Surface-gate-defined single-electron transistor in a MoS2 bilayer
M. Javaid,1, ∗ Daniel W. Drumm,1, 2 Salvy P. Russo,1 and Andrew D. Greentree1, 2
1
arXiv:1608.07894v1 [cond-mat.mes-hall] 29 Aug 2016
Chemical and Quantum Physics, School of Sciences,
RMIT University, Melbourne VIC 3001, Australia
2
The Australian Research Council Centre of Excellence for Nanoscale BioPhotonics,
School of Sciences, RMIT University, Melbourne, VIC 3001, Australia
(Dated: August 30, 2016)
We report the multi-scale modeling and design of a gate-defined single-electron transistor in a
MoS2 bilayer. By combining density-functional theory and finite-element analysis, we design a
surface gate structure to electrostatically define and tune a quantum dot and its associated tunnel
barriers in the MoS2 bilayer. Our approach suggests new pathways for the creation of novel quantum
electronic devices in two-dimensional materials.
I.
INTRODUCTION
Novel two-dimensional (2D) materials such as
graphene and transition-metal dichalcogenides (TMDCs)
have attracted significant interest due to their unique
electronic and optical properties [1–3]. The ultrathin
geometry and dangling-bond-free interfaces of 2D materials make them good candidates to integrate on various substrates [4]. Among these 2D materials, MoS2
(a TMDC) is promising for quantum electronics; it possesses interesting layer-dependent properties. For example, by decreasing the thickness of MoS2 from bulk to
a single layer, its band gap switches from indirect to direct and increases by more than 0.6 eV, leading to strong
photoluminescence from a single layer [2, 5].
The single-electron transistor (SET) is a device where
electrons tunnel one by one to and from a small island
through tunnel barriers [6]. The tunneling of an electron and the quantization of charge are controlled by
gate electrodes. SETs have extensive applications in
nano-electronic devices [7]. They have been proposed
as a future alternative to conventional CMOS transistors [8], and have been used as nano-scale electrometers
[9], capable√of measuring sub-electron charge variations
to 10−6 qe / Hz. They also have many applications in
single-electron logic circuits [10] and single-electron turnstile devices [11].
Many different techniques are used to fabricate SETs
based in materials such as Si [12–14], GaAs [15] and
carbon nanotubes [16]. There is a significant body
of literature on fabricating SETs. For example: Thelander et. al. have designed gold-nanoparticle SETs using carbon nanotubes as leads at 200 K [17], Klein
et. al. reported a colloidal chemistry technique to create cadmium-selenide nanocrystals of varying sizes [18],
and Kim et. al. used a focused-ion-beam technique to
fabricate SETs operating at room temperature [19]. Recently, there has been extensive work to define quantumdot devices in 2D materials including graphene [20] and
∗
maria.javaid@rmit.edu.au
TMDCs [21]. Surface gates have been used to create tunable tunnel barriers in Si interfaces [22], but to date there
have been no reported designs of electrostatically tunable
tunnel barriers and tunable quantum dots in 2D MoS2 .
It is known that the MoS2 monolayer band structure
shows no response to applied electric fields [23]. In contrast, MoS2 bilayers exhibit significant band-structure
modification under perpendicular electric field [23, 24].
Here we present a lithographically appealing design of
a surface-gate-defined SET in a MoS2 bilayer, and model
its physical characteristics. The structure of our proposed device is a MoS2 bilayer sandwiched between a
HfO2 substrate and cap layer, with metallic surface gates,
as shown in Fig. 1(a). We use the surface-gate electric
fields to modify the local band structure in the MoS2 bilayer. We present a fully self-consistent design of the device obtained through multi-scale modeling. Recently, a
more conventional gate design has been used to demonstrate quantum confined structures on a few layers of
WSe2 with tunnel barriers defined by electric fields at a
temperature of 240 mK [21].
We modeled a MoS2 bilayer using density-functional
theory (DFT) to study the effects of electric field on the
band structure of MoS2 bilayers. We performed numerical simulations to calculate the potential at the MoS2
bilayer due to the surface gates and used this potential to study the modification in the lowest unoccupied
molecular orbital (LUMO) of the MoS2 bilayer. We modeled our device geometry (obtained from LUMO bending
in the MoS2 bilayer) with the commercial finite-element
analysis simulation tool comsol [25] to obtain the selfcapacitance of the SET island and the capacitances between the island and each of the electrodes. We used
these capacitances in numerical simulations of transport
through the SET at a temperature of 1 K and studied its
physical characteristics. We iteratively cycled through
comsol and numerical simulations to obtain a consistent picture of the device.
This paper is organized as follows: we begin our discussion with the DFT modeling of the MoS2 bilayer followed
by the comsol modeling. Then we present the influence
of the potential due to surface gates in the MoS2 bilayer
plane and numerical transport simulations of the resulting SET device.
2
(a)
G1
G3
G4
G2
Cap layer
Substrate
(b)
(c)
CisR(Vg )
CidR(Vg )
Island
Cig
Vs
Vd
Vg
−0.16
−0.22
0
−200
−200
(e)
−0.28
Potential (V)
Potential (V)
X (nm)
(d)
200
−0.28
0
Y (nm)
200
−0.34
−0.3
Island
Source
Drain
−0.32
−0.34
−200
0
Y (nm)
200
(f )
X (nm)
Z (nm)
Y (nm)
FIG. 1. (Color online) (a) Schematic showing a MoS2 bilayer sandwiched between HfO2 cap layer and substrate, with
metallic surface gates above the cap layer. Gates G3 and
G4 are shown separated from G1 for visibility, but are actually collinear. (b) Perspective view of the black layer in (a)
showing A-A′ stacked, 2H MoS2 bilayer, Mo (large, green)
atoms, S (small, yellow) atoms. (c) Equivalent circuit diagram of SET showing the capacitance/resistance between the
island and the source/drain and capacitance between the island and the top gates. (d) Top view of the potential created
on top of the the MoS2 bilayer sheet, showing the creation of
well-defined source, drain, and island regions. (e) Potential
slice along bilayer directly underneath gates G1, G3, and G4,
showing source, island, and drain for a surface-gate voltage
of 0.27 V. (f) 3D view of the geometry used in the comsol
modeling. The (blue) highlighted regions are the source, island and drain in the MoS2 bilayer plane. Lengths are in
nm.
II. DFT MODELING OF
MOLYBDENUM-DISULPHIDE BILAYER
STRUCTURES
Although the effects of electric field on the band structure of MoS2 bilayers are available in the literature, the
focus has either been on behavior over extreme field
strength ranges (with sampling too coarse for our device) [23] or on the band-gap response without specific
discussion of the LUMO physics [24]. Thus, to achieve
the insight necessary to design our device, we explore the
LUMO physics directly in the target operating regime.
We investigated the band structure of an optimized,
A-A′ stacked (Mo atoms in the top monolayer above the
S atoms of the bottom monolayer), 2H-phase MoS2 bilayer for varying electric fields. A-A′ is the most-reported
stacking order for bulk MoS2 and it is very close in energy
to the most stable stacking order A-B [24]. A perspective
view of the MoS2 bilayer is shown in Fig. 1 (b).
To model the effects of electric field on the MoS2
bilayers, we used DFT in crystal09 [26, 27]. MoS2
has hexagonal symmetry and belongs to the P 63 /mmc
space group, with lattice parameters a = 3.17 Å and
c = 12.324 Å [28]. We created the bilayer unit slab
by cutting a (001) plane from a bulk MoS2 model, and
including vacuum to a total cell height of c = 500 Å.
The exchange and correlation terms were described via
the PBEsol functional [29], which generally predicts lattice constants more accurately than PBE and LSDA
(thus improving the equilibrium properties of solids), and
also handles the electronic response to potentials better than most GGA functionals [29]. We used Gaussian basis sets; Mo SC HAYWSC311(d31)G cora 1997
[30] for Mo atoms (a Hay-Wadt effective-core pseudopotential [31] combined with a valence-electron basis set)
and S pob TZVP 2012 [32] for S atoms (an all-electron
basis set). We set a 8×16×1 Monkhorst-Pack [33] k-point
mesh.
We optimized the geometry of a MoS2 bilayer unit slab
under zero electric field and then used this optimized geometry to study the effects of electric field on the band
structure of MoS2 bilayers. We calculated the band structure along the high-symmetry path Γ-M-K-Γ. Band
structures (shown in the Appendix) were calculated for
varying electric fields applied perpendicular to the MoS2
bilayers.
The responses of the LUMO and the highest occupied
molecular orbital (HOMO) of the MoS2 bilayer to applied
electric fields are shown in Fig. 2(a). The HOMO remains
static with electric fields of up to ±0.2 V/Å (in agreement with [23]) while there is a significant response in the
LUMO. Fig. 2(a) also shows some inconsistent points in
the LUMO and HOMO under electric field. We excluded
these points from further calculations as the band structure at these points is qualitatively different and we believe that these inconsistent points do not represent the
underlying physics.
We applied a linear fit to the filtered LUMO and
HOMO data, and set zero energy at the LUMO intercept (at zero field). As an electric field is applied up
to ±0.2 V/Å, the LUMO bends through 138 meV. Thus
electron confinement is achievable in a MoS2 bilayer via
electric-field modification of the LUMO, which occurs at
a rate of 690 meV / (V/Å).
III.
CIRCUIT SIMULATIONS IN COMSOL
We modeled our device geometry in comsol [25] to get
the capacitances between all the device components. The
3
to the total capacitance of the island. The ground plane
is 11 nm (thickness of the HfO2 substrate) below the island.
The wire-frame rendering of the geometry used in
comsol modeling is shown in Fig. 1(f), where all the
lengths are in nm. For an island of diameter 12 nm, the
self-capacitance is CiΣ = 8.45 aF which corresponds to
a charging energy of e2 /CiΣ = 18.9 meV, where CiΣ is
defined as
(a)
LUMO
Energy (eV)
0
−0.7
CiΣ = Cis + Cid + Cig1 + Cig2 + Cigp
(1)
HOMO
−1.4
0
0.1
Abs. electric field (V/Å)
0.2
(b)
0.6
Island
0.45
0.45
0.3
0.3
Source
Drain
0.15
0.15
0
0
Barrier
Barrier
−0.15
−0.3
−200
−0.15
−100
0
Position (nm)
100
Relative LUMO bending (meV)
Relative electric Field (mV/Å)
0.6
−0.3
200
FIG. 2. (Color online) (a) Responses of the MoS2 bilayer
LUMO (grey circles) and HOMO (blue squares) to applied
electric field, with first-order fits to LUMO (red line) and
HOMO (black line) data. Negative electric-field values are
folded over to the positive axis. (b) Electric field (solid blue
line) relative to the dashed line in (a), created across the MoS2
bilayer for 0.27 V surface-gate voltage. This electric field is
used to calculate the LUMO bending (dashed red line) via the
line-fit parameters from (a). The electric field and the LUMO
bending corresponding to the intersection of dashed line with
LUMO fit in (a) is set at zero. The Fermi level (black dashed
line), barrier and island regions (blue dashed arrows), and
region boundaries (black vertical bars) are marked.
geometry (obtained through potential simulations discussed later and converged by iteratively cycling through
numerical potential and transport simulations and comsol) used for comsol modeling is an island of diameter
12 nm and the same thickness as the MoS2 bilayer. There
are 16 nm wide MoS2 barriers between the island and inplane source and drain. The surface gates are modeled
as two intersecting wires each of 5 nm radius sitting atop
the 7 nm HfO2 cap layer. We excluded gates G3 and
G4 of Fig. 1 (a) from our comsol modeling as they are
collinear with gate G1 and do not contribute significantly
Here Cis = −0.252 aF is the capacitance between the
island and the buried source, Cid = −0.253 aF is the
capacitance between the island and the buried drain,
Cig1 = −2.2 aF is the capacitance between the island
and gate G1, Cig2 = −2.38 aF is the capacitance between the island and gate G2, and Cigp = −3.35 aF is
the capacitance between the island and the ground plane.
IV.
SINGLE-ELECTRON TRANSPORT
SIMULATIONS
In the design of our proposed MoS2 bilayer SET, there
are two longer surface gates G1 and G2, perpendicular to
each other. Two shorter gates, G3 and G4, are collinear
with the G1 gate as shown in Fig. 1(a). All of these surface gates are electrostatically insulated from each other;
this can be done by thermally growing a thin oxide layer
between them. The equivalent electrical circuit diagram
is shown in Fig. 1(c), where R(Vg ) is the tunnel resistance
between the island and the source/drain. This resistance
is a function of the gate voltage Vg applied to the surface
gates. Vs and Vd are the source and drain potentials applied directly to the buried source and drain in the MoS2
bilayer plane (created by the surface gates).
We modeled our device design through in-house numerical simulations. Gates G3 and G4 were placed on
top of gate G1, with their finite edges 32 nm apart from
each other. The thicknesses of the cap layer and the
substrate were 7 nm and 11 nm respectively. A varying DC potential of 0.2 V to 0.34 V was applied to all
surface gates. All of these parameters and the DC potential range were finalized through iterations between the
comsol modeling and the numerical simulations. The
top view of the potential created in the MoS2 bilayer
plane due to the surface gates is shown in Fig. 1(d). The
intersection of gates G1 and G2 generates a well-defined
island in the MoS2 bilayer plane. Gates G1, G3, and G4
define the source/drain wiring configuration and tunable
tunnel barriers in the MoS2 bilayer plane. An analytical expression for the potential created due to one of the
surface gates is given in the Appendix.
Fig. 1(e) shows a slice of the potential on top of the
MoS2 bilayer plane and underneath gates G1, G3, and
G4. Note that there is a small local minimum in the potential (and the electric field) before the barriers. The
(a)
RT (arb. units)
4
7,500
5,500
3,500
0.2
(b)
0.24
0.28
Vg (V)
0.32
Current (arb. units)
0.02
electron tunnels to the island and current flows through
the SET which is shown by the peaks in Fig. 3(b). On
the other hand, when none of the island states are available to the incoming electron, it cannot tunnel to the
island and the current through the SET drops to zero,
defining the Coulomb-blockade region between the peaks.
The height variation between the current peaks arises
from the changing response of the MoS2 -bilayer density
of states in the source/drain and island, in contrast to an
ideal metallic SET.
V.
0.01
0
0.2
0.24
0.28
Vg (V)
0.32
FIG. 3. (a) Variation of tunnel resistance with gate voltage,
Vg . (b) Characteristic Coulomb blockade peaks for a bias
voltage of 100 µV and at a temperature of 1 K, showing a
variable current through the MoS2 bilayer SET.
potential difference across the MoS2 bilayer is proportional to the electric field (Fig. 2(b)). We used the linear
fit to Fig. 2(a) to obtain the LUMO bending at fields
shown in Fig. 2(b). This LUMO bending as a function
of the position across the MoS2 bilayer plane is shown
in Fig. 2(b). With the above dimensions of the device,
a 12 nm diameter island and 16 nm wide tunnel barriers
are created in the MoS2 bilayer plane. The dimensions of
the SET are comparable with experimental structures in
GaAs/ AlGaAs [34]. The size of the island and the width
of the tunnel barriers can be controlled through the substrate and the cap layer thicknesses. The LUMO bending
is used as an input to numerical transport simulation.
Fig. 3(a) shows the variation of tunnel resistance with
Vg . Increases to the gate voltage cause the barrier height
to rise, reducing the transmission coefficient. Due to scaling of input electron energy, there is a minimal change
in transmission coefficient with gate voltage but density
of states in both source/drain and island increases predominantly and that leads to the observed decrease in
tunnel resistance. (For details of the tunnel resistance
and transport calculations, see the Appendix)
The behavior of the SET can be described by varying
Vg . Fig. 3(b) shows the characteristic current passing
through the SET at a temperature of 1 K and a bias
voltage of 100 µV (In this paper, we are only considering
the low-bias regime.) By varying the surface gate voltage, the overall potential landscape through the MoS2
bilayer changes, thus modifying the barrier height and
alignment between incoming electron energy and island
states. When the energy of an incoming electron resonates with an unoccupied energy level in the island, that
CONCLUSIONS
The surface gate approach to quantum devices is quite
flexible and amenable to scalable integrated devices.
With identical gate voltages, we already have significant
control over barrier height, the energy states in the island and hence, electron tunneling. One could achieve
further independent control of the incoming electron energy or island states by applying different voltages to each
gate. Moreover, due to the small size of the island, our
preliminary modeling holds promise for relatively hightemperature single-electron devices.
The most important feature of the gate design shown
here, as opposed to more conventional gate-defined quantum dot designs (such as, for example, [21]) is that the
creation of potential wells directly underneath and in the
near field of the metallic surface gates should allow moreprecise lithographic definition of the quantum features in
the active layer. In the conventional approach to creating
sub-surface quantum dots, the relatively large distance
from the gate layer to the active layer gives rise to a
complicated potential landscape that must be solved for
self-consistently with Schrödinger-Poisson solvers. Conversely, as our active layer is in the near field of the surface gates, the potential should more precisely follow the
metallisation on the surface layer. This in turn means
that more complicated integrated circuits should be realisable, for example multi-island dot structures or integrated qubit/readout type circuits. This design ethos is
exemplified by the surface gate defined SETs fabricated
at a silicon/silica interface by Angus et al. [22].
Here we have studied an embedded MoS2 bilayer system. However, we would expect our results to be easily applicable to other bilayer two-dimensional materials that exhibit a LUMO response to electric field (e.g.
MoSe2 , MoTe2 and WS2 [23]), with only minimal modification of the applied surface potentials.
VI.
ACKNOWLEDGEMENTS
MJ and ADG acknowledge funding by ARC Discovery Grant No. D130104381. DWD acknowledges the
support of the ARC Centre of Excellence for Nanoscale
BioPhotonics (CE140100003). This work was supported
by computational resources provided by the Australian
5
(a)
Government through the National Computational Infrastructure under the National Computational Merit Allocation Scheme.
E-Ef (eV)
2.72
0
−2.72
Here we present a brief overview of the design of the
MoS2 bilayer single-electron transistor, followed by details of our modeling. We show our analytic analysis of
the electric field across the bilayer plane, the response
of the MoS2 bilayer band structure to electric fields, the
spatial behavior of the LUMO under different gate voltages, and details of our calculations of the resistance of
(and current through) the SET.
Gates G1 and G2 of Fig. 1(a) are approximated as
infinitely long lines of charge, the intersection of which
defines our island via the doubly-strong steep potential.
To establish source and drain leads, we use gates G3 and
G4 to create similarly strong potential landscapes
directly underneath gate G1.
For convenience, gates G3 and G4 are approximated
as one infinite line of charge with a missing, finite segment centred on the intersection of gates G1 and G2.
For our treatment, this dictates V3 = V4 , but that condition is not necessary if the semi-infinite gates G3 and
G4 are treated separately and identical source and drain
behavior is not required. It should also be noted that
V2 ≈ V3 = V4 achieves comparable island and lead depths
below the barrier peaks; V2 = V3 = V4 is not a constraint.
Gate G1 is therefore effectively a plunger gate controlling
the depth of all components, which could in principle be
operated in isolation from the rest (although V1 +V3 > V2
is required to avoid an unintended perpendicular lead
under gate G2). For simplicity here, however, we set
V g = V 1 = V2 = V3 = V 4 .
Defining x as the distance in the bilayer plane transverse to gate G1 (with zero value directly underneath
G1), (y similarly with reference to G2), and z as the
bilayer-ground plane separation, the electric potential
difference created across the MoS2 bilayers is the sum
effect from all gates, which is analytic. For example, the
potential due to gate G1 is:
2
λ
x + (z − d)2
V =
log 2
,
(A1)
2πǫo ǫr
x + (z + d)2
where λ is the linear charge density of gate G1, ǫr is the
dielectric constant of HfO2 [35] and d is the distance of
gate G1 from the ground plane. The other potentials
are similarly derived, and the full potential is the sum
effect from all of the gates. The finite difference between
the top and bottom edges of the bilayer is evaluated to
provide the field strength across the bilayer.
Having obtained a two-dimensional map of the potential difference across the bilayer plane – e.g., Fig. 1(d) in
the main paper – we then consider its effect on the MoS2
band structure at fields from 0 to 0.2 V/Å in steps no
larger than 0.01 V/Å, and smaller as required to verify
−5.44
EF=0.1 V/Å
5.44
2.72
E-Ef (eV)
Appendix A
(b)
EF=0 V/Å
5.44
0
−2.72
Γ
M
K
Γ
−5.44
Γ
M
K
Γ
FIG. 4. (a) Band structure of MoS2 bilayer under zero electric
field. There is an indirect band gap of 1.30 eV. (b) Band
structure of MoS2 bilayer under finite electric field. There is
an indirect band gap of 1.23 eV.
linear behavior. Fig. 4 shows two sample band structures.
We are particularly interested in the minimal substrate
footprint, and the constraint of reasonable applied field
strength. Therefore in the main paper, without loss of
generality in the method, we explicitly consider the high
charging energy case, when the classical occupancy of
the SET island can change by at most one electron at
a time. Thus, the extent of the LUMO bending created
must match well to the charging energy of the system;
∆L ≈ E =
1 T −1
Q C Q,
2
(A2)
where Q is the charge on the SET, defined by
Q=[Qi Qs Qg1 Qg2 Qd Qgp ]T . C is the capacitance matrix between all the device components and is defined
as:
CiΣ Cis Cig1 Cig2 Cid Cigp
Csi CsΣ Csg1 Csg2 Csd Csgp
C
Cg1Σ Cg1g2 Cg1d Cg1gp
C
C = g1i g1s
(A3)
Cg2i Cg2s Cg2g1 Cg2Σ Cg2d Cg2gp
C
di Cds Cdg1 Cdg2 CdΣ Cdgp
Cgpi Cgps Cgpg1 Cgpg2 Cgpd CgpΣ
the terms of which are obtained via comsol modeling as
described in the main paper.
The tunnel resistance cannot be evaluated exactly.
The usual approach to an arbitrary barrier, the WKB
approximation, is not directly applicable if the incoming
energy of the electron matches the barrier energy at any
point. In fact, not only does the approximation not hold
at those points, but also in their vicinities. The connection formulas, using an Airy function ansatz to patch
difficult regions, are the standard way to circumvent this
issue; however, the conditions for their use are not met
by this barrier. Accordingly, we make the approximation
that the barrier is finite and rectangular, with the identical width (16 nm) and height equal to the peak value of
LUMO bending, called Lpeak while setting L(−200) = 0.
Since such a barrier is definitively larger, and no thinner, than the original barrier, its transmission coefficient
is guaranteed to be lower than any experimental device.
6
Island
LUMO bending (meV)
0.45
Vg =0.34V
Vg =0.305V
Vg =0.27V
0.3
Vg =0.235V
Vg = 0.27 V, where µ is set such that ε is equal to the
black-dashed line. There, the island has 12 nm diameter
and 16 nm wide tunnel barriers. L is the barrier height
created by LUMO bending due to surface gate potentials.
The tunnel resistance is given by [37]
Vg =0.20V
0.15
RT−1 =
0
−0.15
−100
0
Position(nm)
100
200
FIG. 5. (Color online) LUMO bending as function of position along the source-island-drain axis for varying gate voltages within the device target operating regime. The barrier
function behaves linearly with Vg . Vertical marks show the
boundaries of the labelled regions (black-dashed lines with
arrow heads). ε is not shown since it varies with Vg , but its
values for each Vg can be identified as the intersections of the
corresponding LUMO data traces with the region boundaries.
The transmission coefficient of a finite square barrier is:
T = e−2γ ,
(A4)
where
1
γ=
h̄
(A6)
where | T (ε) | is the transmission co-efficient, ρ1 is the
density of states in source/drain electrodes and ρ2 is the
density of states in the island. We approximate the density of states by assuming constant band bending over
the leads (and island), giving:
Barrier Barrier
−200
2πe2
ρ1 ρ2 | T (ε) |2 ,
h̄
Za q
q
a
2m(Lpeak − ε)dx =
2m (Lpeak − ε),
h̄
0
(A5)
h̄ = h/2π is the reduced Planck’s constant, m = 0.38mo
is the effective mass of electrons in the MoS2 bilayer [36],
a is the width of the tunnel barrier, and ε is the energy
of an incoming electron in the MoS2 bilayer.
Here ε can also be regarded as ε = Vsd + Eth + µ,
where Vsd is the source-drain bias, Eth is the thermal
energy described by the Fermi smearing incorporated
below, and µ is the chemical potential of the electron
which is set by the absolute source and drain voltages
Vs and Vd . (Recall Vsd = Vs − Vd ; modifying the absolute value of each in tandem has no effect on Vsd .)
We assume that the incoming electron has µ such that
ε ≈ 0.38 [Lpeak − L (−200)] = 0.38Lpeak ; this defines the
points where L = ε (the region boundaries) at consistent
locations in space – enforcing the desired geometry with
large charging energy. (The left edge of L has been set as
zero energy, the absolute value of µ is somewhat different
to the marked ε.) Fig. 5 shows L (x) for several values
of Vg , together with the constant region boundaries. In
principle, this assumption could be relaxed as needed to
describe a particular physical device.
As an example, Fig. 2(b) in the main text shows
the particular case of the optimised structure under
ρ=
Zεf
D(ε) dε =
2
D(ε − εo ) × (ε − εo );
3
(A7)
εo
Here εf is the Fermi level and εo is the energy of the
lowest level in the source/drain leads of MoS2 bilayer
plane. (This corresponds to making the assumption that
electron-hole recombination is negligibly slow compared
to the tunneling rate). D(ε) is the density of states per
unit energy in a solid, given by
√
8π 2 3 √
D(ε) =
(A8)
m 2 ε × η for ε > εo ,
h3
where η is the volume.
Hence,
√
16π 2 3 √
m 2 ε − εo (ε − εo ) × ηs for ε > εo
ρ1 =
3h3
(A9)
and
√
16π 2 3 √
ρ2 =
m 2 ε − εo (ε − εo ) × ηi for ε > εo (A10)
3h3
Here ηs and ηi are the volumes for the source and island
respectively. Volumes were estimated assuming the bilayer height of 0.94 Å, and island and lead in-plane areas
2
of π (6 nm) and 400 × 20 nm2 , respectively which leads
to tunnel resistance of 3500 MΩ to 8500 MΩ for this idealised design in perfect MoS2 in HfO2 . Practically, many
factors involved in experiment (e.g., fabrication imprecision, defects, etc.) often accumulate to affect resistances
by several orders of magnitude.
Although we use a semiconductor material (MoS2 bilayer) in the design of our SET, after connecting it with
the external batteries the Fermi level rises into the conduction band in the source/island/drain regions while
sitting below the conduction band in the tunnel barrier
regions. Thus, the SET behaves effectively like a metalinsulator-metal junction and all the standard approximations for metallic SET are still applicable to our SET.
To determine the current through the SET, we need
the tunneling rates on and off the island for different
7
charge configurations of the SET. These tunneling rates
are given by [7]
Γnχi =
n
∆Eχi
1
,
n
qe2 RT exp( ∆Eχi ) − 1
kB T
(A11)
where χ denotes the source or drain and i the island.
n
∆Eχi
is the energy difference between different charge
configurations of the SET and is given by
Here Vχ is the chemical potential applied to the
source/drain lead.
The current through the SET is sum of the probabilities for all possible tunneling rates on and off the island.
This current is given by
I = qe
∞
X
X
n=−∞ χ=s,d
pn (Γnχi − Γniχ )
(A13)
(A12)
where pn is the probability that island is in a state with
n excess electrons. These probabilities are calculated by
the master equation as described in [38].
[1] K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang,
M. I. Katsnelson, I. V. Grigorieva, S. V. Dubonos, and
A. A. Firsov, Nature 438, 197 (2005).
[2] K. F. Mak, C. Lee, J. Hone, J. Shan, and T. F. Heinz,
Phy. Rev. Lett. 105, 136805 (2010).
[3] G. Fiori, F. Bonaccorso, G. Iannaccone, T. Palacios,
D. Neumaier, A. Seabaugh, S. K. Banerjee,
and
L. Colombo, Nat. Nanotechnol. 9, 768 (2014).
[4] K. Bolotin, K. Sikes, Z. Jiang, M. Klima, G. Fudenberg,
J. Hone, P. Kim, and H. Stormer, Solid State Commun.
146, 351 (2008).
[5] A. Splendiani, L. Sun, Y. Zhang, T. Li, J. Kim, C. Y.
Chim, G. Galli, and F. Wang, Nano Lett. 10, 1271
(2010).
[6] V. Singh, A. Agrawal, and S. Singh, IJSCE 2, 502 (2012).
[7] H. Grabert and M. H. Devoret, (Plenum Press, New
York, 1992).
[8] S. Mahapatra, V. Vaish, C. Wasshuber, K. Banerjee, and
A. M. Ionescu, IEEE Trans. Electron Devices 51, 1772
(2004).
[9] M. Devoret and R. Schoelkopf, Nature 406, 1039 (2000).
[10] S. J. Kim, J. J. Lee, H. J. Kang, J. B. Choi, Y.-S. Yu,
Y. Takahashi, and D. G. Hasko, Appl. Phys. Lett. 101,
183101 (2012).
[11] Z. L. Yuan, Science 295, 102 (2002).
[12] M. Fuechsle, S. Mahapatra, F. A. Zwanenburg,
M. Friesen, M. A. Eriksson, and M. Y. Simmons, Nat.
Nanotechnol. 5, 502 (2010).
[13] M. Hofheinz, X. Jehl, M. Sanquer, G. Molas, M. Vinet,
and S. Deleonibus, Appl. Phys. Lett. 89, 143504 (2006).
[14] S. J. Shin, C. S. Jung, B. J. Park, T. K. Yoon, J. J. Lee,
S. J. Kim, J. B. Choi, Y. Takahashi, and D. G. Hasko,
App. Phys. Lett. 97, 103101 (2010).
[15] R. G. Knobel and A. N. Cleland, Nature 424, 291 (2003).
[16] P. Stokes and S. I. Khondaker, Appl. Phys. Lett. 92,
262107 (2008).
[17] C. Thelander, M. H. Magnusson, K. Deppert, L. Samuelson, P. R. Poulsen, J. Nygard, and J. Borggreen, Appl.
Phys. Lett. 79, 2106 (2001).
[18] D. L. Klein, R. Roth, A. K. L. Lim, A. P. Alivisatos, and
P. L. McEuen, Nature 389, 699 (1997).
[19] T. W. Kim, D. C. Choo, J. H. Shim, and S. O. Kang,
Appl. Phys. Lett. 80, 2168 (2002).
[20] C. Stampfer, E. Schurtenberger, F. Molitor, J. Güttinger,
T. Ihn, and K. Ensslin, Nano Lett. 8, 2378 (2008).
[21] X.-X. Song, D. Liu, V. Mosallanejad, J. You, T.-Y. Han,
D.-T. Chen, H.-O. Li, G. Cao, M. Xiao, G.-C. Guo, and
G.-P. Guo, Nanoscale 7, 16867 (2015).
S. J. Angus, A. J. Ferguson, A. S. Dzurak, and R. G.
Clark, Nano Lett. 7, 2051 (2007).
A. Ramasubramaniam, D. Naveh, and E. Towe, Phys.
Rev. B 84, 205325 (2011).
Q. Liu, L. Li, Y. Li, Z. Gao, Z. Chen, and J. Lu, J. Phys.
Chem. C 116, 21556 (2012).
comsol multiphysics R version 5.1, COMSOL, Inc.,
Burlington, MA, USA, 2008.
R. Dovesi, R. Orlando, B. Civalleri, C. Roetti, V. R.
Saunders, and C. M. Zicovich-Wilson, Zeitschrift für
Kristallographie 220, 571 (2005).
R. Dovesi, V. R. Saunders, C. Roetti, R. Orlando, F. Pascale, B. Civalleri, K. Doll, N. M. Harrison, I. J. Bush,
M. Llunell, C. Science, and A. Technologies, Crystal09
User’s Manual (2013).
R. G. Dickinson and L. Pauling, J. Am. Chem. Soc. 45,
1466 (1923).
J. P. Perdew, A. Ruzsinszky, G. I. Csonka, O. A. Vydrov,
G. E. Scuseria, L. A. Constantin, X. Zhou, and K. Burke,
Phys. Rev. Lett. 100, 136406 (2008).
F. Corà, A. Patel, N. M. Harrison, C. Roetti, and
C. R. A. Catlow, J. Mater. Chem. 7, 959 (1997).
A. Ehlers, M. Böhme, S. Dapprich, A. Gobbi,
A. Höllwarth, V. Jonas, K. Köhler, R. Stegmann,
A. Veldkamp, and G. Frenking, Chem. Phys. Lett. 208,
111 (1993).
M. F. Peintinger, D. V. Oliveira, and T. Bredow, J.
Comput. Chem. 34, 451 (2013).
H. J. Monkhorst and J. D. Pack, Phys. Rev. B 13, 5188
(1976).
W. G. van der Wiel, S. De Franceschi, J. M. Elzerman,
T. Fujisawa, S. Tarucha, and L. P. Kouwenhoven, Rev.
Mod. Phys. 75, 1 (2002).
M. Jones, Y. Kwon, and D. Norton, Appl. Phys. A 81,
285 (2005).
T. Cheiwchanchamnangij and W. R. L. Lambrecht, Phys.
Rev. B 85, 205302 (2012).
P. Koppinen, Bias and temperature dependence analysis of the tunneling current of normal metal-insulatornormal metal tunnel junctions, Master’s thesis, University of Jyvaskyla, department of physics (2003).
V. I. Conrad, A. D. Greentree, D. N. Jamieson, and
L. C. L. Hollenberg, J. Comput. Theor. Nanos. 2, 214
(2005).
n
∆Eχi
= E(n − 1) − E(n) + Vχ qe
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
[34]
[35]
[36]
[37]
[38]