Electronic Transport in Atomically Thin Layered
Materials
by
Britton William Herbert Baugher
MASSACHUSETTS INGTIME
OF TECHNOLOGY
B.A. Physics, Philosophy
University of California at Santa Barbara, 2006
JUL 0 1 2014
LIBRARIES
Submitted to the Department of Physics
in partial fulfillment of the requirements for the degree of
Doctor of Philosophy
at the
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
June 2014
@ Massachusetts Institute of Technology 2014. All rights reserved.
Signature redacted
Author ......
Department of Physics
May 23, 2014
Signature redacted
Certified by...
Pablo Jarillo-Herrero
Associate Professor
/
I
Signature redacted
Thesis Supervisor
..................
Krishna Rajagopal
Chairman, Associate Department Head for Education
Accepted by ..........
Electronic Transport in Atomically Thin Layered Materials
by
Britton William Herbert Baugher
Submitted to the Department of Physics
on May 23, 2014, in partial fulfillment of the
requirements for the degree of
Doctor of Philosophy
Abstract
Electronic transport in atomically thin layered materials has been a burgeoning field
of study since the discovery of isolated single layer graphene in 2004. Graphene, a
semi-metal, has a unique gapless Dirac-like band structure at low electronic energies,
giving rise to novel physical phenomena and applications based on them. Graphene is
also light, strong, transparent, highly conductive, and flexible, making it a promising
candidate for next-generation electronics.
Graphene's success has led to a rapid expansion of the world of 2D electronics,
as researchers search for corollary materials that will also support stable, atomically
thin, crystalline structures. The family of transition metal diclialcogenides represent
some of the most exciting advances in that effort. Crucially, transition metal dichalcogenides add semiconducting elements to the world of 2D materials, enabling digital
electronics and optoelectronics. Moreover, the single layer variants of these materials
can posses a direct band gap, which greatly enhances their optical properties.
This thesis is comprised of work performed on graphene and the dichalcogenides
MoS 2 and WSe 2. Initially, we expand on the family of exciting graphene devices with
new work in the fabrication and characterization of suspended graphene nanoelectromnechanical resonators. Here we will demonstrate novel suspension techniques for
graphene devices, the ion beam etching of nanoscale patterns into suspended graphene
systems, and characterization studies of high frequency graphene nanoelectromechanical resonators that approach the GHz regime. We will then describe pioneering work
on the characterization of atomically thin transition metal dichalcogenides and the
development of electronics and optoelectronics based on those materials. We will
describe the intrinsic electronic transport properties of high quality monolayer and
bilayer MoS 2 , performing Hall measurements and demonstrating the temperature dependence of the material's resistivity, mobility, and contact resistance. And we will
present data on optoelectronic devices based on electrically tunable p-n diodes in
monolayer WSe 2 , demonstrating a photodiode, solar cell, and light enmitting diode.
Thesis Supervisor: Pablo Jarillo-Herrero
Title: Associate Professor
3
4
Acknowledgments
This thesis had a long, complex, and convoluted life, and its successful completion
was only made possible by the ever-present guidance and help of my family, friends,
and colleagues.
I would like to thank my advisor, Pablo, for his guidance and patience, my thesis
committee for their stewardship of my thesis, my group for all of their insight and
assistance, my teammates, Hugh and Yafang, Tchefor and Kevin, and Max, for all
their hard work and their direct contributions to this thesis, my family and friends
for their encouragement, and my wife, Allie, for her untiring love and support.
5
6
Contents
1
Introduction
15
1.1
M otivation . . . . . . . . . . . . .
15
1.2
Outline.
. . . . . . . . . . . . . .
17
1.3
Graphene
. . . . . . . . . . . . .
18
1.4
Dichalcogenides . . . . . . . . . .
20
2 Fabrication of Electronics Based on Atomically Thin Layered Materials
3
4
23
2.1
Introduction . . . . .
23
2.2
Graphene Suspension
24
2.3
Current Annealing
28
2.4
Ion Beam Etching.
33
2.5
Acknowledgements
36
Etching of Graphene Devices with a Helium Ion Beam
37
3.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . .
37
3.2
Motivation . . . . . . . . . . . . . . . . . . . . . . . . . .
38
3.3
R esults . . . . . . . . . . . . . . . . . . . . . . . . . . . .
39
3.4
Conclusions
. . . . . . . . . . . . . . . . . . . . . . . . .
42
3.5
Methods . . . . . . . . . . . . . . . . . . . . . . . . . . .
43
3.6
Acknowledgments . . . . . . . . . . . . . . . . . . . . . .
43
Approaching GHz Resonant Frequencies in Suspended Graphene
7
NEMS Resonators
5
6
45
. . . . . . . . . . . . . . . . . .
4.1
Introduction . . .
4.2
Motivation .
46
4.3
Device Fabrication
46
4.4
Mixing Current
49
4.5
Resonant Modes.
53
4.6
Mode Fitting
55
4.7
Conclusions . . .
58
4.8
Acknowledgments
58
.
45
Intrinsic Electronic Transport Properties of High Quality Monolayer
and Bilayer MoS 2
59
5.1
Introduction . . . . . . . . .
59
5.2
Motivation . . . . . . . . . .
60
5.3
Device Fabrication
. . . . .
61
5.4
Resistivity Measurements. .
62
5.5
Hall Measurements . . . . .
65
5.6
Conclusions . . . . . . . . .
68
Optoelectronic Devices Based on Electrically Tunable p-n Diodes in
a Monolayer Dichalcogenide
69
6.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . .
69
6.2
M otivation . . . . . . . . . . . . . . . . . . . . . . . . .
70
6.3
Device Fabrication
72
6.4
Transport in Gate Controlled p-n Junctions
. . . . . .
72
6.5
Optoelectronics . . . . . . . . . . . . . . . . . . . . . .
76
6.6
Conclusions . . . . . . . . . . . . . . . . . . . . . . . .
80
6.7
M ethods . . . . . . . . . . . . . . . . . . . . . . . . . .
81
A MoS
2
. . . . . . . . . . . . . . . . . . . .
Supplementary Information
A. 1 Fabrication .......
................................
8
83
83
A.2
Metal Insulator Transition
. . . . . . . . . . . . . . . . . . . . . . . .
86
A.3 Hall Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . .
86
A.4 Mobility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
87
A.5 Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
89
B WSe 2 Supplementary Information
B.1 Device Fabricatioi
91
. . . . . . . . . . . . . . . . . . . . . . . . . . . .
91
B.2 M id-Gap Current . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
94
B.3 Schottky Barriers . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
95
B.4 Acknowledgements
99
. . . . . . . . . . . . . . . . . . . . . . . . . . . .
9
10
List of Figures
1-1
Atomically layered materials with diverse electronic and optoelectronic
prop erties
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
17
1-2
Schematic and electronic band structure of graphene
. . . . . . . . .
18
1-3
Overview of a TMD crystal structure and band structure . . . . . . .
20
2-1
Suspended graphene structures
24
2-2
Current annealing of suspended graphene devices
2-3
Etching suspended graphene with a gallium based focused ion beam
(FIB) and a helium ion beam
. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . .
30
. . . . . . . . . . . . . . . . . . . . . .
33
2-4
Etching suspended graphene by electron beam assisted water etching
36
3-1
Schematic of a graphene device
38
3-2
Etching suspended graphene with a helium ion bean
3-3
Electrically isolating suspended graphene with a helium ion beam
. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . .
40
. .
41
3-4
Etching graphene on SiO 2 with a helium ion beam . . . . . . . . . . .
42
4-1
Graphene NEMS device schematics, image, and electrical readout
48
4-2
Resonance maps, avoided crossings, quality factor, and high frequency
m o d es
. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
53
4-3
Fitting resonance modes and mode softening . . . . . . . . . . . . . .
57
5-1
MOS2 device schematics, images, and two-ternminmal transport measure-
m en ts
5-2
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
60
Contact resistance and four-terminal resistivity of monolayer and bilayer M oS 2
. . . .
. . . . . .
. . .
11
. . .
. .
. . .
. . .
. . . .
. .
. .
.
64
5-3
Field-effect and Hall mobilities as a function of back-gate voltage and
temperature for monolayer and bilayer MoS 2 . . . . . . . . . . . . . .
66
6-1
Gate-controlled monolayer WSe 2 p-n junction diodes
. . . . . . . . .
71
6-2
Current through the device as a function of doping configuration . . .
75
6-3
Photodetection in monolayer WSe 2
. . . . . . . . . . . . . . . . . . .
77
6-4
Photovoltaic response and light emission . . . . . . . . . . . . . . . .
79
A-i MoS
flake AFM and step heights . . . . . . . . . . . . . . . . . . . .
84
A-2 Effect of annealing on two-terminal resistance of MoS 2 devices . . . .
85
. . . . . . . . . . . . . . . . . . . . . . .
87
2
A-3 Hall measurements of MoS
2
A-4 Contact resistance and four-terminal resistivity of monolayer device Mi 88
A-5 Room temperature mobilities and leakage current
. . . . . . . . . . .
88
B-1 Optical micrograph and AFM with step height of the main device . .
92
B-2 Optical micrograph the EL device . . . . . . . . . . . . . . . . . . . .
93
B-3 Temperature dependence of mid-gap current . . . . . . . . . . . . . .
94
B-4 Photocurrent and reflected light image . . . . . . . . . . . . . . . . .
97
B-5 Dependence of photovoltaic power generation on laser power . . . . .
98
B-6 Reflected and emitted light image . . . . . . . . . . . . . . . . . . . .
98
12
List of Tables
A. 1 Device Fabrication Parameters . . . . . . . . . . . . . . . . . . . . . .
84
A.2 Metal Insulator Transition . . . . . . . . . . . . . . . . . . . . . . . .
86
13
14
Chapter 1
Introduction
1.1
Motivation
Electronic transport in atomically thin layered materials has been a burgeoning field
of study since the discovery of isolated single layer graphene in 2004 [1]. Single layer
graphene is a sheet of carbon atoms bonded together in a flat, hexagonal, honeycomb lattice, exactly one atom thick. The inherently two dimensional (2D) nature of
graphene, as a one atom thick material, and its highly symmetric honeycomb lattice
combine to give it a myriad of new and interesting properties. Graphene is light [2],
strong [3], transparent [4], highly conductive [5], and flexible [3].
It has a unique
gapless Dirac-like band structure at low electronic energies [6, 7, 8, 9] (see Fig. 1-la).
It has the one of the highest thermal conductivities of any know material [10]. And it
has a Young's modulus five tinmes greater than that of steel [3]. Already, since its discovery, many novel and exciting applications have been realized utilizing graphene's
unique properties [11, 12, 13].
This thesis will expand on the family of exciting graphene devices with new work
in the fabrication and characterization of suspended graphene nanoelectromechanical resonators.
Here we will demonstrate novel suspension techniques for graphene
devices, the ion beam etching of nailoscale patterns into suspended graphene systemus, and characterization studies of high frequency graphene nianoelectromechanical
resonators that approach the GHz regime.
15
Graphene's success has led to a rapid expansion of the world of 2D electronics, as
researchers search for corollary materials that will also support stable, atomically thin,
crystalline structures. Graphene, a semi-metal, is limited in the scope of electronic
device applications it can realize. Insulators and semiconductors are needed to complement graphene and to broaden the capabilities of 2D atomically layered materials
(ALMs) into a fuller range of electronic devices. Hexagonal boron nitride, an insulator
that can be exfoliated down to a single atomic sheet, just as graphene can, was one
of the first ALMs to be added into this growing class of materials [14] (see Fig. 1-1b).
Structurally, it is very similar to graphene, with the carbon atoms replaced by alternating atoms of boron and nitrogen. This alternating structure breaks the sublattice
symmetry of graphene, creating a large band gap of 5.5 eV. The ultra-flat surface,
high crystallinity, inertness, and atomic layer control over thickness conspire to make
hexagonal boron nitride an outstanding dielectric [15, 14]. More recently, transition
metal dichalcogenides (TMDs) were isolated in atomically thin layers, adding semiconducting elements to the world of 2D materials. Semiconducting TMDs, such as
MoS
2
and WSe 2 , are made of three atomic sheets: two sheets of chalcogen atoms,
such as sulfur or selenium, on the top and bottom, and one sheet of transition metal
atoms, such as molybdenum or tungsten, in the middle. This alternating structure
of atoms breaks the sublattice symmetry seen in graphene, creating a band gap, and
enabling TMDs' use in digital electronics and optoelectronics. What is more, the single layer variants of these materials posses a direct band gap, which greatly enhances
their optical properties verses the indirect gap of the bulk (see Fig. 1-1c).
This thesis provides pioneering work into the characterization of these new atomically thin TMDs and into the development TMD based electronics and optoelectronics. First, we will describe the intrinsic electronic transport properties of high quality
monolayer and bilayer MoS 2 , performing Hall measurements and demonstrating the
temperature dependence of the material's resistivity, mobility, and contact resistance.
We will then present data on optoelectronic devices based on electrically tunable p-n
diodes in monolayer WSe 2 , demonstrating a photodiode, solar cell, and light emitting
diode (LED).
16
Semi-Metal:
Graphene
(a)
(b)
Insulator:
Boron Nitride
(c)
Semiconductor:
Dichalcogenide
E
kX
Figure 1-1: Atomically layered materials with diverse electronic and optoelectronic
properties. (a) Graphene, a semi-metal, can serve as a high-mobility conductor and
transparent gate electrode. (b) Hexagonal boron nitride is an ultra-flat insulator useful as a gate dielectric with high breakdown voltage. (c) Transition metal dichalcogenides are semiconductors with a direct band gap in their monolayer forms. They
also posses a valley degeneracy and a spin-split valence band.
1.2
Outline
Here we present a brief introduction to the world of atomically layered materials
(ALMs) and their properties. We will first describe graphene and both its electrical
and mechanical properties as they pertain to suspended resonators.
Next we will
describe the electrical and optical properties of the transition metal dichalcogenides
MoS 2 and WSe 2 . Chapter 2 will discuss device fabrication, specifically emphasizing
graphene suspension techniques, current annealing suspended graphene, and the etching of suspended graphene devices with ion beams. In chapter 3 we will demonstrate
the electrical isolation of suspended graphene by etching it with a helium ion beam.
In chapter 4 we present a study of suspended graphene resonators with resonant frequencies approaching the GHz regime. Chapter 5 elucidates the intrinsic electronic
transport properties of high-quality monolayer and bilayer MoS 2 .
And chapter 6
demonstrates optoelectronic devices based on tunable p-n diodes in monolayer WSe 2 .
17
1.3
Graphene
(a)
(b)
Zr
1
0.5
JK<r
_
-0.5
-1
.
-V
-0.5
-1
Graphene Lattice
(c)
0
0.5
1
kx (T/a)
(d)
3
0.1
1.5
0
0
0
-1.5
-3
-0.1
0
Figure 1-2: Schematic and electronic band structure of graphene. (a) Schematic of
the 2D hexagonal honeycomb lattice of graphene. (b) An energy contour plot of
graphene's conduction band. Energy increases from dark to light. Minima occur at
the K and K' points. (c) A construct of the full band structure of graphene. (d) A
zoomed in view of the band structure near one of the K points. At low energies the
band structure is well described by a Dirac cone. The valence and conduction bands
can be seen to meet at the K-points, creating a gapless material.
Graphene, a one atom thick sheet of graphite, is a 2D array of carbon atoms arranged
in a honeycomb lattice (Fig. 1-2a). The excitement surrounding its electronic prop18
erties stems from its peculiar band structure (Fig. 1-2b and c) [1, 9, 16, 14], which
at low energies can be approximated by two non-equivalent conical valleys where the
charge carrier dynamics are described by the Dirac equation (Fig. 1-2d) [6, 8].
The
symmetric electron arid hole valleys result in charge carriers with linear energy dispersion up to -1eV
and a vanishing band gap energy. This peculiar band structure
naturally leads to exotic electronic behavior [17, 18, 19, 20, 21, 22].
In addition, single layer graphene has been shown to exhibit unmatched mechanical properties.
Recent studies have shown graphene to posses the highest Young's
modulus of any existing material at 1 TPa, a third order elastic stiffness of -2 TPa,
and an intrinsic stress of 130 GPa [3].
Coupled with its low density, graphene's
superior mechanical behavior makes it an ideal material for nanoelectromechanical
systems (NEMS), with properties which compare favorably to those of the current
state of the art NEMS [23, 24, 25, 26]. Moreover, suspended graphene has a high thermal conductivity of ~4000 W/n-K (exceeding that of diamond and graphite)
[10].
This means that graphene resonators may have substantially reduced heat dissipation problems compared to conventional NEMS. Additionally, due to its light mass,
graphene resonators can be designed for high frequency applications.
This can be
achieved, for example, by etching arid suspending subrmicron-sized graphene devices
from exfoliated graphene flakes. Such high frequency resonators can be used as ultrasensitive mass, charge, arid chemical sensors. Any extra weight/charge added to such
a resonator will appreciably change its mass/doping, arid thus its electromechanical
response.
In this thesis we investigate the electromiechanical properties of suspended graphene
based nanodevices, probing the coupling between electronic [1, 9, 16] and mechanical [27, 28, 29, 3, 30, 31] degrees of freedom. We fabricate and characterize ultra-high
quality devices, demonstrating the mechanical motion of graphene based resonators
and probing their electromechanical coupling.
The tunability of these resonators
prove very useful to explore their potential as ultra-sensitive mass, charge, arid chemical sensors. By changing geometry arid size, graphene resonators with varied resonance frequencies are fabricated.
These experiments set the stage for further work
19
aimed at cooling a graphene resonator mechanical mode to its quantum mechanical
ground state. We build the foundation for the integration of the excellent properties
of graphene into next generation nanoelectromechanical systems and sensors.
1.4
Dichalcogenides
(a) side view
(b)
top view
(c)
band structure
K'
Figure 1-3: Overview of a TMD crystal structure and band structure. (a) The structure of a bulk TMD, such as MoS 2 or WSe 2 , consists of layered triangular prisms of a
transition metal atoms (grey) and chalcogen atoms (yellow). (b) Top view of a monolayer TMD. (c) Schematic band structure of MoS 2 near the K points that are relevant
to transport and optical properties at low energy. The hexagonal network of Mo and
S atoms gives rise to K and K' valleys as in graphene, but because the structure lacks
inversion symmetry, TMDs acquire a band gap and spin-orbit coupling.
The most recent additions to the toolbox of ALMs are the transition metal dichalcogenides, a class of compounds of the form MX 2 , where M is a transition metal, such
as molybdenum or tungsten, and X is a member of the chalcogen, or oxygen, group
of the periodic table, such as sulfur or selenium. The TMDs display an astonishing
variety of physical properties, including metals (NbS 2 and VSe 2 ), semi-metals (WTe2
and TcTe2 ), semiconductors (MoS 2 , WSe 2 , and PdS2 ), and superconductors (NbSe
2
and TaSe 2 ) [32]. The most common TMDs have a layered structure in which the
transition metal atom is sandwiched between two layers of chalcogen atoms forming
a triangular prism (Fig. 1-3a). Viewed from above, each X-M-X monolayer forms a
20
hexagonal lattice with alternating M and X atoms (Fig. 1-3b). The band structure of
layered TMDs is similar to that of hexagonal boron nitride, but with a smaller band
gap and strong spin-orbit coupling (Fig. 1-3c). Multi-layer TMDs have an indirect
band gap with the valence band maximum at the F point. However, the gap changes
to a direct band gap for monolayers, with the direct gap at the two inequivalent valleys, K and K' [33, 34, 35]. This indirect-to-direct transition between multi-layer and
imonolayer TMDs significantly enhances the optical properties of monolayer TMDs
relative to their bulk forms [35].
MoS 2 and WSe 2 have been studied in bulk form for several decades [32], but
single atomic layers have only been isolated in the past few years with the advent of
the mechanical exfoliation technique used to isolate graphene [1].
Initial transport
and optoelectronic measurements on few-layer MoS 2 , MoSe 2 , WS2 , and WSe 2 have
been done by several groups [36, 37, 38, 39, 40, 41], and field effect mobilities up to
200 cm 2 /Vs for monolayers [36] and 700 cm 2 /Vs for thicker films [42, 43] have been
demonstrated. These early results show that TMDs are promising materials with all
opportunity to be combined with graphene and hexagonal boron nitride to fabricate
high quality electronic TMD devices.
We expect that combining the strengths of
different ALMs will also be a vital rianotechnology for photonic and optoelectronic
applications.
In this thesis we fabricate and characterize ultrahigh quality TMD devices from
MoS 2 and WSe 2 . The device technologies demonstrated here integrate ALMs into
electronic devices to probe the physics and applications of this family of transparent and flexible materials.
This thesis develops the foundation of an ALM-based
architecture for next generation electronics and optoelectronics.
21
22
Chapter 2
Fabrication of Electronics Based on
Atomically Thin Layered Materials
2.1
Introduction
Here we present studies in the fabrication of nanoscale electronics based on atomically thin layered materials. We discuss various methods of suspending graphene
structures including: suspending graphene films on transmission electron microscope
(TEM) grids, deposition of flakes over trenches, transferring flakes over contacts, and
etching out the substrate from below. We then describe three techniques for current
annealing suspended graphene, and present transport data showing the improvements
in conductivity and mobility that can be achieved. Finally, we demonstrate the patternless etching of suspended graphene using a gallium based focused ion beam, a
helium ion beam, and electron beam assisted water etching.
23
2.2
Graphene Suspension
(a)
8
(b)
10
-4
-8
(c)
-8
-4
-10
0
X (pm)
4
0
4
8
-10
0
X (pm)
10
5 nm (d)
3O0000
-2
-4 '
-4
-2
0
X (pm)
2
4
0
1
2
X (pm)
3
4
Figure 2-1: Suspended graphene structures. (a) An SEM image of a CVD grown
graphene film suspended over an amorphous carbon sheet with 2.5 Pm holes in it.
The amorphous carbon sheet is supported by a TEM grid. Inset: A zoomed out SEM
image of the sample showing the gold TEM grid. (b) an optical image of a graphene
flake deposited on SiO 2 with 1.5 pm holes pre-etched in it. inset: An SEM image
of one hole partially covered by the graphene flake, showing the flake is suspended.
(c) AFM image of a graphene flake transferred over gold contacts. Inset: A zoomed
out optical image of the same device. (d). An SEM image of a device suspended by
etching out the SiO 4 substrate from below it. Insets: A zoomed out optical image of
the device is shown in the upper left. The resistivity and mobility of the device, both
before (blue) and after (yellow) suspension are shown in the upper right.
24
Many promising applications for devices based on suspended graphene have been
proposed in the literature: from resonators and mass sensors to nano-sieves and pressuremeters [30, 2, 5, 27, 44, 28, 29]. Suspending the graphene takes unique advantage
of graphene's mechanical, optical, and electrical properties to form devices with such
new and interesting functionality. Here we investigate four different ways to fabricate suspended graphene devices and compare their results. We suspend graphene by
transferring a chemical vapor deposition (CVD) grown graphene film onto a transmission electron microscope (TEM) grid; we directly deposit exfoliated graphene onto
pre-patterned holes etched in an SiO
2
substrate; we transfer exfoliated flakes on top
of gold electrodes; and we etch out the SiO 2 substrate from below contacted graphene
flakes.
For strictly mechanical or optical devices, where electrical contact is not necessary, it is possible to make a large number of suspended graphene membranes by
transferring a CVD grown graphene film onto a TEM grid. For this method we use a
gold TEM grid with a lattice of 100 pim wide squares that is covered in an amorphous
carbon film with 2.5 pim holes in it. Such TEM grids are commercially available from
suppliers such as Structure Probe Instruments.
The graphene film is grown in a CVD oven on a copper substrate. The filn is
gently released from the copper below with an acid etch, leaving the graphene to float
on the surface of the liquid. We then scoop the film out of the liquid using the TEM
grid, thus suspending the film over the grid. A scanning electron microscope (SEM)
image of graphene suspended over a TEM grid is shown in Fig. 2-la. And a zoomed
out view of the TEM grid, with the graphene film half covering the amorphous carbon,
can be seen in the inset to that panel.
This technique produces a large quantity of suspended graphene membranes very
quickly and easily. It is ideal for testing further processing or measurement schemes,
or obtaining large number statistics. It is also, quite naturally, ideal for TEM studies.
The drawbacks are that it the TEM grid is too delicate for adding electrical contact,
CVD graphene is inherently lower quality than exfoliated graphene, many membranes
will be rippled, torn, or broken, and there can be significant residue left on the film
25
that is difficult to remove. For our purposes, however, these devices offer an excellent
test bed for etching experiments using ion beams.
A second suspension method is to directly deposit exfoliated graphene flakes onto
pre-patterned holes etched in an SiO 2 substrate. For this method we use electron
beam lithography to pattern holes in the oxide ranging from 1 Pm to 5 Pm in diameter. We then etch out these holes using reactive ion etching. After cleaning the
substrate, we deposit exfoliated graphene and scan the wafer under an optical microscope, searching for graphene that happens to cover the holes. An optical image of
a graphene flake covering a few holes is shown in Fig. 2-1b and an SEM image of a
half covered hole is shown in the inset.
This technique provides us with high quality single crystal graphene sheets that
are accessible to mechanical, optical, and electrical measurements.
So long as the
graphene flake fully covers the hole, the membrane can survive resist spinning, lithography, and metal evaporation to make electrical contact. Unfortunately, this is a very
low yield process, as the graphene sticks poorly to the etched substrate and it relies on
a significant amount of chance. It also cannot provide unadulterated electrical access
to the suspended graphene, as there will always be at least a small portion of the
graphene between contacts on the substrate in parallel with the suspended section.
To achieve unfettered electrical access to a suspended graphene sheet, we developed a method to transfer high quality exfoliated graphene flakes onto predefined gold
contacts. Our method for transferring graphene onto various substrates was developed by the Jarillo-Herrero group based on a similar technique pioneered at Columbia
University [14].
The transfer process begins by depositing exfoliated graphite onto a glass slide
covered in a polyvinyl acetate (PVA)- polymethyl methacrylate (PMMA) resist stack.
The slide is also framed with a thick edge of tape that serves as a structural support
for the resist membrane when the glass slide is removed.
The glass slide can be
removed leaving the resist membrane intact thanks to the weak structural strength of
the sacrificial PVA layer. Once separated the membrane is scanned upside down for
candidate graphene flakes under an optical microscope. When a suitable flake is found,
26
a small metal washer (~
2 mm in diameter) is placed on the back of the membrane
surrounding it. When the membrane is fully scanned, a new piece of thick tape is
placed over the back side, sealing the washers in place against the membrane. Each
individual tape-washer-membrane stack can then be carefully cut out for transfer.
The actual transfer utilizes a flip chip bonder, which aligns a substrate and a
transfer object, and then flips the transfer object over onto the substrate in a very
precise motion. With this method, we could regularly achieve an alignment accuracy
of a few microns. The transfer process has since been improved for simplicity, reliability, and higher yield. It is now possible to align transfers within a few hundred
nianometers, and to do so with greater yield.
For these devices the graphene is transferred onto 100 nm tall gold electrodes.
The electrodes were designed with various gaps between them, ranging from 200 nm
to 1 pin. The graphene flake bridges the gaps between the electrodes and remains
suspended. An AFM image of a suspended graphene flake prepared in this manner
is shown in Fig. 2-1c. A zoomed out optical image of the same device is presented in
the inset.
Using this suspension method, electrical contact is made to a suspended device
without any of the flake coming contact with the substrate. This method also reduces
the number of fabrication steps that the graphene sees, possibly reducing contamination, as the graphene transfer is the final step in the process.
There are significant drawbacks to this method, however. First, the transfer process can be relatively low yield. It is much more difficult to find quality single layer
graphene flakes on the transfer slide than it is to find them on an SiO 2 substrate.
Transfer can also fail by misalignment, failure of the PMMA to stick to the substrate, or device collapse or tearing. Additionally, even when a device is successfully
fabricated, it usually has poor electrical contact to the gold. This can be fixed by
annealing the device is forming gas at 250 C. While annealing can lead to very low
contact resistance, it tends to cause rippling in the graphene due to the mismatch
between the thermal expansion coefficients of graphene and gold. These ripples, as
can be seen in the device shown in Fig. 2-1c, can dramatically effect the device's
27
electrical and mechanical properties.
Finally, we show data from devices that are suspended by etching the substrate
out from below a contacted graphene flake. To build such a device, we begin by
exfoliating graphite and depositing it onto an SiO 2 substrate. We scan the wafer
for monolayer flakes, then design, pattern, and evaporate contacts onto them, using
e-beam lithography and thermal evaporation of gold. With the devices held securely
by the gold, we etch out the SiO 2 substrate from below using hydrofloric acid (HF).
Critically the HF will wet through the graphene - SiO 2 boundary but not through
the gold - SiO 2 boundary [5]. This means that the HF will etch evenly below the
graphene surface, freeing the device, but will not etch below the gold contacts, leaving
them as structural support for the graphene membranes. An SEM image of a flake
suspended in this manner is shown in Fig. 2-1d. In this image it can be seen that
the substrate has been evenly removed everywhere except tinder the gold contacts,
where pillars of SiO 2 remain as supporting structure. An optical image of the device
is shown in the inset in the upper left of the panel.
We have found this suspension method to be the most successful for generating
high quality suspended graphene devices with good electrical and mechanical properties. All of the devices described in the graphene nanoelectromechanical resonator
chapter were prepared in this manner. A resistivity plot is presented in the inset to
Fig. 2-1d, showing the Dirac peak of the device in the main panel both before (blue)
and after (yellow) suspension. The mobility of the suspended device is also plotted
within the inset showing a peak above 30,00 cm 2 /Vs. This demonstrates the very
high quality results that can be achieved with successfully suspended graphene.
2.3
Current Annealing
After fabrication of graphene devices, a significant quantity of residue often remains.
Standard graphene fabrication usually involves some kind lithographic resist polymer
(e.g. PMMA, MMA, or SU-8) that will leave contamination behind after processing.
Because graphene, as a one atom thick material, has such a severe surface to volume
28
ratio, it is very susceptible to outside contamination. In these situations, graphene's
robust thermal stability can be taken advantage of to remove the residue.
One way to do this is to anneal the device in an oven with forming gas.
This
will successfully remove most residue. Thermal annealing, however, can be limited
by the other materials that are around the graphene such as contacts, the substrate,
or large amounts of residue in the surroundings. These other materials may limit the
temperatures that can be used or the cleanliness of the environment. For instance,
it is best to keep temperatures lower than 250 C when annealing devices with gold
contacts to reduce migration of the gold. Ideally, only the graphene would be heated
and nothing around it. It would also be best to do this process in high vacuum and
in the final measurement apparatus where the device will be used. Current annealing
is a method that possesses all of these criteria.
Current annealing removes residue by sending a large current through the device.
The graphene's resistance converts the electrical energy into heat, burning off contamination on its surface. The graphene will heat very quickly, due to its minuscule
mass, and can survive very high temperatures, due to its thermal stability. These
properties allow it to survive what can often be a violent process.
There are three main categories of annealing techniques that are used to clean
graphene by various groups: pulse annealing, hold annealing, and wave annealing [45,
46, 47].
In pulse annealing, short pulses of voltage are delivered to the sample driving
bursts of current through the device. The sample is then given time to cool before
the next pulse. The pulses begin at a low magnitude and are gradually increased. A
schematic of this process is shown at the top of Fig. 2-2a in light blue. In theory, the
short pulses concentrate the heating to the device itself and reduce excess heating of
the contacts or the graphene-contact interface, where over heating may reduce device
performance.
In practice, it is quite easy to over step and deliver a fatal current
spike to the device. The line between a sufficiently powerful pulse that will remove
residue, and one that will destroy the graphene is quite narrow. Pulse annealing may
be satisfactory on substrate, but is too violent for reliable use on suspended devices.
29
(a)
Hold
Wave
0
1
2
3
4
5
6
7
8
9
t (min)
(b)
(c)
60
40
-10
200,000
40,000
100,000 E
40
0
0
10
0
-10
0 5
20-
20-
S01
-20
60
80,000
-10
0
Vbg
10
M
-20
-10
0
10
Vbg (V)
Figure 2-2: Current annealing of suspended graphene devices. (a) Schematic time
traces of three different current annealing techniques. At the top, in light blue, pulse
annealing sends short voltage pulses to drive current through the device.
In the
middle, in dark blue, hold annealing ramps the voltage up and holds it constant
for an extended period of time. At the bottom, in purple, wave annealing ramps the
voltage up and back down with one polarity and then does the same with the opposite
polarity. (b) and (c) The evolution of the conductivity of suspended graphene devices
being current annealed by wave annealing. Early traces are in the lightest blue and
progress to later traces in the darkest blue. The final measurement is in yellow.
Insets: Mobility data corresponding to the conductivity measurements shown in the
main panels. The color coding is the same for the insets as it is in the main panels.
30
For hold annealing, a source voltage is slowly ramped up to a desired value and
then held there for an extended period of time, driving a current through the device.
The voltage is then ramped back to zero for a resting period, and the processes is
repeated at increasing voltage levels. A schematic of hold annealing is shown in Fig. 22a in dark blue. This method has the benefit of increasing the user control during
the process. By monitoring the current through the device as the source voltage is
ramped, the user can watch for signs that residue is being removed. As contamination
is ejected, the current will begin to decrease as the graphene becomes more intrinsic
and its Dirac point moves closer to zero. When this signal is seen, the user can stop
ramping and hold the voltage at that value until the current stabilizes. The user can
then repeat the cycle at high voltages.
This also gives the user (or an automated
script) control to stop the annealing process if they see a jump in current. For some
spikes, if they are mild enough and caught quickly enough, cutting the current can
save the device from destruction.
This method does work fairly well on suspended graphene samples, though it can
be very labor intensive. It is difficult to fully automate the procedure, as each device
will have slightly different thresholds and requirements. Therefore, a user must be
present for much of the process, which can take hours depending on the results desired
and the risks the operator is willing to take with their device. Ultimately, however,
it is simply not the best way to anneal a sample. Wave annealing improves upon the
hold annealing method to produce better results.
In wave annealing, the source voltage is ramped slowly up to a given value and
then back down. It then performs the same ramps with the opposite polarity. It
repeats this cycle with increasing power until the desired annealing is achieved. A
schematic of this process is shown in the lower part of Fig. 2-2a in purple. This method
is gentle, like hold annealing, but requires much less monitoring, as it does not hold a
large voltage for a long period of tine. Wave annealing also has the additional feature
of reversing the polarity of the annealing current. Reversing the current significantly
improves the annealing process, possibly due to residue migration, uneven heating
near the contacts, or other general asymmetries of the device.
31
We were able to write an automatic script to run this annealing process with fairly
high yield. After each cycle the program would take a trace of the conductivity of the
sample verse back gate voltage and compare its performance to that of the previous
round. Based on the comparison, the script would choose the amount to increase
the maximal source voltage or whether the annealing was complete.
Progressions
of conductivity traces are plotted in Figs. 2-2b and c. The corresponding mobilities
for these traces is shown in the insets of these panels. The lightest blue traces are
the earliest measurements and the darkest blue are the latest. The yellow trace is
the final measurement. As the figures show, it can take many iterations to complete
an annealing process. In these cases, the annealing took 105 cycles for b and 146
cycles for c. These figures also demonstrate how gains can often come in jumps. It is
common that little will happen for many cycles, and then the mobility can double in
a single round.
32
2.4
Ion Beam Etching
(a) 1
(b)
4
3
P
0
2
1
0
-1.6 M
-1.6
1.6
0
x (pm)
0
E
C
0
-250
-500
-500
4
3
10
250
S
2
x (Pm)
(d)
(c) 500
E
1
0
-10
-250
0
X (nm)
250
500
-10
0
X (nm)
10
Figure 2-3: Etching suspended graphene with a gallium based focused ion beam (FIB)
and a helium ion beam.
(a) Four SEM images of suspended graphene structures
patterned by FIB. (b) SEM image of a contacted, suspended graphene device etched
into a nanoribbon with FIB. (c) An SEM image of a suspended graphene flake etched
into a nanoribbon with a helium ion beam. (d) Transmission electron microscope
(TEM) image of the edge of a suspended graphene film after etching by a helium
ion beam. Inset: A zoomed out TEM image of the slots etched in the suspended
graphene filn.
33
Suspended graphene devices can be of supreme quality due to their separation from
any substrate or contamination, and their ability to be annealed to pristine levels
of cleanliness. Devices to date, however, have been limited in size and shape due
to the strains of the suspension process. It would be beneficial to have a process
for shaping graphene membranes that are already suspended. In this way, a robust
and structurally sound membrane can be suspended safely, and then gently etched
into new and interesting shapes and structures, in a clean, dry etching process. Such
etching would also avoid problems of hardened residue from reactive ion etching and
may be able to create cleaner, straighter etch edges. To investigate the possibilities
of this type of patterning, we tested three post suspension etching techniques. First
we used a standard gallium based focused ion beam (FIB). Second we used a helium
ion beam, a fairly new technology. And finally, we tested the etching of suspended
graphene with electron beam assisted water etching.
Gallium based focused ion beams are the most common type of FIB commercially
available. They are often found in dual beam configurations, with a complementary
SEM that can be used for aligning and imaging of the sample without the destructive
effects of using the FIB. The FIB can be programmed to deliver specific doses in a
precise pattern to an aligned sample. We use this capability to etch various patterns
in suspended graphene samples that would be impossible to etch before suspension.
Fig. 2-3a depicts four examples of such structures. In the upper left there is an SEM
image of a hexagonal drum head supported by nanoribbons smaller than 100 nm in
width. In the upper right is a suspended graphene membrane with the MIT logo
etched into it. The nanoribbons that define the logo are as small as 40 nm. In the
lower left, a square perimeter is etched into the membrane with the central square
shown collapsed on the bottom of the trench. Finally, in the lower right, there is
a spiral pattern showing a spiral etched nanoribbon of ~ 100 nm in width. These
shapes and structures demonstrate the range of devices that can be fabricated in this
method.
The FIB can also be used to etch electrically contacted suspended graphene devices. Fig. 2-3b shows an SEM image of a device after being etched into a ~ 140
34
nm wide nanoribbon. Though the etching was successful, the electrical properties
greatly diminished after the process.
Most likely this is due to gallium implanta-
tion or nanometer scale destruction of the nanoribbon by the outer ions of the beam
waist. It seems that etching with a gallium based focused ion beam is best suited for
patterning larger scale optical or mechanical structures that would be too delicate to
suspend after etching.
The helium ion microscope is a relatively new technology that was only developed
commercially in the last few years.
This microscope uses helium ions in place of
gallium, taking advantage of their lighter mass to produce higher resolution imaging
and more delicate etching with less implantation and contamination. Fig. 2-3c shows
a suspended graphene nanoribbon that has been successfully etched down to ~ 25
nim wide over a 500 nm length for an aspect ratio of 20:1.
We further characterize the helium ion etching process by observing the edge
structure left behind after etching slits in a suspended graphene film. A high resolution transmission electron microscope (TEM) image of such an edge is provided
in Fig. 2-3d. The edge is shown to be very clean and well formed with an average
roughness of only about 2-3 nmi.
A zoomed out image of the slits used in this TEM
study is shown in the inset. This work shows the promise of helium ion etching of
graphene films as it is a more gentle, higher resolution option compared to standard
gallium based focused ion beams. As the technology advances and the commercial
machines become more reliable and easier to use, many new and exciting, previously
unrealizable, devices may come out.
As a final test, we investigated electron beam assisted water etching of graphene
membranes. This process uses a e-bean and a gas injection system to introduce a
small amount of water vapor into the SEM environment at the graphene flake. Where
the electron beam hits the surface, it locally ionizes the water and the free oxygen
selectively etches the carbon of the graphene. This is a well established process for
etching graphitic films [48], but is a novel technique for use oii suspended graphene.
A nano-constriction etched using this process is shown in Fig. 2-4.
The process
has some rastering issues as can be seen in the jaggedness of the cut. Because of
this the resulting constriction was not as small as designed. However, the process
did ultimately succeed in etching the structure into a suspended graphene sample.
E-beam assisted water etching shows real merit and warrants further investigation,
especially for those without access to a focused ion beam. We believe many interesting
structures can be fabricated in this manner.
3
2
0
0
2
1
3
X (pm)
Figure 2-4: Etching suspended graphene by electron beam assisted water etching. An
SEM image of a suspended graphene film with a nano-constriction etched into it by
electron beam assisted water etching. The jagged edges are due to rastering problems
with the process.
2.5
Acknowledgements
We would like to thank the Jarillo-Herrero group for help with all the aspects of this
research, especially Joel Wang and Javier Sanchez-Yamagishi for their development
of the graphene transfer method, Ken Van Tilburg for taking the TEM images, Lewis
Stern, Max Lemme, and David Bell for their assistance with the helium ion microscope, Alfonso Reina-Cecco for help preparing CVD graphene on TEM grids, and
Monica Allen and Tchefor Ndukum for their help with suspension techniques.
36
Chapter 3
Etching of Graphene Devices with
a Helium Ion Beam
with Max Lemme, David Bell, James Williams, Lewis Stern, Pablo JarilloHerrero, and Charles Marcus'
3.1
Introduction
We report on the etching of graphene devices with a helium ion beam, including
in situ electrical measurement during lithography. The etching process can be used
to nanostructure and electrically isolate different regions in a graphene device, as
demonstrated by etching a channel in a suspended graphene device with etched gaps
down to about 10 nn. Graphene devices on silicon dioxide (SiO 2 ) substrates etch
with lower He ion doses and are found to have a residual conductivity after etching,
which we attribute to contamination by hydrocarbons.
'A version of this chapter appeared in the journal ACS Nano [49]
37
VS=OV
Vd
Graphene
Vg
c
soc et
to
trou
Figure 3-1: Schematic of a graphene device. Inset: Photograph of the microscope
chamber with installed chip.
3.2
Motivation
Graphene, a stable two-dimensional carbon crystal, has attracted great interest recently as a model system for fundamental physics as well as for possible nanoelectronics applications [1, 11, 50]. Many experiments in the field are targeted at graphene
devices where artificial confinement in one or two dimensions produces nanoribbons or quantum dots. Typically, such structures are on the ~5-50 nm scale and
have been fabricated by electron beam lithography followed by reactive ion etching [51, 52, 53, 54] and by chemical means such as thermally activated nanoparticles [55] or unfolding of carbon nanotubes [56, 57, 58].
While these methods are
suitable to produce devices near the atomic limit, they also have significant shortcomings.
Reactive ion etching typically erodes the resist mask creating disordered
graphene edges.
Chemical methods can result in irregular shaped and distributed
flakes poorly suited for integrated device applications. It has further been proposed
to etch graphene at the nanoscale with a focused electron beam [59].
This method,
however, requires suspending graphene on specific transmission electron microscope
grids, making it difficult to perform simultaneous electrical measurements.
38
Helium ion microscopy (HeIM) has recently been introduced as high-resolution
imaging technology for nanoscale structures and materials [60, 61, 62]. In this work,
we use a helium ion microscope (Zeiss ORION) as a lithography tool to controllably
modify electrical properties of graphene devices. We further demonstrate in situ electrical measurement during lithography. The HeIM is particularly well suited for this
purpose because it produces a high-brightness, low-energy-spread, subnanomneter-size
beam. The microscope benefits from the short de Broglie wavelength of helium, ~100
times smaller than the corresponding electron wavelength. This gives the beam an
ultimate resolution of 0.5 nm or better [61], making it an attractive tool for precision
lithography of graphene devices. While process details are published elsewhere [63],
this letter focuses on the modification of electrical properties of graphene. Fig. 3-1
shows a schematic of a graphene field effect transistor as used in this work.
Note
that for some experiments, the SiO 2 substrate was removed prior to measurements to
obtain a suspended graphene device (see the Methods). The inset in Fig. 3-1 shows
a photograph of a chip carrier inside the HeIM as used for in situ measurements.
3.3
Results
A suspended graphene device with a length of ~150 nmn and a width of ~1.5
im,
shown in the HeIM microscope image in Fig. 3-2a, was He ion etched by sequential
imaging in high resolution. The graphene was exposed to the He ion beam at a field
of view of 2 pim x 2 pin and an image size of 2048 x 2048 pixels, which resulted
in a pixel spacing of -1
nmn.
The dwell time was chosen to be 50 pis resulting in
an effective line dose of 0.8 nC/cn.
Fig. 3-2b shows such a high-resolution image,
expanded and labeled to distinguish the suspended graphene from the underlying SiO
and the chromium (Cr)/gold (Au) contacts.
2
Fig. 3-1c shows a sequence of images
taken under these conditions (nos. 1-13, where image 1 is identical to Fig. 3-1b). The
red circle indicates the region of the graphene flake where etching occurred initially.
Each scan with the He ion beam resulted in an increase of etched area. After 13 scans,
the dwell time, and hence the image quality, was increased to 500 ps, equivalent to a
39
line dose of 8 nC/cm, still not sufficient to completely etch the device (Fig. 3-1c, 14).
These images indicate that removal of edge atoms is favorable over atoms within in
the graphene crystal.
C)
I)
a)a)
2
3
5
iC
11
I
I
Figure 3-2: Etching suspended graphene with a helium ion beam. (a) HeIM image
of suspended graphene devices. The yellow box indicates the area that was subsequently imaged and etched in high resolution. (b) High-resolution image used to etch
graphene.
(c) Sequence of images of progressive etching of a suspended graphene
sheet. Image 1 corresponds to (b). The red circle indicates the area where etching
occurred initially.
The remaining graphene film was etched using live scanning mode with a 100
to 10 nm field of view. Here, etching was confirmed via the live screen image. A
resultant cut with minimum feature sizes in the 10 nm range is shown in the HeIM
image in Fig. 3-3a. The gap was measured with DesignCAD software after importing
the original image.
40
10
before etching
.- 8I
10
10"~
12
Vd= 0.5 mV
after etching-
10 -4-2 02
V9 M
4
b)
Figure 3-3: Electrically isolating suspended graphene with a helium ion beam. (a)
HeIM image (with false color) of a suspended graphene device after etching with
minimum feature sizes of about 10 nn.
(b) Electrical measurement of the device
before and after etching.
After a trench was etched across the entire graphene flake, the device was removed
from the HeIM and its drain current was measured as a function of back gate voltage
(Fig. 3-3b,
Vd
= 0.5 mV; note that the gate voltage range is limited in suspended
graphene devices [5, 64] and hence, Id changes little with V). The current dropped
to about 15 pA, compared to 1 pA prior to etching. While the latter is typical for a
functional graphene device of the given dimensions, the postetching value corresponds
to the noise level of the measurement setup. Adjacent, nonimaged devices made from
the same graphene flake showed conductivity similar to the investigated device prior
to imaging. These results confirni that the graphene was etched successfully using
the He ion beam.
Next, the drain current of a graphene device on SiO 2 substrate was measured
inside the He ion microscope, while part of it was exposed to the ion beam. A field
of view of 1 pim x 1 pim was chosen, indicated by the top yellow box in Fig. 3-4a.
After about 150 s the current saturated, indicating complete etching of the graphene
inside the field of view (Fig. 3-4b).
At this point, the imaging window was moved
to the next part of the device in the direction of the white arrow in Fig. 3-4a. The
current was again monitored until it saturated. A beam current of 1 pA, dwell time
of 3 ps, and pixel spacing of ~1 nm allowed us to estimate a suitable He ion line dose
41
for etching graphene on SiO 2 : 1.5 nC/cm.
A residual drain current of about 4 nA was measured after etching the entire
device, which could not be reduced further by subsequent He ion beam exposure.
We attribute this residual conductivity to contamination of the SiO 2 surface with
hydrocarbons.
Id
Vd
-1.0
-0.8
d=mV
V =0V
Wg
-
-0.4
-0.2n.0
0
m
/
n
400 800 1200
time [s]
b)
Figure 3-4: Etching graphene on SiO 2 with a helium ion beam. (a) HeIM image of a
graphene device. The boxes indicate the field of view used for etching. The window
was subsequently moved in the direction of the arrow. (b) Drain current vs time
of exposure of the graphene device. The etching window was moved as the current
saturated.
3.4
Conclusions
We have demonstrated etching of graphene devices with a helium ion beam. Suspended graphene has been etched conclusively, with minimum feature sizes in the
10 nm range. Graphene on SiO 2 was etched with a lower dose compared to suspended graphene. However, these devices showed a residual conductivity attributed
to contaminants on the surface. Helium ion etching can be considered an alternative
42
nanofabrication method for suspended graphene devices and, if contamination issues
can be solved, graphene on SiO 2 substrates.
3.5
Methods
Graphene was deposited onto ~300 nm of silicon dioxide on degenerately doped silicon by mechanical exfoliation
[65]
similar to the method described by Novoselov
et al. [1] Next, monolayer and few layer graphene akes were identified with an optical microscope.
Contacts to the graphene were defined by conventional electron
beam lithography, followed by evaporation of chromium/gold (3 nm/150 nin) and
titanium/gold (5 nni/40 urn). Suspension of the graphene sheet was obtained by wet
etching of the underlying SiO 2 in diluted HF, followed by critical point drying. All
devices were measured in a standard field effect transistor (FET)-like configuration,
with the evaporated contacts acting as source and drain and the doped substrate as a
gate electrode (Fig. 3-1). The drain current Id through the flake was then measured
as a function of gate voltage V for a constant drain voltage V.
Electrical data of
suspended devices were made outside the microscope, before and after He ion etching, using two Keithley 2400 source meters in a Desert Cryogenics probe station at
a pressure of ~ 5 - 103 mbar. The second set of graphene devices on SiO 2 substrate
were wirebonded to chip carriers and placed in a chip socket inside the Helium ion
microscope to enable in situ electrical measurements (inset in Fig. 3-1). These were
taken at a pressure of ~1 - 106 inbar with an Agilent 4155B parameter analyzer
connected to the device via a vacuum feedthrough. All measurements were taken at
room temperature.
3.6
Acknowledgments
We would like to acknowledge the support of the Alexander von Humboldt foundation
through a Feodor Lynen Research Fellowship. Research also supported in part by the
NRI INDEX program. We thank S. Nakaharai for fruitful discussions.
43
44
Chapter 4
Approaching GHz Resonant
Frequencies in Suspended
Graphene NEMS Resonators
with Tchefor Ndukum, Kevin Fischer,
and Pablo Jarillo-Herrero
4.1
Introduction
We report the excitation of suspended graphene NEMS resonators with resonance
modes as high as 900 MHz.
We successfully suspend high quality graphene res-
onators with resonance frequencies in the hundreds of MHz, mobilities in the tens of
thousands, arid
Q
values up to 1650. We excite these resonators using a novel mixing
circuit that allows for the signal from the high frequency modes to be read with a
basic low frequency lock-in amplifier. We then demonstrate the tunability of these
modes using a capacitively coupled back gate, and we fit that tunability to a model
of a membrane under tension. Finally, we show how this model can lead to capacitive
spring softening, and present experimental data that demonstrates this phenomenon.
45
4.2
Motivation
Graphene, a strong [3], light, two dimensional material, exactly one atom thick, embodies the ideal properties sought for a nanoelectromechanical system (NEMS) resonator. Its light weight and stiffness [3] make it perfectly suited to mass sensing.
And its unbeatable surface to volume ratio at one atom thick, distinguishes it as
one of the most promising materials for chemical and biological sensing as well [27.
Early incarnations of graphene based NEMS resonators have shown that these devices
can achieve microwave resonance frequencies, sensitive mass detection, and frequency
tunability [30, 27, 29]. Here we demonstrate devices with resonance modes as high as
900 MHz, furthering development toward graphene NEMS that, beyond acute sensing
tasks, can also enter the quantum regime as they are cooled to the ground state, enabling many exotic systems mixing the unique electrical properties of graphene with
quantized mechanical modes [12].
mobility, high
Q suspended
In addition we describe the fabrication of high
graphene devices and an actuation circuit used to excite
them. We then present data on the tunability of the resonance modes and provide
a model to fit the data. Finally we show how capacitive coupling can lead to spring
softening and fit data demonstrating that phenomenon.
4.3
Device Fabrication
The devices in this study are all fabricated from graphene exfoliated from naturally
forming mined bulk graphite. The graphite is exfoliated using a low tack semiconductor grade tape and then deposited onto highly doped silicon wafers covered in a
285 nm thick layer of thermally grown SiO 2 , as is a standard graphene fabrication
process [1]. The wafers are then searched using an optical microscope, and monolayers of graphene are identified by their distinct optical contrast [66]. A circuit is
patterned onto the graphene flakes using electron beam lithography and a polymethyl
methacrylate (PMMA) mask. 100 nm of gold with a .3 nm sticking layer of chromium
is evaporated onto the patterned circuit by thermal evaporation. The gold acts as
46
both the contacts to the graphene and the anchors that will support the membrane
after suspension. The device is next released from the substrate by etching out the
SiO 2 under the graphene using hydrofloric acid (HF). The HF preferentially etches
along the graphene-SiO 2 boundary and does not etch under the gold-SiO 2 boundary [5].
This phenomenon allows the graphene to be fully suspended while keeping
the gold anchors solidly supported. To prevent surface tension forces collapsing the
suspended devices, the wafers are dried through critical point drying. A schematic
of the device geometry is shown in Fig. 4-la and a scanning electron micrograph of
a suspended device is presented in Fig. 4-1b. The completed devices are bonded and
placed into a vacuum chamber. All measurements presented here were performed in
high vacuum (~
10-6 torr) and at room temperature. The high vacuum environment
helps to eliminate damping effects due to air resistance, which for resonators as light
as these, would be substantial. As a final step, the resonators are current annealed
to remove fabrication residue and other contaminants and to improve the mobility of
the devices [45].
47
(a)
(b)
raphene
Reference Line
Device
-D
fSource
(c)
Lock-in
d
12
(d)
60000
60,000.
40,000
9
z
6
E
Mixing Current
0
3
-
20,000
-10
0
10 20
)Vbg(V)
0-
3
-3
o
-20
6
-10
I
1
0
10
Vbg
-6
-20
20
M
I
-10
0
10
20
Vbg (
Figure 4-1: Graphene NEMS device schematics, image, and electrical readout. (a)
Upper: A schematic side view of the suspended graphene device, showing the gold
electrodes (yellow), graphene resonator (light blue), etched SiO 2 (dark blue), and
doped silicon gate (grey).
Lower: A diagram of the actuation and measurement
circuit. A high frequency source sends an amplitude modulated excitation to the
device. The mixed down measurement signal is read at the lock-in amplifier, which
is synced to the source by a reference line. A DC voltage is applied to the back
gate to capacitively tune the resonator.
(b) A scanning electron micrograph of a
suspended graphene resonator. (c) Resistivity of a device as a function of back gate
voltage, showing the characteristic Dirac peak of graphene. Inset: Mobility of the
device shown in (c), demonstrating its high quality. (d) Mixing current of a device
away from any mechanical resonance.
48
Suspended graphlene has been shown to benefit greatly from its separation from the
substrate [5, 64]. Impurities, residue, and unevenness of the SiO 2 substrate have been
shown to limit graphene mobility and conductivity [67].
Substantial improvements
can be made by placing the graphene on hexagonal boron nitride [14], but still the
best performing graphene devices have been suspended, set apart from all substrates
and outside influences.
Our experiments have yielded similar results.
Devices of
mediocre performance on substrate improve in mobility by an order of magnitude
better after suspension.
We typically find that devices with mobilities of ~
100
cm 2 /Vs on substrate will increase to mobilities of ~ 1000 cm 2 /Vs.
Once suspended, current annealing of graphene is also extremely effective [64].
With both sides of the material free to expel residue unimpeded, it may be possible
to get cleaner results than would be allowed on substrate. The downside, however,
is that the device is thermally isolated and far more prone to breakage during the
annealing process.
While current annealing is very effective, it must be done very
carefully. When a device is successfully annealed it can gain an additional order of
magnitude in mobility. It is common for our devices to have mobilities in the tens of
thousands after annealing. A typical device characteristic is shown in Fig. 4-1c with
the conductivity vs. back gate voltage plotted in the main panel and the mobility vs.
back gate voltage plotted in the inset.
4.4
Mixing Current
In order to measure the resonance frequency of our NEMS devices we utilize the
nonlinear response of the graphene to create a mixing circuit. This scheme is similar to
other mixing circuits that have been used to measure graphene and carbon nanotube
resonators in the past [68].
A schematic of the circuit is depicted in Fig. 4-la. A
high frequency source is connected to the source electrode and a lock-in amplifier
is connected to the drain electrode.
The source and the lock-in are synced via a
reference line that allows their frequency and phase to be matched.
Finally, the
doped silicon back gate is connected to a DC voltage source. The resonator is excited
49
at high frequency with an amplitude modulated signal provided by the high frequency
source. The source signal, V has the form:
V, = VamCOS(Wt)(1 + cos(Awt))
where w is the high frequency signal. w will be swept in order to find the device's
resonance frequencies. Aw is the low frequency used for measurement. Aw is matched
to the frequency of the lock-in amplifier. Vain is the amplitude of the signal and t is
time.
The current, I, that flows through the device as a result of this excitation can be
written in a general form as the sum of currents arising at each individual frequency,
Wi.
The signal the lock-in amplifier will measure is the current at the measurement frequency, IA.
To find the expected current IA, we can split the source voltage into its component
pieces.
V = Van [cos(wt)
+-
1
2
cos((w + Aw)t)
+-I COS((
2
- AW)t)
The current will be related to the voltage through Ohm's Law, I = Gj, where G
is the conductance of the device. Crucially, the conductance of the device will not
be constant; it will vary with frequency. Just as with the current, the conductance
can be written in a general form as a sum of conductance arising at each individual
frequency, ws.
G
GS
I + E 6G(w)cos(wt)
From this we see that the current will have two components: a steady state current,
50
ISU,
that will be proportional to the sourced voltage by the steady state conductance,
G.., and will not change with frequency or device response, and a dynamic current,
61(w), that will respond to the source frequency as
61(w) = 6G(w)cos(wt)Vanmcos(wt)(1 + cos(Awt))
Expanding this equation yields five components.
6I(w)
=
G(o)m
-aiCos (0t)
12
+~ICos
(2wt)
1
+
+
+
1
4
2
cos((2w
cos((2w
Aw)t)
-
Aw)t)
cos( Awt)
The last component is at the mixed down frequency of the lock-in amplifier. This is
the component that we measure.
To complete our analysis we can further describe the change in conductance, 6G,
as inversely proportional to the distance, z, between the graphene sheet and the silicon
back gate. This is because the conductance is proportional to the charge carriers on
the sheet, q, which are proportional to the capacitance, C, which, in turn, is inversely
proportional to the distance between the resonator and the back gate, z. That is:
G oc q oc C Oc-
1
z
Making simple linear approxinations, the change in conductance, 6G can be expanded
as:
6G(w)
dG
dq-q(w)
dq
dG
dq-VC(w)
dq
dG dC
dq-K dz(w)
dq
For our purposes the voltage in the capacitor, Vc = Vb9
-
dz
V, ca n be simplified to V
because the mixing current will only depend appreciably on the direct current voltage
51
applied to the back gate. The alternating current components to V will not produce
any significant response at the mixing frequency, Aw. Assembling the measurement
current back together, we find:
I dG dC
I(W)
=
-
4 dq dz
6Z (W) V,V,
This current is proportional to 6z the deflection due to the capacitive force between
the membrane and the back gate. That deflection will be very small for any value of
w off the resonance of the device, and then exponentially larger on the resonance. It
is this phenomenon that delivers the signal of a resonance through the mixing current
measured at the lock-in. The current will also be proportional to the transconductance, dG/dq. Thus, the signal is expected to be large for high mobility suspended
graphene, though it will disappear at the Dirac point. This dependence can be used
to calibrate the system and check the functionality of the device.
Off resonance,
the mixing current should be seen to change sign as it crosses the Dirac point and
to peak at the maximal transconductance along a sweep in back gate voltage.
representative mixing current trace is shown in Fig. 4-1d.
52
A
4.5
Resonant Modes
(a) 130
(b) 125
110
120
90
115
NN
-70
110
50
105
100.
30
-12
-6
0
Vbg (V)
6
12
-10
-9
-7
-8
Vbg (V)
(c) 500
(d) goo
495
490
485
-20
-10
0
10
700L
-30
20
Vbg (V)
-20
-10
0
10
20
Vbg (V)
Figure 4-2: Resonance maps, avoided crossings, quality factor, and high frequency
modes. (a) A map of resonance modes in frequency and gate space. The modes are
seen to be tunable with the back gate voltage. The blue box shows the zoomed in
area in (b). (b) A high resolution view of avoided crossings from the same resonator
in (a) in the region of the blue box. This shows in greater detail the interaction of the
graphene resonance modes with those from the gold contacts. (c) A map of resonance
modes from a second device. The vertical blue line shows the location of a cut in
frequency space. Inset: the data cut in frequency space showing the sharpness of the
resonance peak, giving a quality factor of
Q=
1650. (d) A resonance map of a thrid
device with modes approaching the GHz regime. Inset: Mode data extracted from
the raw data shown in (d).
53
A map of a graphene resonator's resonance modes is shown in Fig. 4-2a. The map
depicts the measurement current IA, as a function of frequency and back gate voltage.
Multiple resonance modes are seen as here quasi-parabolic signals on the map. These
line shapes are due to the tensioning of the the graphene sheet as the magnitude of
the back gate voltage is increased. An attractive force is felt between the charges on
the back gate and the charges on the graphene sheet, pulling the resonator down and
tensioning it. This increases the resonance frequency of the device as the magnitude
of the back gate voltage is increased. Such tensioning allows for these resonators to
be tuned over a fairly substantial range. In this case the resonator has a maximal
tunability of 13 MHz/V.
Also evident in this resonance map are multiple flat resonance modes and avoided
crossings. Fig. 4-2b shows a high resolution view of a section of the map containing
these features. When the tunable graphene resonance approaches a flat resonance of
the system they form a coupled oscillator system that shares energy between the to
resonances and forms the avoided crossing. These flat resonances are suspected to
come from the suspended portion of the gold electrodes. Because hydrofloric acid will
wet under the graphene, the section of the gold electrode over the graphene will be
suspended. This suspended electrode will have a mechanical resonance but will be
too rigid to be tuned with the back gate voltage. We can model such a resonator as
a doubly clamped beam, where the expected resonance frequency is
t
E
with t, f, E, and p the thickness of the electrode, the length of the suspended segment,
the Young's modulus of gold, and the density of gold respectively. Typical numbers for
our devices lead to expected resonance frequencies of around 50 MHz. This matches
well with the lowest order modes that we see. Higher modes are also seen on top of
those.
We can also take a cut of the resonance map data along the frequency axis to
extract the quality factor of the resonance peak. The quality factor of a resonator is
54
defined as the ratio of the resonant frequency to the bandwidth of the resonance peak,
Q
=
fr /Af.
This figure of merit quantifies the energy dissipation in the resonator.
The inset to Fig. 4-2c shows a cut of the resonance data along the frequency axis.
Fitting the resonance peak to a Lorentzian, we find a quality factor of 1650. This is
quite a respectable number for a graphene resonator.
Finally, we can look for the highest frequency modes that our system can measure.
In Fig. 4-2d we see modes that extend as high as 900 MHz. The resonance peaks
of the two modes shown here are extracted and plotted in blue circles in the inset
for clarity. These high frequency miodes have ground states that are nearly achievable in traditional dilution refrigerators. Cooling such a mode to the ground state
would enable the study of novel physical systems that mix graphene's unique electronic properties with quantized mechanical oscillator. This measurement represents
a significant step toward that goal.
4.6
Mode Fitting
We fit a model of tunable resonance modes to our data to further characterize our
graphene resonators.
Ultimately, the devices can be treated as one dimensional
springs [27], though determining the spring constant, k is nontrivial. Their resonant
frequency therefore will be
fr =
where r is the mass of the resonator.
w*
Ideally, m would be described simply as
m = f - w - t - p, where f, w, t, and p are the length, width, thickness, and density
of the graphene sheet. However, this ignores any adsorbates or residues left on the
surface of the graphene. Since the resonator's mass is so small, these additions will not
be negligible. This sensitivity to adsorbed mass is, in fact, exactly what makes these
resonators such promising mass sensors. To account for this, the mass is calculated
as the mass of the graphene plus the mass of the residue,
= ing + m,. The mass
of the residue then, will be one of the two fitting parameter of the model.
55
The spring constant, k, will behave as the spring constant of a tensioned membrane [69].
16 To
3 f
128 Ewt
3
2_
__2
ze
f3
Z3
b
This relation comes from solving the equation of motion of an elastic membrane with
some built in tension, To, and added energy from a capacitively couple back gate
voltage, Vg.
The tension, To, is the second free parameter in the fit. f, w, and t
are the length, width, and thickness of the graphene sheet. E and c are the Young's
modulus of graphene and the permittivity of the graphene gate capacitor. z is the
maximal deflection of the sheet as it is pulled toward the back gate by capacitive
forces.
And z, is the deflection of the sheet under equilibrium as determined by
minimizing the energy of the system. ze has the form
9
-. 870
9
1.7
a
4a3 33 + 27a4
l 28Eivt
3
2
+ 1.7
27472
2.6a
2
16To
3f
4a 3 /3+
_
cfw
Z
2
bg,
-
2
cw
T,
2z2 b
The first term in the equation for k is a simple constant, which describes the intrinsic
spring hardness of the resonator due to to its built in tension. The second term is
dependent on the back gate voltage and is a hardening term. It raises the spring
constant with the increase of the magnitude of the back gate voltage. At high values
of
IVbg
this term causes the spring constant to be proportional to Vbg4/3.
The final
term in the spring constant equation accounts for a softening effect. Here, as the
magnitude of the back gate voltage increases, the spring constant decreases.
This
effect is due to the increased capacitance of the system at higher back gate voltage.
This capacitance acts as a damping mechanism on the resonator, lowering its spring
constant.
Depending on the initial tension in the resonator, this spring softening
term can dominate at low Vg. In frequency space, this would lead to a lowering of
the resonance frequency with increasing
jVbgI
at low values, resulting is a 'W' shape
dependence of the resonant frequency with respect to the back gate voltage.
56
(b) 13 !0
(a) 130
110
N
11
90
C
0
To = 100 nN
m
Mr=150
-
0-
-I
70
70 -
50
50
30
-12
-6
0
Vbg M
6
--
3n
-12
12
-
-6
0
6
12
Vbg (
(c) 240
(d) 240
To= 450 nN
Mr
235
235
2301
2301
225 IN
-12
-6
0
Vbg M
6
12
225 L
-12
-6
0
=81 mg9
6
12
Vbg (
Figure 4-3: Fitting resonance modes and mode softening. (a) A map of resonance
modes in frequency and gate space. The tunability of the graphene modes is evident.
(b) Extracted mode data (light blue circles) from the map in (a) are fit using our
model (dark blue lines), showing excellent agreement. (c) A resonance mode map of
a second device showing capacitive mode softening. (d) Extracted mode data (light
blue circles) from the map in (c) are fit using our model (dark blue line), showing
excellent agreement.
Fits to our resonator data are shown for two devices operating in the two different
regimes. One in the low To regime where the resonant modes rise monotonically with
increasing IVlbg
(data in Fig. 4-3a and fits in Fig. 4-3b).
And one in the high To
regime where the mode has a 'W' shape (data in Fig. 4-3c and fits in Fig. 4-3d).
57
The three modes fit in the low To case are three sequential harmonics of the system.
The fits give an initial tension of 100 nN and a resonator mass 150 times that of
pristine graphene. The high tension device gives an initial tension of 450 nN and a
resonator mass 81 times that of pristine graphene. It is clear from these fits that
contamination is still a significant problem and that attempts at developing higher
frequency resonators and resonators that can be cooled to their ground state should
focus on reducing fabrication residue and improved cleaning processes.
4.7
Conclusions
In conclusion, we successfully suspend high quality graphene resonators with microwave resonance frequencies approaching the GHz range, high mobilities in the
tens of thousands, and high
Q values
in the thousands. We have demonstrated the
tunability of the modes in these devices using a capacitively coupled back gate, and
we have fit that tunability to a model of a membrane under tension. Finally, we have
show how this model can lead to mode softening, and presented experimental data
that demonstrates this phenomenon. These are promising results for the future of
graphene NEMS resonators.
4.8
Acknowledgments
We would like to thank Monica Allen for her help with device fabrication techniques
and Adrian Bachtold and especially Joel Moser for their hospitality and incredible
helpfulness showing us their actuation and measurement systems. We would also like
to thank Hadar Steinberg for many thoughtful discussions.
58
Chapter 5
Intrinsic Electronic Transport
Properties of High Quality
Monolayer and Bilayer MoS 2
with Hugh Churchill, Yafang Yang, and Pablo Jarillo-Herrero'
5.1
Introduction
We report electronic transport measurements of devices based on nonolayers and bilayers of the transition-metal dichalcogenide MoS 2 . Through a combination of in situ
vacuum annealing and electrostatic gating we obtained ohmic contact to the MoS 2
down to 4 K at high carrier densities. At lower carrier densities, low temperature
four probe transport measurements show a metal-insulator transition in both monolayer and bilayer samples. In the metallic regime, the high temperature behavior of
the mobility showed strong temperature dependence consistent with phonon dominated transport.
At low temperature, intrinsic field-effect mobilities approaching
1000 cm 2 /Vs were observed for both monolayer and bilayer devices. Mobilities extracted fromn Hall effect measurements were several times lower and showed a strong
dependence on density, likely caused by screening of charged impurity scattering.
'A version of this chapter appeared in the journal Nano Letters [70]
59
Top View
(a)
Si de View
Monolayer
(b)
Bilayer
(c)
1 ,-
MO
2L
(d)
(e) 106
fVxxq
(f) 10-6
Vds = 0.1 V
10O
10-8
? 10
Vds, Ids -
U,
300K
10
_.
200K
100K
10-
5OK
5K
14
Vbg
Vds = 0.1 V
10 -50
0
Vbg (V)
50
1OF 10
300K
200K
1OOK
1OF2
1- 50
1
0
50K
K
50
( V)45
Vbg M)
Figure 5-1: MoS 2 device schematics, images, and two-terminal transport measurements. (a) Structure of a single sheet of MoS 2 from both a top and side view, showing
the locations of molybdenum (black) and sulfur (yellow) atoms. (b) and (c) Color enhanced AFM height images of six-terminal monolayer (Ml) and bilayer (Bi) devices.
Scale bars are 2 pm. (d) Device schematic and measurement setup. (e) Two-terminal
drain-source current, Is, of monolayer MoS 2 as a function of back-gate voltage, Vg.
Temperatures from 300 K to 5 K are denoted with traces colored from red to black in
(e) and (f). (f) Id, of bilayer MoS 2 as a function of Vg. The DC drain-source voltage,
Vds, was 0.1 V for both (e) and (f).
5.2
Motivation
Molybdenum disulfide (MoS 2 ), a layered transition-metal dichalcogenide (TMD) semiconductor, is attracting increasing interest for its novel nanoelectronic and optoelectronic properties [71]. Bulk MoS 2 is a stack of atomic trilayers composed of a single
atomic layer of molybdenum between two layers of sulfur [Fig. 5-1a]. Strong intralayer covalent bonds lead to high mechanical strength in plane [721, while weak Van der
Waals bonds between layers render the material chemically inert with robust electrical
properties [73]. As with graphite, this weak inter-layer coupling also allows individual
layers to be isolated and studied [9]. Monolayer MoS 2 (one S-Mo-S unit) is of particular interest as a large (1.8 eV), direct-gap semiconductor [35] with strong spin-orbit
interaction leading to a spin- and valley-split valence band [74, 32]. These qualities
60
could lead to novel physics such as an unconventional quantum Hall effect, combined
spin Hall and valley Hall effects [75], and new devices such as high-performance,
ultra-low power transistors [76] and devices integrating spin- and valley-tronics [77].
Further, MoS
2
is compatible with standard semiconductor manufacturing [71], can
be grown in large-scale by chemical vapor deposition [78, 79, 80], and integrated with
other two-dimensional, flexible, and transparent materials [77].
Few-layer MoS 2-based devices in field-effect transistor geometries have demonstrated the promise of these materials [36, 71, 81, 82, 83, 84]. However, in a twoterminal contact configuration mobilities are underestimated by including contact resistance. This point is illustrated by measurements of imulti-terininal devices based on
thick MoS 2 with mobilities up to - 500 cmn2 /Vs, whereas two-terminal measurements
2
showed mobilities on the order of 10-50 cnm
/Vs [73, 85, 86, 87, 88]. Encapsulating
monolayer MoS 2 in high-k dielectric or a polymer electrolyte improved device performance in several ways: increasing miobilities for imonolayer devices from r
up to
r
15 cm 2 /Vs
160 cm 2 /Vs [36, 42, 78, 41, 77, 89], increasing the on/off current ratio to
108 [76], and enabling observation of a metal-insulator transition [89]. Nevertheless,
investigation of the novel quantum transport phenomena expected for TMDs will
require further improvements in device quality. Here we report mnulti-terminal devices based on monolayer and bilayer MoS 2 with sufficiently transparent contacts at
high density to enable access to the intrinsic mobility of MoS 2 at low temperatures.
These measurements highlight the potential for observing novel quantum transport
phenomena in TMDs.
5.3
Device Fabrication
Devices based on monolayer and bilayer MoS 2 were fabricated from bulk MoS 2 (SPI
Supplies) that was exfoliated on highly doped silicon substrates with 285 urn of
thermal oxide using the micronechanical cleavage process standard for graphene [1].
Monolayer and bilayer flakes were identified by optical contrast [90] and confirmed
with AFM measurements (see Appendix A). We present data from two nimonolayer and
61
two bilayer devices, denoted Mi,
M2, B1, and B2. Device contacts were patterned
using PMMA masks and e-beam lithography. To optimize fabrication procedures, we
investigated a number of process variations. Titanium/gold contact metal was evaporated in thicknesses of 0.3-4 nm for Ti and 50-100 nm for Au. Some devices were
annealed in an Ar/H
2
atmosphere at 350 'C for 3 hours, both before and after con-
tacting. All of the devices were annealed in situ at high temperatures (- 120 0 C) for
up to 20 hours in vacuum (~10-6 mbar) before measurement. Variations in contact
metal thickness and whether devices were annealed in Ar/H
2
were not found to have
a substantial effect on mobility or contact resistance. In contrast, vacuum annealing
caused a substantial drop in two-terminal resistivity (see Appendix A) and the nearly
complete elimination of Schottky behavior in the contacts, even at 4 K (Fig. 5-2).
Vacuum annealing substantially doped the sample with n-type carriers, shifting
the threshold gate-voltage for conductance by as much as 100 V toward negative values. The formation of S vacancy donors would cause n-doping [91], but this scenario
is unlikely given the thermal stability of MoS
2
up to 1000 0 C in vacuum [92].
We
speculate that this effect points to intrinsic doping of mined, natural MoS 2 [93]. The
result is a large increase in carrier density and conductivity, and a significant decrease
in contact resistance. After vacuum annealing, both monolayer and bilayer devices
displayed stable, smooth field-effect transistor characteristics with on/off current ratios up to 106 (Fig. 5-1).
Two-terminal measurements such as these reiterate the
viability of MoS 2 as a transistor material, but contact resistance ultimately obscures
the intrinsic material properties.
5.4
Resistivity Measurements
To investigate the intrinsic properties of MoS 2 , we further annealed our devices to
reduce contact resistance, allowing us to use standard current-biased lockin techniques
to measure four-probe resistivity for a second set of monolayer and bilayer devices.
Without vacuum annealing, two-terminal I-V measurements showed evidence of large
Schottky barriers. After sufficient annealing, two-terminal I-V measurements remain
62
nearly linear at small drain-source bias, low back-gate voltage, and low temperature,
with a clear positive slope across zero bias at 5 K and zero Vg (insets to Fig. 5-2 a and
b). Non-linearities indicating small residual Schottky barriers begin to appear at zero
Vbg,
but are small enough to allow reliable four-terminal measurements of resistivity
in the regimes studied here.
To measure the contact resistance, the devices were connected as shown in Fig. 5-1
with the outer contacts serving as the current drain and source and two inner contacts
as voltage probes. These measurements, combined with the two-probe measurements
like those shown in Fig. 5-1, allowed for the separation of the MoS 2 resistivity and the
device contact resistance (both shown in Fig. 5-2). Contact resistance was calculated
from the resistivity as R, = VdjIdS - p. l/w, where / and w are the full sample length
and width, respectively, and p = (Vx/Ids) - (lm,/w), with lj, the length between the
inner contacts. Even after annealing, contact resistance makes up a significant portion
of the total device resistance. Contact resistances of 5-50 kQ could be reliably achieved
at higher densities, though Rc increases to greater than 1 MQ at low temperature and
low density (Fig. 5-2).
The temperature dependence of the resistivity provides useful information such
as whether a sample is metallic or insulating and provides a means of distinguishing nobility-liniting scattering mechanisms. At high gate voltages, p monotonically
decreases with decreasing temperature, showing a consistent metallic state for both
monolayer and bilayer MoS 2 (Fig. 5-2). In an intermediate range, p is non-monotonic
but ultimately increases at low temperature. At the lowest gate voltages, p mnonotonically increases with temperature, characteristic of insulating behavior.
These
observations indicate the presence of a imetal-insulator transition in the samples with
critical resistivities p = 0.8 h/c
2
for the monolayer sample and 0.3 h/e
layer, consistent with previous observations for
2
for the bi-
monolayer MoS2 [89] and theoretical
expectations [94]. Other imonolayer and bilayer samples we measured also showed
critical resistivities of order h/e
2
(see Appendix A).
63
(a)
Monolayer
107
O.3 5K
0
106
a
(b)
-1
1
10
a
0(
0.
-25
-0
-20
105
5K
0
25
4 50K
lOOK
200K
10o3 300,K
-50
75
50
(c) 10 7
1
1o5
0
50
Vbg (V)
(d) 10
5K
50K
100K
100K
200K
300K
1061
7
5K
50K
100K
200K
300K
106
1
CL
300
CL
10 4
103
0
VSV
Vbg M
1
-25
0
10 3
75
50
25
-50
0
Vbg (
Vbg M
10
(e) 107
0.
0 Vbg
-1
10
100K
200K
300K,
10 3
Bilayer
20 5K
7
106
Vds(VM
5K
10
10 6
106
105
10 5
50
-75 Vbg
-50 Vbg
-25 Vbg
0OVbg
50 Vbg
75 V
04
10
-20 Vbg
0
10
Vbg
50
Vbg
75 Vl
10
T(K)
100
300
10
-
10
T(K)
100
300
Figure 5-2: Contact resistance and four-terminal resistivity of monolayer and bilayer
MoS 2 . (a) Contact resistance, Rc, to monolayer MoS 2 as a function of Vg. Curves
colored from red to black show measurements at 300, 200, 100, 50, and 5 K, respectively, across panels (a) through (d). Inset: Two-terminal Ids vs. V, at 5 K
and Vbg = 0 V for the monolayer. (b) Contact resistance, Rc, to bilayer MoS 2 as
a function of Vg.
Inset: Two-terminal Ids, vs. Vd at 5 K and Vbg = 0 V for the
bilayer. (c) Four-terminal resistivity of a monolayer device as a function of Vbg. (d)
Four-terminal resistivity of a bilayer device as a function of Vg. (e) Resistivity of a
monolayer device as a function of temperature. The curves, from purple to yellow,
correspond to Vbg = -20, 0, 25, 50, and 75 V. (f) Resistivity of a bilayer device as a
function of temperature. The curves, from black to yellow, correspond to Vbg = -75,
-50, -25, 0, 25, 50, and 75 V. Data from devices M2 and B2.
64
5.5
Hall Measurements
Hall effect measurements were used to determine carrier density, n, as a function of
Vb,.
The Hall coefficient, RJ
verse resistance, Rx,
= 1/ne, was calculated by fitting the slope of the trans-
as a function of magnetic field up to 1 T (see Appendix A).
We note here that for the full gate voltage and temperature ranges of the Hall ineasurenents, contact resistances of both the monolayer (Ml, with further annealing
relative to data shown in Fig. 5-1) and bilayer (B2) devices remained below 500 kQ.
The nearly linear dependence of rt
On
V
corresponds to a capacitance per unit area
of c = 10 ± 2 nF/cm2 for the monolayer and 13 + 1 nF/cm 2 for the bilayer sample
(Fig. 5-3 a, b, and insets). These values are in rough agreement with the capacitance
expected for a parallel plate geometry, 12 nF/cm2 , which we expect to underestimate
the true capacitance by about 10% due to finite size effects for an MoS 2 flake only
a few times wider than the oxide thickness due to the contribution from fringing
fields [95, 96]. At fixed Vbg, the density varied with temperature significantly for the
monolayer, though not for the bilayer. The monolayer density decreased by a factor
of 1.5 from 300 K to 5 K, whereas the density change in the bilayer was negligible.
A reduction in density is expected in inhonogeneous samples due to localization of
charge carriers at low temperatures. Inhonogeneities in electric potential from different sources such as the substrate, charged impurities, or defects will trap more charge
carriers as the device cools and the ability of carriers to be thermally excited out of
these potential wells decreases.
Combining the measurements of n and p, we extract Hall mobility
IH
=
/nc as a
function of n, where - = i/p is the conductivity and c is the electron charge (Fig. 5-3
a and b).
The Hall mobilities increased with density, reaching 250 cm 2 /Vs for the
monolayer and 375 cn 2 /Vs for the bilayer at high n. Another method commonly used
to estimate carrier mobility is to calculate the field-effect mobility upFE = du/dVbg -1/c,
where c = creo/d is the gate capacitance per unit area (12 nF/cm 2 for 285 nm of
SiO 2 ). The large carrier density of these highly doped samples leads to very high
field effect mobilities.
Both devices show I'FJ1
65
1000
Cm2 /Vs
(Fig. 5-3 a and b),
(a)1000
.
.
Mo~nol-aerN~nn~k~,I~r
.
750
2
V
0
v
E 500500 vVg(V)
50
Bilayer
iIlav.er
1.5x10 13
(b)1000
21.5x1*
750
E
0.5
0
1
500
E 50
250
1FE
50
(M
Vbg(V
PH
250
0
-25
0
25
T=10K
0
-25
50
0
Vbg (V)
(C)
1E
25
T=5K
50
75
Vbg(V)
500
(d) 500
y=1.1
EE
2
y=1. 7
100
50 n=1.9x10 cm
1.8x10 13
i
20
3
100
-
7'
1.6x10 13
0.7x10 13
20
100
10
50 n=1.3x10 cm'
1.1x10 13
300
3
100
10
300
T(K)
T(K)
Figure 5-3: Field-effect and Hall mobilities as a function of back-gate voltage and
temperature for monolayer and bilayer MoS 2. (a) Field-effect mobility, PFE, (yellow)
and Hall mobility, pH, (blue) of a monolayer device as a function of back-gate voltage,
Vbg, at 10 K. Inset: Density of a monolayer device as a function of Vbg at 10 K. Solid
lines in the insets to (a) and (b) are fits to n = mVbg + b, where the slope, m, and the
intercept, b, are free parameters. (b) PEE (yellow) and pH (blue) of a bilayer device
as a function of Vg at 5 K. Inset: Density of a bilayer device as a function of Vbg at
5 K. (c) pH as a function of temperature for a monolayer device. The curves, from
blue to yellow, correspond to n = 1.6, 1.7, 1.8, and 1.9 . 1013 cm-2 . The black line is
a power law fit, PH oc T-7, with y = 1.7 for the high density data from 150-300 K.
(d) Pm as a function of temperature for a bilayer device. The curves, from blue to
yellow, correspond to n = 0.7, 0.9, 1.1, and 1.3- 1013 cm- 2 . The black line is a power
law fit, Pm oc T-1, with -y = 1.1 for the high density data from 150-300 K. Data from
devices MI and B2.
66
the highest field-effect mobilities reported to date for either monolayer or bilayer
MoS 2 .
Encapsulating MoS
2
in a high-k dielectric has shown substantial mobility
improvements [36, 42, 78, 41, 77, 89]. The devices reported here are not in a high-k
environment, however, suggesting the possibility of further mobility improvement in
future devices.
The discrepancy between pH and PFE evident in Fig. 5-3 a and b can be explained
by the density dependence of the Hall mobility. Substituting O- =
into the field effect mobility formula, we find PFE
pHn
=
nepH
and nc
CVbg
pH+n dpH /dn. Thus,
the field effect mobility can differ significantly from the Hall mobility if the Hall
mobility changes with density. Within this model a linear trend of pH with density
would lead to a field effect mobility trend with twice the slope (see Appendix A).
Conversely, when dpH /dn approaches zero, the two mobility values should nearly
match.
These two behaviors roughly match the data: in the gate range where PH
is nearly independent of density, pEE approaches PH (Fig. 5-3b).
increases with density,
PFE
And where PH
does so at about twice the rate. Additionally, we note
that the devices were not operated in a regime of saturated current with drain-source
bias, a situation which would lead to a similar overestimation in the calculation of
PFE.
For high densities in the metallic regime, Hall mobilities at constant density increase rmonotonically from 300 K to 5 K for both the imonolayer and the bilayer saulples (Fig. 5-3 c and d).
At high temperatures (above
follows a power law in temperature, PH oc T--,, with Y=
vice and 1.1 for the bilayer. It is expected that MoS
2
100 K), PH approximately
1.7 for the monolayer de-
devices with mobilities limited
by hioriopolar, optical phonons should follow this form in this temperature range,
with 'l = 1.69 for monolayer [97] and y = 2.6 for bulk MoS 2 [98].
The nonolayer
value agrees well with the prediction, though we note that the fit was obtained over
a limited temperature range. The power law for the bilayer sample is significantly
lower than the prediction.
suppressed
'
This is similar to a previously reported value where a
= 1.4 was attributed to phonon quenching by a top-gate dielectric [89].
Such quenching, however, is not expected in our case.
67
The monolayer and bilayer mobilities begin to saturate below 100 K, a temperature by which scattering from optical phonons is expected to become negligible [97].
At lower temperatures, scattering from phonons should be dominated by acoustic
modes with a linear dependence of mobility on temperature
[98],
in contrast with the
near saturation we observe (Fig. 5-3 c and d). Another candidate explanation is mobility limited by long-range Coulomb scattering. This picture is consistent with our
observation of p-H being roughly linear with carrier density [99], which is not expected
for phonon scattering [97]. Whether the temperature dependence we observed can be
mostly described by long-range Coulomb scattering awaits further experimental and
theoretical study.
5.6
Conclusions
In conclusion, we have demonstrated high quality MoS 2 devices that may open opportunities for measuring novel quantum transport phenomena at low temperature. We
showed that in situ vacuum annealing can dope devices and significantly reduce Schottky barriers and contact resistance, allowing for ohmic contact to MoS 2 down to 4 K
at high densities. A clear metal-insulator transition was evident in both monolayer
and bilayer samples at p ~ h/c 2 , showing that high density MoS 2 devices remain
conducting at low temperature. Additionally, the field effect mobilities were quite
high, indicating good sample quality. Finally, Hall mobilities calculated from density
measurements gave relatively high values, though the field effect equivalents are still
several times larger due to the density dependence of the Hall mobility. Based on the
temperature and density dependence of the mobilities, we infer a crossover from a
regime limited by optical phonon scattering at high temperature to one likely limited
by long-range Coulomb scattering below 100 K.
68
Chapter 6
Optoelectronic Devices Based on
Electrically Tunable p-n Diodes in
a Monolayer Dichalcogenide
with Hugh Churchill, Yafang Yang, and Pablo Jarillo-Herrerol
6.1
Introduction
The p-n junction is the functional element of many electronic and optoelectronic
devices, including diodes, bipolar transistors, photodetectors, light emitting diodes,
and solar cells. In conventional p-n junctions, the adjacent p- and n-type regions of
a semiconductor are formed by chemical doping. Materials with ambipolar conductance, however, allow for p-n junctions to be configured and modified by electrostatic
gating. This electrical control enables a single device to have multiple functionalities. Here we report ambipolar monolayer WSe 2 devices in which two local gates are
used to define a p-n junction within the WSe 2 sheet. With these electrically tunable
p-n junctions, we demonstrate both p-n and n-p diodes with ideality factors better
than 2. Under optical excitation, the diodes show photodetection responsivity of 210
nmA/W and photovoltaic power generation with a peak external quantum efficiency of
'A version of this chapter appeared in the journal Nature Nanotechnology [100]
69
0.2%, promising values for a nearly transparent monolayer material in a lateral device
geometry. Finally, we demonstrate a light emitting diode based on monolayer WSe 2.
These devices provide a building block for ultra-thin, flexible, and nearly transparent
optoelectronic and electronic applications based on ambipolar dichalcogenide materials.
6.2
Motivation
Next generation photodetectors and photovoltaic devices, as well as sensors, displays,
and light emitting diodes, will require new optoelectronic materials with characteristics superior to those currently in use.
Candidate materials must be flexible for
wearable devices, transparent for interactive displays, efficient for solar cells, and
robust for broad distribution. Monolayer semiconducting transition metal dichalcogenides (TMDs) such as tungsten diselenide (WSe 2 ) are flexible [72], nearly transpar-
ent [35, 101], high strength [72], direct band gap [35] materials with the potential to
meet all of these criteria. The crystal structure of tungsten diselenide comprises stacks
of sheets, each made of one atomic layer of tungsten encapsulated by two layers of
selenium. This structure leads to strong intralayer and weak interlayer bonding [72].
Such bonding allows the exfoliation of bulk crystals down to a single Se-W-Se layer,
similar to graphene [9].
Monolayer WSe 2 is a direct gap semiconductor with a band gap of -1.65 eV [102],
a complement to other two-dimensional materials [103] such as graphene, a gapless
semimetal, and boron nitride, an insulator. The direct band gap distinguishes monolayer WSe 2 from its bulk and bilayer counterparts, both indirect gap materials with
smaller band gaps [32, 102]. This sizable direct band gap in a two-dimensional layered material enables a host of new optical and electronic devices. Thin film TMDs
have already demonstrated novel nanoelectronic and optoelectronic devices such as
ambipolar and high quality field effect transistors [73, 89, 70], electric double-layer
transistors [104], integrated circuits [77], and phototransistors [105] with high responsivity [106].
In addition, photoluminescence [35] and electroluminescence [93, 107]
70
(a)
(b) 200
PN NN
PP NP
100
0
V
WSe 2
Vds
-100
ids
Hf02
-200
2
V19
-1
0
1
2
1
2
VdsM
Vrg
(d) 106
(c) 106
S
S"D
VD
10'
10'
PN
n=1.9
1012
-2
-1
0
1
1012
-2
2
NP
n=1.9
-1
0
VdS (M
Vd (
Figure 6-1: Gate-controlled monolayer WSe 2 p-n junction diodes. (a) Optical micrograph of a monolayer WSe 2 device controlled by two local gates. The WSe 2 is
contacted with Au electrodes. The flake and contacts are insulated from the gates by
20 uni of HfO 2 . The scale bar is 2 pm. Below is a schematic side view of the device
including electrical connections. (b) Current-voltage (Id-V,) curves showing transport characteristics of four doping configurations of the device, NN, PP, PN, and NP.
Both gates were set to 10 V for the NN configuration and -10 V for PP. Vig was set
to ±10 V and V, to -F10 V for PN/NP. The NN and PP configurations (yellow and
black curves respectively) are ohmic at low Vd5, while the PN and NP configurations
(blue and green curves, respectively) strongly rectify current in opposite directions.
(c) and (d) Semi-log plots of Ids through the PN (blue circles) and NP (green circles)
diodes as a function of Vd8 , with fits in yellow (see text). The fits give a diode ideality
of n = 1.9 for both the PN and NP configurations. Insets: Schematic band diagram
of the device in forward bias for PN and NP configurations.
71
have revealed effects including giant spin-valley coupling [108] and optical control of
valley polarisation [109, 110 and coherence [111].
P-n junctions in TMDs have, to date, been less explored than field effect transistors. Junctions of p- and n-doped bulk WSe 2 were reported over 30 years ago [91].
More recently, p-n devices have been constructed of low-dimensional semiconductors [112] and thin-film TMDs, such as an ionic-liquid gated bulk MoS 2 device [104],
an InAs/bulk WSe 2 heterojunction [101], and a doped silicon/monolayer MoS 2 heterojunction [107].
Unfortunately, bulk TMDs and heterostructures involving other
materials lack many of the most appealing properties of monolayer TMDs: direct
band gap, flexibility, and transparency. Here we characterise monolayer WSe 2 p-n
junctions defined and controlled by electrostatic gates, demonstrating devices with
diverse functionality, including current rectifying diodes, a photodetector, a photovoltaic device, and a light-emitting diode.
6.3
Device Fabrication
Device fabrication uses exfoliated natural crystals of WSe 2 , with monolayer flakes
transferred [14, 113] onto substrates with a pre-patterned pair of local back gates
separated by 100 nm and covered by 20 nm of HfO 2 . An optical image and schematic
of the device used for all measurements except electroluminescence are shown in Fig. 6la. Comprehensive fabrication details are provided in Appendix B. All measurements
were performed at room temperature and in vacuum (~ 10-
torr), with a voltage
bias, Vd,, applied to the source contact and the DC current through the device, Ids,
measured at the drain.
6.4
Transport in Gate Controlled p-n Junctions
The voltages on the two gates, V, for the left gate and Vg for the right gate, independently control the carrier density in the left and right sides of the monolayer. In
this way the device can be electrostatically doped into various conducting regimes
72
(Fig. 6-1). With both sides of the device n-doped (Vg =
or both sides p-doped (V. =
,.g = 10 V, denoted NN)
V, denoted PP), the device shows a nearly
._,= -10
ohmic current-voltage relation at low V,
(Fig. 6-1b). The device has a significantly
higher conductance in NN, most likely due to lower contact resistance between gold
and n-type WSe 2 , as observed in MoS
2
[36].
The current-voltage relations in these
configurations are slightly nonlinear, due to Schottky barriers that likely dominate
the contact resistance.
By oppositely biasing the two gates, the device rectifies current as a diode. Current
measurements for the two diode configurations (V
and VIg = 10 V,/.g = -10
= -10
V, V. 9 = 10 V, denoted PN
V, denoted NP) are shown in Fig. 6-1b. At low Vd8 , current
in both PP and NN is substantially larger than in NP and PN, demonstrating that
the p-n junction dominates transport in this regime rather than Schottky barriers at
the contacts. The effect of Schottky barriers is discussed further in Appendix B. Ids
versus V, for PN and NP are fit (Figs. ic and id) to the Shockley diode equation [114],
extended [115] to include a series resistance, R,:
n/
Ids =
W
Io R
Vds +Io Rs
xp
-UV0
where VT = kBT/q is the thermal voltage at temperature T, with kB the Boltzmann
constant and q, the electron charge.
1
o is the reverse-bias current, and n is the diode
ideality factor (n = 1 is ideal). W is the Lambert W function [116].
The rapid increase in current under forward bias defines an ideality factor n
1.9 for both PN and NP, indicating current mostly limited by recombination rather
than diffusion [114].
Investigating this recombination, including contributions from
Shockley-Read-Hall or Auger processes, will be a focus of future work. The roll-off of
Ids at high bias constrains R, = 3 MQ for PN and R, = 5 MQ for NP due to contact
resistances. Environmental conditions such as substrate and fabrication residue most
likely determine the contact resistance and the asymmetry between the PN and NP
configurations. An increase in current at high reverse bias indicates a 0.5 TQ shunt
resistance across the junction. This large shunt resistance indicates a high quality
73
p-n interface and is an expected advantage of a lateral device geometry. Though our
data do not strongly constrain the value of Io, because it is below the 1 pA noise floor
of the measurement, uncertainty in 1o has an insignificant effect on the fits to the
slope and roll-off of the current and does not impact the values of n and R,. Both
diodes have rectification factors of 105 and reverse bias currents less than 1 pA up to
IVdI = 1 V, promising characteristics for low-power electronics.
A more complete view of transport through the device is shown in a map of
current as a function of V1, and Vg (Fig. 6-2). The four corners of the map show the
extremes of the four doping configurations (NN, PP, PN, and NP). The off state of
the device can be seen in the dark blue region in the centre of the map, separating the
four conducting regions. Though mid-gap states, most likely due to disorder, form a
conducting region between the NN and NP quadrants, current through these states
is thermally activated and can be eliminated by cooling the device (see Appendix B).
Along the diagonal line defined by Vbg = Vig = Vg, line cuts show the device
operating as an ambipolar field effect transistor (Fig. 6-2c). Along the off-diagonal,
line cuts show gate-controlled rectifying behaviour (Fig. 6-2d) as a function of the
asymmetric gate voltage, 1V =
, = -Vg,
which defines a junction between the p-
and n-type regions. We note a decrease in device performance between the datasets
presented in Fig. 6-1 and Fig. 6-2. We suspect that this decrease stems from local
Joule heating at the contacts when the device is operated at high Vds.
74
=2V
V
(a)
10
VdS=-2V
10
ds(b)
I
s1 pA
V
1nA
5
5
0
-
0
-5
-10
-10
1pA
-5
N
-5
0
Vg (V)
5
10
(d)
(C) 10-6
1P
-10
-5
0
Vg (V)
5
10
1 0 -8
Vds= 2 V
Vds = -2 V
Vds- 2 V
Vds = -2 V
10
-
-
1010
10
-1
0-12-
PN
10-1
-10
- --
-5
0
Vbg M
5
10
-10
-
--
NP
-
1
1
10~12
-5
0
VJ (V)
5
10
Figure 6-2: Current through the device as a function of doping configuration. (a) and
(b) Current magnitude through the device, Ids , as the drain-source bias, Vda, is held
at 2 V (a) and -2 V (b), and the two back gates are varied independently. The colour
scale is logarithmic from 1 pA to 1 pA. (c) Diagonal cuts of U1dJ where both gates
are swept together as one back gate, V, - V1,
and -2
= V,,, for V, = 2 V (yellow curve)
V (blue curve). These curves are characteristic of an ambipolar field-effect
transistor. (d) Off-diagonal cuts of hiasl where the two gates are swept with opposite
polarity, showing the dependence of the current on the asymmetric gate voltage,
Vj = V
-- -- Vg, that defines the junction. The current was measured at V,
V (yellow curve) and -2
= 2
V (blue curve), demonstrating rectification of current in
opposite directions for p-n (Vj < 0) and n-p (Vj > 0) gate configurations.
75
6.5
Optoelectronics
The optoelectronic properties of monolayer WSe 2 p-n diodes were investigated with
scanning photocurrent microscopy (Fig. 6-4a), in which Ids is measured as a laser
(~ 2
spot
pm diameter) is scanned over the sample.
With the gates in the NP
configuration, photons impinging on the junction create electron-hole pairs that are
separated to opposite contacts by the electric field at the junction, generating a
photocurrent. Scanning the beam (power 75 jpW, wavelength 830 nm) over the sample
and measuring Ids at V, = 0 yields a spatial map of the photoresponse of the device
(Fig. 6-4b). Calibrating the photocurrent map with a simultaneously acquired image
of reflected light from the sample (Fig. B-4), we overlay the positions of the contacts
and gates to demonstrate that maximum photocurrent arises when light is incident
on the junction. A line cut through the centre of the photocurrent map is shown in
Fig. 6-4c with the positions of the contacts and the junction illustrated for reference.
The photocurrent is symmetric and centred on the junction, demonstrating that the
photoresponse is dominated by the p-n junction and not Schottky barriers at the
contacts (see Appendix B).
At relatively large V, in the NP configuration, a substantial photocurrent is
generated when the junction is illuminated with laser light at 532 nm.
Vd,
curves at laser powers 0-10 pW are shown in Fig. 6-4d.
Iph =
Ids,light -
Ids,dark, at Vds = -2
The Id,-
The photocurrent,
V and a linear fit of I,, up to 8 PW are
shown in the inset. The slope from the fit gives a responsivity of 210 mA/W, which
is comparable to commercial silicon photodetectors for green light. We note that
phototransistors based on monolayer dichalcogenides can achieve even higher responsivities [106], though the principle of operation and device geometry differ significantly
from the photodiodes presented here.
In addition to photodetection, monolayer WSe 2 p-n diodes are also capable of
photovoltaic power generation. With the device in the NP configuration, current is
measured as a function of V. for various laser powers from a 700 nm band-pass filtered
supercontinuum white-light source. Figure 4a depicts a zoomed-in view of the Ids-Vds
76
(a)
Photodiode
Iph (nA)
8
(b) 4
4
laser
0
2
beam splitter
V V
-,piezo
mirror
lenses
Vds
=S
>-
ds
0
-2
-4
V9
1.5
w
0
2
4
2
4
X (pm)
(d)
1
-2
Vrg
i
(c) 8
-
6
Junc.
2Vds=
8 pW
*9 1 4pW
2pW t~
OpW
,'
nph
,
0
0
0.5
4-
5
4
10
SIaser (pW)
-
2Contacts
0
M....
0
-2
-1
0
Vds (V
1
2
-4
-2
0
X(Pm)
Figure 6-3: Photodetection in monolayer WSe 2 . (a) Schematic of the scanning photocurrent microscopy setup and the device. The laser (solid red line) is focused onto
the sample through a microscope objective at a position set by a piezo-controlled
mirror. As the laser is scanned over the sample (represented by the dashed red line),
photocurrent is recorded simultaneously with the reflected laser power measured by
a photodiode. (b) Photocurrent, 'ph, as a function of laser position (75 puW, 830 nm
diode laser), with the device in the NP configuration and V, = 0 V. The peak in
current corresponds to the location of the junction. Outlines of the contacts (orange
lines) and gates (red lines) are overlaid on the current map based on the reflected
image (see Fig. B-4).
(c) Line cut from (b) at Y=0 (blue curve).
Outlines of the
contacts (orange rectangles) and the position of the junction (red dashed line) are
shown based on the reflected image. (d) Id,-V, characteristics with the device in
the NP configuration at laser powers 0-10 pW from a 532 nm diode laser. Inset: I.,,
(green dots) at Vd, = -2 V as a function of laser power. A linear fit (black dashed
line) to 'ph for powers 0-8 pW gives a responsivity of 210 mA/W.
77
curve, focusing on the quadrant of photovoltaic power generation. The short-circuit
current, I,
which is the zero-bias current through the illuminated device, increases
linearly with power up to at least 10 /uW (Fig. 6-3a inset). The power generated by
the photovoltaic device, P = Id
Vd, is shown for laser powers 0-10 puW in Fig. 6-3b.
The photovoltaic power generation also has a linear dependence on laser power (see
Fig. B-5).
Varying Vj, the asymmetric gate voltage that defines the junction, at different
laser powers, we observe a saturation in the short-circuit current when Vj
-
t5
V
(Fig. 6-3c). The current is higher for NP than for PN due to a difference in contact
resistances between the two contacts. We also note that the photocurrent due to
the photovoltaic effect in these WSe 2 diodes is approximately an order of magnitude
larger than photothermoelectric currents observed at the contacts of a monolayer
MoS 2 field-effect transistor [117].
To obtain spectrally resolved photocurrent in the NP configuration, we measure 1',
as a function of excitation wavelength. To quantify the efficiency of light conversion
into current, we extract the external quantum efficiency, EQE = (Isc/Paser)(hc/eA),
as a function of wavelength, A, at constant laser power, Piaser, where h, c, and e,
are Planck's constant, the speed of light, and electron charge, respectively. We observe three peaks in the spectrally resolved photocurrent at 755, 591, and 522 nm
(Fig. 6-3d), corresponding to energies of 1.64, 2.10, and 2.38 eV, respectively. These
energies match well with the values observed via photoluminescence [102], differential
reflectance [108], and optical absorption spectroscopy [118] for the A, B, and A' transitions of monolayer WSe 2 , as depicted in the band diagram in the inset to Fig. 6-3d.
We measure a maximum EQE of 0.2% at 522 nm.
This number does not take into
account the low absorption of monolayer WSe 2 or the narrow cross-section of the
p-n junction relative to the size of the laser spot, which together suggest an internal
quantum efficiency at least an order of magnitude larger than the EQE reported here.
Finally, we measure the electroluminescence spectrum of a second monolayer WSe 2
p-n diode (Fig. 6-3d). To improve hole injection, this device was fabricated with Pd
instead of Au to contact the p-type WSe 2 (see Fig. B-2). In the PN configuration
78
(a)
(b) 400
2
700 nm
10'
e,,' -1.5
10 p-4w
8 pW
300
-
0-
1
0
5
PIaser (pW)
10 VV
8 pW
(c)
0
100
2
/
-0.75
1
() 200
10
4pW
2 pW
0
4 pW
2 pW
-0.5
Vds M
0
-1
0
-0.25
-0.75
-0.5
Vds (
0
-0.25
(d)
- p
0
1
2
A
4E
Cb0
-
4
-WE
B
-2
-10
PN
-5
-
-0
Vj (V)
5
w
NP
10
10
Ek-
Cn
5
A
C
a)
B
1
-
-1
K
\Q-
A'
0
PN NP
NN PP
NP
500
750
10
0
0)0
A (nm)
Figure 6-4: Photovoltaic response and light emission. (a) Ids as a function of V, in
the NP configuration for laser powers 2 - 10 pW (wavelength 700 rim). Positive Ids
and negative Vds in this regime reflects photovoltaic power generation. Inset: Shortcircuit current, 'Sc (green dots), versus laser power with a linear fit (black dashed
line). (b) Power, P = Ids - V,, produced by the device as a function of Vd. for
different incident laser powers, calculated from the data in (a). (c) IU, as a function
of the the asymmetric gate voltage, Vi, for different laser powers. I,, is nearly linear
with power up to 10 piW in the NP configuration, though it saturates in the PN
configuration beyond 6 pW. As a function of gate voltage, the current saturates in
both configurations for V ~ ±5 V. (d) External quantum efficiency as a function
of wavelength at a constant laser power of 2 pW in the NP configuration (purple
line). Peaks in the external quantum efficiency correspond to exciton transitions A,
B, and A', as labeled. Right axis: Electroluminescence (EL) intensity from a second
monolayer WSe 2 device with one Au and one Pd contact. Vd, = 2 V in the PN
(blue trace), NN (yellow trace), aid PP (black trace) configurations, and Vd =-2
V in the NP (green trace) configuration. NN and PP traces are offset vertically for
clarity. Inset: Diagram of the band structure around the K and Q points, with arrows
indicating the lowest energy exciton transitions for monolayer WSe 2.
79
= 2 V, Ids = 100 nA), the device behaves as a light-emitting diode (LED).
(VS
The emitted light spectrum peaks at 752 un, corresponding to the direct-gap exciton
transition seen in the photocurrent spectrum (Fig. 6-3d). Using a blackbody source for
calibration, we estimate the electroluminescence efficiency, defined as optical output
power divided by electrical input power, to be approximately 1%. Light emission and
a peak at ~ 750 nm are also seen in the NP configuration (Vd,
= -2
V,
Is,
= 4
nA), with a peak height smaller in proportion to Is. No emission is seen in either
NN (with V, = 2 V and Ids = 300 nA) or PP (with V, = 2 V and Ids = 500 nA),
confirming that the gate-defined p-n junction generates the electroluminescence. A
spatial image of light emission from a third device is shown in Fig. B-6.
Based on the device performance presented here, we anticipate a prominent role for
diodes and optoelectronic devices based on monolayer dichalcogenide p-n junctions.
Taking into account the three-atom thickness and low optical absorption of monolayer
WSe 2 [118, 119] the responsivity and EQE reported here are quite substantial. Further, the device geometry could be optimised to significantly enhance photoresponse.
We expect that vertical junctions based on transfer-aligned exfoliated flakes [14] or
large-area dichalcogenides grown by chemical vapour deposition [118] could increase
responsivity and external quantum efficiency by more than an order of magnitude.
Additionally, improved contact resistance, particularly for holes, should dramatically
improve device performance.
6.6
Conclusions
In conclusion, we have demonstrated electrically tunable p-n diodes based solely on
monolayer WSe 2 . These diodes strongly rectify current, in a direction selectable by
the two gates controlling the device. Both the PN and NP configurations have diode
ideality factors of n = 1.9 and a rectification factor of 10'. With laser light incident
on the junction, these diodes produce a large photocurrent with a responsivity of 210
mA/W at high bias. At low bias, the diodes generate power via the photovoltaic
effect, with a peak external quantum efficiency of 0.2% at 522 un. The spectral
80
response of the photocurrent from visible to near infrared wavelengths showed peaks
corresponding to the three lowest excitonic transitions expected for monolayer WSe 2 .
Finally, these devices also function as light-emitting diodes with an electroluninescence peak at 752 urn.
These p-n diodes demonstrate the potential of monolayer
WSe 2 , in addition to other direct gap semiconducting dichalcogenides, for novel electronic and optoelectronic applications. As device quality improves, they also lay the
foundation for more fundamental quantum transport experiments [120].
Note added: During the final preparation of this manuscript we became aware of
similar work on p-n diodes inimonolayer WSe2 [121, 122].
6.7
Methods
Device fabrication begins with exfoliation of bulk, natural WSe 2 (Nanosurf Inc.) down
to few-layered sheets using the mechanical cleavage method pioneered for graphene
[9].
The thin flakes are deposited onto a transfer slide made as a stack of glass, a polymuer (polydimethylsiloxane [PDMS]), and a resist (methyl methacrylate [MMA]), as
described for graphene-boron nitride device fabrication [14, 113].
Single molecular
layers are identified by optical contrast. Layer number is later confirmed with atomic
force microscopy and either photocurrent spectroscopy or electroluminescence (see the
Device Fabrication section of Appendix B). These monolayers are transferred onto a
pair of split gates covered by 20 nm of HfO 2 (grown by atomic layer deposition at 80
C). The gates are separated by a 100 nm gap, patterned using e-beam lithography
on a highly doped Si substrate covered in 285 nm of thermally grown SiO 2 . They
are made from e-beam evaporated gold and are 20 nim thick. The two back gates are
capacitively coupled to the device through the HfO 2 dielectric, which has a dielectric
constant, Er~ 15. The WSe 2 is contacted by two gold electrodes, each approximately
1 /-mt wide and 25 nmi thick with a 0.3 nm chromium sticking layer.
All measurements were performed at room temperature and in vacuum (-
10-
torr) to avoid device degradation from adsorbates present in air, which could be
imitigated by encapsulation of the WSe 2 . Electrolumirnescence (EL) was measured
81
using a liquid nitrogen cooled charged-coupled device with an integration time of 60
seconds. A background measured at Vd, = 0 was subtracted from all four EL traces
in Fig. 6-3d.
82
Appendix A
MoS 2 Supplementary Information
A.1
Fabrication
The devices presented here were fabricated starting from mined, naturally-grown,
flake MoS2 acquired from SPI Supplies. MoS 2 was exfoliated from the bulk material down to atomic layers using a mircomechanical cleavage process standard for
graphene [1].
The flakes were deposited onto highly doped silicon substrates with
a 285 urn, thermally-grown, silicon-oxide dielectric layer. Monolayers and bilayers
were identified by optical contrast [90] using a Zeiss Axio-Inager microscope. Optical identification was calibrated and checked with AFM measurements (Fig. A-1).
Step heights fron the substrate to the flake were often found to be slightly larger
than the predicted inter-layer spacing for MoS 2 (0.62 nmi)
[123]. As is common with
graphene depositions on SiO 2 , the discrepancy was reduced after thermal annealing
in Ar/H 2 , though the step would still remain marginally higher than expected. Layer
to layer measurements, however, match very well with predicted inter-layer spacing
(Fig. A-1).
Once identified, candidate flakes were contacted using PMMA masks patterned
with e-beam lithography and developed in cold (-
5 'C) MIBK:IPA, 1:1. Contacts
were then deposited using either e-beam or thermal evaporators, with gold contacts
50-100 num thick, and titanium sticking layers 0.3-3.5 nm thick (Table A.1).
annealing, device B2 was annealed in an Ar/H
83
2
After
environment at 350 'C for 3 hours,
AFM
(a)
2 nm (b)
Step Height
2
Bilayer
, 1.5
E
1.5
1.0
.
0.5
0.5
0
0
0
0.67 nm
0.85 nm
0.2
0.4 0.6 0.8
Distance (pm)
1
Figure A-1: MoS 2 flake AFM and step heights. (a) Atomic force micrograph of
an MoS 2 flake containing both monolayer and bilayer regions deposited on an SiO 2
substrate. The light blue (monolayer) and dark blue (bilayer) lines added show the
position of the step heights in (b). Scale bar is 2 Am. Inset: Optical image of the
flake in (a) showing the color contrast change with layer number. Scale bar is 2 Pm.
(b) Step height for monolayer (light blue) and bilayer (dark blue) regions of an MoS 2
flake.
though this procedure did not seem to have a substantial effect on quality. Device
B2 was also etched into a Hall bar pattern using oxygen plasma. All other devices
reported on in chapter 5 were unetched.
All of the devices were annealed in situ at high temperatures (-
120 'C) in vac-
uum (~10-6 mbar) for 5-20 hours. This procedure caused a substantial drop in two
terminal resistivity (Fig. A-2) and the elimination of Schottky behavior in the contacts
down to 4 K.
Table A.1: Device Fabrication Parameters
Device
M1
M2
M3
B1
B2
Length
4.0 pm
3.5 pm
3.5 pan
4.6 jim
5.0 pim
Width
2.6 pm
4.2 pm
3.2 pm
5.1 Am
0.6 pm
Sticking Layer (Ti)
0.3 nm
0.3 nm
0.3 nm
0.3 nm
3.5 nm
84
Metal (Au)
100 nm
60 nm
60 nm
100 nm
50 nm
Vacuum
120 0 C,
120 'C,
120 'C,
120 0 C,
100 OC,
Annealing
20 hours
10 hours
10 hours
10 hours
5 hours
(a)
Annealing
107
Monolayer
(b) 1012
T=1200C
50 Vbg
109
-
Before
106
T=20 C
Monolayer
5
10
After
106
Bilayer
0
9
Time (hours)
3
6
12
10
0
10
20
30
40
50
Vbg (
Figure A-2: Effect of annealing on two-terminal resistance of MoS 2 devices. (a)
Two-terminal resistance change over time for monolayer (yellow) and bilayer (blue)
devices being annealed at 120 'C in vacuum. (b) Back-gate sweeps of a monolayer
device before and after vacuum annealing. Data from devices Mi and Bi.
85
A.2
Metal Insulator Transition
The Ioffe-Regel condition for two-dimensional semiconductors states that kF *A
-
at a metal-insulator transition [124], for the Fermi wave vector, kF
and
\/2w7n
1
the mean free path, A = hkF u/nc 2 . This leads to the expectation that the critical
resistivity, pc, will be of order h/c 2 . Our results match this condition, though the
mean,
h/2e2 , is slightly below the predicted value.
Table A.2: Metal Insulator Transition
Device
Critical Resistivity,
pC (h/e 2 )
M1
0.9
M2
0.8
M3
0.7
B1
0.4
B2
0.3
Table A.2: Resistivity of monolayer and bilayer MoS 2 at the metal insulator transition.
Four-terminal critical resistivity, pc, of monolayer and
bilayer devices. The values mark the resistivity at which the lowest temperature traces of p vs. Vbg cross.
A.3
Hall Measurements
Density, n, and Hall mobility, I'H, were calculated from Hall measurements performed
in a magnetic field. The current was sourced longitudinally along the device from one
of the outer contacts. The transverse voltage, Vy, was then measured across the
sample. The Hall resistance was calculated as R
I/Voy. Fitting a linear trend
to RY as it varies with magnetic field yields the Hall coefficient, RH= dR '/dB.
(Fig. A-3). There was a non-zero transverse resistance at zero field due to imperfec86
tions in contact alignment and sample inhomogeneities. This offset was subtracted
from the data as it was small and irrelevant to our fit of the slope.
Monolayer
(a)
(b)
Density
3
25
-/
Monolayer
2
75 Vb.
0[
25 V
o Vbg
E
0:
1
Bilayer
-251
T=10K-
-1
1
0
B (T)
n
00
25 V
o5 Vbg
200
100
300
T (K)
Figure A-3: Hall measurements of MoS 2 . (a) Hall resistance of a monolayer MoS 2
device. Hall resistance, Rx,
of a monolayer device as a function of magnetic field,
B, measured at 10 K. Traces shown are at back-gate voltages of 25 (blue) and 75
(yellow). The dotted lines show linear fits to the traces. (b) Density as a function of
temperature for monolayer and bilayer MoS 2 at 0, 25, and 50 Vbg, colored from dark
to light. Data from devices M1 and B2.
A.4
Mobility
If pH varies linearly with density, the field-effect mobility will vary with density
with twice the slope. To show this, we write pH = A - n + C, where A and C are
constants. Then PFE
dp-rE;/dn
2A
PH
+ n dtH/dn = A -n +
C + n - (A)
2 - dpH/dn, as stated in chapter 5.
87
= 2A . n + C. Then
(a)
Monolayer
10
106
Monolayer
(b) 105
T=5K
-50 Vbg
-25Vbg
0 Vbg
4
,-0C
0
10
10
101K
10
200K
300K
-50
-25
25
0
Vbg (V)
103
50
3
100
10
300
T (K)
Figure A-4: Contact resistance and four-terminal resistivity of monolayer device M.
(a) Four-terminal resistivity of device M1 as a function of Vbg. Curves colored from
red to black show measurements at 300, 200, 100, 50, and 5 K, respectively. Inset:
Contact resistance, Rc, to device M1 as a function of
Vb,
at 5 Kelvin. (b) Resistivity of
device M1 as a function of temperature. The curves, from black to orange, correspond
to V, = -50, -25, 0, 25, and 50 V.
(a)
150
Room Temp. Mobility
(b)
T=20 0 C
Leakage
4
2
1001
C
'0'
E
CU
50
-2
PF
M B
PH
Oh
-25
M2 B2
Monolayer
0
25
50
Vbg (V)
75
-4
-50
0
50
VbgV)
Figure A-5: Room temperature mobilities and leakage current. (a) Hall and field
effect mobilities for monolayer and bilayer devices at room temperature. Data from
devices M1 and B2. (b) Representative leakage currents measured for each device
presented in the text.
88
A.5
Acknowledgments
We would like to thank Lili Yu and Han Wang for their help with early device fabrication and measurements, and Andrea Young for many fruitful discussions. This
work was funded by the ONR GATE MURI and a Packard Fellowship. This work
made use of the MRSEC Shared Experimental Facilities supported by NSF under
award No. DMR-0819762 and of Harvard's CNS, supported by NSF under grant No.
ECS-0335765.
89
90
Appendix B
WSe 2 Supplementary Information
B.1
Device Fabrication
Devices for this paper were fabricated from a natural crystal of WSe 2 from Nanosurf.
The bulk crystal was exfoliated down to few-layered sheets by mechanical cleavage
using a semiconductor grade low-tack tape. The thin flakes were then deposited onto
a glass/PDMS/MMA transfer slide. The transfer slide is built as a stack of a glass
slide, PDMS ~ 2 mm thick, a piece of clear packing tape (the tape improves smooth
MMA adhesion), and a bilayer of MMA. The transfer slide is then searched using an
optical microscope, and single molecular layers are identified by optical contrast
[90].
Flake thickness is later confirmed with AFM (see Fig. B-1) and either photocurrent
spectroscopy or electroluminescence spectroscopy.
The gates were patterned using e-beam lithography onto a highly doped Si substrate covered in 285 urn of thermally grown SiO 2 . The gates are Au for the main
device used for most measurements and AuPd for the electroluminescence (EL) device
(see Fig. B-2). They were evaporated in an e-beam evaporator 20-30 nm thick with
a 0.7 nin Cr sticking layer, and are separated by a 100 inmi gap for the main device
and a 300 rim gap for the EL device. They are both covered in 20 nim of HfO 2 grown
by atomic layer deposition at a terriperature of 80 C.
Monolayer WSe 2 was transferred onto the gates using a mnicromianipulator and
a microscope with a long working distance objective. The WSe 2 sits on top of the
91
Height (nm)
(a)
(b)
5
20
0.9 nm
2.5
10
t~~
--
-.~
.
-2.5
-5
-5
-2.5
0
X (pm)
2.5
5
Figure B-1: Optical micrograph and AFM with step height of the main device. (a)
Optical micrograph of the main device after transfer to pre-patterned split gates. The
consistent faint color of the flake is indicative of a monolayer. The scale bar is 2pm.
(b) Atomic force micrograph of the right side of the device. This AFM image was
taken after the left side of the device was destroyed by electrostatic shock. Inset:
Step height of the WSe 2 flake showing a step of 0.9 nm, which is consistent with one
molecular layer of WSe 2 with some fabrication residue on an uneven surface.
HfO 2 and is contacted by two Au contacts for the main device and one Au and one
Pd contact for the EL device. The contacts are each approximately 1 pm wide. The
main device contacts are 25 nm thick with a 0.3 nm chromium sticking layer, and the
EL device contacts are 50 nm thick with a 0.7 nm chromium sticking layer.
Fig. B-1 shows an optical micrograph of the main device on the split gates after
transfer. An atomic force micrograph of the right side of the device is also shown along
with the step height of the WSe 2 , confirming that it is a monolayer. The measured
thickness of 0.9 nm is consistent with a monolayer with some fabrication residue on
an uneven surface and is appreciably less than the 1.4 nm thickness of a bilayer. Photocurrent spectroscopy and electroluminescence spectroscopy also corroborate this
identification.
92
EL Device
Figure B-2: Optical micrograph the EL device. Optical micrograph of the device
used to measure electroluminescence. The WSe 2 flake sits on two AuPd split gates
separated by 300 nm and covered in 20 nm of HfO2 . It is contacted by Pd on the left
and Au on the right. The scale bar is 2pm.
93
B.2
Mid-Gap Current
It is expected that the band gap of WSe 2 will create an insulating state between the
four conducting quadrants in gate-voltage space (NN, PP, PN, and NP). In Fig. 6-2 of
the main text, however, current through the device does not drop to zero between NN
(upper right corner) and NP (lower right corner) at room temperature. By cooling
the device to 200 K, the conducting states disappear and the gap region becomes fully
insulating (see Fig. B-3). We speculate that this behavior reflects thermal activation
of mid-gap states created by disorder in the sample.
Vds=-2V
(a) 10
(b) 10
(C) 10
5
5
>E 0
>e 0
l-2s
pA
2K22
KInA
2, 0
-5
-10
lpA
-5
-5
-10
0
V (V)
5
10
-10-10
-10
-5
0
V" (V)
5
10
-10
-5
0
V9 V
5
10
(d) 10-6
10"
10
270 K
200 K
10
-10
-5
0
5
10
Figure B-3: Temperature dependence of mid-gap current. (a) A logarithmic color
plot of Ids at V,
= -2 V as the two back gates are varied independently. The map
was taken at room temperature (290 K), and mid-gap current is visible connecting the
NP and NN quadrants with a conducting region. (b) and (c) The same measurement
as in (a), at 270 K in (b) and 200 K (c), showing the diminishing conductance of the
mid-gap states. (d) Line cuts along Vg with Vg fixed at 10 V. By 200 K, the mid-gap
current is completely suppressed.
94
B.3
Schottky Barriers
In this section we discuss the role of Schottky barriers on the performance of the
p-n diodes presented in chapter 6.
Schottky barriers likely dominate the current
through the device near zero gate voltage. However, the measurements presented in
this paper were almost all obtained near the maximum gate voltage allowed before
dielectric breakdown, minimizing the effects of any barriers present. In this regime of
high electrostatic doping, several aspects of our measurements demonstrate that the
transport characteristics of these devices are dominated by the p-n junction defined
by the gates and not by residual Schottky barriers:
First, Figure lb shows that nearly ohmic contact is achievable at high carrier
density for both contacts at both n-type and p-type doping (NN and PP configurations). The Id,-V, curves for the NN and PP configurations are slightly asymmetric,
indicating the presence of residual Schottky barriers even at high carrier density, but
it is important to note that the currents for both the PP and NN configurations
are much larger at low source-drain bias than the NP and PN configurations. This
demonstrates the dominant role of the p-n junction. For instance, at Vds = 0.5 V
we see that the relatively weak PP current is four orders of magnitude larger the the
current in either PN or NP (~20 nA vs. 1 pA). Because of the small cross-coupling of
gate capacitances (which can be seen in Fig. 6-2 of the main text), the widths of the
Schottky barriers in the PN and NP configurations are expected to be the same as in
the PP and NN configurations for the same gate voltage magnitudes. Thus, current
through the device in PN and NP is dominated by the junction.
In addition to this, the presence of a strong photocurrent signal in the PN and NP
configurations further supports the conclusion that Schottky barriers at the contacts
have only a weak influence on the device characteristics. Though band bending from
Schottky barriers generally produces photocurrent, the addition of a p-n junction
between the contacts suppresses the photocurrent contribution from the Schottky
barriers [125].
In such a configuration, the carriers generated at the contacts are
blocked by the depletion region at the p-n junction. A p-n junction capable of blocking
95
photocurrent from the contacts is certainly formed in our devices, as demonstrated
by the suppression of the drain-source current at low bias, presented in Fig. 6-1 of the
main text. Similarly, photocurrent generated by the p-n junction could be blocked by
Schottky barriers at the contacts. If Schottky barriers dominated the flow of current
through our devices, we would expect little or no photocurrent in the PN and NP
configurations, contrary to what is observed.
Finally, the emission of light in the PN and NP configurations and not the NN
and PP configurations (Figure 4d) provides another indication that the p-n junction
and not residual Schottky barriers dominates the performance of these devices. We
note in chapter 6 that asymmetric contacts (one Au, one Pd) were used specifically
to reduce the influence of contacts on the performance of the device. If the device
behavior were, nevertheless, Schottky-dominated, we would expect light emission
from the Schottky barriers for all four doping configurations [93], in contrast with
our observations. In principle, some electroluminescence from the residual Schottky
barriers in our devices should still be observable for NN and PP. But the fact that
these emissions are below the sensitivity of our measurement indicates that they are
at least two orders of magnitude weaker than the electroluminescence peak in the PN
configuration.
Contact resistance is expected to reduce the magnitude of photocurrent generated
by the p-n junction, negatively impacting the photovoltaic EQE and the maximum
current that limits electroluminescence yield. For the photovoltaic performance of
the device, we note in chapter 6 that the EQE of 0.2% is limited by the two dominant
factors of the 10% light absorption by monolayer WSe2 and the narrow p-n junction
formed by our lateral gate geometry. Both of these effects reduce the efficiency of the
device by approximately a factor of 10. Further losses may be limited by the residual
Schottky barriers, recombination of electron-hole pairs, or other factors, but are not
the dominant effects presently limiting PV efficiency.
96
Iph (nA)
(a) 4
Vph (MV)
30
8 (b) 4
4
2
10
0
2
=0
0
-2
-4
-4
15
-2
-2
0
X (pm)
2
-4
-4
4
-2
0
X (pm)
2
4
Figure B-4: Photocurrent and reflected light image. (a) Color plot of the photocurrent, Iph, as the laser spot (power 75 pW, wavelength 830 nin) scans over the device.
The peak in current corresponds to the location of the p-n junction.
Outlines of
the contacts and gates taken from the reflected image are overlaid for context. The
scan was taken with the device in the NP configuration (ig
= 10 V, 1V g = -10
V)
at zero bias. (b) The corresponding reflected image taken simultaneously with the
photocurrent map. The outlines of the contacts and gates are drawn over the image.
97
PV Power
Annfl
,0
300
-
~0
Si
0.
0/
200
-
0
0~
0'
100
-
0
0
.
2
.
6
4
8
10
Plaser (pW)
Figure B-5: Dependence of photovoltaic power generation on laser power. Plot of the
maximum power output from the device as a function of laser power. The data is
plotted in green circles and the black dashed line shows a linear fit to the data. Laser
wavelength: 700nm.
Figure B-6: Reflected and emitted light image. A third monolayer WSe 2 device,
configured as an LED (PN configuration. V
= 2 V, Id, = 25 nA) and simultaneously
illuminated with a halogen lamp. The emitted and reflected light are imaged with
a liquid nitrogen cooled CCD (integration time 10 s, color scale represents counts
recorded by the monochrome CCD). This device has one Au and one Pd contact.
The scale bar is 5 pym.
98
B.4
Acknowledgements
We would like to acknowledge experimental assistance from Kaiphol Akkaravarawong, Trond Andersen, Nathaniel Gabor, Qian Lin, Qiong Ma, Wenjing Fang, and
Javier Sanchez-Yamagishi and discussions with Patrick Brown. This work has been
primarily funded by ONR Young Investigator Award N00014-13-1-0610, and partly by
the ONR GATE MURI and a Packard Fellowship. This work made use of the MRSEC
Shared Experimental Facilities supported by the NSF under award No. DMR-0819762
and of Harvard's CNS, supported by the NSF under grant ECS-0335765.
99
100
Bibliography
[1] K S Novoselov, A K Geim, S V Morozov, D Jiang, Y Zhang, S V Dubonos, I V
Grigorieva, and A A Firsov. Electric field effect in atomically thin carbon films.
Science, 306(5696):666-669, 2004.
[2] J. Scott Bunch, Scott S. Verbridge, Jonathan S. Alden, Arend M. van der Zande,
Jeevak M. Parpia, Harold G. Craighead, and Paul L. McEuen. Impermeable
atomic membranes from graphene sheets. Naro Letters, 8(8):2458-2462, 2008.
[3] Changgu Lee, Xiaoding Wei, Jeffrey W. Kysar, and James Hone. Measurement
of the elastic properties and intrinsic strength of monolayer graphene. Science,
321(5887):385--388, 2008.
[4] Kin Fai Mak, Matthew Y. Sfeir, Yang Wu, Chun Hung Lui, James A. Misewich,
and Tony F. Heinz. Measurement of the optical conductivity of graphene. Phys.
Rev. Lett., 101:196405, Nov 2008.
[5] K.I. Bolotin, K.J. Sikes, Z. Jiang, M. Klima, G. Fudenberg, J. Hone, P. Kim,
and H.L. Stormer. Ultrahigh electron mobility in suspended graphene. Solid
State ComrUnications,146(910):351 - 355, 2008.
[6] P. R. Wallace. The band theory of graphite. Phys. Rev., 71:622-634, May 1947.
[7] J W McClure. Diamagnetism of graphite. Physical Review, 104(3):666, 1956.
[8] J. C. Slonczewski and P. R. Weiss.
Band structure of graphite.
Phys. Rev.,
109:272-279, Jan 1958.
[9] K S Novoselov, D Jiang, T Booth, V V Khotkevich, S M Morozov, and A K
Geim.
Two Dimensional Atomic Crystals.
Proc. Nati. Acad. Sci. U. S. A.,
102(cond-mat/0503533):10451, March 2005.
[10] Jae Hun Seol, Insui Jo, Arden L. Moore, Lucas Lindsay, Zachary H. Aitken,
Michael T. Pettes, Xuesong Li, Zhen Yao, Rui Huang, David Broido, Natalio
Mingo, Rodney S. Ruoff, and Li Shi. Two-dimensional phonon transport in
supported graphene. Science, 328(5975):213 -216, 2010.
[11] A. K. Geim and K. S. Novoselov.
The rise of graphene.
6:183-191, 2007.
101
Nature Materials,
[12] Andre Konstantin Geim.
Graphene:
status and prospects.
science,
324(5934):1530-1534, 2009.
[13] Phaedon Avouris and FN Xia. Graphene applications in electronics and photonics. MRS Bull, 37(12):1225-34, 2012.
[14] C. R. Dean, A. F. Young, I. Meric, C. Lee, L. Wang, S. Sorgenfrei, K. Watanabe,
T. Taniguchi, P. Kim, K. L. Shepard, and J. Hone. Boron nitride substrates for
high-quality graphene electronics. Nat Nano, 5(10):722-726, 2010.
[15] Kenji Watanabe, Takashi Taniguchi, and Hisao Kanda. Direct-bandgap properties and evidence for ultraviolet lasing of hexagonal boron nitride single crystal.
Nat Mater, 3(6):404-409, 2004. 10.1038/nmatll34.
[16] Yuanbo Zhang, Yan-Wen Tan, Horst L. Stormer, and Philip Kim. Experimental
observation of the quantum hall effect and berry's phase in graphene. Nature,
438:201-204, 2005.
[17] T Ohta, A Bostwick, T Seyller, K Horn, and E Rotenberg. Controlling the
electronic structure of bilayer graphene. Science, 313(5789):951-954, 2006.
[18] SY Zhou, DA Siegel, AV Fedorov, and A Lanzara. Metal to insulator transition
in epitaxial graphene induced by molecular doping. Physical review letters,
101(8):086402, 2008.
[19] SY Zhou, G-H Gweon, AV Fedorov, PN First, WA De Heer, D-H Lee, F Guinea,
AH Castro Neto, and A Lanzara. Substrate-induced bandgap opening in epitaxial graphene. Nature materials, 6(10):770-775, 2007.
[20] Jeroen B Oostinga, Hubert B Heersche, Xinglan Liu, Alberto F Morpurgo, and
Lieven MK Vandersypen. Gate-induced insulating state in bilayer graphene
devices. Nature materials, 7(2):151-157, 2008.
[21] Fengnian Xia, Damon B Farmer, Yu-ming Lin, and Phaedon Avouris. Graphene
field-effect transistors with high on/off current ratio and large transport band
gap at room temperature. Nano letters, 10(2):715-718, 2010.
[22] Yuanbo Zhang, Tsung-Ta Tang, Caglar Girit, Zhao Hao, Michael C Martin,
Alex Zettl, Michael F Crommie, Y Ron Shen, and Feng Wang. Direct observation of a widely tunable bandgap in bilayer graphene. Nature, 459(7248):820823, 2009.
[23] HJ Mamin and D Rugar. Sub-attonewton force detection at millikelvin temperatures. Applied Physics Letters, 79(20):3358-3360, 2001.
[24] MD LaHaye, Olivier Buu, Benedetta Camarota, and KC Schwab. Approaching
the quantum limit of a nanomechanical resonator. Science, 304(5667):74-77,
2004.
102
[25] YT Yang, Carlo Callegari, XL Feng, KL Ekinci, and ML Roukes. Zeptogramscale nanomechanical mass sensing. Nano letters, 6(4):583---586, 2006.
[26] KL Ekinci and ML Roukes. Nanoelectromechanical systems. Review of scientific
instrmernts, 76(6):061101, 2005.
[27] Changyao Chen, Sami Rosenblatt, Kirill I. Bolotin, William Kalb, Philip Kim,
Ioannis Kyrissis, Horst L. Stormer, Tony F. Heinz, and James Hone. Performance of monolayer graphene nanomechanical resonators with electrical readout. Nature Nanotechnology, 4(12):861-867, 2009.
[28] Vibhor Singh, Shaiashis Sengupta, Hari S Solanki, Rohan Dhall, Adrien Allain,
Sajal Dhara, Prita Pant, and Mandar M Deshmukh. Probing thermal expansion
of graphene and modal dispersion at low-temperature using graphene nanoelectromechanical systems resonators. Nanotechnology, 21(16):165204, 2010.
[29] Arend M. van der Zande, Robert A. Barton, Jonathan S. Alden, Carlos S. RuizVargas, William S. Whitney, Phi H. Q. Pham, Jiwoong Park, Jeevak M. Parpia,
Harold G. Craighead, and Paul L. McEuen. Large-scale arrays of single-layer
graphene resonators. Nano Letters, 10(12):4869-4873, 2010.
[30] J. Scott Bunch, Arend M. van der Zande, Scott S. Verbridge, Ian W.
Frank, David M. Tanenbaum, Jeevak M. Parpia, Harold G. Craighead, and
Paul L. McEuen. Electromechanical resonators from graphene sheets. Science,
315(5811):490-493, 2007.
[31] D. Garcia-Sanchez, A. M. van der Zande, A. San Paulo, B. Lassagne, P. L.
McEuen, and A. Bachtold. Imaging mechanical vibrations in suspended
graphene sheets. Nano Letters, 8(5):1399-1403, 2008.
[32] J.A. Wilson and A.D. Yoffe. The transition metal dichalcogenides discussion
and interpretation of the observed optical, electrical and structural properties.
Advances in Physics, 18(73):193-335, 1969.
[33] T Li and G Galli. Electronic properties of MoS2 nanoparticles. The Journal of
Physical Chemistry C, 111(44):16192-16196, 2007.
[34] S Lebegue and 0 Eriksson. Electronic structure of two-dimensional crystals
from ab initio theory. Physical Review B, 79(11):115409, 2009.
[35] Kin Fai Mak, Chaniggu Lee, James Hone, Jie Shan, and Tony F. Heinz. Atomically thin muos 2 : A new direct-gap semiconductor. Phys. Rev. Lett., 105:136805,
Sep 2010.
[36] B Radisavljevic, A Radenovic, J Brivio, V Giacometti, and A Kis. Single-layer
MoS2 transistors. Nature Nanotechnology, 6(3):147--150, 2011.
103
[37] Daniele Braga, Ignacio Gutierrez Lezama, Helmuth Berger, and Alberto F Morpurgo. Quantitative Determination of the Band Gap of WS 2with Ambipolar
Ionic Liquid-Gated Transistors. Nano Letters, 12(10):5218-5223, October 2012.
[38] H Fang, S Chuang, T C Chang, K Takei, T Takahashi, and A Javey. HighPerformance Single Layered WSe 2 p-FETs with Chemically Doped Contacts.
Nano Letters, 12(7):3788-3792, 2012.
[39] Stefano Larentis, Babak Fallahazad, and Emanuel Tutuc. Field-effect transistors and intrinsic mobility in ultra-thin mose 2 layers. Applied Physics Letters,
101(22):-, 2012.
[40] Hee Sung Lee, Sung-Wook Min, Youn-Gyung Chang, Min Kyu Park, Taewook
Nam, Hyungjun Kim, Jae Hoon Kim, Sunmin Ryu, and Seongil Im. MoS 2
Nanosheet Phototransistors with Thickness-Modulated Optical Energy Gap.
Nano Letters, 12(7):3695-3700, July 2012.
[41] M W Lin, L Liu, Q Lan, X Tan, K S Dhindsa, P Zeng, V M Naik, M M C
Cheng, and Z Zhou. Mobility enhancement and highly efficient gating of monolayer MoS2 transistors with polymer electrolyte. Journal of Physics D: Applied
Physics, 45(34):345102, 2012.
[42] Han Liu and Peide D Ye. MoS 2 Dual-Gate MOSFET With Atomic-LayerDeposited A1203 as Top-Gate Dielectric. IEEE Electron Device Letters,
33(4):546-548, 2012.
[43] Saptarshi Das, Hong-Yan Chen, Ashish Verma Penumatcha, and Joerg Appenzeller. High performance multilayer mos 2 transistors with scandium contacts.
Nano Letters, 13(1):100-105, 2013.
[44] De-en Jiang, Valentino R. Cooper, and Sheng Dai. Porous graphene as the
ultimate membrane for gas separation. Nano Letters, 9(12):4019--4024, 2009.
PMID: 19995080.
[45] Joel Moser, Amelia Barreiro, and Adrian Bachtold. Current-induced cleaning
of graphene. Applied Physics Letters, 91(16):163513, 2007.
[46] DC Elias, RV Gorbachev, AS Mayorov, SV Morozov, AA Zhukov, P Blake,
LA Ponomarenko, IV Grigorieva, KS Novoselov, F Guinea, et al. Dirac cones reshaped by interaction effects in suspended graphene. Nature Physics, 7(9):701704, 2011.
[47] Monica T Allen, Jens Martin, and Amir Yacoby. Gate-defined quantum confinement in suspended bilayer graphene. Nature communications, 3:934, 2012.
[48] PS Spinney, DG Howitt, SD Collins, and RL Smith. Electron beam stimulated
oxidation of carbon. Nanotechnology, 20(46):465301, 2009.
104
[49] Max C. Lemnie, David C. Bell, James R. Williams, Lewis A. Stern, Britton
W. H. Baugher, Pablo Jarillo-Herrero, and Charles M. Marcus. Etching of
graphene devices with a helium ion beam. ACS Nano, 3(9):2674-2676, 2009.
PMID: 19769403.
[50] Max C. Leinme, T.J. Echterineyer, M. Baus, and H. Kurz. A graphene fieldeffect device. Electron Device Letters, IEEE, 28(4):282-284, April 2007.
[51] Melinda Y. Han, Barbaros Ozyilnaz, Yuanbo Zhang, and Philip Kim. Energy
band-gap engineering of graphene nanoribbons. Phys. Rev. Lett., 98:206805,
May 2007.
[52] Zhihong Chen, Yu-Ming Lin, Michael J. Rooks, and Phaedon Avouris.
Graphene nano-ribbon electronics. Physica E: Low-dimensional Systems and
Nanostructures, 40(2):228 - 232, 2007. International Symposium on NanometerScale Quantum Physics.
[53] L. A. Ponomarenko, F. Schedin, M. I. Katsnelson, R. Yang, E. W. Hill, K. S.
Novoselov, and A. K. Geii. Chaotic dirac billiard in graphene quantum dots.
Science, 320(5874):356-358, 2008.
[54] C. Stampfer, J. Gttinger, F. Molitor, D. Graf, T. Ihn, and K. Ensslin. Tunable
coulomb blockade in nanostructured graphene. Applied Physics Letters, 92(1):-,
2008.
[55] Sujit S. Datta, Douglas R. Strachan, Sainuel M. Khamnis, and A. T. Charlie Johnson. Crystallographic etching of few-layer graphene. Nano Letters,
8(7):1912-1915, 2008.
[56] Xiaolin Li, Xinran Wang, Li Zhang, Sangwon Lee, and Hongjie Dai. Chemically derived, ultrasmooth graphene nanoribbon semiconductors. Science,
319(5867):1229-1232, 2008.
[57] D. V. Kosynkin, A. L. Higginbotham, A. Sinitskii, J. R. Lomeda, A. Dimiev,
B. K. Price, and J. M. Tour. Longitudinal unzipping of carbon nanotubes to
form graphene nanoribbons. Nature, 458:872876, 2009.
[58] L. Jiao, L. Zhang, X. Wang, G. Diankov, and H. Dai. Narrow graphene nanoribbons from carbon nanotubes. Nature, 458:877880, 2009.
[59] Michael D. Fischbein and Marija Drndi. Electron beam rnanosculpting of suspended graphene sheets. Applied Physics Letters, 93(11):-, 2008.
[60] J. Morgan, J. Notte, R. Hill, and B. Ward. An introduction to the helium ion
microscope. Microscope Today, 14:24-31, 2006.
[61] L. Scipioni, L. Stern, and J. Notte. Applications of the helium ion microscope.
Microscope Today, 15:12-14, 2007.
105
[62] David C. Bell. Contrast mechanisms and image formation in helium ion microscopy. Microscopy and Microanalysis, 15:147-153, 4 2009.
[63] D. C. Bell, M. C. Lemme, J. R. Williams, L. A. Stern, and C. M. Marcus.
Helium ion etching of graphene. Nanotechnology, 2009.
[64] X Du, I Skachko, A Barker, and E Y Andrei. Approaching ballistic transport
in suspended graphene. Nature Nanotechnology, 3(8):491-495, 2008.
[65] J R Williams, L DiCarlo, and C M Marcus. Quantum Hall effect in a gatecontrolled pn junction of graphene. Science, 317(5838):638-641, 2007.
[66] P Blake, EW Hill, AH Castro Neto, KS Novoselov, D Jiang, R Yang, TJ Booth,
and AK Geim. Making graphene visible. Applied Physics Letters, 91(6):063124063124, 2007.
[67] Jian-Hao Chen, Chaun Jang, Shudong Xiao, Masa Ishigami, and Michael S
Fuhrer. Intrinsic and extrinsic performance limits of graphene devices on sio 2 .
Nature nanotechnology, 3(4):206-209, 2008.
[68] Vera Sazonova, Yuval Yaish, Hande Ustiinel, David Roundy, Tomis A Arias,
and Paul L McEuen. A tunable carbon nanotube electromechanical oscillator.
Nature, 431(7006):284-287, 2004.
[69] C. Chen. Graphene NanoElectroMechanicalResonators and Oscillators. PhD
thesis, Columbia University, 2013.
[70] Britton W. H. Baugher, Hugh 0. H. Churchill, Yafang Yang, and Pablo JarilloHerrero. Intrinsic electronic transport properties of high-quality monolayer and
bilayer mos 2 . Nano Letters, 13:4212, 2013.
[71]
Q H Wang, K Kalantar-Zadeh, A Kis, and J N Coleman. Electronics and
optoelectronics of two-dimensional transition metal dichalcogenides. Nature
Nanotechnology, 7:699, 2012.
[72] Simone Bertolazzi, Jacopo Brivio, and Andras Kis. Stretching and breaking of
ultrathin mos 2 . ACS Nano, 5(12):9703-9709, 2011.
[73] V Podzorov, M E Gershenson, Ch Kloc, R Zeis, and E Bucher. Highmobility field-effect transistors based on transition metal dichalcogenides. Applied Physics Letters, 84(17):3301, 2004.
[74] B L Evans and P A Young. Optical Absorption and Dispersion in Molybdenum
Disulphide. Proceedings of the Royal Society of London A, 284:402, 1965.
[75] D Xiao, G B Liu, W Feng, X Xu, and W Yao. Coupled Spin and Valley Physics
in Monolayers of MoS_ 2 and Other Group-VI Dichalcogenides. Physical Review
Letters, 108(19):196802, 2012.
106
[76] Han Wang, Lili Yu, Yi-Hsien Lee, Wenjing Fang, Allen Hsu, Patrick Herring,
Matthew Chin, Madan Dubey, Lain-Jong Li, Jing Kong, and Tomas Palacios.
Large-scale 2D Electronics based on Single-layer MoS 2 Grown by Chemical
Vapor Deposition. arXiv:1302.4027, February 2013.
[77] Han Wang, Lili Yu, Yi-Hsien Lee, Yumeng Shi, Allen Hsu, Matthew L Chin,
Lain-Jong Li, Madan Dubey, Jing Kong, and Tomas Palacios. Integrated Circuits Based on Bilayer MoS 2 Transistors. Nano Letters, 12(9):4674 4680, Sep
2012.
[78] Y H Lee, X Q Zhang, W Zhang, M T Chang, C T Lin, K D Chang, Y C Yu,
J T W Wang, C S Chang, and L J Li. Synthesis of Large Area MoS 2 Atomic
Layers with Chemical Vapor Deposition. Advanced Materials, 24:2320, 2012.
[79] Keng-Ku Liu, Wenjing Zhang, Yi-Hsien Lee, Yu-Chuan Lin, Mu-Tung Chang,
Ching-Yuan Su, Chia-Seng Chang, Hai Li, Yumeng Shi, Hua Zhang, Chao-Sung
Lai, and Lain-Jong Li. Growth of large-area and highly crystalline mnos 2 thin
layers on insulating substrates. Nano Letters, 12(3):1538-1544, 2012.
[80] Y Zhan, Z Liu, S Najmaei, P M Ajayan, and J Lou. LargeArea VaporPhase
Growth and Characterization of MoS 2 Atomic Layers on a Si0 2 Substrate.
Small, 8:966, 2012.
[81] Subhanoy Ghatak, Atindra Nath Pal, and Arindam Ghosh. Nature of electronic
states in atomically thin mos 2 field-effect transistors. ACS Nano, 5(10):77077712, 2011.
[82] A K M Newaz, D Prasai, J I Ziegler, D Caudel, S Robinson, R F Haglund Jr,
and K I Bolotin. Electrical Control of Optical Properties of Monolayer MoS 2 .
arXiv:1211.0341, November 2012.
[83] Hao Qiu, Lijia Pan, Zongni Yao, Junjie Li, Yi Shi, and Xinran Wang. Electrical
characterization of back-gated bi-layer MoS 2 field-effect transistors and the effect of ambient on their performances. Applied Physics Letters, 100(12):123104,
2012.
[84] Masihhur R Laskar, Lu Ma, ShanthaKumar K, Pil Sung Park, Sriran Krishnamoorthy, Digbijoy N Nath, Wu Lu, Yiying Wu, and Siddharth Rajan. Large
Area Single Crystal (0001) Oriented MoS 2 Thin Films. arXiv:1302.3177, February 2013.
[85] A Ayari, E Cobas, 0 Ogundadegbe, and M S Fuhrer. Realization and electrical characterization of ultrathin crystals of layered transition-metal dichalcogenides. Journal of Applied Physics, 101(1):014507, 2007.
[86] Yijin Zhang, Jianting Ye, Yusuke Matsuhashi, and Yoshihiro Iwasa. Anbipolar
mos 2 thin flake transistors. Nano Letters, 12(3):1136-1140, 2012.
107
[87] Wenzhong Bao, Xinghan Cai, Dohun Kim, Karthik Sridhara, and Michael S
Fuhrer. High Mobility Ambipolar MoS 2 Field-Effect Transistors: Substrate
and Dielectric Effects. arXiv:1212.6292, December 2012.
[88] NR Pradhan, D Rhodes, Q Zhang, S Talapatra, M Terrones, PM Ajayan, and
L Balicas. Intrinsic carrier mobility of multi-layered mos 2 field-effect transistors
on sio 2 . Applied Physics Letters, 102(12):123105, 2013.
[89] B Radisavljevic and A Kis. Mobility engineering and a metalinsulator transition
in monolayer mos 2 . Nature Materials, 12:815, 2013.
[90] M M Benameur, B Radisavljevic, J S Hron, S Sahoo, H Berger, and A Kis.
Visibility of dichalcogenide nanolayers. Nanotechnology, 22(12):125706, 2011.
[91] R Spah, U Elrod, M Luxsteiner, E Bucher, and S Wagner. Pn Junctions in
Tungsten Diselenide. Applied Physics Letters, 43:79-81, 1983.
[92] T Spalvins. A Review of Recent Advances in Solid Film Lubrication. Journal of
Vacuum Science & Technology A- Vacuum Surfaces and Films, 5:212-219, 1987.
[93] R. S. Sundaram, M. Engel, A. Lombardo, R. Krupke, A. C. Ferrari, Ph.
Avouris, and M. Steiner. Electroluminescence in single layer mos 2 . Nano Letters, 13(4):1416-1421, 2013.
[94] D C Licciardello and D J Thouless. Conductivity and mobility edges for twodimensional disordered systems.
Journal of Physics C: Solid State Physics,
8(24):4157, 1975.
[95] W C Chew and Kong J A. Effect of Fringing Fields on the Capacitance of Circular Microstrip Disk. IEEE Transactions on Microwave Theory and Techniques,
MTT-28(2):98, 1980.
[96] E M Purcell. Electricity and Magnetism. McGraw-Hill, 2nd ed. edition, 1985.
[97] K Kaasbjerg, K S Thygesen, and K W Jacobsen. Phonon-limited mobility in ntype single-layer MoS 2 from first principles. Physical Review B, 85(11):115317,
2012.
[98] R Fivaz and E Mooser. Mobility of charge carriers in semiconducting layer
structures. Physical Review, 163(3):743, 1967.
[99] Tsuneya Ando, Alan B Fowler, and Frank Stern. Electronic properties of twodimensional systems. Reviews of Modern Physics, 54:437-672, 1982.
[100] Britton WH Baugher, Hugh OH Churchill, Yafang Yang, and Pablo JarilloHerrero. Optoelectronic devices based on electrically tunable pn diodes in a
monolayer dichalcogenide. Nature nanotechnology, 9(4):262- -267, 2014.
108
[101] Steven Chuang, Rehan Kapadia, Hui Fang, Ting Chia Chang, Wen-Chun Yen,
Yu-Lun Chueh, and Ali Javey. Near-ideal electrical properties of InAs/WSe 2
van der Waals heterojunction diodes. Applied Physics Letters, 102:242101, 2013.
[102] Weijie Zhao, Zohreh Ghorannevis, Leiqiang Chu, Minglin Toh, Christian Kloc,
Ping-Heng Tan, and Goki Eda. Evolution of electronic structure in atomically
thin sheets of ws2 and wse 2 . ACS Nano, 7(1):791-797, 2013.
[103] A. K. Geirn and I. V. Grigorieva.
499:419, 2013.
Van der waals heterostructures.
Nature,
[104] Y. J. Zhang, J. T. Ye, Y. Yomogida, T. Takenobu, and Y. Iwasa. Formation of
a stable pn junction in a liquid-gated 1mlos 2 ambipolar transistor. Nano Letters,
13(7):3023-3028, 2013.
[105] Zongyou Yin, Hai Li, Hong Li, Lin Jiang, Yuneng Shi, Yinghui Sun, Gang Lu,
Qing Zhang, Xiaodong Chen, and Hua Zhang. Single-layer mos 2 phototransistors. A CS Nano, 6(1):74-80, 2012.
[106] Oriol Lopez-Sanchez, Dominik Lembke, Metin Kayci, Aleksandra Radenovic,
and Andras Kis. Ultrasensitive photodetectors based on monolayer MoS 2 . Nature Nanotechnology, 8:497, 2013.
[107] Y. Ye, Z. Ye, M. Gharghi, H. Zhu, M. Zhao, X. Yin, and X. Zhang. Excitonrelated electroluminescence from monolayer MoS 2 . arXiv:1305.4235, may 2013.
[108] Hualing Zeng, Gui-Bin Liu, Junfeng Dai, Yajun Yan, Bairen Zhu, Ruicong
He, Lu Xie, Shijie Xu, Xianhui Chen, Wang Yao, and Xiaodong Cui. Optical
signature of symmetry variations and spin-valley coupling in atomically thin
tungsten dichalcogenides. Sci. Rep., 3, 2013.
[109] Kin Fai Mak, Keliang He, Jie Shan, and Tony F. Heinz. Control of valley
polarization in monolayer rmos 2 by optical helicity. Nature Nanotechnology,
7(8):494-498, 2012.
[110] G. Sallen, L. Bouet, X. Marie, G. Wang, C. R. Zhu, W. P. Han, Y. Lu, P. H. Tan,
T. Anand, B. L. Liu, and B. Urbaszek. Robust optical emission polarization in
111OS2 inonolayers through selective valley excitation. Phys. Rev. B, 86:081301,
Aug 2012.
[111] Aaron Mitchell Jones, Hongyi Yu, Nirmal Ghimire, Sanfeng Wu, Grant
Aivazian, Jason Solomon Ross, Bo Zhao, Jiaqiang Yan, David Mandrus, and
Di Xiao. Optical Generation of Excitonic Valley Coherence in Monolayer WSe 2.
Nature Nanotechnology, 8:634, 2013.
[112] J.U. Lee, P. P. Gipp, and C.M. Heller. Carbon nanotube p-n junction diodes.
Applied Physics Letters, 85(1):145-147, Jul 2004.
109
[113] P J Zomer, S P Dash, N Tombros, and B J van Wees. A transfer technique
for high mobility graphene devices on commercially available hexagonal boron
nitride. Applied Physics Letters, 99(23):232104-232104, 2011.
[114] Chih-Tang Sah, Robert N Noyce, and William Shockley. Carrier generation and
recombination in p-n junctions and p-n junction characteristics. Proceedings of
the IRE, 45(9):1228-1243, 1957.
[115] T.C. Banwell and A. Jayakumar. Exact analytical solution for current flow
through diode with series resistance. Electronics Letters, 36(4):291-292, 2000.
[116] J.H. Lambert. Observationes variae in mathes in puram. Acta Helvetica,
Physico-mathematico-anatomico-botanico-medica,
3:128-168, 1758.
[117] M Buscema, M Barkelid, V Zwiller, and HSJ van der Zant. Large and tunable
photo-thermoelectric effect in single-layer MoS 2 . Nano Letters, 13:358, 2013.
[118] Jing-Kai Huang, Jiang Pu, Chih-Piao Chuu, Chang-Lung Hsu, Ming-Hui Chiu,
Zhen-Yu Juang, Yong-Huang Chang, Wen-Hao Chang, Yoshihiro Iwasa, and
Mei-Ying Chou. Large-Area and Highly Crystalline WSe 2 Monolayers: from
Synthesis to Device Applications. arXiv:1304.7365, 2013.
[119] M Bernardi, M Palummo, and Jeffrey C Grossman. Exraordinary Sunlight
Absorption and One Nanometer Thick Photovoltaics Using Two-dimensional
Monolayer Materials. Nano Letters, 13:3664, 2013.
[120] Xiao Li, Fan Zhang, and Qian Niu. Unconventional Quantum Hall Effect and
Tunable Spin Hall Effect in Dirac Materials: Application to an Isolated MoS 2
Trilayer. Physical Review Letters, 110(6):066803, February 2013.
[121] Andreas Pospischil, Marco M Furchi, and Thomas Mueller. Solar energy conversion and light emission in an atomic monolayer p-n diode. arXiv:1309.7492v1,
2013.
[122] J. S. Ross, P. Klement, A. M. Jones, N. J. Ghimire, J. Yan, D. G. Mandrus,
T. Taniguchi, K. Watanabe, K. Kitamura, W. Yao, D. H Cobden, and X. Xu.
Electrically Tunable Excitonic Light Emitting Diodes based on Monolayer WSe 2
p-n Junctions. ArXiv c-prints, December 2013.
[123] Tingmei Wang, Weimin Liu, Jun Tian, Xin Shao, and Dacheng Sun. Structure characterization and conductive performance of polypyrrol-molybdenum
disulfide intercalation materials. Polymer Composites, 25(1):111-117, 2004.
[124] M. Gurvitch. Ioffe-regel criterion and resistivity of metals.
24:7404-7407, Dec 1981.
Phys. Rev. B,
[125] Gilles Buchs, Maria Barkelid, Salvatore Bagiante, Gary A. Steele, and Val
Zwiller. Imaging the formation of a p-n junction in a suspended carbon nanotube with scanning photocurrent microscopy. Journal of Applied Physics,
110(7):-, 2011.
110