS
interface. When Z is equal to zero, there is no tunneling barrier at the interface,
such as a perfect normal metal-superconductor junction, the electrical transport is
dominated by the Andreev reflection process. A very large Z indicates a substantial
classical tunneling barrier, such as in a normal metal-insulator-superconductor
(NIS) junction, where the electrical transport is dominated by electron tunneling.
Figure 4.14(c) shows that the Andreev reflection probability (curve A) is not a
monotonic function of electron energy [32]; the probability increases as the
electron energy increases and reaches a maximum at Δ (where 2Δ is the BCS
superconducting gap), then the probability decreases dramatically with the
increase of electron energy. Similar behavior is found in the differential
conductance as shown in Figure 4.15(d), which shows a peak at Δ.
Our observations may be well explained by this model. In our SN junction, Z is
neither zero nor very large, thus there is a finite probability for the Andreev
reflection, as shown in Figure 4.14 (a) and (b). According to the BCS theory, the
effective superconducting gap shrinks with an increasing magnetic field. As a result,
the probability of Andreev reflection increases and resistance decreases. The
resistance reaches a minimum when Δ is equal to the electron energy. As the
magnetic field further increases, the Cooper pair breaking effect dominates the
transport, and the resistance increases monotonously. Overall, the experiment fits
qualitatively with the theoretical calculations.
Superconductor-normal metal junction in gated MoS2
97
4.6 Conclusion
We fabricated an SN junction at the surface of a few-layer MoS2 flake on the h-
BN substrate, with the middle of the MoS2 flake covered by a top h-BN flake. For
the h-BN covered MoS2 channel, ionic gating had a small effect thus the
semiconducting properties were preserved, providing an ideal device platform for
probing the intrinsic properties of 2D layered materials. High quality and
homogeneity in the h-BN protected MoS2 channel are strongly supported by the
observation of pronounced SdH oscillations. The Landau level degeneracy and
effective mass obtained from the oscillation period confirm that the electronic
transport of few-layer MoS2 is dominated by the Q/Q’ valleys of the conduction
band.
Superconductivity was induced in the exposed MoS2 channel by ionic gating,
while the back gate induced a metallic state in the h-BN covered MoS2 channel,
forming a superconductor-normal metal (SN) junction. Due to the huge difference
in the doping levels between the exposed and h-BN covered MoS2 channel, a
Schottky-like barrier formed at the interface. An unusual magnetoresistance (MR)
was observed in this region at the low magnetic fields and low temperatures. This
unusual MR behavior can be explained by the Andreev reflection with an
intermediate barrier. If the barrier strength is very high, the transport is
dominated by single particle tunneling, like a tradition metal-insulator-
superconductor (MIS) tunneling junction. If the tunneling barrier is zero,
corresponding to a perfectly clean interface, the transport is determined by the
Andreev reflection process. With an intermediate barrier, the probability of
Andreev reflection shows a non-monotonic behavior. The probability is maximized
when the electron energy is equal to Δ (where 2Δ is the BCS superconducting gap
size). Negative MR is observed because the superconducting gap decays with the
increasing magnetic field. As the magnetic field further increases, the Cooper pair
breaking effect dominates the transport hence positive MR is observed.
Chapter 4
98
References
[1] A.M. Marsh, D.A. Williams, H. Ahmed, Multiple Andreev reflection in buried
heterostructure-alloy superconductor devices, Phys. B Condens. Matter. 203 (1994) 307–309.
[2] J. Kutchinsky, R. Taboryski, T. Clausen, C.B. Sørensen, A. Kristensen, P.E. Lindelof, J.
Bindslev Hansen, C. Schelde Jacobsen, J.L. Skov, Decay Lengths for Diffusive Transport
Activated by Andreev Reflections in Al/n-GaAs/Al Superconductor-Semiconductor-
Superconductor Junctions, Phys. Rev. Lett. 78 (1997) 931–934.
[3] D. Gunnarsson, J.S. Richardson-Bullock, M.J. Prest, H.Q. Nguyen, A.V. Timofeev, V.A. Shah,
T.E. Whall, E.H.C. Parker, D.R. Leadley, M. Myronov, M. Prunnila, Interfacial Engineering
of Semiconductor-Superconductor Junctions for High Performance Micro-Coolers, Sci. Rep. 5
(2015) 17398.
[4] N. Kim, H.-J. Lee, H.-S. Chang, J.-J. Kim, J.-O. Lee, J.W. Park, K.-H. Yoo, Reentrance of
conductance in mesoscopic normal-metal/superconductor junctions, Phys. B Condens. Matter.
284–288, Part 2 (2000) 1866–1867.
[5] G.J. Strijkers, Y. Ji, F.Y. Yang, C.L. Chien, J.M. Byers, Andreev reflections at
metal/superconductor point contacts: Measurement and analysis, Phys. Rev. B. 63 (2001)
104510.
[6] M. Eschrig, J. Kopu, J.C. Cuevas, G. Schön, Theory of Half-Metal/Superconductor
Heterostructures, Phys. Rev. Lett. 90 (2003) 137003.
[7] J. Cayssol, Crossed Andreev Reflection in a Graphene Bipolar Transistor, Phys. Rev. Lett. 100
(2008) 147001.
[8] X. Du, I. Skachko, E.Y. Andrei, Josephson current and multiple Andreev reflections in
graphene SNS junctions, Phys. Rev. B. 77 (2008) 184507.
[9] C. Ojeda-Aristizabal, M. Ferrier, S. Guéron, H. Bouchiat, Tuning the proximity effect in a
superconductor-graphene-superconductor junction, Phys. Rev. B. 79 (2009) 165436.
[10] R.M. Lutchyn, J.D. Sau, S. Das Sarma, Majorana Fermions and a Topological Phase
Transition in Semiconductor-Superconductor Heterostructures, Phys. Rev. Lett. 105 (2010)
77001.
[11] L.P. Rokhinson, X. Liu, J.K. Furdyna, The fractional a.c. Josephson effect in a semiconductor-
superconductor nanowire as a signature of Majorana particles, Nat. Phys. 8 (2012) 795–799.
[12] T. Van Duzer, S. Kumar, Semiconductor-superconductor hybrid electronics, Cryogenics. 30
(1990) 1014–1023.
[13] A.F. Andreev, Thermal Conductivity of the Intermediate State of Superconductors, Sov. Phys.
JETP. 19 (1964) 1228-1231
[14] C.W.J. Beenakker, Quantum transport in semiconductor-superconductor microjunctions, Phys.
Rev. B. 46 (1992) 12841–12844.
[15] C. Lee, H. Yan, L.E. Brus, T.F. Heinz, J. Hone, S. Ryu, Anomalous Lattice Vibrations of
Single- and Few-Layer MoS2, ACS Nano. 4 (2010) 2695–2700.
[16] J. Chang, L.F. Register, S.K. Banerjee, Ballistic performance comparison of monolayer
transition metal dichalcogenide MX2 (M = Mo, W; X = S, Se, Te) metal-oxide-semiconductor
field effect transistors, J. Appl. Phys. 115 (2014) 84506.
[17] J.T. Ye, Y.J. Zhang, R. Akashi, M.S. Bahramy, R. Arita, Y. Iwasa, Superconducting Dome in
a Gate-Tuned Band Insulator, Science. 338 (2012) 1193–1196.
Superconductor-normal metal junction in gated MoS2
99
[18] Y. Zhang, J. Ye, Y. Matsuhashi, Y. Iwasa, Ambipolar MoS2 Thin Flake Transistors, Nano Lett.
12 (2012) 1136–1140.
[19] B.W.H. Baugher, H.O.H. Churchill, Y. Yang, P. Jarillo-Herrero, Intrinsic Electronic Transport
Properties of High-Quality Monolayer and Bilayer MoS2, Nano Lett. 13 (2013) 4212–4216.
[20] B. Radisavljevic, A. Kis, Mobility engineering and a metal–insulator transition in monolayer
MoS2, Nat. Mater. 12 (2013) 815–820.
[21] X. Cui, G.-H. Lee, Y.D. Kim, G. Arefe, P.Y. Huang, C.-H. Lee, D.A. Chenet, X. Zhang, L.
Wang, F. Ye, F. Pizzocchero, B.S. Jessen, K. Watanabe, T. Taniguchi, D.A. Muller, T. Low, P.
Kim, J. Hone, Multi-terminal transport measurements of MoS2 using a van der Waals
heterostructure device platform, Nat. Nanotechnol. 10 (2015) 534–540.
[22] K. Kaasbjerg, K.S. Thygesen, K.W. Jacobsen, Phonon-limited mobility in n-type single-layer
MoS2 from first principles, Phys. Rev. B. 85 (2012) 115317.
[23] R. Fivaz, E. Mooser, Mobility of Charge Carriers in Semiconducting Layer Structures, Phys.
Rev. 163 (1967) 743–755.
[24] Z. Wu, S. Xu, H. Lu, G.-B. Liu, A. Khamoshi, T. Han, Y. Wu, J. Lin, G. Long, Y. He, Y. Cai,
F. Zhang, N. Wang, Observation of Valley Zeeman and Quantum Hall Effects at Q Valley of
Few-Layer Transition Metal Disulfides, ArXiv151100077 Cond-Mat. (2015).
[25] J.-H. Chen, C. Jang, S. Xiao, M. Ishigami, M.S. Fuhrer, Intrinsic and extrinsic performance
limits of graphene devices on SiO2, Nat. Nanotechnol. 3 (2008) 206–209.
[26] T. Ando, A.B. Fowler, F. Stern, Electronic properties of two-dimensional systems, Rev. Mod.
Phys. 54 (1982) 437–672.
[27] S. Das Sarma, S. Adam, E.H. Hwang, E. Rossi, Electronic transport in two-dimensional
graphene, Rev. Mod. Phys. 83 (2011) 407–470.
[28] A.V. Kretinin, Y. Cao, J.S. Tu, G.L. Yu, R. Jalil, K.S. Novoselov, S.J. Haigh, A. Gholinia, A.
Mishchenko, M. Lozada, T. Georgiou, C.R. Woods, F. Withers, P. Blake, G. Eda, A. Wirsig,
C. Hucho, K. Watanabe, T. Taniguchi, A.K. Geim, R.V. Gorbachev, Electronic Properties of
Graphene Encapsulated with Different Two-Dimensional Atomic Crystals, Nano Lett. 14
(2014) 3270–3276.
[29] A. Gerber, A. Milner, G. Deutscher, M. Karpovsky, A. Gladkikh, Insulator-Superconductor
Transition in 3D Granular Al-Ge Films, Phys. Rev. Lett. 78 (1997) 4277–4280.
[30] N. Hadacek, M. Sanquer, J.-C. Villégier, Double reentrant superconductor-insulator transition
in thin TiN films, Phys. Rev. B. 69 (2004) 24505.
[31] R.P. Barber, S.-Y. Hsu, J.M. Valles, R.C. Dynes, R.E. Glover, Negative magnetoresistance,
negative electroresistance, and metallic behavior on the insulating side of the two-dimensional
superconductor-insulator transition in granular Pb films, Phys. Rev. B. 73 (2006) 134516.
[32] G.E. Blonder, M. Tinkham, T.M. Klapwijk, Transition from metallic to tunneling regimes in
superconducting microconstrictions: Excess current, charge imbalance, and supercurrent
conversion, Phys. Rev. B. 25 (1982) 4515–4532.
Chapter 4
100
101
Chapter 5. Transport properties of
multi-layer germanane
Abstract
The studies of MoS2 presented in previous chapters revealed the typical
electronic properties of semiconducting two-dimensional (2D) transition metal
dichalcogenides (TMDs). Compared to widely studied graphene and TMDs, other
new 2D materials are much less explored. Among them, 2D hexagonal structures
composed of silicon or germanium are very important candidates for future
transistors because they are highly compatible with current silicon- or germanium-
based electronic devices. In this chapter, we study hydrogen-terminated multilayer
germanene, known as germanane (GeH). GeH is a layered material that has a
lattice structure similar to graphite. GeH is semiconducting with a direct band gap
about 1.6 eV. Since the Fermi level lies inside the band gap, pristine GeH flakes
show zero conductivity. After annealing in Ar atmosphere, the conductivity is
greatly improved, and the enhancement is found to be positively related to the
annealing temperature. After low temperature annealing, GeH shows p-type
transistor characteristics with an on/off ratio higher than 30. A highly conductive
state can be obtained by high temperature annealing. Raman spectra suggest that
the layered structure is preserved after annealing, the huge conductivity
enhancement is due to the dehydrogenation. After annealing, hydrogen atoms
escape from the lattice hence germanane becomes multilayer germanene. In
addition, strong spin-orbit interaction is confirmed by the observation of weak
antilocalization in the magnetoresistance measurement. The experimental
observations of the insulator-metal transition and spin-orbit coupling in GeH pave
the way to a better understanding and potential application of this novel 2D
material.
In preparation
“Metal-insulator transition and strong spin-orbit interaction in annealed germanane”
Q.H. Chen, J.M. Lu, L. Liang and J.T. Ye
Chapter 5
102
5.1 Introduction
Currently, silicon-based transistors are the most important building blocks in
integrated circuits, and transistors of even smaller sizes are in demand in order to
make portable and powerful electronic devices. The scaling of silicon-based
transistors will eventually fail due to direct source-to-drain tunneling and loss of
gate electrostatic control on the channels [1,2]. To overcome these difficulties, the
exploration of new channel materials has attracted tremendous interest. Since the
discovery of graphene [3], two-dimensional (2D) materials are regarded as one of
the most promising candidates for replacing silicon. Although graphene shows
exceptional physical properties, the lack of native band gap limits its application in
electronic devices. Many other 2D materials, represented by transition metal
dichalcogenides (TMDs), are semiconducting hence are suitable for making high-
performance transistors. TMDs are a large family of layered compounds that have
crystal structures similar to graphite. Bulk TMDs are stacked by monolayers with
weak interlayer interaction, and can be easily exfoliated into uniformly thin flakes
using the simple scotch tape method. As mentioned in previous chapters, MoS2 is
one of the most widely studied 2D materials besides graphene. MoS2 has an
indirect band gap of 1.2 eV in bulk and a direct band gap of 1.9 eV in monolayer,
showing great potential for applications in high-speed electronics, optoelectronics
and sensors. Very recently a MoS2 transistor with a 1-nm gate length has been
reported, in which a single-walled carbon nanotube is used as the gate
electrodes [4].
While research on TMDs is still being carried out, the exploration of new 2D
structures is desirable to enrich the present study. 2D monolayer materials
composed of other group IV elements such as silicene [5,6] and germanene [7] are
very attractive because they are highly compatible with current silicon- and
germanium-based transistors, so their fabrication and integration will benefit from
the well-developed silicon technology. The structures of silicene and germanene
are similar to graphene, but the hexagonal lattice is periodically buckled instead of
being flat. Regarding the electronic properties, silicene has been reported to be
almost equivalent to graphene from both theoretical calculations [8–10] and
experimental observations [5,6,11,12]. That is, the band structure exhibits a
crossing at the K/K’ points of the Brillouin zone with a linear energy dispersion
near the crossing, and the charge carriers behave like massless Dirac fermions. A
single-layer silicene field effect transistor has been reported [13], showing
ambipolar transistor operation characteristics but the on/off ratio is low.
Compared with graphene and silicene, germanene possesses slightly different band
structure, in which higher conductivity is predicted by ab initio calculations.
However, there are two disadvantages to silicene and germanene: (1) there is no
Transport properties of multi-layer germanane
103
intrinsic gap in the band structure and (2) they are not stable when exposed to
ambient conditions.
To improve the stability, hydrogen termination on both sides of the 2D film is
introduced. Similar to the formation of hydrogen-terminated graphane [14],
germanane (GeH) is created by adding covalently bonded hydrogen atoms on both
sides of germanene. The pz orbitals of the Ge atoms are occupied by the covalent
bonding to the H atoms. A large band gap opens at the K/K’ points of the Brillouin
zone, and the electron transport from the contribution of the pz orbitals is
significantly suppressed. Instead, a conduction band derived from the s-orbitals
and a valence band derived from the px and py orbitals near the Γ point determine
the transport properties [15]. The electron mobility is predicted to be as high as
18000 cm2V-1s-1, which is five times higher than that of bulk germanium and ten
times higher than that of silicon. Theoretical calculations using the LDA method
predict that the energy gap of GeH is about 1.5 eV and that it is a direct band
gap [16,17]. Considering that the LDA method normally underestimates the gap
size, it is expected that the correct value is close to 3 eV [17–20]. The large band
gap, high mobility and low dimensionality make GeH very promising for making
short-channel field effect transistors. Furthermore, since the σ bond (composed of
px and py orbitals) that dominates the electron transport has a stronger spin-orbit
coupling (SOC) than the original π bond (composed of pz orbitals), a non-trivial
spin-orbit gap is expected in the band structure which is in the order of ~ 0.2 eV.
Combined with the large direct band gap nature, the traditional spin-selective
optical process is also expected to be present in germanane, providing an attractive
potential for optoelectronic applications in the blue/violet spectral range.
Despite the interesting properties germanane has shown, the number of
publications regarding its electrical transport properties is very limited [21,22]. As
an extension of the quantum transport study in 2D materials, as well as to search
for new device applications, we fabricated field effect transistors based on thin
germanane flakes and measured the transport properties at low temperatures and
under magnetic fields. Pristine germanane shows zero conductivity, consistent
with the large band gap predicted by theoretical calculations. Upon annealing in an
Argon atmosphere, insulator to metal transition is observed. Previous reports have
argued that the structure gradually becomes amorphous and hydrogen atoms
escape from the lattice during heating [23]. In our measurements, however, no
significant change of the Raman signal is detected after annealing, which suggests
that the structure is preserved during the heating process. Hence the metallic
behavior is very likely to be associated with the dehydrogenation, leaving behind
the multilayer germanene crystal structure, which is highly conductive according to
previous reports [9]. At the base temperature T = 2 K, clear evidence for strong
spin-orbit coupling is observed, manifesting as weak antilocalizaton. These
Chapter 5
104
preliminary explorations open new opportunities for quantum and spin transport
studies of GeH for both basic research and device applications.
5.2. Structure and characterization
The bulk GeH crystals were synthesized by our collaborators, following the
recently developed method [23]. In this method, large-scale multilayer germanane
is synthesized by topochemical deintercalation from the CaGe2 precursor, where
CaGe2 was prepared by the reaction of high purity germanium (Ge) and calcium
(Ca) in vacuum at high temperatures. Following an ion-exchange process between
CaGe2 and HCl, Ca atoms in CaGe2 are substituted by H atoms. This step can be
described by the following chemical equation:
CaGe2 + HCl → GeH + CaCl2
By this method, millimeter-sized crystallites and gram-scale multilayer GeH
flakes are obtained. It has been proven that the synthesized germanane is resistive
to oxygen and thermally stable at ambient conditions.
The lattice structure of GeH is shown in Figure 5.1. From the top view (Figure
5.1(a)), germanium atoms form a honeycomb lattice similar to graphene. Instead of
being flat, however, they have a buckled structure, as seen in the side view (Figure
5.1(b)). Each germanium atom is bonded with three other germanium atoms in the
ab-plane and one hydrogen atom in the c-direction for sp3 hybridization. Lattice
constants are a = 3.880 Å and c = 11.04 Å, with an interlayer distance of 5.5 Å.
With hydrogen termination, GeH becomes stable in air.
Figure 5.1 Top (a) and side (b) view of monolayer germanane. Red spheres represent Ge atoms
and black spheres represent H atoms.
Prior to transport measurement, we performed optical microscopy, atomic
force microscopy (AFM) and Raman spectroscopy to characterize the GeH crystal.
Figure 5.2(a) shows the as-grown GeH crystals, which were synthesized by
Transport properties of multi-layer germanane
105
topochemical deintercalation from the CaGe2 precursor. The dimensions of the
flakes are typically ~1 mm in diameter and ~200 μm in thickness.
Figure 5.2 (a) Synthesized bulk GeH crystals, black flakes with typical dimensions of ~1 mm.
The diameter of the glass bottle is ~1 cm. (b) Optical image of a thin GeH flake exfoliated on a
silicon wafer with 300 nm SiO2. Scale bar: 2 μm.
Similar to other layered materials, bulk GeH is held together by a weak van der
Waals force, which means it can be mechanically exfoliated into thin flakes. For the
exfoliation process, GeH crystals were first thinned down by peeling with scotch
tape. After repeating this step a few times, the sticky side of the scotch tape with
GeH flakes was pressed onto a silicon wafer with 300 nm SiO2, and then the tape
was slowly peeled off. Thin GeH flakes were randomly deposited on the silicon
wafer. Optical microscope and AFM were used to select uniform and thin flakes.
Figure 5.2(b) shows a typical optical image of the exfoliated thin GeH flakes. In
terms of convenience for processing, we chose flakes with thicknesses in the range
of 20-50 nm for device fabrication and electrical measurements.
Raman spectroscopy is a very useful tool to study the structure of 2D
materials [24,25]. We have characterized the thin GeH flakes using micro-Raman
spectroscopy, Figure 5.3 shows the typical spectra. The main peak appears at ~300
cm-1 (inset of Figure 5.3), which corresponds to the E2 mode of Ge-Ge vibration in
the ab-plane direction [23]. It is slightly blue-shifted but very close to the Ge-Ge
stretching mode of crystalline germanium at 297 cm-1, which suggests that the light
H atoms have a negligible influence on the vibration mode of Ge-Ge. It has been
reported that the peak position of the Ge-Ge stretching mode is greatly influenced
by stress, with the application of high pressure, this vibrational peak shifts to a
higher frequency [26]. The vibration mode of silicon at 520 cm-1 is also observed,
due to the scattering of laser light. It should be noted that in the range of 300 to
800 cm-1 there is a broad band background. A similar feature has also been found
Chapter 5
106
in the previous report [23], but it is weaker compared to our observation. The
reason is not clear at this moment, further study is needed to make a concrete
conclusion.
Figure 5.3 Experimental Raman spectra of GeH measured in ambient conditions. The inset
shows the magnification of the featured peak around 300 cm-1, corresponding to the E2
vibration mode of the Ge-Ge.
5.3 Electrical transport
After the characterization of GeH, thin flakes with thicknesses of 20-50 nm
were selected to fabricate the FET device. We used the traditional electron beam
lithography method to pattern Hall bar geometry, and electrodes composed of 5
nm Ti and 65 nm Au were deposited by a thermal evaporator. To exclude any
thermal decomposition of GeH, high temperature treatments were avoided during
the fabrication process. Figure 5.4(a) shows the optical image of the measured
device. Transfer curves were obtained by applying a constant AC voltage across the
source-drain electrode and measuring the current that flows through the channel,
using a standard lock-in amplifier (Stanford Research SR830). Two other lock-in
amplifiers were used to measure the voltage drop across the channel. Gate voltages
were set by a DC source meter (Keithley 2450).
As expected, the as-prepared device shows no electrical conductivity (below the
detection limit of our setup). When the back gate voltage is applied through 300
nm SiO2, no significant change in the source-drain current is observed until 50V,
as shown in Figure 5.4(b). Similar behavior is observed with ionic gating, even
though it provides a much stronger gating effect, as shown in Figure 5.4(c). This
result is reasonable since pristine GeH has a large band gap, it is a good insulator.
Transport properties of multi-layer germanane
107
It should be noted that although the device is in an off state, there is a finite
current contributed by the capacitance in the circuit due to the AC measurement.
This is evidenced by the 90-degree phase of the lock-in amplifier during the
measurement. The difference between Figure 5.4 (b) and (c) is due to the different
applied source-drain voltages.
Figure 5.4 (a) Optical image of the as-prepared GeH FET device with Hall bar geometry. Scale
bar: 2 μm. (b) Transfer curve by SiO2 back gate with VDS = 0.3 V. (c) Transfer curve by ionic
gating (DEME-TFSI) with VDS = 0.1 V. Temperature is maintained at T = 220 K.
Annealing was tested to improve the electrical conductivity. The GeH device
was sealed in a quartz tube with flowing argon gas. The temperature was ramped
up to the target within 1-2 hours and kept for 12 hours before naturally cooling
down. The Ar flow rate was maintained at 100 sccm for the whole process.
Different heating temperatures were tested, and the results are categorized in three
representative regimes.
5.3.1 Low annealing temperature, ~ 170 °C
Figure 5.5 shows the electrical characteristics of the device after annealing at
170 °C. Before annealing, the device behaved the same as that shown in Figure 5.4,
namely, totally insulating that cannot be gated by either solid or ionic gating. After
annealing, the sample became conductive. Furthermore, the conductivity of the
channel can be tuned by the back gate like a normal transistor, with the ratio
between high and low conductivity higher than 5. The FET operation shows a p-
type semiconductor behavior, which means accumulating holes by the negative
gate voltage leads to an increase in conductivity. Correspondingly, the positive gate
reduces the conductivity. The temperature dependence of the conductivity is
shown in Figure 5.5(b). The sample shows insulating behavior for all back gate
voltages, where conductivity decreases as the temperature goes down.
Chapter 5
108
Figure 5.5 (a) Conductance of the GeH channel as a function of back gate voltage at T = 300 K
and VDS = 0.1 V. A p-type transistor operation is observed and the ratio between high and low
conductance is 5. The inset shows the optical image of the measured device. (b) Temperature
dependence of the conductance for different back gate voltages.
Since ionic gating has a much stronger gating effect, we also performed a test
for it on this sample. The ionic gating was conducted at T = 220 K, consistent with
previous measurements. As shown in Figure 5.6, a consistent p-type transistor
behavior is observed with a current on/off ratio of ~20. Overall, the conductivity of
the sample after annealing at 170 °C is still very low.
Figure 5.6 Source-drain current as a function of the ionic liquid gate voltage at T = 220 K, with
VDS = 0.3 V. A p-type transistor operation is observed similar to back gating. The on/off ratio
between high and low current is about 20.
Transport properties of multi-layer germanane
109
5.3.2 Intermediate annealing temperature, ~190 °C
Figure 5.7 (a) Transfer curve by ionic gating at T = 220 K with VDS = 4 mV. The black curve
corresponds to the source-drain current (left axis), while the red curve corresponds to the four-
probe channel resistance (right axis). Inset: optical image of the measured device. (b)
Conductance as a function of the back gate voltage for temperatures at 170, 150 and 80 K, from
top to bottom. On/off ratio between high and low conductance is higher than 30. Ionic gate
voltage was fixed at VLG = 6 V during cooling down. From the linear fitting, hole mobility is
determined as 24 cm2V-1s-1. (c) and (d) are the same set of data showing the temperature
dependence of the channel resistance and conductance, respectively, for different back gate
voltages.
With the annealing temperature increased to 190 °C, a significant increase in
the conductivity can be seen from the transfer curve by ionic gating, as displayed in
Figure 5.7(a). Both the source-drain current and channel resistance change by a
factor of five with liquid gate voltage varying from -6 to 6 V. The behavior of the
current and resistance are inversely related with each other, suggesting that the
contact resistance is very small hence the change in the source-drain current is
dominated by the change in the channel resistance. The EDLT operation shows p-
Chapter 5
110
type semiconductor behavior, in line with the previous device annealed at a lower
temperature.
The ionic gate voltage was fixed at VLG = 6 V, and the sample was cooled to
below 180 K, where the liquid is frozen and ions cannot move. Figure 5.7(b) shows
the transfer curves by the back gate at different temperatures. Consistent with the
ionic gating, the conduction is also dominated by holes. Therefore, adding
electrons to the system leads to a decrease in conductivity. The mobility can be
roughly estimated from the geometric capacitance with 𝜇FET =1
𝐶g
d𝜎
d𝑉g , where 𝜎 is
the conductivity and Cg is the gate capacitance per unit area. Here for 300 nm SiO2,
𝐶g = 11 nF/cm2. By the linear fitting of the gate dependence of conductivity, the
extracted FET mobility is 24 cm2V-1s-1 at T = 170 K. The temperature dependence
of the conductivity of this sample state is shown in Figure 5.7(d). The sample still
behaves like an insulator because dG/dT > 0, but the overall conductivity is higher
than the sample annealed at 170 °C, showing a positive correlation between
conductivity and heating temperature.
5.3.3 High annealing temperature ~ 210 °C
Figure 5.8 (a) Temperature dependence of the resistance for two different devices after
annealing at ~210 °C. (b) Source-drain current (black curve, left axis) and channel resistance
(red curve, right axis) as a function of back gate voltage, at T = 10 K and VDS = 1 mV,
corresponding to the lower curve shown in panel (a). The inset shows the optical image of the
measured device. (c) Source-drain current (black curve, left axis) and channel resistance (red
curve, right axis) as a function of ionic gate voltage, at T = 220 K and VDS = 4 mV.
Given the previous positive correlation between conductivity and heating
temperature, the annealing temperature was further raised to ~ 210 °C. The results
are shown in Figure 5.8. As expected, the conductivity increases in a huge amount.
Contrary to the previous insulating state, a metallic behavior is observed (Figure
5.8(a)). However, the conductivity can no longer be easily tuned by the field effect.
Transport properties of multi-layer germanane
111
The back gate has no influence on the source-drain current or channel resistance
(Figure 5.8(b)). While stronger ionic gating can change the conductivity a little, the
gating effect is also very weak, as shown in Figure 5.8(c). The annealed GeH
behaves similarly to a piece of metal.
5.3.4 Discussion
As a summary of the previous results, we plot the temperature dependence of
the resistance for samples annealed at different temperatures in Figure 5.9. Clear
insulator to metal transition is observed, providing strong evidence for the positive
correlation between the conductivity and annealing temperature.
Figure 5.9 Temperature dependence of the channel resistance after annealing at different
temperature ranges. The top two dashed circles represent two devices annealed at the
respective temperatures. Different curves in each circle are measured with different back gate
voltages. The curves highlighted by the bottom circle are data from three different devices after
annealing at 210 °C. GeH remains insulating with annealing temperatures of 170 and 190 °C,
whereas a metallic state is observed when the annealing temperature is raised to 210 °C.
It is surprising that highly conductive states can be obtained in GeH through
annealing in an inert gas atmosphere. Since GeH possesses a large band gap, it is
unlikely that GeH itself can be tuned to a metallic state. Previous studies of its
thermal properties [23,27] showed that GeH starts to amorphize above 75 °C,
evidenced by a significant change in the absorption spectra. From the XRD pattern,
however, no obvious change occurs until 150 °C; above 150 °C, the peak for the 002
Chapter 5
112
plane disappears, corresponding to the completion of the amorphization. From the
Raman spectra, the peak at 300 cm-1 corresponding to the Ge-Ge vibrational mode
also disappears. In the thermogravimetric analysis (TGA), a mass loss of 1.1% is
observed between 200-250 °C, which is attributed to the loss of hydrogen.
In our experiment, however, a different behavior is observed. In the Raman
spectra before and after the annealing as shown in Figure 5.10, the Ge-Ge
vibrational mode at 300 cm-1 always exists. This is reasonable because the Ge-Ge
vibrational mode should exist in any allotropes of Ge, such as GeH, amorphous or
crystalline germanium. For pure or hydrogenated amorphous germanium [28–30],
the vibrational mode of Ge-Ge is around 270 cm-1 and it is very broad, with a
FWHM of about 50 cm-1. In our spectra after annealing, the peak at 300 cm-1
becomes slightly blue-shifted and narrower compared to that before annealing, the
FWHM is about 10 cm-1. It is therefore unlikely that the sample becomes
amorphous due to annealing. Rather, the narrower peak may indicate a better
crystal structure after annealing. Furthermore, the broad background between 300
and 800 cm-1 disappears after heating, which is not fully understood at this
moment; further study is needed to understand this behavior. Therefore, the
layered structure of GeH is expected to be preserved during the annealing. Since
the crystal structure did not change, the significant increase in conductivity
observed is very likely to be associated with the dehydrogenation. Upon heating,
hydrogen atoms escape from the lattice, leaving behind the germanene multi-layer
structure. The germanene multi-layer has a very similar structure to bulk graphite,
the conductivity is predicted to be very high.
Figure 5.10 Comparison of the Raman spectra before (black curve) and after (red curve)
annealing. Inset: blowup of the region for the Ge-Ge vibration mode. While the main features
are preserved in the spectra, the broad background between 300 and 800 cm-1 disappears after
annealing.
Transport properties of multi-layer germanane
113
From the electrical transport measurement, we are able to calculate the
resistivity of the GeH after annealing. The most conductive sample shows a
resistance of ~ 3 Ω at room temperature, with channel dimensions of 1.2 μm
(length) × 2.5 μm (width) × 40 nm (thickness). The calculated resistivity is 4×10-7
Ω·m. Table 5.1 lists the room-temperature resistivity for different materials that
are related to our analysis. For bulk crystalline and amorphous germanium, the
resistivity at room temperature is relatively high, and the temperature
dependences are dominated by the semiconducting gap and variable range
hopping, respectively. Resistivity increases significantly with the decrease in
temperature [31–34]. Combined with Raman data, these two different structures
composed of Ge are clearly not viable candidates. Quasi-metallic behavior can be
obtained in highly doped germanium [35], but the minimum reported resistivity is
~5×10-5 Ω·m corresponding to a very high doping level, which is more than two
orders of magnitude higher than the annealed GeH. Combined with the Raman
spectra, it is less likely that the layered structure changes to the bulk germanium
structure at this heating temperature. Multi-layered germanene remains the most
reasonable candidate. This conclusion can be further supported by comparing it
with graphite: annealed GeH is even more conductive than graphite, and the
resistivity is higher than monolayer graphene. Multilayer germanene can clearly
reach such high conductivity, which is in line with the theoretical calculations.
Table 5.1 Resistivity for different materials at room temperature. Any allotropes of Ge can be
excluded due to the large resistivity compared with annealed GeH.
5.4 Strong spin-orbit interaction in annealed GeH
Theory predicts that both germanene and germanane have strong spin-orbit
interaction [15,36,37], which is attractive for potential applications in spin-
selective optoelectronics. At low temperatures, we measure the magnetoresistance
(MR) with the perpendicular magnetic fields. Clear weak antilocalization (WAL) is
observed, manifesting as a characteristic sharp MR dip at B = 0 T, as shown in
Chapter 5
114
Figure 5.11(a). WAL provides strong evidence for strong spin-orbit interaction (SOI)
in our device.
According to the 2D localization theory [34–36], by assuming that the elastic
and spin-orbit scattering time are much shorter than inelastic scattering time, the
magnetoconductivity (MC) can be described by the following equation:
∆𝜎2D = 𝜎2D(𝐵) − 𝜎2D(0) = −𝛼𝑒2
2𝜋2ℏ[ln
ℏ
4𝐵𝑒𝑙Φ2 − ψ (
1
2+
ℏ
4𝐵𝑒𝑙Φ2 )]
where ℏ is the reduced Planck constant, e is the charge of electron, ψ(𝑥) is the
digamma function and 𝑙Φ is the phase coherence length. 𝛼 is a fitting parameter
equal to 1, 0 and -1/2 for the orthogonal, unitary and symplectic cases,
respectively [38,41]. The MC at T = 2.5 K is plotted in Figure 5.11(b), and the solid
curve is the fitting with the above equation. We can see that the fitting matches
well with the experimental observations, and the fitting parameters are 𝛼 = −0.53
and 𝑙Φ = 79 nm. As the temperature increases, the WAL gradually disappears due
to the enhanced electro-phonon scattering, which leads to the decrease in phase
coherence.
Figure 5.11 (a) Normalized magnetoresistance of an annealed GeH transistor at different
temperatures, showing clear weak antilocalization (WAL) characteristics at low field region. (b)
The black empty squares are the magnetoconductance defined as ΔG = G(B) - G(0) for T = 2.5 K.
The red solid line is the best fit to ΔG using the WAL equation (1), with α = -0.53 and phase
coherence length lΦ = 79 nm.
As expected, the WAL can be suppressed by applying a large current, as shown
in Figure 5.12(a). With the current increase from 5 μA to 80 μA, the WAL dip at B
= 0 T gradually fades out. The angle (θ) dependence of magnetoresistance was
measured, as shown in Figure 5.12(b). θ is defined as the angle between the
Transport properties of multi-layer germanane
115
magnetic field B and the ab-plane of annealed GeH, e.g. θ = 90° means the
magnetic field B is perpendicular to the ab-plane, θ = 0° means the magnetic field
B is parallel to the ab-plane. However, no obvious change is observed for different
angles. This is likely to be related to the relatively large thickness of the GeH flakes.
The coupling between layers is also quite strong after annealing, and therefore the
behavior shows 3D behavior. Stronger angle dependence may be observed by
reducing the thickness of GeH sample.
Figure 5.12 (a) Normalized MR measured with different currents, at T = 2 K. The curves are
vertically shifted for comparison. (b) WAL measurement at different angles.
5.5 Conclusion
We measured the transport properties of multilayer hydrogen-terminated
germanane, which has a layered structure similar to hydrogen-terminated
graphene. Pristine germanane without further treatment is insulating due to the
large gap in the band structure. After annealing in an Ar atmosphere, GeH
becomes conductive. For low temperature annealing, a field effect transistor
operation is realized with the highest on/off ratio of 30. The conduction is
dominated by holes, corresponding to a p-type semiconductor. For high
temperature annealing, the conductivity becomes very high and the temperature
dependence of the resistance shows metallic behavior. The resistivity is 4×10-7 Ω·m
at room temperature, which is lower than any allotropes of germanium, e.g.
crystalline germanium, amorphous germanium and highly doped germanium.
Furthermore, the conductivity is even higher than bulk graphite and only slightly
lower than monolayer graphene. Hence, what we obtained after annealing is very
likely to be multilayer germanene, which has a structure similar to graphite yet
Chapter 5
116
much higher conductivity. The interlayer coupling in multilayer germanene is
much stronger than graphite. In Raman spectra, the Ge-Ge vibration mode at 300
cm-1 becomes slightly sharper compared to that before annealing. This result also
provides strong evidence that the layered structure is preserved instead of
becoming amorphous during heating, since both pure and hydrogenated
amorphous germanium have remarkably red-shifted peaks and FWHMs in the
order of 50 cm-1, which is much larger than our observation of 10 cm-1.
Clear weak antilocalization is observed at low temperatures with a phase
coherence length of 80 nm, which proves the presence of strong spin-orbit
interaction in annealed GeH. As far as we know, this is the first experimental
observation of strong spin-orbit coupling in GeH or germanene, which indicates
the possibility of developing spintronic devices and optoelectronic devices based
on annealed GeH. The weak angle dependence of the observed WAL is possibly
associated with the relatively large sample thickness and strong interlayer coupling
in annealed GeH.
Transport properties of multi-layer germanane
117
References
[1] H. Kawaura, T. Sakamoto, T. Baba, Observation of source-to-drain direct tunneling current in
8 nm gate electrically variable shallow junction metal–oxide–semiconductor field-effect
transistors, Appl. Phys. Lett. 76 (2000) 3810–3812.
[2] M. Lundstrom, Moore’s Law Forever?, Science. 299 (2003) 210–211.
[3] K.S. Novoselov, D. Jiang, F. Schedin, T.J. Booth, V.V. Khotkevich, S.V. Morozov, A.K.
Geim, Two-dimensional atomic crystals, Proc. Natl. Acad. Sci. U. S. A. 102 (2005) 10451–
10453.
[4] S.B. Desai, S.R. Madhvapathy, A.B. Sachid, J.P. Llinas, Q. Wang, G.H. Ahn, G. Pitner, M.J.
Kim, J. Bokor, C. Hu, H.-S.P. Wong, A. Javey, MoS2 transistors with 1-nanometer gate
lengths, Science. 354 (2016) 99–102.
[5] B. Aufray, A. Kara, S. Vizzini, H. Oughaddou, C. Léandri, B. Ealet, G.L. Lay, Graphene-like
silicon nanoribbons on Ag(110): A possible formation of silicene, Appl. Phys. Lett. 96 (2010)
183102.
[6] B. Lalmi, H. Oughaddou, H. Enriquez, A. Kara, S. Vizzini, B. Ealet, B. Aufray, Epitaxial
growth of a silicene sheet, Appl. Phys. Lett. 97 (2010) 223109.
[7] M.E. Dávila, L. Xian, S. Cahangirov, A. Rubio, G.L. Lay, Germanene: a novel two-
dimensional germanium allotrope akin to graphene and silicene, New J. Phys. 16 (2014)
95002.
[8] G.G. Guzmán-Verri, L.C. Lew Yan Voon, Electronic structure of silicon-based nanostructures,
Phys. Rev. B. 76 (2007) 75131.
[9] S. Lebègue, O. Eriksson, Electronic structure of two-dimensional crystals from ab initio
theory, Phys. Rev. B. 79 (2009) 115409.
[10] Z. Ni, Q. Liu, K. Tang, J. Zheng, J. Zhou, R. Qin, Z. Gao, D. Yu, J. Lu, Tunable Bandgap in
Silicene and Germanene, Nano Lett. 12 (2012) 113–118.
[11] L. Chen, C.-C. Liu, B. Feng, X. He, P. Cheng, Z. Ding, S. Meng, Y. Yao, K. Wu, Evidence for
Dirac Fermions in a Honeycomb Lattice Based on Silicon, Phys. Rev. Lett. 109 (2012) 56804.
[12] P. Vogt, P. De Padova, C. Quaresima, J. Avila, E. Frantzeskakis, M.C. Asensio, A. Resta, B.
Ealet, G. Le Lay, Silicene: Compelling Experimental Evidence for Graphenelike Two-
Dimensional Silicon, Phys. Rev. Lett. 108 (2012) 155501.
[13] L. Tao, E. Cinquanta, D. Chiappe, C. Grazianetti, M. Fanciulli, M. Dubey, A. Molle, D.
Akinwande, Silicene field-effect transistors operating at room temperature, Nat. Nanotechnol.
10 (2015) 227–231.
[14] D.C. Elias, R.R. Nair, T.M.G. Mohiuddin, S.V. Morozov, P. Blake, M.P. Halsall, A.C. Ferrari,
D.W. Boukhvalov, M.I. Katsnelson, A.K. Geim, K.S. Novoselov, Control of Graphene’s
Properties by Reversible Hydrogenation: Evidence for Graphane, Science. 323 (2009) 610–
613.
[15] W. Amamou, P.M. Odenthal, E.J. Bushong, D.J. O’Hara, Y.K. Luo, J. van Baren, I. Pinchuk,
Y. Wu, A.S. Ahmed, J. Katoch, M.W. Bockrath, H.W.K. Tom, J.E. Goldberger, R.K.
Kawakami, Large area epitaxial germanane for electronic devices, 2D Mater. 2 (2015) 35012.
[16] X. Luo, E. Zurek, Crystal Structures and Electronic Properties of Single-Layer, Few-Layer,
and Multilayer GeH, J. Phys. Chem. C. 120 (2016) 793–800.
[17] L.C.L.Y. Voon, E. Sandberg, R.S. Aga, A.A. Farajian, Hydrogen compounds of group-IV
nanosheets, Appl. Phys. Lett. 97 (2010) 163114.
[18] M. Houssa, E. Scalise, K. Sankaran, G. Pourtois, V.V. Afanas’ev, A. Stesmans, Electronic
properties of hydrogenated silicene and germanene, Appl. Phys. Lett. 98 (2011) 223107.
[19] J.O. Sofo, A.S. Chaudhari, G.D. Barber, Graphane: A two-dimensional hydrocarbon, Phys.
Rev. B. 75 (2007) 153401.
Chapter 5
118
[20] S. Lebègue, M. Klintenberg, O. Eriksson, M.I. Katsnelson, Accurate electronic band gap of
pure and functionalized graphane from GW calculations, Phys. Rev. B. 79 (2009) 245117.
[21] N.G. Sahoo, R.J. Esteves, V.D. Punetha, D. Pestov, I.U. Arachchige, J.T.M. Jr, Schottky
diodes from 2D germanane, Appl. Phys. Lett. 109 (2016) 23507.
[22] J.R. Young, B. Chitara, N.D. Cultrara, M.Q. Arguilla, S. Jiang, F. Fan, Ezekiel Johnston-
Halperin, J.E. Goldberger, Water activated doping and transport in multilayered germanane
crystals, J. Phys. Condens. Matter. 28 (2016) 34001.
[23] E. Bianco, S. Butler, S. Jiang, O.D. Restrepo, W. Windl, J.E. Goldberger, Stability and
Exfoliation of Germanane: A Germanium Graphane Analogue, ACS Nano. 7 (2013) 4414–
4421.
[24] A.C. Ferrari, J.C. Meyer, V. Scardaci, C. Casiraghi, M. Lazzeri, F. Mauri, S. Piscanec, D.
Jiang, K.S. Novoselov, S. Roth, A.K. Geim, Raman Spectrum of Graphene and Graphene
Layers, Phys. Rev. Lett. 97 (2006) 187401.
[25] A.C. Ferrari, D.M. Basko, Raman spectroscopy as a versatile tool for studying the properties
of graphene, Nat. Nanotechnol. 8 (2013) 235–246.
[26] Y. Ma, Y. Dai, W. Wei, B. Huang, M.-H. Whangbo, Strain-induced quantum spin Hall effect
in methyl-substituted germanane GeCH3, Sci. Rep. 4 (2014) 7297.
[27] S. Jiang, E. Bianco, J.E. Goldberger, The structure and amorphization of germanane, J. Mater.
Chem. C. 2 (2014) 3185–3188.
[28] M. Wihl, M. Cardona, J. Tauc, Raman scattering in amorphous Ge and III–V compounds, J.
Non-Cryst. Solids. 8 (1972) 172–178.
[29] D. Bermejo, M. Cardona, Raman scattering in pure and hydrogenated amorphous germanium
and silicon, J. Non-Cryst. Solids. 32 (1979) 405–419.
[30] L.W. Veldhuizen, C.H.M. van der Werf, Y. Kuang, N.J. Bakker, S.J. Yun, R.E.I. Schropp,
Optimization of hydrogenated amorphous silicon germanium thin films and solar cells
deposited by hot wire chemical vapor deposition, Thin Solid Films. 595, Part B (2015) 226–
230.
[31] A.H. Clark, Electrical and Optical Properties of Amorphous Germanium, Phys. Rev. 154
(1967) 750–757.
[32] S.K. Sharma, S.C. Jain, S.S. Aggarwal, V.G. Bhide, Structure and electrical resistivity of thin
amorphous germanium films, J. Non-Cryst. Solids. 7 (1972) 285–294.
[33] M. Pollak, M.L. Knotek, H. Kurtzman, H. Glick, dc Conductivity of Amorphous Germanium
and the Structure of the Pseudogap, Phys. Rev. Lett. 30 (1973) 856–859.
[34] V. Čápek, S. Koc, J. Zemek, DC conductivity of amorphous germanium, J. Non-Cryst. Solids.
18 (1975) 95–106.
[35] W. Thieme, H. Diletti, H. Coufal, W. Gierisch, E. Lüscher, The electrical resistivity of highly
doped p-type germanium from 40 mK to 300 K, J. Common Met. 62 (1978) 405–406.
[36] L. Seixas, J.E. Padilha, A. Fazzio, Quantum spin Hall effect on germanene nanorod embedded
in completely hydrogenated germanene, Phys. Rev. B. 89 (2014) 195403.
[37] C. Si, J. Liu, Y. Xu, J. Wu, B.-L. Gu, W. Duan, Functionalized germanene as a prototype of
large-gap two-dimensional topological insulators, Phys. Rev. B. 89 (2014) 115429.
[38] S. Hikami, A.I. Larkin, Y. Nagaoka, Spin-Orbit Interaction and Magnetoresistance in the Two
Dimensional Random System, Prog. Theor. Phys. 63 (1980) 707–710.
[39] E. McCann, K. Kechedzhi, V.I. Fal’ko, H. Suzuura, T. Ando, B.L. Altshuler, Weak-
Localization Magnetoresistance and Valley Symmetry in Graphene, Phys. Rev. Lett. 97 (2006)
146805.
[40] H.-T. He, G. Wang, T. Zhang, I.-K. Sou, G.K.L. Wong, J.-N. Wang, H.-Z. Lu, S.-Q. Shen, F.-
C. Zhang, Impurity Effect on Weak Antilocalization in the Topological Insulator Bi2Te3, Phys.
Rev. Lett. 106 (2011) 166805.
[41] F.J. Dyson, Statistical Theory of the Energy Levels of Complex Systems. I, J. Math. Phys. 3
(1962) 140–156.
119
Appendix.
Ultra-small carbon nanotubes in zeolite
template
Abstract
Besides the widely explored two-dimensional (2D) materials, one-dimensional
(1D) structures represented by carbon nanotubes are also regarded as one of the
basic units for the future electronic industry. Here we discuss new developments
on the chemical vapor deposition growth of 0.4 nm single-walled carbon
nanotubes (SWCNTs) inside the linear channels of the aluminophosphate zeolite,
AlPO4-5 (AFI), single crystals (0.4 nm-SWCNTs@AFI). Ethylene (C2H4) and
carbon monoxide (CO) were used as the feedstock. Polarized Raman spectroscopy
was used to analyze the structure and quality of SWCNTs, both the radial breathing
mode and G band are much clearer and stronger than the samples grown by the
old process which used template tripropylamine molecules for growing
SWCNTs@AFI. From the Raman spectra, it is clearly seen that the RBM is
composed of two peaks at 535 and 551 cm-1. By using the pseudopotential module
in Material Studio to calculate the Raman lines, the 535 cm-1 peak is attributed to
the (5,0) SWCNTs and the 551 cm-1 peak to the (3,3) SWCNTs. The abundance of
(4,2) is relatively small. Thermal gravity analysis showed that while the samples
grown by CO display less than 1 wt% of carbon, for the samples heated in C2H4
atmosphere the weight percentage of SWCNTs is around 10%, which implies ~30%
of the AFI channels are occupied with SWCNTs, a significant increase compared
with the previous samples.
Published in:
Q. Chen, Z. Wang, Y. Zheng, W. Shi, D. Wang, Y.-C. Luo, B. Zhang, J. Lu, H. Zhang, J. Pan,
C.-Y. Mou, Z. Tang, P. Sheng, “New developments in the growth of 4 Angstrom carbon
nanotubes in linear channels of zeolite template”, Carbon. 76 (2014) 401–409.
Appendix
120
1 Introduction
Electronic circuits with reduced dimensions have always been the long pursuing
goal for making small and powerful devices. MoS2 and graphene represent the
family of two-dimensional (2D) materials that show promising properties for
applications. On the other hand, one-dimensional (1D) material such as carbon
nanotubes (CNTs), are also very attractive since they are even smaller compared
with 2D structure. CNTs have been intensively studied since their discovery [1, 2].
Owing to their unique physical properties and the 1D character, they are regarded
as one of the basic materials of the nanotechnology industry [3-6]. Much effort has
been devoted to the synthesis of single-walled carbon nanotubes (SWCNTs) with
desirable chirality at a given location, with well-controlled direction and length [7-
11]. Compared with large-sized SWCNTs, ultra-small SWCNTs have aroused
special interests because of the strong curvature effect that can lead to a significant
hybridization between the sigma ( * ) and pi ( * ) orbitals [12]. The curvature
effect enriches the electronic properties by redistributing the energy in the
electronic states, and opens new channels for electron-phonon coupling which can
lead to special transport characteristics. However, SWCNTs with very small
diameter can become unstable, as the system energy increases rapidly with
decreasing diameter [13]. Theoretical calculations predicted the existence of
SWCNTs with the diameter as small as 0.4 nm [14]. Later it was reported that the
ultra-small SWCNTs can be produced inside the channels of aluminophosphate,
AlPO4-5 (AFI), zeolite single crystals [15], and the diameter was determined to be
0.4 nm by high-resolution transmission electron microscopy [16]. With such a
small diameter and extreme curvature, these well-aligned and mono-sized SWCNT
arrays showed interesting physical properties, e.g. superconductivity has been
observed in this system [17-21].
The host AFI zeolite crystal is a type of microporous aluminophosphate zeolites.
It is widely used in the host-guest chemistry because of its thermal stability up to
1200 °C, optical transparency over a broad range from ultraviolet to near infrared,
and electrically insulating characteristics [22-26]. Figure 1 shows the structure of
the AFI crystal viewed along the [001] (c-axis) direction. The framework is
composed of alternating tetrahedral (AlO4)- and (PO4)+ units, forming parallel
open channels that are arranged in a triangular lattice structure [27]. The inner
diameter of the channel is 0.734 nm and the center-to-center distance between two
neighboring channels is 1.374 nm. The tripopylamine [(CH3CH2CH2)3N (TPA)]
molecules are aligned head-to-tail along the c-axis, serving as the precursor
template for synthesizing the AFI crystals. In the old method of producing the 0.4
nm CNTs, the SWCNTs were obtained by pyrolyzing the TPA molecules [15]. The
AFI crystals were heated to 580 °C in vacuum for several hours, leading to the
Ultra-small carbon nanotubes in zeolite template
121
decomposition of the TPA molecules that are encapsulated inside the AFI channels,
with a small fraction forming the SWCNTs. There are limitations for this process:
in the Raman spectra the intensity of the radial breathing mode (RBM) was weak
and the peaks are broad; the thermal gravity analysis (TGA) data showed the
weight percentage of carbon over the total mass of SWCNTs@AFI crystal to be ~1.5
wt%, translating into a filling factor of SWCNTs inside the AFI channels of ~4.5%.
This is a relatively small number, indicating that the samples are not uniform and
only a small fraction of AFI channels are occupied with SWCNTs. In this approach,
the carbon atoms that formed the SWCNTs came from the pyrolysis of the TPA
molecules inside the AFI channels. Since the total amount of TPA molecules are
limited in quantity and some of them may escape during the heating process (the
AFI crystals were heated in vacuum), the low filling factor is understandable.
Figure 1 The (a) framework, (b) optical and (c) SEM images of AFI zeolite crystals. The
framework is composed of (AlO4)- and (PO4)+, forming linear pores packed in a triangular
lattice. The diameter of the channel is 0.73 nm and the distance between two neighboring
channels is 1.37 nm. The as-grown AFI crystals are transparent, with channels occupied by the
TPA template.
To improve this situation, we have developed a new chemical vapor deposition
(CVD) process by first burning off the TPA molecules in an oxygen atmosphere,
and then introducing carbon-containing gas as the feedstock. Two gases were
tested: C2H4 and CO. In this approach, continuous carbon source can enter the
empty channels of the AFI crystals. Compared with the traditional CVD method
for growing CNTs, no catalyst is used in the present process. The fact that CNTs
can still be formed suggests that the framework of AFI crystal must play a weak
catalytic role in the pyrolysis and conversion of ethylene, or CO, to SWCNTs.
Appendix
122
2 Experimental
The AFI zeolite crystals are synthesized by the conventional hydrothermal
method. In the synthesis procedure, aluminum tri-isopropoxide [(iPrO)3Al 99 wt%]
and phosphoric acid (H3PO4 85 wt%) were used as aluminum and phosphorus
sources, respectively. The TPA molecules served as the template during the
synthesis process. The details are described in Ref. 28. The as-grown AFI crystals,
with TPA molecules capsulated in the channels, are optically transparent with
typical dimensions of 300 μm in length and 100 μm in diameter.
The as-made AFI crystals are first heated in a flowing oxygen atmosphere at
900 °C, to remove the organic template TPA molecules through oxidization.
Raman spectroscopy of the AFI crystals after this step showed no RBM or G band,
which means that the channels of the AFI crystals are mostly empty and ready for
the growth of SWCNTs. Subsequently, a flowing gas mixture of C2H4 (or CO) and
nitrogen was introduced for growing the SWCNTs, while the heating temperature
was controlled within the range of 600-950 °C. In theory, perfect AFI crystal is
stable up to 1200 °C, so we have tried even higher heating temperature with CO.
The resulted quality of the sample was poor, however, because in reality the AFI
crystals have defects, which tend to degrade the crystals before the stability
temperature is reached. Hence the heating temperature was controlled to be below
950 °C. After the heating treatment, the AFI crystals appear dark brown in color
and exhibit strong optical anisotropy; i.e., almost transparent (black) when the
electric field is perpendicular (parallel) to the c-axis.
The SWCNTs@AFI were characterized by polarized Raman spectroscopy using
a Jobin Yvon T6400 micro-Raman system and 514.5 nm Ar laser excitation at
room temperature. Thermo-gravimetry analyses were carried out by a thermal
analysis apparatus (Netzsch, STA 449C Jupiter) to characterize the carbon content
inside the AFI channels.
Ultra-small carbon nanotubes in zeolite template
123
3 Results and discussion
3.1 Raman spectra
Figure 2 The Raman spectra of the 0.4 nm-SWCNTs@AFI samples fabricated by the old process
(I), shown as the black curve (a); new CVD process with CO (II), shown as the red curve (b); and
new CVD process with C2H4 (III), shown as the blue curve (c).
Several hundreds of heating processes were carried out by varying the heating
temperature, gas pressure, gas flow rate, and carbon source. Overall, they can be
categorized into three groups: (I) the old process which used TPA molecules as the
carbon source for growing SWCNTs; (II) new CVD process with CO being the
carbon source for SWCNTs; and (III) new CVD process with C2H4 being the carbon
source. The results can vary significantly. No direct Transmission Electron
Microscope (TEM) image of 0.4 nm-SWCNTs is obtained at this stage because
such small-diameter SWCNTs are not stable when exposed to an electron beam, it
is quite difficult and time-consuming to observe 0.4 nm-SWCNTs by electron
microscope. But they were clearly observed by high-resolution TEM in the previous
report [16]. The method to prepare a sample for TEM observation and the images
are shown in Ref. 29. In contrast, Raman spectroscopy is more convenient as well
as providing abundant information about the 0.4 nm-SWCNTs@AFI. In Figure 2,
we choose one representative sample in each group to display their Raman spectra.
The associated heating processes of the three batches of SWCNTs@AFI are shown
in Figure 3.
Appendix
124
The P-O and P-O-Al vibrations of the AFI framework give rise to signals at 450
and 1100 cm-1 in the Raman spectra, but they are normally very weak compared
with the signals from the SWCNTs [30]. Hence the effect of the AFI framework can
be neglected in the Raman spectra. It can be seen that although the SWCNTs were
grown from different sources, their Raman spectra have very similar features.
Especially for the samples grown by the new CVD process with ethylene or CO
source, their spectra coincide quite well with each other. They differ only in some
small but notable features. Below we comment on the various features of the
Raman spectra.
Figure 3 Detailed heating processes of the three batches of samples whose Raman spectra are
shown in Figure 2. For the old process, TPA template molecules are used as the carbon source
for growing CNTs. For new CVD process, two steps are employed: first remove TPA molecules
by heating in oxygen, and then introduce CO or C2H4 as carbon source for growing CNTs.
(1) The radial breathing mode is in the low frequency region of 500-600 cm-1.
The RBM serves as a unique signature of CNTs because it corresponds to the
coherent vibration of the carbon atoms along the radial direction. Planar
structures such as graphene or graphite do not have such Raman-active modes.
The RBM of SWCNTs fabricated by the new CVD process (II) or (III) is much more
Ultra-small carbon nanotubes in zeolite template
125
evident than that of (I), thereby they provide us more information about the
structure of the SWCNTs. According to the high-resolution TEM characterization
[16], the diameter of the SWCNTs@AFI was determined to be 0.4 nm. There are
only three possible SWCNTs with such a small diameter: the zigzag (5,0), the
armchair (3,3) and the chiral (4,2).
Figure 4 Statistics of RBM/G ratio for samples fabricated by the three different processes.
Raman measurements are performed on different samples from each method, in each spectra
we analyze the RBM/G ratio and calculate the frequency. The ratio of the CVD process is
evidently higher than the old process.
Since not all the carbon atoms in the AFI channels are in the nanotube form,
some of them can be in the form of amorphous carbon which will give rise to G
band signal but no RBM signal. So the ratio of RBM relative to the G band (RBM/G
ratio) serves as a simple yet intuitive indication of the amount of carbon atoms that
is in the nanotube form. However, it should be noted that the RBM/G ratio is not
necessarily related to the TGA result, since the latter indicates the total amount of
carbon, rather than its distribution among the amorphous form of the nanotube
form. Figure 4 shows the results of a statistical study on the RBM/G ratio, for the
three batches of samples shown in Figure 2. We can see that the RBM/G ratio
increased greatly from ~5% (sample I) to almost ~20% (sample III). This clearly
suggests that for the new fabrication process, there is less amorphous carbon in the
AFI channels.
(2) G band is an important feature of CNTs. For SWCNTs, the G band consists
of two main components as shown in Figure 5(a): G+ and G-, corresponding to the
Appendix
126
movement of carbon atoms along the tube axis and the circumferential direction,
respectively. From Figure 5(b), we can see that the G+ band of the 0.4 nm-
SWCNTs@AFI is exactly at 1590 cm-1 and can be fitted with the Lorentzian line
shape very well, in good agreement with the larger diameter SWCNTs reported
elsewhere [31, 32]. There is a small peak around 1614 cm-1. This may be due to the
tangential mode of semiconducting nanotubes composed of a few Lorentzians [31].
This also shows that the G+ peak is independent of the tube diameter, even down to
0.4 nm. Besides, the G+ band is very sharp compared with that from samples
fabricated by process (I).
Figure 5 (a) Two types of vibrations for the G band, corresponding to the longitudinal and
transverse vibrations of carbon atoms. (b) Fitting of the G band of sample III at 1590 cm-1 by
the Lorentzian line shape; (c) Fitting of the G𝑀− band at 1170 cm-1 by the BWF line shape; (d)
Fitting of the G𝑆− band at 1370 and 1480 cm-1 by the Lorentzian line shape.
In contrast to the G+ peak, the frequency and line shape of the lower frequency
G- peak are strongly related to the chirality and diameter of the nanotubes. The G-
band of armchair carbon nanotube have a Breit-Wigner-Fano (BWF) line shape (a
broad and asymmetric peak), while the G- band of the semiconducting carbon
nanotubes remains Lorentzian [33, 34]. In accordance with the density functional
theory (DFT) calculation of the electronic band structure, the (5,0) SWCNTs are
metallic because of its small radius [35]. However, from the zone folding approach
Ultra-small carbon nanotubes in zeolite template
127
[36] or the tight binding calculations [37], the zigzag (5,0) nanotubes are
semiconducting in character. For the Raman characteristics, we should therefore
examine both the metallic and semiconducting characteristics of the G- band.
For armchair SWCNTs, the frequency of G- band is related to its diameter by
the relation [31]:
2
t
GGd
C , (1)
with 𝐶 = 𝐶M = 79.5 cm−1nm2. According to Eq. (1), the frequency of the G- band
should be around 1100 cm-1 for the metallic armchair 0.4 nm-SWCNTs@AFI. In
the Raman spectra of SWCNTs@AFI there is a broad peak around 1170 cm-1, and
in Figure 5(c) we see that it fits well with the BWF line shape. So this peak can very
well be the G- band of armchair 0.4 nm-SWCNTs@AFI. The deviation from what is
predicted by Eq. (1) can be due to the strong curvature effect of the 0.4 nm tube
structure.
For the non-armchair SWCNTs, six Raman-active modes of A+E1+E2 symmetry,
along each of the two orthogonal (the axial and circumferential) directions, can be
present in the G band [38]. The frequency of the G- band for the semiconducting
SWCNTs can be well fitted by the relation:
1592G
t
C
d , (2)
With 𝛽 = 1.4 , 𝐶E2 = 64.6 cm−1nm1.4 , 𝐶A = 41.4 cm−1nm1.4 , and 𝐶E1 =
32.6 cm−1nm1.4 [39]. According to Eq. (2), the G- peaks should be at 1359 (E2), 1443
(A), and 1474 cm-1 (E1) for the non-armchair 0.4 nm-SWCNTs@AFI. From the
Raman spectra fitting of sample (III) in Figure 5(d), four peaks centered at 1352,
1377, 1460, 1483 cm-1 are identified. The 1352 cm-1 should be associated with the
disorder induced D band which will be discussed in the next section. It is tempting
to attribute the remaining three peaks to the G- band of semiconducting 0.4 nm-
SWCNTs@AFI. However, from the polarized Raman spectra shown in Figure 6,
The 1377 cm-1 mode not only exhibit strong intensity at (ZZ) and (XX) polarized
spectra associated with the A1(A1g) symmetry, but also has intensities in cross-
polarized (ZX) and (XZ) spectra indicating that it should be assigned the E1(E1g)
symmetry, but the frequency is too far from the predicted E1 mode at 1474 cm-1.
The 1460 and 1483 cm-1 peaks are observed in the (ZZ) and (XX) parallel polarized
spectra, hence they should be assigned A1(A1g) symmetry. Whereas 1460 cm-1 is
close to the 1443 cm-1 G- mode with the correct symmetry, the other one, at 1483
cm-1, cannot be easily identified so far. It should be noted that Eq. (2) is obtained
by measuring the Raman spectra of isolated SWCNTs [39], but the Raman spectra
of 0.4 nm-SWCNTs@AFI are measured within the framework of the AFI crystals.
Appendix
128
Thus some frequency difference may be attributed to the interactions between the
nanotube and the AFI framework.
Figure 6 Polarized Raman spectra of the SWCNTs@AFI. Y direction denotes the incident light
direction and the Z direction denotes the c-axis direction of the SWCNTs. The XZ, XX and ZX
configurations are enlarged and vertically shifted for comparison. This batch of sample was
heated with the C2H4 under the condition of pressure = 400 Torr and temperature = 700 °C for
4 hours, but it showed similar Raman behavior with sample (III) fabricated under higher
temperature and pressure, even though the carbon content is much lower as evidenced by the
TGA.
(3) As mentioned above, there is a D band around 1352 cm-1, which might be
due to defects in the tube structure. The D band signal cannot be coming from the
amorphous carbon layer on the surface of AFI crystal because the Raman spectra
were measured with half of the crystal milled away by mechanical polishing, which
means the surface amorphous carbon was removed. A higher D band peak would
correspond to a higher concentration of defects in the sample. Different fabrication
conditions have been tested to minimize the D band, but the intensity is still
relatively large. It is speculated that the D band might also arise from the curvature
effect, owing to the small radius.
(4) There is a sharp, asymmetric peak at 986 cm-1 which is still puzzling. From
Figure 6 we can see that the 986 cm-1 peak always accompanies the RBM: It is
weak when the RBM is weak (both the XX and ZX configurations), and it
disappears when the RBM is absent (XZ configuration). Previous study on the
intermediate frequency region modes (between 600 -1100 cm-1) of larger diameter
SWCNTs (𝑑t = 1.5 ± 0.3 nm) has revealed that this peak can be related to the
combination of one optical and one acoustic phonon that originate from a zone
Ultra-small carbon nanotubes in zeolite template
129
folding procedure of two 2D phonon branches, associated with the out-of-plain
vibrations of carbon atoms [40]. Such mode is Raman-inactive for large carbon
nanotubes, but can be active for small diameter carbon nanotubes. However, the
physical origin of this mode still remains an open question.
We note that another feature in Figure 6 is the peak at 500 cm-1. It is
independent of the RBM and is clear and sharp for the XZ configuration. It also
appears in the Raman spectra of the empty AFI crystals after the TPA molecules
were removed by heating in an oxygen atmosphere. So the peak at 500 cm-1 should
be associated with the AFI framework, not related to the SWCNTs.
Figure 7 RBM of the three processes from Figure 2: (I) old process; (II) CVD process with CO;
(III) CVD process with C2H4. Two RBM peaks at 535 cm-1 and 551 cm-1 are identified from the
CVD process, associated with (5,0) and (3,3) CNTs, respectively.
Figure 7 shows the RBM of the samples from three different processes. From (II)
and (III) it is seen that the RBM is composed of two peaks at 535 and 551 cm-1. The
broad peak around 500 cm-1, which is very weak compared with other peaks, is
attributed to the framework of AFI crystal as discussed above. For our template-
grown carbon nanotubes, there are only three candidates with suitable diameters:
(5,0), (3,3) and (4,2), with the respective diameters of 0.393, 0.407 and 0.414 nm.
Since the diameter of the AFI channels is 0.73 0.01 nm and the distance between
the wall of SWCNTs and the AFI channels should be close to the separation
between graphite sheet (0.34 nm), the allowed diameter is 0.39 0.01 nm. So the
diameters of the (5,0) and (3,3) are more favorable than (4,2). In addition, it is
reported that by using CO, the growth of the (10,0) zigzag tubes is more favorable
Appendix
130
than the (5,5) armchair tubes [41]. This may suggest that zigzag tubes can be
preferred in the growth using CO gas. In Figure 7 we can see that for the 0.4 nm-
SWCNTs@AFI fabricated by CO gas, the peak at 535 cm-1 is stronger than that at
551 cm-1. So the peak at 535 cm-1 is attributed to (5,0), and the peak at 551 cm-1 is
attributed to the (3,3) carbon nanotubes. This assignment is further supported by
theory calculations, shown below. The signal from (4,2), if it exists, must be very
weak.
3.2 Simulation of the RBM
We have carried out the calculation of RBM for the (5,0), (3,3) and (4,2)
nanotubes by using the CASTEP pseudopotential module in Material Studio. The
results are listed in Table 1.
Table 1 The RBM of (3,3), (5,0), (4,2) as calculated by the CASTEP pseudopotential module in
Material Studio. The cartoon pictures show the crystal structures for each CNT, and the green
arrows indicate the movement of carbon atoms for the RBM mode.
From Table 1, the RBMs of (3,3), (5,0), (4,2) are at 556.7, 534.9, and 526 cm-1,
respectively. Although the frequency of RBM may change a few wave numbers
when different pseudopotential is used, their differences are fixed relative to each
other. The calculated results therefore agree well with our previous assignment
that the peaks at 551 and 535 cm-1 belong to (3,3), (5,0), respectively, and that (4,2)
is hardly seen in the 0.4-nm SWCNTs@AFI system. Hence it may be possible to
develop a chirality-selective process of growing SWCNTs using CO.
3.3 Thermal-gravimetric analysis (TGA)
Thermal-gravimetric analysis was carried out to characterize the quantity of
SWCNTs inside the channels of AFI crystals. The TGA result of sample (III) is
shown in Figure 8. The weight loss, starting at around 570 °C, is attributed to the
oxidation of carbon. We see that by using ethylene as the carbon feedstock and
heating under high pressure condition the total weight percentage of carbon can
Ultra-small carbon nanotubes in zeolite template
131
reach 25.8 wt%. However, the weight percentage of SWCNTs should be smaller
than the total weight loss because there is amorphous carbon on the surface of the
AFI crystals. Normally the oxidization temperature of amorphous carbon should be
lower than that of SWCNTs, but for our case they are very close. We have measured
the TGA of amorphous carbon which was collected from the quartz tube after
growing 0.4 nm-SWCNTs@AFI. They confirm that in our case, the oxidization
temperature of amorphous carbon is higher than what might be usually expected
[29], most probably owing to the relatively high heating pressure and temperature
under which the amorphous carbon was formed, leading to a more tightly-packed,
highly-linked structure that is more robust against oxidation. The thickness of the
amorphous carbon layer was measured to be ~2.5 μm. By taking into account the
density of amorphous carbon — 2.0 g/cm3, density of AFI crystals —2.6 g/cm3, and
the average size of the crystals — 300 μm in length and 100 μm in diameter; the
weight percentage of amorphous carbon is estimated to be 15%. That means the
contribution from SWCNTs is approximately 10%, which is much higher than the
value obtained the old process (I).
Figure 8 Thermogravimetric data from 18 mg of 0.4 nm-SWCNTs@AFI obtained from process
(III) in flowing oxygen, with a ramp rate of 5 °C/min. The weight loss of 26% at 550 °C is
attributed to the decomposition of SWCNTs.
Although the carbon content is greatly increased, by calculation it is inferred
that not every channel is filled with CNTs. We define the filling factor as the
percentage of the channel space occupied by SWCNTs over the total amount of
channel space. The carbon content and filling factor are listed in Table 2. From this
table we can see that when we use ethylene as the carbon source and heat under
Appendix
132
high pressure (III), the carbon content can be increased from 1.47% to 10% as
compared to the old process (I), with the filling factor increasing from 4.5% to 30%.
Table 2 Carbon content and pore filling factors of samples from the three processes.
We note that the carbon content of SWCNTs@AFI heated by process (II), using
CO, is about 0.7%, much lower than both process (III) and (I). This may be due to
the fact that CO is more stable than either ethylene or TPA, and no catalyst was
used in the heating process. Without a catalyst, it may be difficult for CO to
decompose and form SWCNTs in the AFI channels. By using CO as the carbon
source and heating under high pressure, there are some good SWCNTs in local
areas of AFI crystals, but the overall amount of SWCNTs is still small.
We show in Figure 9 a two-dimensional summary of results, with ethylene gas
as the feedstock. Here pressure is the vertical axis and heating temperature serves
as the horizontal axis. The old process is noted to yield samples with only very
weak RBMs and small amount of nanotubes in the channels. The use of CO without
catalyst, on the other hand, leads to good RBM signals but very small pore filling.
Thus the optimal approach is to use ethylene as the carbon source, with the
conditions delineated by the dark red area in Figure 9. The difficulty of using
ethylene as the source, however, is that at temperatures higher than 800oC the
polymerizing tendency of the ethylene molecules can lead to the proliferation of
byproducts in both liquid and solid form. When that happens, the AFI crystals can
be difficult to extract from the resulting (fused) mixture.
4 Conclusion
By taking advantage of the small channels of AFI crystals, we have developed an
efficient CVD process of growing uniform 0.4 nm-SWCNTs using ethylene and CO.
This new process is better than the old process as evidenced by the Raman spectra
showing that both the RBM and G band to be clearer and stronger than the old
process. The two clear RBM peaks at 535 and 551 cm-1 indicate that the 0.4 nm-
Ultra-small carbon nanotubes in zeolite template
133
SWCNTs@AFI are mainly composed of (5,0) and (3,3) nanotubes. This conclusion
is supported by theoretical calculations. Our results also suggest the possibility of
developing a chirality-selective growing process of 0.4nm-SWCNTs@AFI using CO.
By using ethylene as the feedstock and heating under high pressure with a
temperature of 700 oC, both the quantity and quality of 0.4 nm-SWCNTs in the
AFI channels are significantly improved.
Figure 9 Overall result of different heating conditions for process (III), using ethylene as the
feedstock. The darkness of the red color represents the overall quality of the grown SWCNTs.
The optimal condition is found to be at 750 °C heating temperature and 3 atm pressure.
Appendix
134
References
[1] Oberlin A, Endo M, Koyama T. Filamentous growth of carbon through benzene decomposition. Journal of Crystal Growth. 32 (1976) 335-349.
[2] Iijima S. Helical microtubules of graphitic carbon. Nature. 354 (1991) 56-58.
[3] Saito R. Physical properties of carbon nanotubes. London: Imperial college press. 35 (1998).
[4] Charlier J-C, Blase X, Roche S. Electronic and transport properties of nanotubes. Reviews
of modern physics. 79 (2007) ; 677-732.
[5] Dresselhaus MS, Dresselhaus G, Eklund P, Rao A. Carbon nanotubes. Springer; 23 (2000)
331-379.
[6] Baughman RH, Zakhidov AA, de Heer WA. Carbon nanotubes--the route toward applications. Science. 297 (2002) 787-792.
[7] Dai H. Carbon nanotubes: synthesis, integration, and properties. Accounts of chemical research. 35 (2002) 1035-1044.
[8] Hanus MJ, Harris AT. Synthesis, characterisation and applications of coiled carbon nanotubes. J Nanosci Nanotechno. 10 (2010) 2261-2283.
[9] Awasthi K, Srivastava A, Srivastava O. Synthesis of carbon nanotubes. J Nanosci Nanotechno. 10 (2005) 1616-1636.
[10] Tan L-L, Ong W-J, Chai S-P, Mohamed AR. Growth of carbon nanotubes over non-metallic
based catalysts: A review on the recent developments. Catalysis Today. 217 (2013) 1-12.
[11] Azam MA, Manaf NSA, Talib E, Bistamam MSA. Aligned carbon nanotube from catalytic
chemical vapor deposition technique for energy storage device: a review. Ionics. 19 (2013) 1455-1476.
[12] Blase X, Benedict LX, Shirley EL, Louie SG. Hybridization effects and metallicity in small radius carbon nanotubes. Phys Rev Lett. 72 (1994) 1878-1881.
[13] Robertson D, Brenner D, Mintmire J. Energetics of nanoscale graphitic tubules. Phys Rev B. 45 (1992) 12592.
[14] Sawada S, Hamada N. Energetics of Carbon Nanotubes. Solid State Commun. 83 (1992) 917-919.
[15] Tang ZK, Sun HD, Wang J, Chen J, Li G. Mono-sized single-wall carbon nanotubes formed
in channels of AlPO4-5 single crystal. Applied Physics Letters. 16 (1998) 2287-2289.
[16] Wang N, Tang Z-K, Li G-D, Chen J. Materials science: Single-walled 4 Å carbon nanotube
arrays. Nature. 408 (2000) 50-51.
[17] Wang Z, Shi W, Xie H, Zhang T, Wang N, Tang Z, et al. Superconducting resistive transition in coupled arrays of 4Å carbon nanotubes. Phys Rev B. 81 (2010) 174530.
[18] Lortz R, Zhang Q, Shi W, Ye JT, Qiu C, Wang Z, et al. Superconducting characteristics of 4-Å carbon nanotube–zeolite composite. Proc. Natl. Acad. Sci. 106 (2009) 7299-7303.
[19] Zhang T, Sun MY, Wang Z, Shi W, Sheng P. Crossover from Peierls distortion to one-
dimensional superconductivity in arrays of (5, 0) carbon nanotubes. Phys Rev B. 84 (2011) 245449.
[20] Wang Z, Shi W, Lortz R, Sheng P. Superconductivity in 4-Angstrom carbon nanotubes—A
short review. Nanoscale. 4 (2012) 21-41.
[21] Sun M, Hou Z, Zhang T, Wang Z, Shi W, Lortz R, et al. Dimensional crossover transition
in a system of weakly coupled superconducting nanowires. New Journal of Physics. 14 (2012) 103018.
Ultra-small carbon nanotubes in zeolite template
135
[22] Terasaki O, Yamazaki K, Thomas J, Ohsuna T, Watanabe D, Sanders J, et al. Isolating
individual chains of selenium by incorporation into the channels of a zeolite. Nature. 330 (1987) 58-60.
[23] Cox S, Gier T, Stucky G. Second harmonic generation by the self-aggregation of organic
guests in molecular sieve hosts. Chemistry of Materials. 2 (1990) 609-619.
[24] Wu C-G, Bein T. Conducting polyaniline filaments in a mesoporous channel host. Science. 264 (1994) 1757-1759.
[25] Li IL, Tang Z. Structure study of Se species in channels of AlPO4-5 crystals. Applied physics letters. 80 (2002) 4822-4824.
[26] Marlow F, Hoffmann K, Caro J. Photoinduced switching in nanocomposites of azobenzene and molecular sieves. Advanced Materials. 9 (1997) 567-570.
[27] Wilson ST, Lok BM, Messina CA, Cannan TR, Flanigen EM. Aluminophosphate
molecular sieves: a new class of microporous crystalline inorganic solids. J Am Chem Soc. 104 (1982) 1146-1147.
[28] Zhai J, Li Z, Liu H, Li I, Sheng P, Hu X, et al. Catalytic effect of metal cations on the
formation of carbon nanotubes inside the channels of AlPO4-5 crystal. Carbon. 44 (2006) 1151-1157.
[29] See "http://dx.doi.org/10.1016/j.carbon.2014.04.094" for (i) TEM images of 0.4 nm SWCNTs@AFI (ii) Characterization of the by-product amorphous carbon.
[30] Tang Z-K, Wang N, Zhang X, Wang J, Chan C-T, Sheng P. Novel properties of 0.4 nm
single-walled carbon nanotubes templated in the channels of AlPO4-5 single crystals. New Journal of Physics. 5 (2003) 146.
[31] Jorio A, Souza Filho A, Dresselhaus G, Dresselhaus M, Swan A, Ünlü M, et al. G band
resonant Raman study of 62 isolated single-wall carbon nanotubes. Phys Rev B. 65 (2002) 155412.
[32] Dresselhaus M, Eklund P. Phonons in carbon nanotubes. Advances in Physics. 49 (2000) 705-814.
[33] Dresselhaus MS, Dresselhaus G, Saito R, Jorio A. Raman spectroscopy of carbon nanotubes. Physics Reports. 409 (2005) 47-99.
[34] Barnett R, Demler E, Kaxiras E. Electron-phonon interaction in ultrasmall-radius carbon nanotubes. Phys Rev B. 71 (2005) 035429.
[35] Liu H, Chan C. Properties of 4 Å carbon nanotubes from first-principles calculations. Phys Rev B. 66 (2002) 115416.
[36] Dresselhaus M, Dresselhaus G, Jorio A, Souza Filho A, Saito R. Raman spectroscopy on
isolated single wall carbon nanotubes. Carbon. 40 (2002) 2043-2061.
[37] Saito R, Dresselhaus G, Dresselhaus M. Trigonal warping effect of carbon nanotubes. Phys
Rev B. 61(2000) 2981.
[38] Jorio A, Dresselhaus G, Dresselhaus M, Souza M, Dantas M, Pimenta M, et al. Polarized
Raman study of single-wall semiconducting carbon nanotubes. Phys Rev Lett. 85(2000) 2617.
[39] Jorio A, Pimenta M, Souza Filho A, Samsonidze GG, Swan A, Ünlü M, et al. Resonance
Raman spectra of carbon nanotubes by cross-polarized light. Phys Rev Lett. 90 (2003)
107403.
[40] Fantini C, Jorio A, Souza M, Ladeira L, Souza Filho A, Saito R, et al. One-dimensional
character of combination modes in the resonance Raman scattering of carbon nanotubes. Phys Rev Lett. 93 (2004) 087401.
[41] Seo K, Kim C, Kim B, Lee YH, Song K. Growth energetics of single-wall carbon
nanotubes with carbon monoxide. The Journal of Physical Chemistry B. 108 (2004) 4308-4313.
Appendix
136
137
Summary
In the past decade, we have witnessed rapid developments in electronic
technology. Semiconductor devices are getting smaller and more powerful,
following the Moore’s Law. Nevertheless, silicon-based transistors are rapidly
approaching their scaling limit, and thus exploration of new channel materials as
well as novel device architectures is highly demanded for the further
miniaturization of devices. To discover next-generation device technology,
atomically thin two-dimensional (2D) materials have been intensively explored as
potential channel materials for replacing silicon. Among them, semiconducting
layered compounds called transition metal dichalcogenides (TMDs) are widely
regarded as very promising candidates due to their exceptional electrical and
optical properties. One of the most studied compounds in the TMDs family is MoS2,
which has shown attractive properties such as exceptional field effect transistor
(FET) characteristics, superconductivity and valley Zeeman effect. Other new 2D
materials have also been theoretically proposed and experimentally explored. One
of these is germanane (or GeH), which is just now beginning to be studied; because
of this, physical properties of GeH remain largely unknown.
As part of my four years of PhD study, I spent two years at the University of
Groningen, working on understanding the physical properties of the 2D materials
MoS2, WS2, graphene and GeH. This thesis focuses only on electrical transport,
especially the quantum transport properties of MoS2 and GeH. Quantum transport
in most of the studies in this thesis is related to a special device structure called
electrical double layer transistor (EDLT). Chapter 2 and 3 studied the vertical
hetero-states formed in a single few-layer MoS2 flake, in which two parallel
conducting channels were identified and the interaction between these two
channels can be electrically tuned by the field effect. Chapter 4 discussed the
properties of a lateral superconductor-normal metal (SN) junction formed at the
surface of MoS2. Chapter 5 discussed the structure and transistor characteristics of
thin germanane flakes. The project in the first half of my study in HKUST involved
studying the physical properties of one-dimensional single-walled carbon
nanotubes (SWCNTs) and was presented in the Appendix of this thesis.
Vertical hetero-states in multi-layer MoS2 (Chapters 2 and 3)
In previous studies of multi-layer MoS2-EDLT, superconductivity was observed
and it showed strong 2D characteristics. Theoretical calculations revealed that the
ionic gating confines carriers mostly to the topmost layer. In our transport study of
Summary
138
a dual-gate MoS2 transistor on an h-BN substrate, two parallel conducting
channels were distinguished. Superconductivity is induced at the top surface of the
few-layer MoS2 flake by employing the well-known ionic gating technique. The
carrier density induced by ionic gating is in the order of ~1014 cm-2. Due to the
strong Thomas-Fermi screening effect, charge carriers are mainly confined to the
topmost layer; the bottom layers remain intrinsically semiconducting. The
conductivity of the bottom layers is well controlled by the back gate. Pronounced
Shubnikov-de Haas (SdH) oscillations are observed in high magnetic fields after
the superconductivity is suppressed. The carrier densities that account for the SdH
oscillations can be calculated from the oscillation period, and they turn out to be
more than one order of magnitude lower than the minimum carrier density for
achieving superconductivity. This result suggests that the superconductivity and
SdH oscillations originate from the top and bottom surfaces, respectively, which
unambiguously confirms the two parallel conducting channels model. There are
two reasons for the high carrier mobility of the bottom channel: (1) compared to
SiO2, the atomically flat h-BN substrate effectively reduces the scattering that
originates from the substrates, and (2) the top highly doped channel acts as a
screening layer for the bottom channel and thus improves the mobility.
The Landau level degeneracy g = 3 can be obtained from the SdH oscillations,
the result is consistent with previous reports. Another important parameter is
effective mass, m* = 0.6 me, obtained from the temperature dependence of the SdH
oscillations. These two parameters provide important information about the
energy state of the bottom channel. They suggest that the conducting electrons in
this channel are from the Q/Q’ valleys, because for the K/K’ valleys at the corners
of the Brillouin zone the electron effective mass is m* = 0.5 me, and the Landau
level degeneracy g should be 1 if the Fermi level resides at the K/K’ valleys. The
obtained Landau level degeneracy g = 3, which is equal to half of the 6 Q/Q’ valleys,
indicating that the valley degeneracy is lifted by the magnetic field, corresponding
to the valley Zeeman effect.
The top and bottom conducting channels in multi-layer MoS2 form a vertical
junction, analogous to the conventional van der Waals heterojunction. New device
functionalities have been developed by controlling the interaction between the two
channels. A similar device configuration was reproduced for a MoS2-EDLT on a
high-κ dielectric HfO2 substrate. The ionic gating induces superconductivity in the
top surface and the back gating induces a metallic state in the bottom surface. With
the HfO2 back gate, the conductivity of the bottom channel can be tuned to a larger
extent compared to a SiO2 back gate. Amazingly, the superconductivity in the top
channel can be effectively controlled in different manners according to the
electronic state of the bottom channel. For negative back gate voltages, the bottom
channel remains insulating and acts like an additional gate dielectric layer for the
Summary
139
back gate tuning of superconductivity in the top channel. A continuous
superconductor-insulator (SI) quantum phase transition is realized based on this
configuration. For positive back gate voltages, on the other hand, the bottom
channel becomes metallic due to electron accumulation. The strong anti-proximity
effect between the bottom metallic channel and the top superconducting channel
can suppress the superconductivity significantly. A “bipolar-like” superconducting
transistor is created with different working principles in different back gate
regimes. For negative back gate bias, the superconductivity is continuously
switched off by field effect depletion of carriers. For positive back gate bias, the
anti-proximity effect between the top superconducting channel and the bottom
metallic channel suppresses the superconductivity.
Lateral superconductor-normal metal junction in MoS2 (Chapter 4)
After studying the vertical interactions in few-layer MoS2 flakes, we then
focused on the lateral transport at the surface of MoS2. We fabricated another
MoS2-EDLT device, also on an h-BN substrate but with the middle part of the
MoS2 flake covered by a top h-BN flake. At the h-BN covered MoS2 surface, the
ionic gating has a very weak doping effect because the h-BN separates ions from
the MoS2 surface; hence it is in the normal state. In the exposed MoS2 surface,
superconductivity can be easily induced by the ionic gating, hence it becomes
superconducting. The conductivity of the h-BN covered MoS2 channel can be
controlled by the back gate similar to a traditional transistor. With positive back
gate voltages, the h-BN covered MoS2 channel becomes metallic, and an SN
junction forms. From the transfer curves of the h-BN covered MoS2 channel, the
high mobility indicates the high quality and homogeneity in the MoS2. This is
further proven by the observation of the SdH oscillations under strong magnetic
fields. Both the Landau level degeneracy and effective mass obtained from this
measurement are in line with the values obtained previously by measuring the
bottom high-mobility channel of the multi-layer MoS2 flake.
Due to the huge doping level variation between the exposed and h-BN covered
surface of MoS2, an energy barrier forms at the SN interface. The strength of the
barrier is intermediate (in between a zero barrier and a strong tunneling barrier),
inferred from the intermediate resistance which is in the order of 1 kΩ. In this case,
when the current flows from the normal to the superconducting channel, both the
Andreev reflection and normal reflection occur at the interface with a finite
probability. From the theoretical calculations, the probability of Andreev reflection
is not a monotonic function of electron energy. As the electron energy increases,
the Andreev reflection probability first increases and then decreases, and the
maximum is reached when the electron energy is equal to Δ (where 2Δ is the BCS
Summary
140
superconducting gap). In our case when we apply a perpendicular magnetic field,
the energy gap of the superconducting channel can be effectively reduced, hence
the probability of the Andreev reflection increases at the small fields due to the
decay of the superconducting gap. This manifests as a negative magnetoresistance
(MR) observed in the small field region. With further increase in the magnetic field,
the breaking of Cooper pairs dominates and a positive MR appears.
Germanane transistor (Chapter 5)
While the TMDs family represented by MoS2 shows great potential for
application, other new 2D materials are also being explored. Silicene and
germanene attract a lot of interest because they are compatible with current
silicon- and germanium-based transistors. However, silicene and germanene lack
intrinsic band gaps and are not stable in ambient conditions. These difficulties can
be overcome by hydrogen-termination on both sides of the 2D structure, yielding
silicane and germanane. As an extension of the exploration of the quantum
phenomenon in 2D materials, we studied the transport properties of germanane
(GeH) transistors. The large band gap is corroborated by the insulating state
observed in pristine GeH. Without further treatment, the conductivity cannot be
tuned by either back gate or ionic liquid gate. Thermal annealing in an inert gas
atmosphere was tested. After annealing, the conductivity is significantly improved
and the enhancement is positively correlated with the annealing temperature, with
higher temperatures generally leading to more conductive states. In particular, p-
type FET characteristics are observed with an on/off ratio of ~30 after annealing
with intermediate temperature. With high temperature annealing, a metallic state
is obtained. By comparing the resistivity and Raman spectra with other allotropes
of Ge, we believe that what we obtained after annealing is multilayer germanene.
At low temperatures, weak antilocalization is observed in the magnetoresistance of
annealed GeH, indicating the presence of strong spin-orbit interaction. These
behaviors are the first experimental observations in multilayer GeH.
Ultra-small carbon nanotubes (Appendix)
In addition to 2D materials, one-dimensional (1D) materials are regarded as
another basic unit component for the nanotechnology industry. As a typical
representative of 1D materials, carbon nanotubes (CNTs) have been intensively
studied because of their unique properties. 1D-gated 2D semiconductor field effect
transistors (1D2D-FETs) have been reported very recently. The 1D2D-FET is
composed of a MoS2 channel and a single-walled CNT gate with length ~1 nm, it
exhibits high current on/off ratios, low subthreshold swing and small leak current.
Summary
141
To achieve device applications, synthesis of desirable chirality (n,m) CNTs at a
given location with well-controlled direction and length is highly demanded.
In the first half of my PhD studies, I focused on the synthesis of ultra-small
SWCNTs in the linear channels of AlPO4-5 (AFI) zeolite crystals. The diameter 0.4
nm is confirmed by the Raman spectra and high-resolution transmission electron
spectroscopy (HRTEM). While the old method used the organic template inside
AFI crystal channels as the carbon source for growing CNTs, we developed a new
CVD method that can greatly improve the efficiency of the growth. Different
heating processes were tested to optimize the conditions for growing
SWCNTs@AFI. By using ethylene as the carbon feedstock and heating at 700 oC,
the quantity as well as the quality of SWCNTs were significantly improved. From
the Raman spectra, both the RBM and G bands are stronger compared to those
from the old process. Two observable RBM peaks at 535 and 551 cm-1 indicate that
the 0.4 nm-SWCNTs@AFI are primarily composed of (5,0) and (3,3) nanotubes,
which is in line with the theoretical calculations. Interestingly, it was found that by
using CO as the carbon stock, the SWCNTs were mainly composed of (5,0)
nanotubes. This result also provides the possibility of developing a chirality-
selective process for growing 0.4nm-SWCNTs@AFI using CO.
Outlook
With the combination of ionic liquid gating and back gating, various interesting
phenomena have been observed in dual-gate MoS2 transistors. Vertical electronic
hetero-states form at the top and bottom surface of a single few-layer MoS2 flake.
These two hetero-states can interact with each other through the anti-proximity
effect. SdH oscillations are observed in the bottom channel. By further improving
the device quality, we can explore the possibility of developing a well-defined
quantum Hall state in MoS2, which have never been observed so far. Another
interesting direction is to make use of the vertical junction, if superconductivity is
induced in the electron conduction band of MoS2 and the bottom channel is tuned
to hole conduction, a vertical p-n junction forms with the n side being
superconducting. This device configuration can be used to generate entangled
photon pairs.
While preliminary results show evidence for the production of multilayer
germanene, more characterization techniques, such as XRD, PDF and XPS, are
necessary to confirm this conclusion. Since the present samples are relatively thick,
making FETs based on ultra-thin layers of GeH and studying its thermal and
transport properties remain challenging. Furthermore, it is very interesting to
study the optical properties of annealed GeH. Due to the observed large spin-orbit
Summary
142
coupling, GeH is attractive for developing optoelectronics based on annealed thin
GeH flakes.
0.4nm-SWCNTs in the linear channels of AFI crystals provide a platform for
studying quasi-one-dimensional (1D) material systems, which have shown
interesting electronic properties. Evidence for superconductivity has been
observed in this system. According to the BCS theory, the superconducting
transition temperature Tc is given by 𝑘𝐵𝑇𝑐 = 1.13𝐸𝐷𝑒−1/𝑁(0)𝑉 , where kB is the
Boltzmann constant, ED is the Debye cutoff energy and V is the electron-phonon
coupling potential. N(0) is the density of state (DOS) at the Fermi level. It is well-
known that there are Van Hove singularities in the band structure of CNTs, where
the DOS is very large. If the Fermi level is shifted to the Van Hove singularities,
N(0) becomes very large, and therefore a high transition temperature may be
realized. With the quality of the 0.4 nm-SWCNTs@AFI significantly improved, it is
interesting to explore the electronic properties of these ultra-small nanotubes.
Combined with the EDLT technique, valuable results may be obtained.
143
Samenvatting
In de afgelopen decennia zin er veel ontwikkeling geweest in de elektronische
technologie. Op silicium gebaseerde halfgeleiders worden kleiner en sneller zoals
voorspeld in de wet van Moore. Hoewel de wet van Moore nu nog geld voor deze
transistoren is het einde van deze trend in zicht. Daarom word er veel onderzoek
gedaan naar nieuwe halfgeleider materialen en andere structuren om de wet van
Moore door te zetten. Potentiele vervangers waar veel onderzoek naar gedaan word
is de groep van 2 dimensionale materialen. Deze materialen zijn potentiele opties
om de silicium geledingskanalen in transistoren te vervangen. Een materiaalgroep
die als veel belovend word gezien door zijn elektrische en optische eigenschappen
is een structuur van gelaagde halgeleiders genaamd transitie metalen
dichalcogenides (TMDs). Een specifiek voorbeeld is MoS2 vanwege zijn bijzondere
eigenschappen zoals zijn veldeffecttransistor (FET) eigenschappen, supergeleiding
en splitsing van de bandstructuur onder invloed van een magneetveld het zo
genoemde bandstructuur Zeeman effect (valley Zeeman effect). Andere nieuwe 2D
materialen zijn voorgesteld door theoreten en later onderzocht. Een nieuwe veel
belovende kandidaat waar onderzoek naar gedaan wordt is Germanaan (GeH). Op
dit moment zijn veel van de eigenschappen van GeH nog onbekend.
Twee van de vier jaren van mijn PhD onderzoek heb ik aan de Rijksuniversiteit
van Groningen gedaan. In deze jaren heb ik de fysische eigenschappen van 2D
materialen MoS2, WS2, grafeen en GeH onderzocht. In mijn thesis focus ik alleen
op de elektrische geleiding en met name op de kwantum geleidingseigenschappen
van MoS2 en GeH. Kwantum geleiding is in de meeste studies gerelateerd aan
speciale structuren, elektronische twee-laags transistor (EDLT). In hoofdstuk 2 en
3 worden de verticale heterotoestanden van één enkele lagen dik MoS2 vlok, waarin
twee parallelle geleidingskanalen zijn vastgesteld en waarvan de interactie tussen
deze twee kanalen elektronisch kan worden aangepast door het veldeffect. In
hoofdstuk 4 worden de eigenschappen van een laterale supergeleider-normaal
metaal (SN) gevormd op de aaneenkoppeling van MoS2 met een metaal. Hoofstuk
5 behandeld de eigenschappen van germanaan vlokken. Het project in het eerste
deel van mijn onderzoek dat ik gedaan heb in HKUST bestaat uit de bestudering
van de fysische eigenschappen van enkelwandige koolstofnanobuizen, dit word
beschreven in de Appendix van deze thesis.
Verticale heterotoestanden in meerlaagse MoS2 (Hoofdstuk 2 en 3)
Samenvatting
144
In eerdere onderzoeken naar meerlaagse MoS2-EDLT is supergeleiding met
sterke 2D karakteristieken geobserveerd. Theoretische berekeningen lieten zien
dat met ionische gating de ladingsdragers alleen door de bovenste laag konden
stromen. In ons onderzoek naar de geleidingseigenschappen van een dubbele gate
MoS2 transistor op een h-BN substraat hebben we bewezen dat de geleiding door
twee parallelle kanalen gaat. Supergeleiding word opgewekt in de eerste paar lagen
van de oppervlakte van de enkele lagen dik MoS2 vlok doormiddel van de bekende
ionische gating techniek. De ladingsdrager dichtheid die is opgewekt door de
ionische gating is in de orde grote van ~1014 cm-2. Door het sterke Thomas-Fermi
afschermingseffect zijn de ladingsdragers voornamelijk beperkt tot de bovenste
laag van de MoS2 vlok. De onderste laag blijft een intrinsieke halfgeleider, dit word
beïnvloed door de back gate. Shubnikov-de Haas (SdH) oscillaties zijn
geobserveerd onder invloed van een hoge magnetisch veld nadat de supergeleiding
is onderdrukt. De ladingsdrager dichtheid die voor de SdH oscillatie zorgt kan
worden berekend uit de oscillatie periode. Dit blijkt een order van grootte kleiner
te zijn dan de laagst mogelijke ladingsdichtheid die nodig is voor supergeleiding.
Dit resultaat laat zien dat supergeleiding en SdH oscillaties ontstaan in de bovenste
en onderste oppervlakte laag, respectievelijk. Dit is onomstotelijk bewijs voor het
twee parallelle geleidingskanalen model. Er zijn twee redenen voor een hoge
ladingsbewegelijkheid van het onderste kanaal: (1) in vergelijking met SiO2, het
atomische vlakke h-BN substraat verminderd de verstrooiing door de ruwheid van
het substraat en (2) de bovenste hoog gedoteerde laag werkt als een
afschermingslaag voor de onderste laag en verbeterd daardoor de bewegelijkheid
van de ladingsdragers.
De Laundau niveau ontaarding g = 3 kan worden verkregen uit de SdH
oscillaties, dit resultaat is consistent met eerdere observaties. Een andere
belangrijke parameter is de effectieve massa, m* = 0.6 me welke is verkregen uit de
temperatuursafhankelijkheid van de SdH oscillaties. Deze twee parameters geven
belangrijke informatie over de energie niveaus van het onderste kanaal. Deze
niveaus suggereren de geleidende elektronen uit het kanaal Q/Q’ bandstructuur
komen, omdat de K/K’ bandstructuur van de hoeken van de Brilouin zone een
effectieve massa van m* = 0.5 me heeft, ook zou hier de ontaarding van het Landau
niveau g = 1 moeten zijn als het Fermi niveau zich in de K/K’ bandstructuur
bevindt. De verkregen ontaarding is gelijk aan de helft van 6 Q/Q’ bandstructuur.
Dit laat zien dat de bandstructuur ontaarding word opgeven door het magnetische
veld, zoals beschreven in het bandstructuur Zeeman effect.
De bovenste en onderste geleidingskanalen in de meerlaagse MoS2 vlok vormen
een verticale verbinding gelijkend aan de conventionele van der Waals
heteroverbinding. Structuren met nieuwe functionaliteiten zijn ontwikkeld door de
interactie tussen de twee geleidingskanalen te veranderen. Een vergelijkbare
Samenvatting
145
structuur was gemaakt voor een MoS2-EDLT op een hoge-κ diëlectrisch HfO2
substraat. De ionische gating induceert supergeleiding in de bovenste oppervlakte
laag en de back gating induceert een metallische staat in de onderste oppervlakte
laag. De HfO2 back gate kan de geleiding van het onderste kanaal meer veranderen
in vergelijking met een SiO2 back gate. Verrassend genoeg kan de supergeleiding in
het bovenste kanaal veranderd worden met andere variabelen. Voor negatieve back
gate voltages, blijft het onderste kanaal isolerend en gedraagt zich als een extra
gate diëlektrische laag voor de back gate controle van de supergeleiding in het
bovenste kanaal. Een continue supergeleiderisolator (SI) kwantum fase transitie is
gebaseerd op deze configuratie. Voor positieve back gate voltages wordt het
onderste geleidingskanaal metallisch door een elektronen accumulatie. Het sterke
anti-nabijheidseffect tussen het onderste metallische kanaal en in het bovenste
kanaal kan de supergeleiding worden onderdrukt. Een ‘bipolair’ achtig
supergeleidende transistor is gecreëerd in verschillende regimes door verschillende
mechanismes. Voor een negatieve back gate spanning is de supergeleiding niet
mogelijk door de verwijdering van ladingsdragers door het elektrisch veldeffect.
Voor positieve spanningen, word de supergeleiding onderdrukt door het anti-
nabijheidseffect tussen de supergeleidende laag en de metallische laag.
Laterale supergeleider-normaal metaal verbinding in MoS2
(Hoofdstuk 4)
Na het bestuderen van de verticale interacties in de enkele lagen dikke MoS2
vlok, focuste het onderzoek zich daarna op het laterale transport op de oppervlakte
van MoS2. We fabriceerden een MoS2-EDLT structuur, op een h-BN substraat, het
midden van de MoS2 vlok werd bedekt met een h-BN vlok. Ionische gating heeft
een zwak effect op de overgang tussen h-BN en het MoS2 oppervlak, omdat h-BN
de ionen van het MoS2 oppervlak weghoudt, vandaar dat het in de normale staat is.
De niet bedekte MoS2 oppervlakte laat door de ionische doping makkelijk
supergeleiding zien. De geleiding van ladingsdragers door de h-BN bedekte MoS2
oppervlak kunnen worden beïnvloed door back gateing, vergelijkbaar met een
klassieke transistor. Met positieve back gate spanningen, word het met h-BN
bedekte MoS2 geleidingskanaal metallisch, en een SN verbinding vormt. Uit de IV
karakterisatie blijkt dat het bedekte deel van de MoS2 vlok een hoge
bewegelijkheid heeft voor de ladingsdragers, dit laat tevens zien dat de structuur
homogeen en van goede kwaliteit is. Dit is later bewezen door de observatie van
SdH oscillaties in hoge magneet velden. Daaruit zijn de Landau niveau ontaarding
en een effectieve massa bepaalt. Deze waardes komen overeen met eerdere
observaties in het onderste geleidingskanaal van de meerlaagse MoS2 vlok.
Samenvatting
146
Door het hoge doteringsverschil tussen de onbedekte- en bedekte deel van de
MoS2 vlok vormt een energie barrière in de SN overgang. De hoogte van deze
barrière blijkt tussen een nul energie barrière en een sterke tunnel barrière te
zitten, dit blijkt uit weerstand die in de order van 1 kΩ ligt. In dit geval, wanneer er
stroom loopt tussen het normale en het supergeleidend kanaal kunnen zowel de
Andreev reflectie en de normale reflectie plaats vinden op het scheidingsvlak met
een eindige waarschijnlijkheid. Uit theoretische berekeningen blijkt dat de
waarschijnlijkheid van een Andreev reflectie geen lineaire functie van elektronen
energie is. Wanneer de elektronen energie toeneemt neemt de waarschijnlijkheid
van een Andreev reflectie eerst toe en daarna af. De maximum waarschijnlijkheid
word bereikt wanneer de elektronen energie gelijk is aan Δ (waar 2Δ gelijk is aan
een BCS supergeleiding energie bandkloof). Door het toepassen van een zwak
loodrecht magnetisch veld word de waarschijnlijkheid van een Andreev reflectie
vergroot door het kleiner worden van de energie bandkloof in de BCS. Deze kloof
manifesteert zichzelf in een negatieve magnetischeweerstand (MR) in het zwakke
magnetische veld regio. Met het sterker worden van het magnetische veld worden
de Cooper paren gesplitst wat resulteert in een positieve MR.
Germanaan transistoren (Hoofdstuk 5)
Hoewel de TMDs verzameling gerepresenteerd door MoS2 een goede potentie
heeft voor applicaties, word er naar nieuwe 2D materialen gezocht. Siliceen en
germaneen zijn interessant omdat ze goed passen in huidige op silicium en
germanium gebaseerde transtoren. Hoewel, siliceen en germaneen geen intrinsieke
bandkloof hebben en niet stabiel zijn in normale omgevingsomstandigheden. Deze
problemen kunnen worden voorkomen door waterstof-terminatie aan beide zeiden
van het 2D materiaal, resulterend in silicanaan en germanaan. Als uitbreiding op
het onderzoek naar kwantum fenomenen in 2D materialen, bestudeerden we de
transport eigenschappen van germanaan (GeH) transistoren. De grote energie
kloof is bevestigd door de isolerende toestand van ongerepte GeH. Zonder enige
behandeling kan de geleiding kan niet worden veranderd door een back gate of een
vloeibare ionische gate. Thermische behandeling in een inert gas is geprobeerd. Na
de behandeling is de elektrische geleiding significant verbeterd. De versterking van
de geleiding is positief gecorreleerd met de temperatuur van de thermische
behandeling, hogere temperaturen leiden tot een betere geleiding. In het bijzonder
de p-type FET karakteristieken zijn geobserveerd met een aan/uit verhouding van
~30, na thermische behandeling met medium temperaturen. Door een thermische
behandeling met hoge temperaturen word een metallische staat is verkregen. Bij
het vergelijken van de weerstand en Raman spectra met andere allotropen van Ge
geloven we dat wat we meerlaags germanaan na de thermische behandeling
Samenvatting
147
hebben verkregen. Op lage temperaturen is zwakke antilokalisatie waargenomen in
de magnetische weerstand van thermische bewerkte GeH. Dit gedrag zijn de eerste
experimentele observaties in meerlaags GeH.
Ultra-kleine koolstof nanobuizen (Appendix)
Naast de 2D materialen worden één dimensionale (1D) materialen gezien als
een elementaire onderdeel voor de nanotechnologie industrie. Een typisch
voorbeeld van een 1D materiaal zijn koolstofnanobuizen (CNTs).
Koolstofnanobuizen zijn veel bestudeerd vanwege hun unieke eigenschappen. 1D-
gegate 2D halfgeleider veldeffecttransistoren (1D2D-FETs) zijn recent gepubliceerd.
De 1D2D-FET bestaat uit een MoS2 kanaal en een enkelwandige CNT gate met een
lengte van ~1 nm. De transistor heeft een hoge aan/uit ratio, een lage subtreshold
verandering en een lage lek stroom. Om applicaties in elektronische appraten
mogelijk te maken is een synthese van een gewenste chiraliteit (n,m) CNTs op een
gegeven locatie met gecontroleerde oriëntatie en lengte gewenst.
In het eerste deel van mijn PhD studie heb ik besteed aan onderzoek naar de
synthese van ultra-kleine SWCNTs in lineaire kanalen van AlPO4-5 (AFI) zeoliet
kristallen. De diameter (0,4 nm) is bevestigd door gebruik te maken van Raman
spectroscopie en hoge resolutie transmissie elektronen spectroscopie (HRTEM).
De oude methode gebruikt een organisch voorbeeld in een AFI kristal kanaal als de
koolstof bron om CNTs te maken. Wij hebben een nieuwe CVD methode
ontwikkeld die de efficiëntie van de groei van CNTs verbeterd. Verschillende
verhittingsprocessen zijn getest om de groei condities te optimaliseren voor
SWCNTs@AFI. Door gebruik te maken van ethyleen als koolstof bron en te
verhitten tot 700 ˚C werden hoogwaardige SWCNTs gegroeid. Uit de Raman
spectra blijkt dat de RBM en G banden beter zijn dan in het oude proces. Twee
observeerbare RBM pieken op 535 en 5651 cm-1 laten zien dat de gesyntiseerde 0,4
nm-SWCNTs@AFI voornamelijk (5,0) en (3,3) koolstofnanobuizen zijn, dit is wat
de theorie voorspeld had. Door het gebruik van CO als koolstof bron bij de
synthese resulteert in SWCNTs die voornamelijk (5,0) zijn. Dit resultaat laat zien
dat het mogelijk is om chiraal selectieve processen voor 0,4 nm-SWCNTs@AFI
mogelijk zijn.
Vooruitzicht
Door de combinatie van een ionische vloeistof gate en een back gate zijn
meerdere interessante fenomenen geobserveerd in dubbele gate MoS2 transitoren.
Verticale elektronische heterotoestanden worden gevormd in de bovenste en
onderste laag van de enkele laag dikke MoS2 vlok. Deze twee heterotoestanden
Samenvatting
148
kunnen interacteren doormiddel van het anti-nabijheidseffect. SdH oscillaties zijn
waargenomen in het onderste kanaal. Door het verbeteren van de structuur
kwaliteit, kunnen we mogelijk een kwantum Hall toestand maken in MoS2, tot nu
toe is dit nog niet geobserveerd. Een andere interessante richting is het maken van
een verticale structuur, waarin supergeleiding is geïnduceerd in de elektron
geleidingsband van MoS2 en het onderste kanaal is gedoteerd met gaten
resulterend in een verticale p-n structuur. Waarvan de n kant supergeleidend is.
Deze configuratie kan gebruikt worden om verstrengelde photonen paren te
genereren.
Hoewel de eerste resultaten laten zien dat we meerlaags germaneen
geproduceerd hebben zal dit nog bewezen moeten worden met andere
karakterisatie technieken zoals XRD, PDF en XPS. Omdat de huidige structuren
relatief dik zijn is het maken van FETs gebaseerd ultra-dunne lagen van GeH en
het onderzoek naar thermische en geleidingseigenschappen nog erg lastig.
Daarnaast is het erg interessant om de optische eigenschappen van thermisch
behandeld GeH te onderzoeken. Door de grote geobserveerde spin-orbitaal
koppeling is GeH ook erg interessant voor de ontwikkeling van optische
elektronica.
0,4 nm-SWCNTs in lineaire kanalen van AFI kristallen zijn goede materialen
om quasi-één-dimensionale (1D) materiaal systemen met interessante
elektronische eigenschappen te onderzoeken. Bewijs voor supergeleiding is
geobserveerd in dit systeem. Uit de BCS theorie volgt dat de
supergeleidingstransitie temperatuur TC word gegeven door 𝑘B𝑇c = 1,13𝐸D𝑒−1/𝑁(0)𝑉
waar 𝑘B de Boltzman constante is, 𝐸D is de Debye cutoff energie en V is de
elektron-phonon koppeling potentiaal. 𝑁(0) is de dichtheid van toestanden (DOS)
op het Fermi niveau. Het is bekend dat er van Hove singulariteiten in de
bandstructuur van CNTs zijn, dit zijn punten waar de DOS groot is. Wanneer het
Fermi niveau word verplaats naar een Van Hove singulariteit, word de dichtheid
van toestanden 𝑁(0) groot. Dit resulteert in een hoog transitie materiaal. Met de
verbetering van de kwaliteit van de 0,4 nm-SWCNTs@AFI zijn er interessante
elektronische eigenschappen te onderzoeken van deze ultra-kleine nanobuizen.
Gecombineerd met de EDLT techniek kunnen waardevolle resultaten worden
verkregen.
149
Acknowledgements
It was not until I entered my PhD program that it occurred to me how fast time
travels, yet I am already at this point, finishing my PhD studies. When I look back
there are so many people for whom I am grateful; I could not have finished this
thesis without your help. Now is a good chance to express my gratitude to some of
you.
First of all, I would like to express my deepest gratitude for my supervisors, Prof.
Ping Sheng and Prof. Jianting Ye, for your patient guidance throughout my PhD
studies. Prof. Sheng, thank you so much for giving me the chance to study in your
group. I still remember our first call six years ago; I was very nervous but you were
so kind that everything turned out to be very smooth for me. During all our
discussions, I was always amazed by your deep understanding of fundamental
physics and your enthusiasm for scientific research. You always gave me valuable
advice when I ran into trouble. No matter how naïve my questions may have
sounded to you, you always explained them patiently to me. I enjoyed those
dinners with you during which you shared with us many interesting anecdotes
about the greatest scientists in history. I am also very appreciative of your support
– sometimes the experimental results were not positive, but you always supported
me in trying different ideas. In 2014, you encouraged me to take the opportunity to
study at the University of Groningen, which turned out to be a great experience for
me. Thank you so much for everything, and I wish the best to you!
Prof. Ye, thank you for offering me this great opportunity to join your group;
my wonderful experience at the University of Groningen will be the treasure of my
whole life. To be honest, I was a little bit worried when I first arrived here two
years ago, because I was entering a field with which I was not familiar, and I did
not know what was laying ahead of me. However, it turned out this worry was not
necessary, as your patient guidance has made things much easier for me. I am truly
astonished by your extraordinary intuition and understanding of physics as well as
the broadness of your knowledge – it seems to me that you know almost everything.
I appreciate that many times you explained things to me in detail, which inspired
me a lot. Building a new laboratory is not easy, but you have an amazing capability
to handle many things at the same time while keeping high standards for
everything. I admire your passion about scientific research. We have built many
small functional setups by ourselves – some of which are already working, some
not yet – and I hope I can finish all of them. Sometimes I feel depressed that things
did not progress as fast as we expected, as it is very difficult to reach the same level
Acknowledgements
150
or even half as much as you do. Thank you for your understanding and support; I
cannot learn enough from you. I wish the best to you and your family and hope
your two lovely daughters grow up with health and happiness.
I am very grateful to my reading committee, Prof. Beatriz Noheda Pinuaga, Prof.
Hans Hilgenkamp, Prof. Dario Daghero and Prof. Rolf Lortz. Thank you for
spending your precious time to read this thesis. During the writing, sometimes I
did not think from the readers’ perspective, so some contents were not well
organized and explained in a good manner. Thank you for pointing this out and
some changes have been made according to your suggestions. I benefit greatly
from all your valuable comments. I also sincerely thank the defense committee
members, Prof. Beatriz Noheda Pinuaga, Prof. Bart van Wees, Prof. Maxim
Mostovoy, Prof. Hans Hilgenkamp and Prof. Uli Zeitler. Thank you very much for
your willingness to be the defense committee members; especially Prof. Hans
Hilgenkamp and Prof. Uli Zeitler will travel a long distance.
Sannie, I am truly grateful for your help with many things, from registration to
student scholarships, visa applications, defense arrangements and more. You
organized things so well that everything was in good order for me.
I want to thank Dr. Wang Zhe, Dr. Shi Wu, Dr. Zheng Yuan and Dr. Zhang
Haijing. All of you have taught me many experimental skills and helped me a lot in
HKUST. Prof. Jiang Fuyi, thank you for teaching me how to grow high-quality AFI
crystals; I was so impressed by your talents in chemistry. Dr. Wang Dingdi, Dr.
Zhang Ting, Dr. Sun Mingyuan, Dr. Pan Jie and Dr. Zhang Bing, thank you for all
the useful discussions and suggestions. I learned a lot from you. Special thanks go
to Ulf, Walter, Patrick, Roy, Golden, Cai Yuan, Carrie and Jenny for the technical
support.
Prof. Xixiang Zhang, you are really an expert in the field of magnetic materials.
Thank you for hosting me in KAUST, and I really enjoyed the visit. Prof. Lai
Zhiping, thank you for the collaboration and the great crystals you provided. Dr.
Zhang Bei, Dr. Zhang Qiang, Dr. Li Peng, Dr. Zhao Lan and Dr. Yang Bai, thank
you for helping me in KAUST; I had a great time there.
I also want to thank my friends in HKUST: Ma Guancong, He Ying, Sun Ke,
Wang Xiansi, Chen Xiaolong, Guan Dongshi, Zhang Yilin, Zheng Jianying, Liu
Chang, Ding Yuhui, Shen Junying. You made my life rich and colorful in HKUST,
and I had a great time with you. Thank all my friends from the football team: Ma
Jian, Zou Xinbo, Fan Zhiyong, Li Diandian, Zhuang Fuxing, Lu Xing, Jason, Hua
Tao, Peng Siyuan, Lei Shiwei, Lei Yong, Liu Kai, Zhu Ke, Wu Zhaojun, Lv Junde,
Kalaze, Xiao Songwen, Jia Luheng, Liang Haobo, Wang Meng, etc. I enjoyed
football time on Saturday evenings.
Acknowledgements
151
Jan, thank you for all the technical work you have done for us. I was not very
familiar with the mechanical drawings in the beginning, so thank you for the
patient discussions and making perfect parts for all the projects. Ursula and
Henriet, you are really nice persons, thank you for your help with resident
permission, housing, contracts and many other things – you have made things easy
for me.
I would like to express my deep gratitude for all the colleagues in DPCM group,
Jianming, Jie, Lei, Sasha, Ali, Wytse. When I first arrived here, you helped me get
familiar with the lab and the experiment setups, and also helped me live and
survive in Groningen. Special thanks are given to Jianming and Lei for being my
paranymphs. Francesco, it is my pleasure to have worked with you, thank you for
the discussions and dinners, wish you good luck in your PhD study. Erik, you are
an excellent scientist and you learned things really fast. I also benefit greatly from
the discussions with you. Thank you for the manuscript you prepared. Hope it will
get published in a nice scientific journal.
I would like to give my special thanks to Arjon. Thank you very much for
translating the summary to Dutch, it is not an easy task even for Dutch people
because some special terminologies are difficult to translate. I can imagine how
tough it would be to translate it into Chinese. Nevertheless, you did a very nice job
and I really appreciate it. Wish you good luck in your future study!
我想感谢所有的朋友们。路师兄，跟你认识很长时间了，许多方面都得到了你的
帮助。尤其佩服你渊博的知识面和对于科学的深刻见解，有些实验上的想法是通过
跟你的讨论成形的。谢谢你的支持，我从你那里学到了很多东西。祝你和春蕊将来
一切顺利，更加成功！王喆师兄，小瑛子，海婧，丁迪，很庆幸认识了你们，也很
怀念在香港跟你们一起度过的时光，没想到在荷兰还能相聚，也让我更加珍惜这短
暂的重逢。祝你们今后事事顺利，一切如意！还有小马哥（期待在荷兰见到你），
孙坷，小思，跟你们的交谈很开心也很长见识，期待再次相见。
此外，要感谢在格罗宁根认识的很多新的朋友。梁磊和单娟，跟你们相处的时间
很长，无论实验上还是生活上都得到了许多你们的帮助，非常感谢。特别感谢梁磊
设计的论文封面，设计的非常漂亮！很羡慕你们对待事物的从容态度，把生活过得
很充实。相信你们今后的生活也一定是多姿多彩，祝你们永远幸福快乐！感谢两个
特别友好的师妹许瑾和莹芬，帮了我很多忙，祝你们学习顺利，开开心心的，少一
些烦恼。感谢宏华哥和杨洁，蹭了很多你们的饭吃，希望你们一切顺利，心想事成！
彦曦，作为同级的校友在这里相遇很有缘，我们一起做过一些有趣的实验，期待将
来能继续合作。国伟，毕业这段时间问了你很多关于答辩的问题，多谢你的耐心解
答，不然我要走很多弯路。此外要感谢很多的朋友，刘静，陈思，石朕，杨旭，邱
Acknowledgements
152
立，叶刚，张萍，家文蓓蓓和 Robin，朔哥婧祎和团团，凯哥娟娟和格格荷荷，跟你
们相处的时光很愉快，祝你们今后一切都好。
尤其要感谢在格村一起踢球的好朋友，司韵(队长)，雪哥(政委)，涛哥(刚去踢
球时的带头大哥)，国伟(多亏引路，不然现在可能还混迹于阿克楼踢野球)，超哥
(防线铁闸，鲜有失手)，家文(凌空高手)，朔哥(最稳健的中场)，石头(冲击对手防
线的重型利器)，晨玺(突袭奇兵)，李攀(重炮手)，森(盘带王)，老陈(后防中坚)，
伟立(前锋的噩梦)，范笠(剑走偏锋，防线奇才)，子旭(重型坦克)，伟哥(攻守兼
备)，敏珉(锋线尖刀)，震辰(多面手，伤缺较多)，启珉(门神，伤缺较多)，庆凯
(进攻利器)，还有去年夏天经常一起踢球的，彦曦，梅西，子云，培亮，飞哥，小
明，刘凌，李军，博群，老曹，还有很多没有提到的朋友，很享受跟大家在一起踢
球的时光，有球踢的日子是快乐的，让我有了勇气去面对生活的挑战。祝大家一切
顺利，不受伤，多踢球！
特别感谢我的几个好朋友，乐，潇，舒璨，认识你们有十几年了，跟你们在一起
的日子是最开心最难忘的。转眼你们都成家了，祝愿今后一切都好！不管走到哪里，
期待能经常相聚。
最后，要特别感谢我的家人。父亲和母亲，感谢你们的养育之恩，你们的支持和
鼓励使让我有勇气去面对一切！祝愿你们身体健康，一切如意！
Chen Qihong
2017-1-14