Investigation of Competing Correlated Electronic States in
Quasi-2D Systems
Vol 446 | 1 March 2007 | doi:10.1038/nature05555
David
1
Lioi ,
Goran
LETTERS
1
Karapetrov
of Physics, Drexel University 2Department of Mathematics, Drexel University
Hubert B. Heersche1*, Pablo Jarillo-Herrero1*, Jeroen B. Oostinga1, Lieven M. K. Vandersypen1
& Alberto F. Morpurgo1
Abstract
Induced Superconductivity in
Layered Materials
density only in the bottom one or two layers. For single layers, the
Graphene—a recently discovered form of graphite only one
1
position of the resistance maximum corresponds to the gate voltage
atomic layer thick —constitutes a new model system in condensed
at which the Fermi energy is located at the Dirac point, VD, and we
matter physics, because it is the first material in which charge
typically find that jVDj , 20 V. We unambiguously determine the
carriers behave as massless chiral relativistic particles. The anom2,3
single layer character of a device by quantum Hall effect (QHE)
alous quantization of the Hall conductance , which is now under4,5
measurements. Because the superconducting proximity effect
stood theoretically , is one of the experimental signatures of the
requires two closely spaced electrodes, we can only perform magnepeculiar transport properties of relativistic electrons in graphene.
toconductance measurements in a two terminal configuration. In
Other unusual phenomena, like the finite conductivity of order
2
general, the conductance, G, measured in this way is a mixture of
4e /h (where e is the electron charge and h is Planck’s constant) at
NATURE | Vol 446 | 1 March 2007
2
longitudinal and Hall signals, but at high fields G < jGHallj (this
the charge neutrality (or Dirac) point , have come as a surprise and
approximation is exact at the Hall plateaus19). Indeed, the measureremain to be explained5–13. Here we experimentally study the
ment of G versus VG at B 5 10 T shows clearly identifiable Hall plaJosephson effect14 in mesoscopic junctions consisting of a grab
/ha (Fig.
1d), characteristic of the
teaus at half-integer multiples of 4e2(b)
phene layer contacted by two closely spaced superconducting
(c)
75
QHE in single layer graphene2,3. This demonstrates that, even in 30
electrodes15. The charge density in the graphene layer can be controlled by means of a gate electrode. We observe a supercurrent
20
that, depending on the gate voltage, is carried by either electrons in
0
I
a
b
(a)
(a)conduction
AFM image
graphene
with transparent
1 µm
the
bandoformonolayer
by holes in the
valence band.
More imporVG = –50 V
10
Al contacts.
(b) Hysteretic
current-voltage
curves with
tantly,
we find that
not only the
normal state conductance
of
VG = –7.5 V
graphene
is finite,critical
but alsocurrent.
a finite supercurrent
can flow
at zero
–7 5
gate-tunable
(c) Modulation
of critical
VG = +1 V
charge
density.
Our
observations
shed
light
on
the
special
role
of
0
current with magnetic field, showing characteristic
– 250
0
250
–10
0
10
B (mT)
time reversal symmetry in graphene, and demonstrate phase
I (n A )
Fraunhofer pattern. (d) Multiple Andreev Reflections in
coherent electronic transport at the Dirac point.
c 1.2
d 60
100
(d)
(e)
14,16
sub
gap
region.
(e)
Shapiro
steps
under
microwave
Owing to the Josephson effect , a supercurrent can flow through
Ref. [5].
2 ∆/3
a irradiation.
normal conductor
placed between two closely spaced supercon30
0
.
0
1
ducting electrodes. For this to happen, transport in the normal con0
ductor must be phase coherent and time reversal symmetry (TRS)
–100
d 0.8
must be present. In graphene, the Josephson effect can be investigated
f = 4 .5 G H z
1/2
c
– 30 P (a.u.)
B = 10 T
B = 35 mT
15
2.0
in the ‘relativistic’ regime , where the supercurrent is carried by
2 .5
∆
T
=
3
0
m
K
Dirac electrons. However, it is not clear a priori that graphene can
T = 100∆ m2 K
– 60
– 150
75
– 75
0
–
7
0
0
3
5
0
350
700
–
0
2 .0
support supercurrents, because other quantum interference pheV (µV)
1.5
I (n A)
nomena that require both phase coherence and TRS were found to
1 .5
Figure 2 | Josephson effect in graphene. a, Voltage, V, versus current bias,
17
be absent or strongly suppressed in previous experiments . Below we
I, characteristics at various VG, showing a modulation of the critical current.
1 .0
1.0
show experimentally that the Josephson effect in graphene is a robust
Inset, current bias sweeps from negative to positive (red) and vice versa
(blue), showing that the asymmetry in the main panel is due to hysteretic
phenomenon, and argue that its robustnessLetter
is intimately linked to
REPORTS
0 .5
behaviour (the retrapping current is smaller thantransistor
theoperations
switching
current,
as
(15) suggests that interesting
graphene’s unique electronic structure.
Nano Letters
Letter b, Colour-scale
0.5
Superconducting
Dome
in
a
Gate-Tuned
basic physical properties
may be revealed by usingof
representation
is
typical
for
an
underdamped
junction).
(a) Single- and few-layer
0 .0
the EDL dielectrics (11, 16–18). To make our de(b) graphene Josephson junctions are fabricated
isolated
flakes of MoS from a
dV/dI(I,B)
is zero, vices,
thatweis,
thethinsupercurrent
2030 mK 0(yellow-orange
– 40 at T–5
20
2superconducting
– 40 – 20
0
0
40
Band
Insulator
proximity effect between Pb and Bi2Te3
bulk 2H-type single crystal (Fig. 1A) by the
1
(d)
region,
and
red
corresponds
to
finite
dV/dI).
The
critical
current
Scotch
tape
method
widely
usedexhibits
in graphene a
VG (V) established along
VG (V)
on oxidized Si substrates by mechanical exfoliation of bulk graphite ,
J. T. Ye, the
* Y. J. Zhang,
R. Akashi,direction
Bahramy,
Arita,
Y. Iwasa
thickness
ofR.normal
the Bi
Te3*(Au)
flake.
research (19, 20) and transferred
them onto the
(a) Superconducting
(Al)M. S.and
contacts
coupled
c
,
Differential
series of oscillations
described
by2 a Fraunhofer-like
pattern.
surface of HfO grown by atomic layer deposiThese observations
were
interpreted
of ofamanywidely
followed by optical microscope inspection to locate the thinnest
A dome-shaped
superconducting
region appearsin
in theterms
phase diagrams
unconventional
tion on a Nb-doped SrTiO substrate. We selected
resistance,
dV/dI,
versus
V,
showing
multiple
dips
22by theAndreev
superconductors.
In doped
band
insulators,
however, reaching
optimal insulator
superconductivity
to extended
surface
states
of
topological
Bi1.5reflection
Sb
Teby1.7
Sebelow
atomically
flat thin0.5
flakes
examining
their
1.3
proximity
effect, including the bulk of the flakes.
fine-tuning of carriers has seldom been seen. We report the observation of a superconducting
graphitic flakes, and electron beam lithography to define electrical
optical occur
micrographsat
(21)values
and patternedof
them into
Figure 1 | Sample characterization.
a
,
Atomic
force
microscope
image
of
a
the
superconducting
energy
gap.
The
dips
in
dV/dI
dome in
the temperature–carrier
density
phase diagram ofin
MoSthese
, an archetypal
band insulator. By a Hall bar configuration (Fig. 1B), which acts as
Becasue
the
bulk
transport
can
be
substantial
materials,
quasi-continuous
electrostatic
carrier doping
through
a combination
of liquid and
(BSTS).
Josephson
effect
confirmed
between
(b)
transistorleads
channel (11,
16).by
A droplet
of ionic
d,solid
a.c. aAl
Josephson
effect.
The
5plausible.
2D/en,
where
nachieved
iselectrodes.
an
integer
number.
contacts. Figure 1a shows an atomic force microscope image of a
single layer graphene device
twoV
superconducting
We
gating,
weis
revealed
a large enhancement
in the transition
temperature T occurring
at optimal
thisbetween
interpretation
However,
considering
that
the
liquid (DEME-TSFI) (21) was applied onto the
doping Shapiro
in the chemicallysteps
inaccessible
low–carrier
density
regime. This observation appear
indicates thatwhen
in
the
I–V
characteristics
is
irradiated
surfacethe
of the sample
thin flake covering
the side gate
have fabricated deviceshysteretic
with
electrode
separations
incoupling
the
100–500
nm.
extremely
extended
Josephson
in band
these
reports
is not
supercurrent
(c)
Fraunhofer
pattern
in
typical device. We use as superconducting contacts a Ti/Al bilayer
the
superconducting
dome may
arise
even range
inand
doped
insulators.
electrode (Fig.
1C). A voltage applied between
with
microwaves.
In
the
example,
we
applied
4.5 GHz
microwaves,
the thus formed
liquid gate (LG) andresulting
the channel
well
understood,
the
nonlocal
effect
we
report
here
may
have
b
,
Schematic
representation
of
graphene
between
superconducting
n many unconventional superconductors (4, 5). Except in certain cases (4, 6), however, drives either anions or cations onto the channel
(10/70 nm). Titanium ensures good electrical contact to graphene,
magnetic
field
ofInset,
Ic.doping
(d) Nonlocal
Fraunhofer
inthedependence
9.3
mV
voltage
colour-scale
plot
the characteristic
chemical
to achieve low carrier
den- showing
transition
temperature
T has asteps.
max- using
surface under positive
or negative bias, respecbeenininvolved
in(1–3),
the
observation.
(e)
electrodes.
The
two
electrons
a
Cooper
pair
entering
graphene
into
sities1/2
results in go
nonuniformity
or phase separation. tively. The ions and induced carriers (~10 cm )
imum as
a function
of external parameters
;
a.u.,
arbitrary
units)
dependence
of the
a.c.
microwave
amplitude
(P
and Al establishes a sufficiently high critical temperature to enable the
In
summary,
we
demonstrated
a
surface-dominant
Josephson
Other
systems
exhibiting
optimal
doping
of
the
such
as
chemical
doping
or
pressure;
in
cuprate
right
beneath
form
an
equivalent
capacitance of
patternby between
normal
leads
when
biased
between
superconducting
state
include
two-dimensional
families,
this
so-called
superconducting
dome
~10 mF/cm , largeto
enough
for inducing
superc
,
Twothe
red
and
blue
cones
(see
text).
(c)observation of supercurrents in a dilution refrigeration set-up18. different K-valleys, represented
Josephson
effect
(orange
and
red
correspond
respectively
zero
and
finite
effect in S−TI−S
constitutes
a direct
(2D) electron systems
at surfaces and
and interfaces, conductivity at the interface (10–14). In addiarises upon junctions.
the chemical dopingThis
of the parent
whose
phase
diagrams
may
be
explored
by
apMott
insulator
(1).
In
band
insulators,
similar
tion,
we
were
able
to
modulate
the
carrier
density
superconducting
and
between
superconducting
Tof5the30existence
mK
and
adominant
small
terminal resistance versus
gate voltage,
VdV/dI).
unequivocal
confirmation
of a(e)
surface
G, at leads,
behavior was observed at low carrier densities plying electric fields (7–9). In recent years, this (to ~10 cm ) using a high-k dielectric (HfO )
Before discussing their superconducting properties, we first char- REPORTS
28
electrostatic
carrier
doping
has
been
effectively
(4)
where
superconductivity
is
usually
not
favorback gate (BG), which remains effective after
conduction
in
BSTS
TI
flakes
and
Josephson
coupling
magnetic
B 5is35
mT,
drive
the
electrodes
into
the
normal
state.
vices
(15),normal
the
more pronounced
saturation
ambipolar
transport
could
be found
flakes
with
regime where field,
chemical doping
plagued
by to
leads
when
biased
between
leads.
Ref
[6].
implemented
by using
ionic
liquidsattohigh
form
an the
able
because
of thein
low
density
of states
(DOS)
freezing
of the ion motion at a temperature
acterize the devices with the superconducting electrodes in the norVchannels
less intrinsicto
electron
(sulfur
deficiency),
nonuniformity.
indicates
well-behaved
transistor
operation.
electrical
double the
layer
(EDL)
high capacitance below ~200 K. For carriers induced at the top
confined
the doping
surface
conducting
when
TI offlakes
The
aperiodic
conductance
fluctuations
are
due
to
random
quantum
(10–12).
This
method
has
produced
carrier
densurface
of the MoS flake by the LG, the effecFigure 1D shows the transfer curves of a typ- where the barely metallic state in the p-channel
After introducing carriers onto the channel
arewasin
proximity
tosuperconductivity
superconductors.
The
magnetic
field
mal state. Figure 1c shows the two-terminal resistance, R, versus gate
sities
that
span
the superconducting
in high- tive BG capacitance is affected by two layers of
still far from
reaching
hole
ical double-gate device
atelectron
220 K with a sourcesurface at 220
Kconductance,
with
different
liquid-gatedome
biases
d
,
Two
terminal
interference
of
waves
(see
also
Fig.
4).
T cuprates (13, 14) and may be an effective tool dielectrics: HfO and the bulk of MoS flake.
(16). To confirmof
the the
electrostatic
operation of ofV the
drain voltage V = 10 mV. For the n-channelmodulations
, we measured
the four-terminal
sheet resistsupercurrent
Al−BSTS−Al
junction
to access exotic superconducting states in other Using this double gating method, we could acV
-dependence
of
R
voltage, VG, for one of our samples. The strong
LG,
we
performed
a
transfer
curve
measurement
conduction,
an
on/off
ratio
of
>10
ance
R
was
reached
as
a
function
of temperature Tawhen the
G
at
high
magnetic
field,
B
5
10
T,
and
T
5
100
mK,
showing
G,
versus
V
materials.
cess a large
of carrier densities n quasiG
REPORTS
–2
remained
up
to
the
superconducting
critical
field
of
the
AlMoS
(1012 range
cm
)
for biasing with either the liquid ionic gate (V ) with fast gate bias cycles2(fig. S1). The possibility device was being
cooled
downband
to 2insulator,
K (Fig.
2A)., be- ncontinuously
We chose
a typical
and precisely,
thereby avoiding
a
provides a first indication that the device consists of at most a few
/h,
of
the
series
of steps
values
4ewas
of
a chemical of
reaction
ruled
out
by repeatAt gate
biases
V the
or the high-k
back gateat
(V half-integer
), with a channelelectrodes.
<anomalous
1 V,Al−BSTS−Al
we observed
a negative
the staging effect
cause
high
mobility
found in
its–solid-state
A
voltage
incharacteristic
part
of the TI
outside
the
2
0 (22), even in the low-density
2
vices
(15),
the
more
pronounced
saturation
at
high
ambipolar
transport
could
be
found
in
flakes
with
regime
where
chemical
doping
is
plagued
by
resistance R > 1 gigohm in the off state. Com- ability and a negligible (~1 nA) leak current I , temperature derivative of R (dR /dT) for insu1
junction
was
nonlocally
triggered
along
with
modulations
of
the
0 /dT with V indiQHE
in
single
graphene.
VG 4.affects
the carrier
layers of graphene , since, owing to screening,Figure
less
intrinsic
electron
(sulfur deficiency), VDSoff
nonuniformity.
indicates
pared
BG, the
LGlayer
not doping
only had
state (> 1well-behaved
gigohm). The transistor
lating states. operation.
The increase 2
of0
dR
2
(a) Nonlocal modulation voltage
vwith thecorresponding
to 10 times as well as a persistent
ARTICLES
x = 0.10
2 × 10 – 5
x = 0.08
x = 0.06
x = 0.05
1 × 10 – 5
x = 0.04
x = 0.03
x = 0.02
0
ARTICLES
– 1 × 10 – 5
• Competition between collective
electronic states is common in
Transition Metal
Dichalcogenides (TMD).
100
150
200
250
300
b)
b
240
CuxTiSe2
Semi metal
10
x=0
200
0
3
M/H (300 K )
– 40
2
160
Cu
Se
c
b
a
1
x = 0.01
x = 0.02
Resistivity (mΩ cm)
γ
120
Ti
T (K)
4
x = 0.03
x = 0.04
• Intercalation of Cu or Pd atoms
in TiSe2 suppresses Charge
Density Wave (CDW) and a new
Tc = 1.8 K (Fig. 1(b)). Authors
that the mechanism of superconductivity is fundamentally
superconducting
state argue
emerges.
CDW
80
x = 0.06
0.1
x = 0.08
1
40
– 80
Metal
x = 0.10
40
0
0.02
0.04
0.06
x in CuxTiSe2
0.08
x = 0.01
– 40
0.10
S (µV K – 1)
0
x=0
0
4
3
0.01
Figure 5 Summary of the composition-dependent properties in Cux TiSe2 .
M/H (300 K), ρ (300 K)/ρ (6 K), electronic specific heat coefficient γ and Seebeck
coefficient S(300 K) as a function of Cu composition x. The solid lines are guides to
the eye showing the linear variation of the latter two quantities, whereas for the
former two the lines reflect steep changes at low Cu content (0 < x < 0.02) and
through the superconducting state (0.04 < x < 0.08).
– 160
1
0
0
x = 0.03
– 120
SC
2
x = 0.08
– 80
50
100
100
150
200
150
250
T (K)
200
300
350
250
400
300
0.02
x
0.04
0.06
x in CuxTiSe2
0.08
0.10
T (K)
Figure 2 Magnetization and transport properties of Cux TiSe2 . a, Cux TiSe2 M(T )
curves measured in a constant H = 0.5 T applied field, for 0 < x < 0.10. The solid
lines illustrate how the CDW transition temperatures have been determined. (A small
peak is seen around 60 K in a few of the measurements, which is attributed to an
oxygen impurity trapped in the measurement system.) b, H = 0 temperaturedependent resistivity data for 0 < x < 0.10. Inset: Seebeck coefficient for x = 0,
0.01, 0.03 and 0.08.
2
546
c1
Device Fabrication Results
c2
x
in TiSe .
c
c
−1
2
c2
0
c2
B
B
2
I
0
B
c2
c
c2
∗
c2
c2
Tc
c2
∗
c2
c
4+
2−
c2
c2
0.08
2
c2
c2
• Optical Lithography
• Ohmic contact using Nb (Tc ≈ 9K)
• CDW transition confirmed
• Device thickness = 25 nm ≈ 40
layers
Chiral CDW in TiSe2
c
a)
DS
b)
549
nature physics VOL 2 AUGUST 2006 www.nature.com/naturephysics
©2006 Nature Publishing Group
Untitled-1 6
21/7/06, 4:44:03 pm
The 1T polytype crystal structure of TiSe2. c = 6.004 Å, a =
Fig.
2. Å
(a) Schematic crystal structure ofM. 1T
polytype
of TMDs where the transition metal atoms
octahedrally
IAVARONE
et al.
PHYSICALare
REVIEW
B 85, 155103 (2012)
3.536
coordinate by chalcogen atoms and the sample
thickness
of the unit cell in c direction is one sandwich [26]. (b) Crystal
slightly increases by decreasing the temperature from
300 K, structure
shows a broadofhump
with
a maximumthe
at about
ais1 trigonal prismatic and (a)
a3
2H
polytype,
coordination
the
structure of TiSe2 [44]. (c) Schematic crystal
150 K. The maximum resistance of the sample at 150 K
three times
larger as that
roompolytype
temperature, indicating
unit cell is two sandwiches thick [26]. (d)isOne
sandwich
ofat2H
[71].
a
• Exhibits unique chiral CDW state
B
below T ≈ 200K [2-4].
P H Y Sorder
I C A transition
L R E V Iinto
E W a Lcommensurate
ETTERS
undergoes a second
At ~105,
202 K
TiSe2(2010)
PRL
176401
a good stoichiometry. The resistivity at low temperature is
≈1 m! cm. For the Cu0.05 TiSe2 and Cu0.06 TiSe2 crystals,
the in-plane resistivity is metallic with a residual resistivity
ratio ρab (300 K)/ρab (4 K) = 4 and 4.5 for the two samples,
respectively. The resistivity just above the superconducting
transition is 100 µ! cm for the Cu0.05 TiSe2 crystal and
70 µ! cm for the Cu0.06 TiSe2 crystal. The corresponding superconducting critical temperatures are Tc = 2.0 and
3.0 K.
Low-temperature scanning tunneling microscopy (STM)
and spectroscopy (STS) have been performed at T = 4.2 K
using a Unisoku UHV STM system, with a base pressure of
1 × 10−10 Torr. The samples were cleaved just before cooling
down. We used Pt-Ir tips in all of our experiments, therefore
the tunneling conductance between a normal electrode (tip)
and a sample provides, in the limit of low voltages, the
electronic density of states of the sample. All STM and STS
measurements reported in this paper have been performed at
4.2 K and therefore the samples are in the CDW state but not
in the superconducting state.
week ending
22 OCTOBER 2010
aobserved
a3
2a0 × 2a0 × 2c0 CDW state without an intermediate incommensurate state,
by, neutron
1
rectly the
interplay
between chirality and transport in
diffraction [11], electron diffraction [27], and scanning tunneling microscopy
(STM)
[28].
A
A
a2
electron
systems.
Our findings could also pave the way
driving force of the instability has been under debate for decades,
during
this
time
several
Ia1
Ia2
(b)
I
towards
theorinvestigation
of charge helicity and external
hypotheses were proposed: (i) the formation of an excitonic insulator [12],[16],[29]
evena3 (ii)
fields.
an excitonic condensate, (iii) an antiferroelectric mechanism [17]-[18], (iv) the involvement of a
We thank K. Yamaya and T. Toshima for providing
band-type Jahn-Teller effect [19]-[20], (v) an indirect Jahn-Teller effect
[21]
(c)
Ia2 combination
Ia1 and the
advice on sampleIa3preparation. We also thank K. Inagaki,
of (i) and (iv) [22]-[23],[26]. Up to now, no unambiguous and conclusive explanation has been
T. Matsuura, and H. Nobukane for measuring the electriIII.
STM
CHARACTERIZATION
OF
THE
CDW
STATE
given. However, there is a general consensus that the CDW transition in cal
TiSeproperty
by fruitful discussions. And we thank
2 is not driven
and for
2a0
Atomically resolved images were acquired in the constant
current mode with a constant voltage between sample and
2 STM topography
(a) 12.1 xFIG.
8.91. nm
pure TiSe2 at
(Color online)
(a) 12.1 × 8.9 nm2 STMimage
topographyof
image
tip. Figure 1(a) shows a 12.1 × 8.9 nm2 STM image of pure
of pure TiSe2 acquired at T = 4.2 K. Scanning parameters are I =
TiSe2 . All surface Se atoms of the hexagonal surface layer
Tare= 4.2K.
The dashed line separates two domains of
300 pA and V = −0.2 V. The dashed dark line shows the transition
clearly resolved with a strong superlattice modulation 2a0 ×
between
two different
domains. (b)(b),
Line profiles
along the
unit vectors along unit
different
CDW
chirality.
(c) Line
profiles
2a0 clearly seen in direct space.
a1 , a2 , and a3 in the region B of image (a). Iai is the normalized
A closer
Visualization of the two CDW chiralities.
Reflook
[3].at the amplitude of the modulation along
vectorsamplitude
for regions
and A,along
respectively.
Ref
[4].
along
of the STMBcorrugation
ai . (c) Line profiles
the three directions reveals that the amplitude along the
the unit vectors a1 , a2 , and a3 in the region A of image (a).
crystallographic direction a2 has higher intensity than that
2
along a3 and a1 , in the top region B of the topography
image. The same analysis reveals instead Ia1 > Ia3 > Ia2
The same conclusion can be drawn from the twoin the bottom of the image [portion A of Fig. 1(a)]. We
dimensional Fourier transform (2DFT) of the region A [see
name the phase in region A clockwise and that in region B
Fig. 2(a)] and region B [see Fig. 2(b)].19 The 2DFT shows
anticlockwise, where clockwise and anticlockwise indicates
in both cases two sets of peaks. The outer peaks corresponds
[1] Morosan, E., Zandbergen, H. W., Dennis,
B.
S.,
Bos,
J.
W.
G.,
Onose,
Y.,
Klimczuk,
T., etpeaks
al. (2006).
Nature
Physics,
2,is544.
the direction of increasing amplitude. These two domains
to the Bragg
of the Se lattice
whose
wavelength
a0 ,
cannot
be
superimposed
on
each
other
with
simple
rotational
the
inner
peaks
correspond
to
the
CDW
superlattice
peaks
[2] Di Salvo, F. J., Moncton, D. E., & Waszczak, J. V. (1976). PRB, 14, 4321.
transformation, but they are mirror images of each other. This
whose wavelength is 2a0 . The three sets of CDW peaks have
11
[3] Ishioka, J., Liu, Y. H., Shimatake, K.,
Kurosawa,
T., Ichimura,
K., Toda,
Y.,theet al.different
(2010).
PRL,in105,
176401
observation
in agreement
with previous reports
indicates
intensities
the two
images. The 2DFT of region
of the CDW inM.,
TiSe2Moore,
. The line profiles
the
A in the topography
image
reported
Fig. 1 shows that the
[4] Iavarone,
X., state
Golalikhani,
S. A.,along
& Karapetrov,
G. (2012).
PRB,
85,in155103
0 M., Di Capua, R., Zhang,chiral
three directions in region B are reported in Fig. 1(b) while
CDW peaks increase in intensity clockwise with a normalized
[5] Heersche, H. B., Jarillo-Herrero, P.,those
Oostinga,
J. B.,
Vandersypen,
F.: (2007).
related to the
region
A are reported in L.
Fig.M.
1(c).K.,
The& Morpurgo,
amplitude of Iq1A.
: Iq2
Iq3 = 1.0 :Nature,
0.33 : 0.45,446,
while56.
in the
intensities
of theShin,
modulations
areet
Ia1 al.
: Ia2(2014).
: Ia3 = 0.3Nano
:
2DFT
of region
increase in intensity anticlockwise with
[6] Lee, J. H., Lee, G.-H., Park, J., Lee, relative
J., Nam,
S.-G.,
Y.-S.,
Letters,
14,B they
5029.
1 : 0.7 in region B and Ia1 : Ia2 : Ia3 = 1 : 0.5 : 0.7 in region
a normalized amplitude of Iq1 : Iq2 : Iq3 = 0.46 : 1.0 : 0.7.
[7] Ye, J. T., Zhang, Y. J., Akashi, R., Bahramy,
M. S.,have
Arita,
& Iwasa,
(2012). Science,
1193. and analysis have been performed
A. These profiles
been R.,
normalized
to the Y.
maximum
The same338,
measurements
amplitude alongA.
each
direction. The
profiles
can slightly
for the Cu0.05 TiSe2 and Cu0.06 TiSe2 crystals. In Fig. 3(a), a
[8] Jo, S., Costanzo, D., Berger, H., & Morpurgo,
F. (2015).
Nano
Letters,
15, 1197.
change from location to location especially when approaching
15 × 6.3 nm2 STM image for Cu0.05 TiSe2 is reported. The
4
FIG. 4 (color). Schematic representation of (a) left-handed
chiral CDWs and (b) right-handed chiral CDWs in a TiSe
CDW unit cell in real space and (c) typical 2D CDW. Charge
References
concentration is indicated by the deeply colored part. The colors
correspond to the colored CDW q vector in the inset. In a CDW
unit cell, the density peaks of three CDW are shifted at intervals
of 2c =3. If we look at one layer, the different intensities of three
CDWs form a low symmetry structure. There are two phases
caused by difference in stacking direction (red-blue-green or redgreen-blue). As with cholesteric liquid crystals, the stacking
−2
*Corresponding author.
tanda@eng.hokudai.ac.jp
images of two sub-micron devices. Metal contacts
[1] A. D. Dolgov, Phys. Rep. 222, 309SEM
(1992).
3 He
(a) thermally
evaporated
(World Al (55nm)/Au (5nm), and
[2] G. E. Volovik, Exotic Properties ofare
Superfluid
(b) sputtered Nb (60 nm).
Scientific, Singapore, 1992).
[3] A. P. Mackenzie and Y. Maeno, Rev. Mod. Phys. 75, 657
2
2
*To whom correspondence should be addressed. E-mail:
yejianting@ap.t.u-tokyo.ac.jp (J.T.Y.); iwasa@ap.t.u-tokyo.ac.
jp (Y.I.)
2D
2
LG
G
s
s
LG
D2efficiency
C2 the barely
gate
(the change
of channel
IDS versus supercurrent.
the gradual
formation
of degenerate carriers
VDS characteristic
in the
outputvoltage
op- catesonto
junction
The
across
the
Al−BSTS−Al
metallic
statecurrent
in the
p-channel
Figure 1D shows the transfer curves of a typ- the where
After introducing
carriers
the channel
versus gate voltage DI
), but also created an eration (Fig. 1E) of a typical MoS2 EDL tran- and enhanced mobility at low temperatures. The
DV reaching
1
i
,
as
a
function
of
the
bias
current
I
applied
between
A
and
was
still
far
from
hole
superconductivity
ical
double-gate
device
at
220
K
with
a
sourcesurface
at
220
K
with
different
liquid-gate
biases
wascorroborates
also modulated
by thechannel
nonlocal
at the
A1B1the
D1CPO
A1additional
BThe
1
1
Kavli
of Nanoscience,
Delft University
Technology,
Box 5046, 2600 GA, Delft,
Netherlands.
surface bias
shows current
metallic transport
(posp-channel when a 1negative VG wasjunction
sistor (EDLT)
the more pronounced
Figure
2. Institute
(a) A scanning
electron micrograph
of theofdevice
and
(16).This
To
confirm
the electrostatic
drain
voltage
VDS = field
10 mV.
For
the
n-channel
, 1weV)measured
the opfour-terminal
sheet resist-V. The enhancement
LG<junction.
itive dR
n-channelof
(0 < VVLG
applied.
ambipolar
transport
indicates thatoperation
than p-channel
of
s/dT) at VLG ≥ 1effect
Au−BSTS−Au
This
mirage
Fraunhofer
can
be
B
,
and
magnetic
B.
i
,
is
the
fraction
of
the
modulation
*These
authors
contributed
equally
to
this
work.
1
A
B
D
C
measurement setup used in this study, which consisted ofconduction,
Au an on/off ratio of >10
1 14 1 1
LG iswe
effective
in shifting
the Fermicurve
level EFmeasurement
eration (–0.6 < V
metallicity continues
–0.2V),
performed
a transfer
ance
was reached theLG,
function with
of temperature
T whenwith
thefurther increase of VLG,
LG < R
s as a consistent
explained
only
in terms
of
existence
of robust
the transfer characteristics.
Compared
with the
the down
and
superconductivity
emerges atsurface
V = 4 V.
bothCgate
the
and
conduction
bands
iA1with
10 the
nA)
flowing
between
andfast
vbias
exhibits
electrodes (yellow) and Al electrodes (blue), overlaid on a BSTScurrent
flake.
(fig. S1).
The possibility
for
biasing
either
liquid
ionic gate
(VLG) toDaccess
device was
being cooled
to 2 K (Fig. 2A).
B1 (=
1with
1. valence
D2C2 cycles
10T0c shows LG
n-channel operation
observed in
monolayer
de- The
(16). An enhanced p-channel with more balancedconducting
transition
temperature
clear
56
channels,
which
extend
to
the
sides
of
the
TI.
TheVLG
of
a
chemical
reaction
was
ruled
out
by
repeatAt
gate
biases
V
or
the
high-k
back
gate
(V
),
with
a
channel
<
1
V,
we
observed
a
negative
BG
LG
Fraunhofer-type
variations.
The
blue-colored
(red-colored)
inside
IA1B1 (IC1D1) and VA2B2 (VC2D2) are the bias current and voltage
1
local
and
nonlocal
Fraunhofer
diffraction-type
modulation
of
ability
and
a
negligible
(~1
nA)
leak
current
I
resistance
R
temperature
derivative
of
R
>
1
gigohm
in
the
off
state.
Com,
(dR
/dT)
for
insuNature
Publishing
Group (resistive)
DS ©2007
G
s
s
region
of the pattern
denotes
the pair-conducting
difference between a pair of the Al (Au) electrodes, respectively. (outside)
x
and
pared
with
the
BG,
the
LG
not
only
had
10
times
as
well
as
a
persistent
off
state
(>
1
gigohm).
The
lating
states.
The
increase
of
dR
/dT
with
V
indis
LG
the
junction
critical
current
and
differential
resistance
with
state of the Al−BSTS−Al junction. (b) Nonlocal modulation voltage
the gate efficiency (the change of channel current IDS versus VDS characteristic in thepersistent
catesmodulation
the gradual formation
degenerate shows
carriers nice fits to
output op- field
y denote the spatial coordinates viewed from above. (b) Current−
of theofenvelope
vversus
to
DS iC1D1,B1A1 as a function of the bias current IC1D1
B2A2 corresponding
gate voltage DI
eration (Fig. 1E) of a typical MoS2 EDL tran- and enhanced mobility at low temperatures. The
DVG ), but also created an
voltage characteristics of the Al−BSTS−Al junction at the bath
model of a edge-stepped nonuniform supercurrent density
additionalbetween
p-channelCwhen
negative
sistor
pronounced channel surface shows metallic transport (pos- 0
applied
magnetic
field (EDLT)
B. iC1Dcorroborates
,
is thethe morethe
G was
1 anda D
1, and V
1 B1A1
temperature of 10 mK, where the blue/red curve and arrow denote
the This ambipolar
Fig. 1. MoS2-based EDLT device and its transistor properties.
on the oproughitive
side
BSTS
flakes. of
This strongly (A)
dRssurfaces
applied.
transport indicates that n-channel (0 < VLG < 1 V) than p-channel
/dT) at VLGof≥ 1the
V. The
enhancement
Ball-and-stick model of the layered 2H-type MoS2 single
of iC1D1 (= 100 nA) flowing between B1 and A1. vB2A2 also
r
crystal.
(B) Optical micrograph of a typical MoS2 device un- 0
suggests
that
the
Josephson
coupling
in
a
TI
is
established
sweeping direction of IA1B1 from left/right to right/left. IcA1B1 andfraction
theIALG
is
effective
in
shifting
the
Fermi
level
E
eration
(–0.6
<
V
metallicity
continues
with
further
increase
of
V
<
–0.2V),
consistent
with
,
F
LG
LG
1B1
transmission light illumination. (C) Double-gate device
exhibits
variations.
Temperature
increasecharacteristics.
due to Joule Compared
with the
and
superconductivity
emerges
at Vthat
to accessFraunhofer-type
both the valence and
conduction
bands the transfer
4 V.topologically der
LG =are
through
the
surface
conducting
channels
and measurement configuration.
VD (D) Transfer curve of trandenote the critical and retrapping currents, respectively. (c)heating
The altered the
de- This
(16). An enhanced
p-channel
more
balancedpeakn-channel
Thestudy
transition
temperature
Tc shows
clear Vfor
around Bobserved
= 0. in monolayer
shape with
of the
primary
of vB2A2 operation
LG confirming sistor operations by accumulating carriers by EDL top liquid
protected.
provides
a unique
method
DS
G
• Gate-induced superconductivity in other TMDs
Figure 1. A false-colored schematic diagram of devices with nonlocal
measurement configurations, where the electrodes A and B are used as
a source and a drain, respectively, for the bias current IAB and the
voltage difference VCD are monitored between electrodes C and D.
The normal conductor of the device consists of (a) conducting bulk,
(b,c) insulating bulk, and conducting surface. The conducting surface
in (c) is superconducting between the electrodes
A and B. The arrows
15 µm
denote the current directions in the normal conductors. The crosssection between x1 and x2 corresponds to the dotted x1 −x2 line in
Figure 2a.
• Fabricated sub-micron devices with Al and Nb using e-beam
lithography.
• The Si/SiO2 substrates will be used for applying a backgate
voltage for tuning electronic properties.
Observations
• Extended exposure to solvents tend to accelerate growth of
insulating film.
• Ion milling of a few nanometers of material is usually
necessary to make good contact.
• Better success with sputtering versus thermal evaporation.
(b)
(a)
Nano Letters
(d)
gate (red: ramping V up; blue: ramping V down) and HfO
differential resistance dVA2B2/dIA1B1 of the Al−BSTS−Al junction,
bottom solid gate (green), both at 220 K. (E) Output curve of
Josephson coupling via the topological surface
– 10conducting
0
the thin-flake MoS EDLT with both electron (0 < V < 1 V)
and hole (–0.6 < V < –0.2 V) channel.
obtained Other
by numerically
differentiating
the Isuperconductivity
20
– 40Well-behaved–saturation
20 at large V 0was found in the dominating
TMDs
exhibit
when heavily
A1B1−VA2B2 characteristics at
channels, which in turn provides a solid
basis for exploring
electron transport. For each V , we measured the two overlapping I curves with forward and backward
distribution of the bias current IC1D1 via the BSTS TI surface.
(
V
)
V
scans
of
V
.
G
Majorana Fermionic excitation states by adopting TI/superdifferent doped
values of perpendicular
field(a)-(c)
B, as a function
B
by cionic magnetic
gating.
MoSv2Bof2Adisplays
metalvanished if ICa
< IcA1B1 (= −IcA1B1). For IC1D1,B1A1 < IcA1B1,
conductor
2
1D1,B1A1
and IA1B1 normalized by IA1B1(0).
(c) heterostructures.
1193
www.sciencemag.org
SCIENCE
VOL 338 30
NOVEMBER 2012
Figure 3 | Bipolar
supercurrent
transistor
behaviour
and finite
became
i
×
R
as
shown
in
Figure
4b,
where
R
insulator transition at lowervB2Adoping,
and
a
C1D1,B1A1
B2A2
B2A2
2
supercurrent at the Dirac point. a, Colour-scale plot of dV/dI(VG,I). Yellow
ASSOCIATED
CONTENT
= 80 Ω was the normal-state resistance between B2 and A2. The
means
zero, that is, the supercurrent region, and finite dV/dI increases via
superconducting
dome
at
higher
doping,
with
max
T
c
the Josephson coupling occurs dominantly via the surfaceappearance of this nonlocal feature
indicates the existence of a
S Supporting Information
orange
to dark red. The current is swept from negative to positive values, and
*
conducting
layer. Ref.
One would
not WS
expect
this nonlocal
mirage
topologically
protected conducting surface on our BSTS flake25
≈ 10K.
[7]. (d)
becomes
superconducting
2 also
is
asymmetric
owing crystalline
to the hysteresis
associated with an underdamped
Normal-state properties of the BSTS
flake, overFraunhofer effect for the case of N, sandwiched betweenwith
theJosephson coupling across it.
junction
(seeJosephson
also Fig. 2ajunction,
inset). The
top axis shows the electron density, n, as
damped characteristics
of the
differential
with T ≈ 4K [8].
WS2 (positive) VG
two S electrodes Ac and B with finite bulk conduction becauseThe
in primary lobe of vB2A2 in Figure 4b exhibits a typical
obtained fromjunction
geometrical
resistance of the Al−BSTS−Al
by acconsiderations
measurements,1. For large negative
c
this case the current IAB (>IAB) would mostly be confined
Fraunhofer-type modulation but differs from that of vD2C2 in
Fraunhofer interference
pattern for aisnonuniform
supercurrent
the supercurrent
carried by hole
(electron) Cooper pairs. The
density,
fits
to
the
uniform-current
Fraunhofer
diffraction
for
between A and B as in Figure 1a. Only when the conductance
Figure 4a; the differential resistance is highly enhanced along
25
different
field
periodicity,
estimation
of
the
nonlocal
current,
edge of the critical current. This difference resulted from
in a TI is predominated by the surface conducting layer, thethe
©2007 Nature Publish
MoS
and suppression of Ic due to bias-induced Joule heating. This
the
temperature
increase due to Joule heating generated
by2the
situation depicted in Figure 1c is expected.
material is available free of charge via the Internet at http://
biasitscurrent IC1D1(≃ 43 IB1A1). IcC1D1 ≃ 3 μA in Figure 4b
A scanning electron micrograph of the actual device and
pubs.acs.org.
c
to IB1A1 ≃ 70 nA. The Joule heating due to IC1D1
measurement configuration is shown in Figure 2a. The corresponds
setup
Fig. 2. Transport properties of the thin-flake MoS2 EDLT. (A) Temperature responds to the logarithm of the sheet resistance Rs (W). (D) Normalized
was Al
transferred to the BSTS flake between
B2 and
A2 with
consisted of a 115 nm-thick BSTS crystal flake, overlaid with
dependence
of the channel
sheetthe
resistance Rs at different VLG gate biases superconducting transition Rs/Rs (15K) as a function of temperature for difAUTHOR
ranging from 0 to 6 V (indicated on the right). (B) Temperature
dependenceINFORMATION
ferent gate voltages. The Tc is marked as a circle at 90% of the total transition.
carrier
temperature
in
the
BSTS
flake
reaching
a
temperature
of
and Au electrodes. The BSTS flake, grown using the self-flux
of the channel sheet resistance Rs at VLG = 1 V and different VBG’s showing a Both VLG and VBG are varied; for a given VLG, we show the evolution of the
n2D =Ic6.7 × 1012 cm−2Corresponding
transition with increasing VBG. All data corresponding to the same VLG are
. For each VBG, we marked Author
∼115 mK (see Supporting Informationmetal-insulator
7). Thetransition
valueat of
B A1
method,26,27 was mechanically exfoliated onto a Si substrate
the corresponding n2D measured by1 the
Hall effect at 20 K. (C) Phase dia- shown with the same color. The dashed arrows indicate the order of increasing
*E-mail:
gram nA
showing
phases as a hjlee@postech.ac.kr.
function of VBG from –4 to 2 V in ∆VBG = 2 V. (Upper panel) For VLG between 4 and 5.5 V,
(≃ 70 nA) was suppressed below 110
attheTevolution
= 10of different
mK electronic
capped with a 300 nm thick SiO2 layer. Details of the BSTS
carrier density n2D. The phase diagram shows an insulating (n2D < 6.7 × Tc increased with increasing VBG; (bottom panel) for VLG = 6 V, Tc decreased
28
(without Joule heating) as the carrier temperature
rose
115
(6.7 ×to
1012
< n2D < 6.8 ×Notes
1013 cm−2), and a dome- with increasing VBG because the corresponding n2D had reached the peak of
1012 cm−2), a metallic
crystal growing are described elsewhere. The Au electrodes
c(n2D > 6.8 × 1013 cm−2), where the color cor- the superconducting dome.
like
superconducting
phase
The
authors declare no competing financial interest.
mK
due
to
Joule
heating,
which
led
to
the
condition
ΔI
(0)
B1A1
were deposited on the BSTS flake by combining electron(e)< iB(5
, so that enhancement of the1194
differential resistance at the30 NOVEMBER 2012 VOL 338 SCIENCE www.sciencemag.org
1A1
beam patterning, sequential e-gun evaporation of a Ti/Au
ACKNOWLEDGMENTS
edge of the critical current recurred, as discussed in Supporting
nm/350 nm thick and 400 nm wide) bilayer, and lift-off.
Information
3 in association with Supporting Information
This work was supported by the National Research Foundation
Immediately prior to the Ti/Au deposition, the surface of
the
Figure S3.
(NRF) through the SRC Center for Topological Matter (Grant
BSTS flake was Ar ion-beam cleaned for 20 s with a beam
Fraunhofer modulation has been reported over an extended
2011-0030046 for H.J.L., Y.S.S., and J.S.K.), the GFR Center
power of 4 W (= 400 V × 10 mA). The superconducting
Al in Bi2Se3 from the Pb−Bi2Se3 interface by the proximity
volume
for Advanced Soft Electronics (Grant 2011-0031640 for H.J.L.),
electrodes were prepared in a similar fashion as theeffect.
Au19 The same group has also reported a strong
the Basic Science Research Program (Grant
LG
LG
2
2
LG
LG
DS
LG
DS
DS
sectional view, along with the probable current profile via the
surface and the bulk. The cross-section is drawn through the
middle of the N conductor along the x-axis, which corresponds
to the dotted x1−x2 line in Figure 2a. Electrodes A and B are
used as the source and drain, respectively, for biasing current
IAB, and the potential difference VCD is monitored between the
electrodes C and D. Figure 1a depicts the case in which the N
layer is normal-conducting with finite bulk conductivity. In
Figure 1b, the N material consists of an insulating bulk and a
conducting surface layer that uniformly covers the entire bulk.
This mimics the conduction in a TI flake with normal-metallic
contact leads. The situation of Figure 1c is identical to that of
Figure 1b with the exception that the electrodes A and B are
replaced by superconductors.
In Figure 1a, with the electrodes C and D widely separated
(>1 μm) from the electrodes A and B VCD is expected to be
negligible as the diffusive current flow is confined primarily
between A and B. In Figure 1b, IAB injected from A spreads out
across the entire surface, that is, top, bottom, and four sides,
which results in a nonvanishing value of VCD.25 If the electrodes
A and B are superconducting and arranged sufficiently close to
one other as in Figure 1c, Josephson coupling is established
10 µm
through the surface conducting layer between
electrodes A and
B. For IAB less than the superconducting critical current IcAB, the
entire current IAB injected from A is confined only between A
showing
CDW
T ≈c 150K.
andcharacteristic
B, showing vanishing
VCDpeak
. For Iat
AB > IAB, however, the
surface conduction between A and B becomes resistive so that
IAB injected from A spreads out across the entire surface before
being recovered at the drain B, as depicted in Figure 1b, and
induces a nonvanishing value of VCD. Thus, in this case, VCD
versus B reveals a replica of IcAB versus B, which confirms that
(b)
N. Hatakenaka for helpful comments. This work was supported by the 21COE program on ‘‘Topological Science
and Technology’’ from Ministry of Education, Culture,
Science and Technology of Japan.
2
2
c
s
Resistance vs. Temperature for Al-contacted TiSe2 flake,
(a)
−2
s
DS
• Electron beam lithography
• Ohmic contact using Al (Tc ≈ 1.2K)
• CDW transition confirmed
• Device thickness = 20 nm ≈ 33
layers
2
4
LG
d)
• TiSe is a transition metal
dichalcogenide (TMD), part of
layered MX2 family (e.g. M = Ti,
Mo, Nb, W, Ta. X = S, Se, Te).
Quantum-Phase Electronics Center and Department of Applied Physics, University of Tokyo, 7-3-1 Hongo, Bunkyo-ku,
Tokyo 113-8656, Japan. 2Correlated Electron Research
LGGroup,
RIKEN, Hirosawa 2-1, Wako 351-0198, Japan.
BG
Resistance vs. Temperature for Nb-contacted TiSe2 flake, showing characteristic CDW peak at T ≈
150K. The peak is suppressed possibly due to some local doping
impurities
contacts.
fourby
electrodes
labeledorA,the
B, C,
and D shown with cross-
c1
c)
14
DS
c
where superconductivity and CDW behaviour seem to coexist. The
reason why superconductivity arises from the CDW state in TiSe2
on Cu doping has not yet been determined. It may be that Cu
doping results in a tendency towards increasing the dimensionality
of the Fermi surface, destabilizing the CDW and allowing for
correlations to build in a third dimension, tipping the balance
in favour of superconductivity. Otherwise, superconductivity may
emerge from the CDW state due to the change in electron count on
Cu doping. Further study of Cu
2 x TiSe2 will determine which of these
c
1
c2
penetration depth, and ξ represents the coherence length), it can
be concluded that Cu0.08 TiSe2 is in the extreme type-II limit, as
κ ≈ 1 T/0.01 T = 100.
The variation of the transport, magnetic and thermodynamic
qualities of the normal state in the Cux TiSe2 series is summarized
in Fig. 5. The resistivity, specific heat, magnetic susceptibility
and Seebeck coefficient data taken together indicate that the Cu
atoms contribute electrons to the conduction band on doping.
This electron doping suppresses the CDW and induces metallic
1,2
13
2
c2
1,2
2
−2
B
c
2
c
c2
c
1
2
21/7/06, 4:43:44 pm
−4
2
1
2
2
c
c1
1
3
Goal: Investigate proximity induced superconductivity
c
2
nature physics VOL 2 AUGUST 2006 www.nature.com/naturephysics
Untitled-1 3
c1
Nano Letters
©2006 Nature Publishing Group
x
c
• Nonlocal Josephson effect in Topological Insulator
G (mS)
different in these two approaches, owing to different signs of the Hall effect and an extreme
sensitivity of pressure-induced superconductivity to magnetic
compared
to CuxTiSe
TiSe
.
(a) Phase field
diagram
for Cu-doped
2. 2(b) Resistance versus
• Max TThe
Cu0.08TiSe2, ,and
Figure 6 The Cu TiSe T–x electronic phase diagram. Open circles represent the
c = 4.15K
temperature
for group
several Cu-doping
concentrations,
ofand thethe
transition
metalshowing the
parent compound–TiSe
CDW transition temperature,
filled circles
correspond toIVb
the superconducting
2 a representative
evolution
from
toindicates
superconducting
state. Ref [1].
transition temperature.
The shaded
circleCDW
at x = 0.04
that the transition
Tc
≈
2K
for
Pd
TiSe
to
temperatures
above
1.8
K:
as
the
inset
in
Fig.
4a
shows,
the
0.11
2
diselenides,
has been known since 1960s and
has
been
extensively
temperature
is just below
our minimum
available temperature, andstudied
the dashed circle due to the
magnetization is linear in field at low H values; H is estimated
at x = 0.06 marks the barely visible CDW transition at x = 0.06. Inset: Crystal
as the field values where departures from linearity occurred at each
unconventional
nature
of
CDW
state
[11]-[24],
for the
first
time examined in detail by Di Salvo
.
structure
of Cu TiSe
temperature. The anticipated linear temperature dependence close
T is evident for both H and H (Fig. 4c), which also results
[11]. At into room
temperature, TiSe2 is either a semimetal [11]-[15],[24] or a semiconductor
a linear
√ thermodynamic critical field H (Fig. 4c), calculated as behaviour in Cu TiSe with a resistivity near 10 % cm at room
temperature in the metallic phase. As the carriers are introduced,
H = (H H ).
[20],[22],[25]
with
a
small
indirect
gap
of
the
order contribution
of 10-150
meVheat,[22],[25].
The possible
γ , increases from
to the specific
Close to T = 0, BCS theory predicts that the upper critical field the electronic
decreases with temperature as H (T ) ≈ H (0)[1 − 1.07(T /T ) ] approximately 1 to approximately 4 mJ mol K at the optimal
reasons for
ambiguous
results
beH(i) superconducting
an insufficient
resolution
composition.
Estimates ofof
theARPES
Wilson ratioexperiments,
(ref.2these
27). The dashed
line in Fig. 4c
representscould
a fit of the
data to this expression, yielding a H (0) value of approximately R = χ /(3γ )(πk /µ ) (where χ is the temperature-independent
(ii) a tiny1.39value
of theT gap
combined
the strong
CDW
toisroom
k is the fluctuations
Boltzmann’s constant up
and µ
the Bohr temperature
susceptibility,
(3.2 K) estimated
from thiswith
fit is smaller
T. (The zero-field
than the measured value of 4.15 K, but the value of H (0) is magneton) based on the measured susceptibilities alone yield
• well
TiSe
unique
CDW to state
with
anexpected
instability
to a
2 Thedisplays
[26] and
(iii)
a high
sensitivitydataof(close
theto crystal
growing
conditions
R=1
T ) can also structure
be values between
defined.)
high-temperature
0.3 and 0.4, much
smaller than the [11].
used to estimate H , based on the equation H (0) = 0.693H (0) value for the free-electron approximation. The small susceptibilities
superconducting
yet
less
attention
paid
TiSe
(ref.
28),
where H (0in
) = −(
dH /normal
dT ) T . state,
The phase
dotted line
in observed
inof
this
system
must be corrected
for core to
diamagnetism;
TiSe
the
consists
Ti layers
sandwiched
by octahedrally
2 compared
2 lattice
Fig. 4c represents the extrapolation to T = 0 of the linear fit at however, correcting for the core contributions of Ti and Se
(ref. 29) Se-Ti-Se.
results in R valuesThe
that seem
to be too high configuration
(between 2.5
high
temperatures,
yielding
an estimate
for H (a
0) =
1.27 T. On block
coordinated
Se
atoms,
thus
creating
trilayer
described
is the
to
other
TMDs.
the basis of these measurements, we conclude that the upper and 5). This suggests that additional contributions to the observed
(0) = 1.33 ± 0.06 metal
T. Despite dichalcogenides.
susceptibility need to be considered
to fully understand this system.
field H (0) of Cu TiSe
so-calledcritical
1T-polytype
foris Htransition
Quasi-two-dimensional
crystal
is left to a future study.devices based on TMDs
the expected quadratic
temperature dependence at low This analysis
• exhibiting
In
general,
superconducting
proximity
values the
determined
from field-dependent
temperatures,
the Hfrom
Finally, the overall
behaviour of this
system is summarized
in the trilayer
structure magnetization,
originates
differences
in the nature
of chemical
bonding
within
M(H ), measurements are probably overestimates the electronic phase diagram presented in Fig. 6. Using Cu doping
have not
been
explored
in detail.
of the actual values.
This could be
a result of using
polycrystalline
as a finely
parameter,
CDW transition
in TiSe
blocks Se-Ti-Se
(covalent
bonding)
and
between
the controlled
blockstuning
(van
dertheWaals
bonding).
The latter
pellets rather than single crystals, particularly if the critical field is is driven down in temperature, and a new superconducting state
emerges. The superconducting state appears for x > 0.04, going
anisotropic: the polycrystalline samples yield an average value H
facilitatesthat
the
incorporation
of
foreign
atoms
into
the van der Waals gap, being a wonderful tool
is intermediate between the values corresponding to H ∥ ab through a maximum T of 4.15 K at x = 0.08, followed by a decrease
and H ∥ c . Using the above critical field values to estimate the of T before the chemical phase boundary is reached at x = 0.11.
for tuningGinzburg–Landau
the electronic
of(where
compounds
inisaa small
controlled
way [6].
boundary composition
region (0.04 < x < 0.06)
κ = l/ξ ≈ H /H
l is the There
parameterproperties
0
• Josephson effect in graphene SNS junctions
I (nA)
M/H (300 K) (e.m.u. per mol Ti) × 105
50
S (300 K)
S (300 K) ( µV K–1)
γ (mJ per mol Ti K2)
ρ (300 K) /ρ (6 K)
a)
40
ρ (300 K)/ρ (6 K)
5
0
T (K)
SC
CDW
x = 0.01
x=0
We have fabricated field effect devices based on the layered dichalcogenide titanium diselenide
(TiSe2). This material is a member of the transition metal dichalcogenides exhibiting a unique
chiral charge density wave (CDW) state. This state has an onset at a temperature of T≈200K.
Small amounts of copper or palladium dopants induce a superconducting state with
superconducting critical temperature of Tc = 4.15K and 2K, respectively, which competes with the
CDW state. By using field effect doping on single and few-unit-cell-thick layers we explore the
interplay between superconducting and CDW states in proximity coupled superconducting
devices. We have fabricated multi-terminal devices by mechanically exfoliating TiSe2 flakes from
single crystals of TiSe2 onto SiO2 substrates, and defining Nb and Al contacts using optical and
electron beam lithography techniques. We have confirmed that the CDW state persists in these
devices when scaled to few-unit-cell thick layers. Future efforts include transport measurements
of sub-micron devices down to milliKelvin temperatures.
V
Susceptibility (e.m.u. per Oe mol Ti)
CuxTiSe2
H = 0.5 T
V (µV)
Interplay of Charge Density Wave and
Superconducting States
3 × 10 – 5
pure TiSe2 at low temperatures show the presence of reflections
corresponding to the basic trigonal structure and also the 2a,2c
superstructure reflections associated with the CDW state3,19 .
Figure 1b shows an electron diffraction pattern of Cu0.03 TiSe2 taken
at approximately 120 K with the crystal tilted away from the [001]
zone such that several higher-order Laue zones, with reflections
hkl (l = −1, 0, 1 and 2) are visible. The superreflections are
only observed for l = 2n + 1, and are not visible in the zeroorder Laue zone in the [001] orientation (where l = 0). The 2a,2c
superstructure reflections, as indicated in the figure, are clearly
seen, as they are in TiSe2 . Therefore, the charge density wave is still
present at 120 K at this composition. Significantly, the characteristic
CDW wavevector is unchanged by doping. Apart from the 2a,2c
superstructure reflections, which are very sharp and only occur in
the higher-order zones where l = 2n + 1, more-streaked reflections
can be seen. Raising the temperature by approximately 20 K
results in the disappearance of the 2a,2c superstructure reflections.
The more-streaked reflections are still present above the CDW
transition temperature and remain visible in diffraction patterns
taken at room temperature. Furthermore, they are present in the
diffraction pattern for Cu0.08 TiSe2 , the optimal superconducting
composition, at 120 K (Fig. 1c) and also in pure TiSe2 at room
temperature. The positions of these diffuse peaks are only in part
the same as those of the 2a,2c superstructure, as shown in the
overlays. The diffuse superreflections are not confined to a small
band at l = 2n + 1 but are present everywhere, indicating that
they are also streaked along c ∗ (perpendicular to the TiSe2 planes):
the streaking seems to be continuous in that direction. It would
be of interest to characterize the diffuse scattering as a function
of temperature and composition and determine its origin. It is
probably associated with the soft phonon believed to accompany
the CDW transition25 .
Figure 2a shows the temperature dependence of the magnetic
susceptibilities for Cux TiSe2 over the range of Cu solubility. The
normal-state susceptibility (for example, at 300 K) increases with
Cu content. This suggests that the Cu doping introduces carriers
into the conduction band in TiSe2 , increasing the electronic density
of states and therefore the Pauli paramagnetism. This is further
confirmed by specific heat measurements, described below. A drop
in the susceptibility of pure TiSe2 is seen as the temperature is
lowered below the CDW transition at 200 K, consistent with the
decrease in electronic density of states that occurs on opening a
gap at the Fermi level (the susceptibility becomes negative because
the core diamagnetism is larger than the Pauli contribution). On
doping with increasing amounts of Cu, the CDW state in Cux TiSe2
exists until x = 0.06, as seen in the drops in the susceptibilities. The
susceptibility drop decreases with increasing Cu content, implying
that fewer states are gapped at the CDW transition. The CDW
transition temperatures can be determined from the onsets of the
susceptibility drops, and decrease continuously with increasing Cu
content. For x = 0.06, the CDW transition, marked by a very small
change in susceptibility, is reduced below 60 K and is no longer
visible for higher x . The fact that local moment magnetism is not
generated by Cu doping indicates that the intercalated Cu has a
formal oxidation state of +1, a 3d 10 electron configuration.
A systematic change in the transport properties of Cux TiSe2
occurs on increasing x . The resistivity of our pure TiSe2 (Fig. 2b) is
very similar to that previously reported3 : a broad maximum occurs
around 150 K, with the ratio ρ(150 K)/ρ(300 K) = 4, comparable
to that of the stoichiometric crystals3 . However, unlike the single
crystals where the ratio ρ(300 K)/ρ(6 K) is 3–4, in our sample
this ratio is smaller than unity, probably due to the fact that
it is a polycrystalline pellet. As shown in Fig. 2b, the resistivity
maximum in Cux TiSe2 associated with the CDW state broadens
and moves towards lower temperatures with increasing Cu doping,
dV/dI (kΩ)
4 × 10 – 5
R (kΩ)
a
I (nA)
Background & Motivation
Bipolar supercurrent in graphene
V (µV)
James
2
Curtis ,
I (nA)
1Department
Sam
1
Ciocys ,
G (4e2/h)
Joseph G.
1
Lambert ,
■
■
Conclusions & Future Work
■
• We have fabricated sub-micron TiSe2 devices with good Ohmic contact
electrodes,to
using superconductors
a Ti/Al (5 nm/350 nm thick and 400 nmAl and Nb, using electron beam and optical
wide) bilayer in place of the Ti/Au bilayer. The thin Ti layers
were addedlithography.
to create a good ohmic contact at the BSTS/Al and
BSTS/Au interfaces. The electrical leads A and B (C and D )
Fig. 2. Transport properties of the thin-flake MoS2 EDLT. (A) Temperature
dependence of the channel sheet resistance Rs at different VLG gate biases
ranging from 0 to 6 V (indicated on the right). (B) Temperature dependence
of the channel sheet resistance Rs at VLG = 1 V and different VBG’s showing a
metal-insulator
transition at n2D = 6.7 × 1012 cm−2. For each VBG, we marked
1
1
1
1
the corresponding n2D measured by the Hall effect at 20 K. (C) Phase diawere the source and drain for current biasing the Al−BSTS−Al
gram showing the evolution of different electronic phases as a function of
carrier density n2D. The phase diagram shows an insulating (n2D < 6.7 ×
(Au-BSTS-Au) junction by IA1B1 (IC1D1), while the potential
1012 cm−2), a metallic (6.7 × 1012 < n2D < 6.8 × 1013 cm−2), and a domesuperconducting phase (n2D > 6.8 × 1013 cm−2), where the color corusing
difference VA2B2 (VC2D2) across the junction was measured like
5030
responds to the logarithm of the sheet resistance Rs (W). (D) Normalized
5033
| Nano for
Lett.dif2014, 14, 5029−5034
superconducting
transition Rs/Rs (15K)dx.doi.org/10.1021/nl501481b
as a function of temperature
ferent gate voltages. The Tc is marked as a circle at 90% of the total transition.
Both VLG and VBG are varied; for a given VLG, we show the evolution of the
transition with increasing VBG. All data corresponding to the same VLG are
shown with the same color. The dashed arrows indicate the order of increasing
VBG from –4 to 2 V in ∆VBG = 2 V. (Upper panel) For VLG between 4 and 5.5 V,
Tc increased with increasing VBG; (bottom panel) for VLG SD
= 6 V, Tc decreased
with increasing VBG because the corresponding n2D had reached the peak of
the superconducting dome.
Figure 2. Four-probe transport properties recorded at VG
current I . The observed temperature dependence of t
superconducting transition at lower T (b). The inset of
dependence of the square resistance for different values of
www.sciencemag.org which makes it possible to determine T more precisely. E
C
function of B for T = 0.25 K (e). (f) Two-dimensional
• Charge density wave transition observed in both types of devices.
dx.doi.org/10.1021/nl501481b | Nano Lett. 2014, 14, 5029−5034
1194
30 NOVEMBER 2012
VOL 338
SCIENCE
• Millikelvin electronic transport experiments are ongoing.
Upon further cooling we see that below approximat
the device exhibits a sharp, large decrease of resista
2b zooms-in on the behavior of R□(T) below T = 10
it apparent that at T ≃ 4 K, the device starts un
transition to a zero resistance state, which is attained
0.4 K. This is the first observation of a superconduc
WS2. It demonstrates that the possibility of elec
inducing superconductivity in semiconducting TM
confined to Mo-based compounds27−29 (the only on
the phenomenon had been reported 2
so far).
To further substantiate that the zero-resistance
manifestation of superconductivity, we investigated
the presence of an applied perpendicular magnetic
magnetoresistance measured at T = 0.25 K is shown
of Figure 2b. We find that a truly zero-resistance s
only up to B = 10 G, and that at higher value
resistance increases rapidly, reaching the value meas
• This work is sponsored by the US Army Research Office under contract
normal state at B = 0.14 T (i.e., the critical field at T
W911NF-14-1-0567
BC = 0.14 T). For large field values the magnetor
essentially negligible. Figure 2c further shows the e
• Fabrication facilities provided by:
R□ as a function of T, in the presence of differen
applied perpendicular field, which confirms h
• Dr. Xiaoxing Xi, Physics Dept., Temple University
conductivity is suppressed on a magnetic field sc
• Dr. Chris Lobb, CNAM and Physics Dept., University of Maryland
0.1 T. The same data show that above 4 K virtually
of resistance with magnetic field is observed, confi
• Future experiments include:
• Use superconducting and normal contacts to investigate nonlocal
superconducting transport properties.
• Measure device characteristics down to single and few layer TiSe .
Acknowledgements