ARTICLE

Received 1 Sep 2015 | Accepted 11 Jan 2016 | Published 12 Feb 2016 DOI: 10.1038/ncomms10671 OPEN

Structural semiconductor-to-semimetal phase

transition in two-dimensional materials induced by
electrostatic gating
Yao Li1, Karel-Alexander N. Duerloo2, Kerry Wauson3 & Evan J. Reed2

Dynamic control of conductivity and optical properties via atomic structure changes is of
technological importance in information storage. Energy consumption considerations provide
a driving force towards employing thin materials in devices. Monolayer transition metal
dichalcogenides are nearly atomically thin materials that can exist in multiple crystal struc-
tures, each with distinct electrical properties. By developing new density functional-based
methods, we discover that electrostatic gating device configurations have the potential to
drive structural semiconductor-to-semimetal phase transitions in some monolayer transition
metal dichalcogenides. Here we show that the semiconductor-to-semimetal phase transition
in monolayer MoTe2 can be driven by a gate voltage of several volts with appropriate choice
of dielectric. We find that the transition gate voltage can be reduced arbitrarily by alloying, for
example, for MoxW1  xTe2 monolayers. Our findings identify a new physical mechanism, not
existing in bulk materials, to dynamically control structural phase transitions in two-dimen-
sional materials, enabling potential applications in phase-change electronic devices.

1 Department of Applied Physics, Stanford University, Stanford, California 94305, USA. 2 Department of Material Science and Engineering, Stanford University,

Stanford, California 94305, USA. 3 Klipsch School of Electrical and Computer Engineering, New Mexico State University, Las Cruces, New Mexico 88003, USA.
Correspondence and requests for materials should be addressed to E.J.R. (email: [email protected]).

NATURE COMMUNICATIONS | 7:10671 | DOI: 10.1038/ncomms10671 | www.nature.com/naturecommunications 1

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10671

S
tructural phase transitions yielding a change of electrical monolayer MoTe2 using a capacitor structure. While the
conductivity are a topic of long-standing interest and required field magnitudes are large and may be challenging to
importance1,2. Two of the most studied phase-change achieve, we find that the transition gate voltage may be reduced to
material classes for electronic and optical applications are metal 0.3–1 V and potentially lower by substituting a specific fraction of
oxide materials3,4 and GeSbTe alloys5, both having a large W atoms within MoTe2 monolayers to yield the alloy
electrical contrast. For example, the metal oxide material MoxW1  xTe2. To accomplish these calculations, we have
vanadium dioxide (VO2) is reported to exhibit a structural developed a DFT-based model of the electrostatically gated
metal–insulator transition near room temperature at ultrafast structure (Supplementary Figs 1 and 2; Supplementary Note 1).
timescales, which can be triggered by various stimuli including This approach is validated by comparing to direct DFT
heating6, optical7 excitations and strain8. GeSbTe alloys can simulations in Supplementary Fig. 3 and Supplementary Note 2.
undergo reversible switching between amorphous and crystalline
states with different electrical resistivity and optical properties.
This is usually achieved by Joule heating employed in phase- Results
change memory applications9,10. These materials are Crystal structures. TMDs are a class of layered materials with the
distinguished from the myriad materials that exhibit atomic formula MX2, where M is a transition metal atom and X is a
structural changes by the proximity of a phase boundary to chalcogen atom. Each monolayer is composed of a metal layer
ambient conditions. sandwiched between two chalcogenide layers, forming a X–M–X
Another group of materials that can undergo phase transitions structure16 that is three atoms thick. The weak interlayer
are layered transition metal dichalcogenides (TMDs), which have attraction of TMDs allows exfoliation of these stable three-
received recent attention as single- and few-layer materials, atom-thick layers. Given the crystal structures reported in the
although research on bulk TMDs dates back decades11,12. Early bulk, we expect that exfoliated monolayer TMDs have the
attention has been focused primarily on electronic transitions potential to exist in the crystal structures shown in Fig. 1. Figure 1
between incommensurate and commensurate charge density shows the X atoms with trigonal prismatic coordination,
wave13,14 phases and superconducting phases15. Some TMDs octahedral coordination or a distorted octahedral coordination
have been found to exist in multiple crystal structures16, and around the M atoms16,20,25,26. We will refer to these three
transitions between them have been demonstrated in group V structures of the monolayer as the 2H phase, 1T phase and 1T0
TMDs (TaSe2 and TaS2) utilizing an scanning tunnelling phase, respectively. Symmetry breaking in the 1T0 leads to a
microscope (STM) tip17,18. These reported transitions in TaSe2 rectangular primitive unit cell.
and TaS2 are between two metallic phases. Recently, group VI Among these 2D TMDs, the Mo- and W-based materials have
TMDs have attracted increasing attention because they can exist attracted the most attention because their 2H crystal structures
in a semiconducting phase19. Recent computational work are semiconductors with photon absorption gaps in the 1–2 eV
indicates that structural transitions between phases of large (ref. 27) range, showing potential for applications in ultrathin
electrical contrast in some exfoliated two-dimensional (2D) group flexible and nearly transparent 2D electronics. Radisavljevic
VI TMDs can be driven by mechanical strain20. Excess charges et al.28 fabricated single-layer MoS2 transistors of high mobility,
transferred from chemical surroundings are also reported to large current on/off ratios and low standby power dissipation.
induce structural phase transitions in 2D group VI TMDs21–24. Unlike group IV and group V TMDs (for example, TaSe2 and
One would like to know the threshold charge density required to TaS2), which have been observed in the metallic 1T crystal
induce these transitions and whether these transitions could be structure16, DFT calculations on the group VI TMDs (Mo and W
dynamically controlled by electrostatic gating, utilizing standard based) freestanding monolayers indicate that the 1T structure is
electronic devices. unstable in the absence of external stabilizing influences20.
Here we show the potential of phase control in some However, group VI TMDs do have a stable octahedrally
monolayer TMDs using electrostatic gating device configurations. coordinated structure of large electrical conductivity, which is a
In this work, we use density functional theory (DFT) to determine distorted version of the 1T phase and referred to as 1T0 structure
the phase boundaries of single-layer MoS2, MoTe2, TaSe2 and the (Fig. 1). On the basis of DFT calculation results, Kohn–Sham
alloy MoxW1  xTe2. We consider MoS2 because it has received
considerable attention as an exceptionally stable semiconductor,
and MoTe2 because DFT calculations indicate that its energy 2H 1T 1T′
difference between semiconducting and semimetallic phases is
exceptionally small among Mo- and W-TMDs20. We calculate the M
X
phase boundaries at conditions of constant charge and constant
voltage, the electrical analogues to mechanical conditions of
constant volume and constant pressure, respectively. We find that
a surface charge density of less than  0.04 e or greater than
0.09 e per formula unit is required to observe the semiconductor- a
to-semimetal phase transition in undoped monolayer MoTe2
b
under constant-stress conditions (e is the elementary electric
charge) and a much larger value of approximately  0.29 e or Figure 1 | Three crystal structures of monolayer TMDs. The top
0.35 e per formula unit is required in the undoped monolayer schematics show cross-sectional views and the bottom schematics show
MoS2 case. The charge densities discussed in this work refer to basal plane views. The grey atoms are transition metal atoms and the red
excess charge density and should not be misinterpreted as the atoms are chalcogen atoms; in all three phases, a layer of transition metal
electron or hole density in a charge-neutral material that one atoms (M) is sandwiched between two chalcogenide layers (X). The
might obtain from chemical doping. We also study the potential semiconducting 2H phase has trigonal prismatic structure, and the metallic
of phase control in monolayer MoTe2 and TaSe2 through 1T and semimetallic 1T0 phases have octahedral and distorted octahedral
electrostatic gating using a capacitor structure. We discover that structures, respectively. The grey shadow represents a rectangular
a gate voltage as small as a few volts for some choices of gate computational cell with dimensions a  b, and the red shadow represents
dielectric can be applied to drive the phase transition in the primitive cell.

2 NATURE COMMUNICATIONS | 7:10671 | DOI: 10.1038/ncomms10671 | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10671 ARTICLE

states of this 1T0 crystal structure have metallic or semimetallic example, side contacts. In Fig. 2b, the charge is assumed to be
characteristics, consistent with previous experiments16. This stored in the monolayer rather than the metal contact. Layer I is a
octahedral-like 1T0 crystal structure has been observed in WTe2 monolayer TMDs with a Fermi level mIf , and plate II has a Fermi
under ambient conditions16,29, in MoTe2 at high temperature29 level of mIIf . A dielectric medium of thickness d and capacitance C
and in lithium-intercalated MoS2 (ref. 25). There is recent is sandwiched between monolayer TMDs and plate II. This
experimental evidence that few layer films of the T0 phase of dielectric medium can be vacuum. Distance sI is the separation
MoTe2 exhibit a bandgap that varies from 60 meV to zero with between the centre of monolayer TMDs I and the right surface of
variations in number of layers30. the dielectric medium, while sII is the separation between the
The relative energies of Mo- and W-based TMDs monolayer surface atoms of plate II and the left surface of the dielectric
crystals shown in Fig. 1 have been calculated using semilocal DFT medium. (See Supplementary Figs 5–7, Supplementary Table 1
with spin–orbit coupling, shown in Supplementary Fig. 4. These and Supplementary Note 3 for more details about distance
results are consistent with experimental evidence that the bulk parameters.)
form of WTe2 is stable in the metallic 1T0 phase, while other Mo- When the charge Q on the monolayer is fixed, the total energy
and W-dichalcogenides are stable in the semiconducting 2H of the system E(Q) is the sum of three parts: energy stored in the
phase16. These calculations indicate that the switch from dielectric medium (Ec), energy of the plate II (EII) and energy of
semiconducting 2H phase to semimetallic 1T0 phase in the charged monolayer TMDs (EI), as shown in Fig. 2.
monolayer MoTe2 requires the least energy (31 meV per


EðQÞ ¼ EI Q; sI þ EII  Q; sII þ Ec
formula unit), suggesting the potential for a transition that is
exceptionally close to ambient conditions. Therefore, we choose

Q2
to focus on determining the phase boundary of monolayer ¼ EI Q; sI þ EII  Q; sII þ ; ð1Þ
2C
MoTe2. While the computed energy difference between 2H and
1T0 is considerably larger for MoS2 (548 meV per formula unit), where C is the capacitance of the dielectric medium. EI(Q ¼ 0, sI)
we also compute phase boundaries for this monolayer at constant is the ground-state energy of the electrically neutral monolayer
charge because it has received more attention in the laboratory to TMDs and EI(Q, sI)  EI(Q ¼ 0, sI) is the energy required to move
date. Among 2D group VI TMDs, monolayer MoS2 has attracted electrons Q from the Fermi level of the monolayer TMDs to the
the most experimental attention for its stability and relative ease dielectric surface. EII(  Q, sII) is defined analogously. We take the
of exfoliation and synthesis. Monolayer MoTe2 has also been monolayers to be undoped in this work.
exfoliated31,32 and its synthesis is a fast-developing field. The first term in equation (1), EI(Q, sI), is calculated using DFT
for each phase of the monolayer TMDs to yield a E(Q) for each
monolayer phase (see Supplementary Figs 1 and 2, and
Energy calculations for charged monolayers. We examine two Supplementary Note 1 for calculation details). The phase change
distinct thermodynamic constraints for a system containing a does not enter into the third term in equation (1) or change the
charged monolayer. In one scenario, the monolayer is constrained capacitance of the dielectric medium C.
to be at constant excess charge, as shown in Fig. 2a; in the other, We take plate II to be a bulk metal with a work function W so
the monolayer is constrained to be at constant voltage, as shown that the second term in equation (1) can be approximately
in Fig. 2b. These are the electrical analogues to mechanical con- written as:
ditions of constant volume and constant pressure, respectively.

The electrical contact depictions in Fig. 2b and subsequent figures EII  Q; sII ¼  QW: ð2Þ
are schematic and could be accomplished in other manners, for When the voltage is fixed rather than the charge, the grand
potential FG(Q, V) becomes the relevant thermodynamic energy
defined as:
a II I: MX2
b II I: MX2 FG ðQ; V Þ ¼ EðQÞ  QV; ð3Þ
where E(Q) is computed using equation (1). The QV term in this
V
expression represents external energy supplied to the system
when the charge Q flows through an externally applied voltage V.
The equilibrium charge Qeq can be calculated through minimiza-
I
E I(Q)
II
E II(–Q) Ec = Q 2/2C E (–Qeq) Ec = Qeq2/2C E (Qeq) tion of the grand potential at a given gate voltage V.
 fII  fII @FG ðQ; V Þ
¼0 ð4Þ
V @Q Q¼Qeq
II II
S S
 fI  fI Applying the computed Qeq(V) to equation (3), we can obtain
d
I
d
I
the equilibrium grand potential as a function of gate voltage
S S eq
FG ðV Þ.
eq


Figure 2 | Energy calculations for systems containing a charged FG ðV Þ ¼ FG Qeq ðV Þ; V ¼ E Qeq ðV Þ  Qeq ðV ÞV ð5Þ
monolayer. Layer I is a monolayer TMDs with a Fermi level mIf, and plate II
has a Fermi level of mIIf . Charge Q on the monolayer TMDs is fixed in a, Hereafter, we omit the superscript ‘eq’ for the equilibrium grand
eq
whereas the voltage V is fixed in b giving rise to an equilibrium charge Qeq. potential FG ðV Þ.
A dielectric medium of thickness d and capacitance C, which can be In addition to the electrical constraint, the nature of the
vacuum, is sandwiched between the monolayer and the plate. Distance sI is mechanical constraint on the monolayer is also expected to play a
the separation between the centre of the monolayer and the right surface of role in the phase boundary, discussed in Supplementary Note 4.
the dielectric medium, and sII is the separation between the surface atoms
of plate II and left surface of the dielectric medium. The total energy in the Phase boundary at constant charge. The distinction between the
fixed charge case is the sum of three parts: energy stored in the dielectric constant charge and voltage cases is most important when a phase
medium Ec, energy of the plate II EII, and energy of the charged monolayer transformation occurs. We discover that the transition between
monolayer TMDs EI. semiconducting 2H-TMDs and semimetallic 1T0 -TMDs can be

NATURE COMMUNICATIONS | 7:10671 | DOI: 10.1038/ncomms10671 | www.nature.com/naturecommunications 3

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10671

driven by excess electric charge (positive or negative) in the might be expected to hold when there is a strong frictional
monolayer. A constant charge condition exists when the charge interaction between the monolayer and substrate preventing the
on the monolayer remains constant during the phase transition as monolayer from relaxing freely.
if it is electrically isolated. An approximate condition of constant Figure 3b shows that semiconducting 2H-MoTe2 has lower free
charge could exist when adsorbed atoms or molecules donate energy and is the equilibrium state when the monolayer is
charge to the monolayer. electrically neutral or minimally charged. For the stress-free case
Figure 3 presents the energy difference between the 2H and 1T0 (blue line), when the charge density is between  0.04 e and
phases as a function of the charge density in the monolayer. 0.09 e per formula unit, 2H-MoTe2 is the thermodynamically
Figure 3a is a schematic of the system. The monolayer TMD is a stable phase. These charge densities correspond to  3.7  1013
distance d away from the electron reservoir (metal electrode). and 8.2  1013 e cm  2, respectively. Outside this range, semi-
Because the dielectric medium is vacuum in this schematic, EI metallic 1T0 -MoTe2 will become the equilibrium phase and a
and Ec in equation (1) can be combined, which can be understood transition from the semiconducting 2H phase to the semimetallic
from Fig. 2a, and equation (1) can be rewritten as: 1T0 phase will occur.


In the constant-area case (red line) in Fig. 3b, a considerably
EðQÞ ¼ EI Q; sI þ EII  Q; sII þ Ec larger charge density is required to drive the phase transition.



¼ EI Q; sI þ d þ EII  Q; sII ð6Þ This suggests that the precise transition point may be sensitive to
the presence of a substrate and that the detailed nature of the
When computing the energy difference between a system where mechanical constraint of the monolayer may play a substantive
the monolayer is in the 1T0 phase and another system where role in the magnitude of the phase boundaries. The higher
the monolayer is in the 2H phase ET0 ðQÞ  EH ðQÞ, the terms transition charge in this case can be understood by considering
EII(  Q, sII) in equation (6) cancel, leading to, that the energy of the strained T0 phase is higher than that of the


 zero stress T0 phase, pushing the phase boundary to larger charge
ET0 ðQÞ  EH ðQÞ ¼ EI Q; sI þ d þ EII  Q; sII T0


 states.
 EI Q; sI þ d þ EII  Q; sII H ð7Þ Figure 3c shows that the transition in monolayer MoS2 requires


much larger charge density than the MoTe2 case. If the negative
¼ETI 0 Q; sI þ d  EHI
Q; sI þ d
charge density is 40.29 e per MoS2 formula unit, semimetallic
Equation (7) shows that the energy difference depends on sI þ d 1T0 -MoS2 will have lower free energy and be more stable. For
rather than sI and d independently. Variation of the results of negative charge densities o0.29 e per formula unit, semiconduct-
Fig. 3b,c with the separation sI þ d (chosen to be 15 Å in Fig. 3) is ing 2H-MoS2 will be energetically favourable. This is consistent
weak or none as shown in Supplementary Fig. 7. with previous experimental reports that adsorbed species
The blue lines are constant-stress (stress-free) cases, in which donating negative charge to monolayer MoS2 can trigger a
both phases exhibit minimum energy lattice constants and atomic trigonal prismatic to octahedral structure transformation24,33.
positions. This condition is expected to hold when the monolayer MoS2 single layer is reported to adopt a distorted octahedral
is freely suspended or is not constrained by friction on a structure when bulk MoS2 is first intercalated with lithium to
substrate. The red lines represent constant-area cases, where the form LixMoS2 with xE1.0 and then exfoliated by immersion in
monolayer is clamped to its 2H lattice constants. This condition distilled water25. This is qualitatively consistent with our

a
Constant stress
sI
d Constant area

b c 0.8
MoTe2 MoS2
E1T′ – E2H (eV per f.u.)
E1T′ – E2H (eV per f.u.)

0.06 0.6

0.04 0.4

0.02 0.2

0 0

–0.02 –0.2

–0.04 –0.4
–0.1 –0.05 0 0.05 0.1 0.15 –0.4 –0.2 0 0.2 0.4
 (e per f.u.)  (e per f.u.)

Figure 3 | Phase boundary at constant charge in monolayer MoTe2 and MoS2. (a) Schematic representation of a monolayer TMDs separated by vacuum
from an electron reservoir, for example, the surface of a metal. The internal energy difference between 2H and 1T0 phases E1T0  E2H changes with respect to
the charge density s as shown in b and c. The units are per formula unit (f.u.). The blue line represents constant-stress (stress-free) case, in which both 2H
and 1T0 are structure relaxed. The red line represents the constant-area case, in which the monolayer is clamped to its 2H lattice constants. (b)
Semiconducting 2H-MoTe2 is a stable phase and semimetallic 1T0-MoTe2 is metastable when the monolayer is charge neutral. However, 1T0-MoTe2 is more
thermodynamically favourable when the monolayer is charged beyond the positive or negative threshold values. The charge thresholds exhibit a significant
dependence on the relaxation of lattice constants, indicating that the precise transition point may be sensitive to the presence of a substrate.
(c) MoS2 is stable in the 2H structure when charge neutral. The magnitude of charge required for the transition to 1T0 is larger than for MoTe2. In both
cases, transition at constant stress is more easily induced than the transition at constant area.

4 NATURE COMMUNICATIONS | 7:10671 | DOI: 10.1038/ncomms10671 | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10671 ARTICLE

prediction that a negative charge density 40.29 e per MoS2 may shows that a transition gate voltage of  1.6 V or 4.4 V can be
trigger the phase transition from 2H phase to 1T0 phase MoS2. See applied to drive the phase transition in monolayer MoTe2 using
Supplementary Fig. 8 and Supplementary Note 5 for an intuitive the capacitor in Fig. 4a. The experimental breakdown voltage for
discussion of the mechanism for the charge-induced structural a 4.5-nm-thick HfO2 is reported to be as large as 3.825 V (ref. 35),
phase transition. which is larger than twice the magnitude of the negative
transition voltage. This breakdown field in HfO2 is larger than
Phase boundary at constant voltage. Another relevant type of some other reports and may depend on the details of growth36,37.
electrical constraint is fixed voltage or electron chemical potential. Therefore, employing an appropriate dielectric is likely to be
This constraint is most applicable when the monolayer is in an critical here in observing the phase change. Ionic liquids may be
electrostatic gating structure similar to field-effect transistors employed to help address the challenge of achieving large
made using monolayers. Such a device structure enables a voltages. Ionic gating has been applied to a variety of TMDs to
dynamical approach to achieve semiconductor/semimetal phase investigate superconductivity38 by measuring I–V curves.
control in monolayer TMDs, suggesting intriguing applications However, when a large voltage is applied, it may be challenging
for ultrathin flexible 2D electronic devices including phase- to probe structural phase transitions from I–V curves alone due
change memory. to a large density of charge in the TMDs. Structural
Many distinct electrostatic gating device structures can be characterization approaches, such as Raman spectroscopy, may
utilized to realize this dynamic control through a change in provide a more direct probe of electrically induced structure
carrier density or electron chemical potential of the monolayer. phase transitions in monolayer TMDs.
Here we consider a capacitor structure shown in Fig. 4a. While the curves in Fig. 4 assume that the monolayer is at a state
A monolayer of MoTe2 is deposited on top of a dielectric layer of zero stress across the transition, Fig. 5 presents calculations for
of thickness d, which we take to be HfO2 with a large dielectric MoTe2 at constant stress (Fig. 5a) and constant area (Fig. 5b)
constant of 25 (ref. 34). Monolayer and dielectric are sandwiched utilizing the capacitor structure shown in Fig. 4a. These phase
between two metal plates between which a voltage V is applied. diagrams predict the thermodynamically favoured phase as a
High-dielectric constant material HfO2 is chosen to increase the function of voltage V and thickness d of the HfO2 dielectric
capacitance and hence increase the charge density in the medium. In each phase diagram, there exist two phase boundaries,
monolayer. The metal plate is chosen to be aluminum with a the positions of which vary with the work function W of the
work function of 4.08 eV. The curves in Fig. 4 assume the capacitor plate. The 2H semiconducting phase of MoTe2 is stable
monolayer to be at a state of constant stress, with both 2H and between the two phase boundaries, and metallic 1T0 -MoTe2 is
1T0 phases structurally relaxed. We compute the total energy and stabilized by application of sufficiently positive or negative gate
equilibrium grand potential of this system using equations (1–5). voltages. The transition voltages increase with the thickness of the
Plotted in Fig. 4b is the total energy (equation 1) of the dielectric layer. For a capacitor containing a HfO2 dielectric layer of
capacitor shown in Fig. 4a as a function of charge density in thickness o5 nm, a negative gate voltage of approximately  2 V
monolayer MoTe2. Two black dashed lines depict common may be applied to drive the semiconductor-to-semimetal phase
tangents between 2H and 1T0 energy surfaces, the slopes of which transition at constant stress (Fig. 5a) but the required voltage
are defined by the set of equations, increases to approximately  4 V at constant area in Fig. 5b. In
    analogue with the changes in charge density phase boundaries
@EH @ET0 EH ðQH Þ  ET0 ðQT0 Þ shown in Fig. 3b, the voltage magnitudes for the transition are
¼ ¼ ð8Þ larger in constant-area conditions (Fig. 5b) than at constant stress
@Q QH @Q QT0 QH  QT0
(Fig. 5a). If the substrate constrains the area of the monolayer across
   
the transition through friction, the voltages in Fig. 5b are expected
where Vt ¼ @E H
¼ @E T0
is the transition gate voltage.
@Q QH @Q QT0 to be applicable. The figure also shows a reported experimental
Plotted in Fig. 4c is the equilibrium grand potential breakdown voltage of a 4.5-nm-thick HfO2 film35.
(equation 5) as a function of the gate voltage. Two transition Field-effect transistors based on few-layered MoTe2 have been
voltages are labelled also using black dashed lines. Figure 4b,c reported in ref. 39 using a 270-nm-thick SiO2 gate dielectric layer

a b c
V 0.4 0.1
1T′-MoTe2
ΦG (eV per f.u.)

2H-MoTe2 0
E (eV per f.u.)

MoTe2 0.3 2H-MoTe2

s = 3.78 Å –0.1
0.2 1
Vt = –1.8 V
HfO2 d = 4.5 nm –0.2
1T′-MoTe2 Vbd
0.1 –0.3
Al: W = 4.08 eV 1 2 (exp.) 2
Vt = 4.4 V
Vt = –1.8 V Vt = 4.4 V
0 –0.4
–0.1 –0.05 0 0.05 0.1 –4 –2 –0 2 4 6
 (e per f.u.) V (V)

Figure 4 | Phase boundary at constant voltage and stress. (a) Monolayer MoTe2 deposited on top of a HfO2 layer of thickness d ¼ 4.5 nm, which is on top
of an aluminum plate of work function W ¼ 4.08 eV. Voltage V is applied between the monolayer and the aluminum plate. (b) Plotted is the total energy
E of the capacitor shown in a as a function of the charge density s on monolayer MoTe2. (c) Plotted is the grand potential FG as a function of the gate
voltage V. The blue line represents a capacitor containing 2H-MoTe2, whereas the green line represents a capacitor containing 1T0-MoTe2. The two black
dashed lines in b depict common tangents between the 2H and 1T0 energy surfaces, and in c represent intersections of the 2H and 1T0 grand potentials
indicating two transition voltages Vt1 and Vt2. Between the two transition voltages, semiconducting 2H-MoTe2 has a lower grand potential and is
thermodynamically stable. Outside this range, 1T0 will be more stable. The red dashed line in c represents a breakdown voltage35 obtained experimentally
for a HfO2 film of thickness 4.5 nm. The separation between MoTe2 centre and the surface of HfO2 is assumed to be s ¼ 3.78 Å, and both 2H and 1T0 are
structurally relaxed (constant stress) in b and c.

NATURE COMMUNICATIONS | 7:10671 | DOI: 10.1038/ncomms10671 | www.nature.com/naturecommunications 5

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10671

a Constant stress b Constant area Constant area

16 5
6 eV eV V
1T′-MoTe2 4 1T′-MoTe2 eV 4e

Gate voltage V (V)

4
Gate voltage V (V)

W= eV W= 5.5 4 eV
4 W=
5.5 12 W= 1T′ W= 5.5
8 3 W=
2

Gate voltage V (V)

2H-MoTe2 4 2H-MoTe2 2
0
0
–2 W = 4 eV Vbd (exp.) 1
Vbd (exp.) –4 W = 4 eV 2H-Mo0.67W0.33Te2
–4 1T′-MoTe2 W = 5.5 1T′-MoTe2W = 5.5 eV 0
eV
–8
2 4 6 8 10 2 4 6 8 10 –1
d (nm) d (nm) W = 4 eV
–2
Figure 5 | Phase control of MoTe2 through gating at constant stress and –3 1T′ W = 5.5 eV
Vbd (exp.)
constant area. (a,b) Plotted are phase stabilities of monolayer MoTe2 with
–4
respect to gate voltage V and dielectric thickness d using the capacitor 2 4 6 8 10
structure shown in Fig. 4a. In each phase diagram, there exist two phase d (nm)
boundaries that vary with the work function W of the capacitor plate.
Figure 6 | Reducing transition gate voltages with the alloy MoxW1  xTe2.
Between the two phase boundaries, semiconducting 2H is more stable, and
Plotted is the phase stability of a representative alloyed monolayer
outside 1T0 is the stable structure. The required transition gate voltage is
Mo0.67W0.33Te2 with respect to the gate voltage V and the dielectric
smaller in the constant-stress case (a) than in the constant-area case (b).
thickness d using the capacitor structure as shown in Fig. 4a. The transition
For a constant-stress scenario, a negative gate voltage as small as  1 to
is assumed to occur at constant monolayer area. The magnitudes of the
 2 V can trigger the semiconducting-to-semimetallic phase transition in
transition gate voltages in this alloy are smaller than those of pure MoTe2
monolayer MoTe2. The red triangle represents the breakdown voltage of a
monolayer indicating the potential for phase boundary engineering.
4.5-nm-thick HfO2 film obtained experimentally35.

computed phase diagram can occur with the choice of

(3.9 dielectric constant) with gate voltages as large as  50 V. For configuration. Detailed cluster expansion calculations for these
monolayer MoTe2 (rather than few layers), our model predicts monolayer alloys are presented in ref. 44.
that a gate voltage 4200 V is required to drive the phase For the alloy configuration we employed in this work (assumed
transition for this device configuration. Both the increase of at constant area), the 2H phase is a semiconductor with a
dielectric thickness (from 5 to 270 nm) and the decrease of semilocal quasiparticle Kohn–Sham bandgap of B0.9 eV, and its
dielectric constant (from 25 for HfO2 to 3.9 for SiO2 (ref. 35)) will free energy is 15 meV lower than the 1T0 phase at constant
result in larger transition gate voltages than shown in Fig. 5. To monolayer area, which is metallic or semimetallic. Figure 6 shows
observe the 2H-1T0 phase transition in a device, choosing a that the 2H-1T0 phase transition in this alloy can be driven by
dielectric medium of large dielectric constant and dielectric negative gating of a smaller gate voltage than pure MoTe2
performance will be critical. monolayer. For example, assuming HfO2 medium of 4.5-nm
thickness and capacitor plate of 4.0 eV work function, the
Reducing transition gate voltages with the alloy MoxW1  xTe2. magnitude of negative transition gate voltage can be reduced
Monolayer alloys present the possibility for reducing the required from 3.6 V (MoTe2, constant-area case) to 0.4 V in the constant-
gate voltage by varying the chemical composition. Recently, area case of Mo0.67W0.33Te2 monolayer.
monolayer alloys of Mo- and W-dichalcogenides have attracted One might expect that the transition gate voltage in
increasing attention for their tunable properties40–44. We monolayers can be tuned and reduced potentially arbitrarily by
hypothesize that the 2H-1T0 transition gate voltage can be controlling the chemical composition of this and other potentially
tuned to lower values by alloying MoTe2-WTe2 monolayers. This alloys. To enable a structural phase transition driven by a small
is because in monolayer MoTe2, the 2H phase is energetically gate voltage, elements should be selected for alloying so that the
favourable by 31 meV per formula unit relative to the 1T0 phase, energy difference between charge-neutral 2H and 1T0 phases can
whereas in monolayer WTe2, the energy of the 1T0 phase is be tuned through zero with alloy composition. Alternative
123 meV per formula lower than the 2H phase, as shown in mechanical constraints placed on the alloy monolayer (for
Supplementary Fig. 4. Therefore, one might expect the energy example, constant stress) can also be expected to shift the phase
difference between the two charge-neutral phases to be tunable boundary and transition gate voltage.
through zero with alloy composition. The smaller the energy
difference is, the closer the phase boundary is to ambient Phase transition in Ta-based TMDs. Electrically induced
condition and the smaller the external force required to drive the structural phase changes between 2H and 1T phases in the Ta-
phase transition in monolayer TMDs20. Therefore, controlling based TMDs, TaSe2 and TaS2, have been reported in experiments
alloy composition is likely to enable tuning of the transition gate using a STM tip17,18, although the mechanism for this reported
voltage. effect may differ from the charge-induced effect reported in the
Earlier experimental reports of the synthesis of the bulk alloy present work. As a supplement to the previous calculations on
MoxW1  xTe2 (ref. 45) and detailed calculations on monolayers44 group VI TMDs, we have computed the constant-stress phase
indicate that the phase changes from 2H to 1T0 with increase diagram of monolayer TaSe2 in the capacitor gating structure
in W fraction 1  x. This indicates that the free energy shown in Fig. 7a. Figure 7b shows that the phase diagram of
difference between the 2H and 1T0 phases can be made TaSe2 has a phase boundary at only positive gate voltage,
arbitrarily small by varying x, enabling an arbitrary reduction qualitatively different from MoTe2 and MoS2. We find that this
of the gate voltage. However, the precise value of x required to difference results from the metallic nature of both 2H- and 1T-
achieve a particular transition voltage is likely to depend on a TaSe2, further discussed in Supplementary Note 5.
number of factors including synthesis conditions and mechanical A qualitative difference between these calculations and the
constraints44,45. Here we study an approximate representative STM experiment17 is the observation of the transition at both
atomic configuration for this alloy for x ¼ 0.67, displayed in signs of STM bias, suggesting that other effects could be at play in
Supplementary Fig. 9, with the knowledge that some variation of the experiment. Further quantitative comparison with the

6 NATURE COMMUNICATIONS | 7:10671 | DOI: 10.1038/ncomms10671 | www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10671 ARTICLE

a b projector augmented-wave47 method and the plane-wave basis set with a kinetic
Constant stress
10 energy cutoff of 350 eV. Electron exchange and correlation effects were treated
V using the Generalized Gradient Approximation (GGA) functional of Perdew,
4e

Gate voltage V (V)

V 8 1T-TaSe2 W= Burke and Ernzerhof48. An 18  18  1 Monkhorst-Pack49 k-point mesh was
eV
5.5
TaSe2 6 W= utilized to sample the Brillouin zone. The convergence thresholds for electronic
s = 3.55 Å and ionic relaxations were chosen to be 0.5  10  8 eV per MX2 formula unit and
4 0.5  10  7 eV per MX2 formula unit, respectively. A Gaussian smearing of 50 meV
HfO2 d
2 was used. The computational cell length is 36 Å along the c axis. Spin–orbit
2H-TaSe2
W coupling is employed in all DFT calculations. The ionic relaxations were performed
0 using conjugate gradient algorithm.
2 4 6 8 10 All calculations in this work were performed at zero ionic temperature, omitting
d (nm) the vibrational component of the free energy. Reference 20 has shown that
inclusion of vibrational free energy and temperature would shift the phase
Figure 7 | Phase control of monolayer TaSe2 through gating at constant boundaries closer to ambient conditions and lower the energy required to switch
stress. (b) Plotted is the computed phase diagram of monolayer TaSe2 in the phases. Therefore, one would expect inclusion of these effects to decrease the
magnitude of the transition charge density and gate voltage calculated in this work.
the capacitor gating structure as shown in a. Unlike MoTe2, the phase Also, the change of bandgap width is expected to affect 2H-1T0 phase boundary, as
diagram of TaSe2 only has one phase boundary, which corresponds to a further discussed in Supplementary Fig. 10 and Supplementary Note 6. See
positive gate voltage. Below the phase boundary, 2H-TaSe2 is more stable; Supplementary Note 7 for a discussion on vacuum electronic states.
above the boundary, 1T has lower energy and is more stable. Intuition for
this qualitative difference from MoTe2 is provided in Supplementary Note 5.
References
1. Lencer, D. et al. A map for phase-change materials. Nat. Mater. 7, 972–977
(2008).
experiment is made challenging by the small separation between 2. Wong, H.-S. P. et al. Phase change memory. Proc. IEEE 98, 2201–2227 (2010).
the monolayer and the STM tip (Supplementary Fig. 3; 3. Wong, H.-S. P. et al. Metal-oxide RRAM. Proc. IEEE 100, 1951–1970 (2012).
Supplementary Note 2). 4. Zhou, Y. & Ramanathan, S. Correlated electron materials and field effect
transistors for logic: a review. Crit. Rev. Solid State Mater. Sci. 38, 286–317
(2013).
Discussion 5. Shu, M. J. et al. Ultrafast terahertz-induced response of GeSbTe phase-change
The electrical dynamical control of structural phase in monolayer materials. Appl. Phys. Lett. 104, 251907 (2014).
TMDs has exciting potential applications in ultrathin flexible 2D 6. Aetukuri, N. B. et al. Control of the metal-insulator transition in vanadium
electronic devices. If the kinetics of the transformation are dioxide by modifying orbital occupancy. Nat. Phys. 9, 661–666 (2013).
suitable, nonvolatile phase-change memory9 may be an 7. Cavalleri, A. et al. Femtosecond structural dynamics in VO2 during an ultrafast
solid-solid phase transition. Phys. Rev. Lett. 87, 237401 (2001).
application. One might expect 2D materials to have energy 8. Kikuzuki, T. & Lippmaa, M. Characterizing a strain-driven phase transition in
consumption advantages over bulk materials due to their small VO2. Appl. Phys. Lett. 96, 132107 (2010).
thickness. If the kinetics is sufficiently fast, another potential 9. Wuttig, M. & Yamada, N. Phase-change materials for rewriteable data storage.
application may be subthreshold swing reduction in field-effect Nat. Mater. 6, 824–832 (2007).
transistors to overcome the scaling limit of conventional 10. Lee, S.-H., Jung, Y. & Agarwal, R. Highly scalable non-volatile and ultra-low-
transistors4. In addition, the change in the transmittance of power phase-change nanowire memory. Nat. Nanotechnol. 2, 626–630 (2007).
11. Bulaevskiı̆, L. N. Structural transitions with formation of charge-density waves
light due to the phase transition of monolayer TMDs may be
in layer compounds. Sov. Phys. Uspekhi 19, 836–843 (1976).
employed in infrared optical switching devices, such as infrared 12. McMillan, W. L. Theory of discommensurations and the commensurate-
optical shutters and modulators for cameras, window coating and incommensurate charge-density-wave phase transition. Phys. Rev. B 14,
infrared antennas with tunable resonance. 1496–1502 (1976).
To summarize, we have identified a new mechanism, 13. Thomson, R. E., Burk, B., Zettl, A. & Clarke, J. Scanning tunneling microscopy
electrostatic gating, to induce a structural semiconductor-to- of the charge-density-wave structure in 1T-TaS2. Phys. Rev. B 49, 16899–16916
semimetal phase transition in monolayer TMDs. We have (1994).
14. Scruby, C. B., Williams, P. M. & Parry, G. S. The role of charge density waves in
computed phase boundaries for monolayer MoTe2, MoS2 and
structural transformations of 1T TaS2. Philos. Mag. 31, 255–274 (1975).
TaSe2. We discover that changing carrier density or electron 15. Castro Neto, A. H. Charge density wave, superconductivity, and anomalous
chemical potential in the monolayer can induce a semiconductor- metallic behavior in 2D transition metal dichalcogenides. Phys. Rev. Lett. 86,
to-semimetal phase transition in monolayer TMDs. We find that 4382–4385 (2001).
a surface charge density less than  0.04 e or greater than 0.09 e 16. Wilson, J. A. & Yoffe, A. D. The transition metal dichalcogenides discussion
per formula unit is required to observe the semiconductor-to- and interpretation of the observed optical, electrical and structural properties.
semimetal phase transition in monolayer MoTe2 under constant- Adv. Phys. 18, 193–335 (1969).
17. Zhang, J., Liu, J., Huang, J. L., Kim, P. & Lieber, C. M. Creation of nanocrystals
stress conditions, and a significantly larger value of approximately through a solid-solid phase transition induced by an STM tip. Science 274,
 0.29 e or 0.35 e per formula unit is required in the monolayer 757–760 (1996).
MoS2 case. A capacitor structure can be employed to dynamically 18. Kim, J.-J. et al. Observation of a phase transition from the T phase to the H
control the semiconductor-to-semimetal phase transition in phase induced by a STM tip in 1T-TaS2. Phys. Rev. B 56, R15573–R15576
monolayer MoTe2 with a gate voltage B2–4 V for MoTe2. These (1997).
transition charges and voltages are expected to vary considerably 19. Mak, K. F., Lee, C., Hone, J., Shan, J. & Heinz, T. F. Atomically thin MoS2:
with the nature of the mechanical constraint of the monolayer a new direct-gap semiconductor. Phys. Rev. Lett. 105, 136805 (2010).
20. Duerloo, K.-A. N., Li, Y. & Reed, E. J. Structural phase transitions in
and also potentially the presence of dopants or Fermi level two-dimensional Mo- and W-dichalcogenide monolayers. Nat. Commun. 5,
pinning. While the gate voltages required to observe the 4214 (2014).
transition in MoTe2 are likely near breakdown and could be 21. Eda, G. et al. Coherent atomic and electronic heterostructures of single-layer
challenging to realize in the lab, we find that the voltage MoS2. ACS Nano 6, 7311–7317 (2012).
magnitudes can be reduced arbitrarily by alloying Mo atoms with 22. Gao, G. et al. Charge mediated semiconducting-to-metallic phase transition in
substitutional W atoms to create the alloy MoxW1  xTe2. molybdenum disulfide monolayer and hydrogen evolution reaction in new 1T0
phase. J. Phys. Chem. C 119, 13124–13128 (2015).
23. Kan, M. et al. Structures and phase transition of a MoS2 monolayer. J. Phys.
Methods Chem. C 118, 1515–1522 (2014).
Electronic structure calculations. All periodic DFT calculations were performed 24. Kang, Y. et al. Plasmonic hot electron induced structural phase transition in a
within the Vienna Ab Initio Simulation Package46, version 5.3.3, using the MoS2 monolayer. Adv. Mater. 26, 6467–6471 (2014).

NATURE COMMUNICATIONS | 7:10671 | DOI: 10.1038/ncomms10671 | www.nature.com/naturecommunications 7

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10671

25. Jiménez Sandoval, S., Yang, D., Frindt, R. F. & Irwin, J. C. Raman study and 45. Champion, J. A. Some properties of (Mo, W) (Se, Te) 2. Br. J. Appl. Phys. 16,
lattice dynamics of single molecular layers of MoS2. Phys. Rev. B 44, 3955–3962 1035 (1965).
(1991). 46. Kresse, G. & Furthmüller, J. Efficient iterative schemes for ab initio total-
26. Kan, M., Nam, H. G., Lee, Y. H. & Sun, Q. Phase stability and Raman vibration energy calculations using a plane-wave basis set. Phys. Rev. B 54, 11169–11186
of the molybdenum ditelluride (MoTe2) monolayer. Phys. Chem. Chem. Phys. (1996).
17, 14866–14871 (2015). 47. Blöchl, P. E. Projector augmented-wave method. Phys. Rev. B 50, 17953–17979
27. Wang, Q. H., Kalantar-Zadeh, K., Kis, A., Coleman, J. N. & Strano, M. S. (1994).
Electronics and optoelectronics of two-dimensional transition metal 48. Perdew, J. P., Burke, K. & Ernzerhof, M. Generalized gradient approximation
dichalcogenides. Nat. Nanotechnol. 7, 699–712 (2012). made simple. Phys. Rev. Lett. 77, 3865–3868 (1996).
28. Radisavljevic, B., Radenovic, A., Brivio, J., Giacometti, V. & Kis, A. Single-layer 49. Monkhorst, H. J. & Pack, J. D. Special points for Brillouin-zone integrations.
MoS2 transistors. Nat. Nanotechnol. 6, 147–150 (2011). Phys. Rev. B 13, 5188–5192 (1976).
29. Brown, B. E. The crystal structures of WTe2 and high-temperature MoTe2.
Acta Crystallogr. 20, 268–274 (1966).
30. Keum, D. H. et al. Bandgap opening in few-layered monoclinic MoTe2. Nat.
Acknowledgements
Our work was supported in part by the US Army Research Laboratory, through the
Phys. 11, 482–486 (2015).
Army High Performance Computing Research Center, Cooperative Agreement
31. Ruppert, C., Aslan, O. B. & Heinz, T. F. Optical properties and band gap of
W911NF-07–0027. This work was also partially supported by NSF grants EECS-1436626
single- and few-layer MoTe2 crystals. Nano Lett. 14, 6231–6236 (2014).
and DMR-1455050, Army Research Office grant W911NF-15-1-0570, Office of Naval
32. Lezama, I. G. et al. Indirect-to-direct band gap crossover in few-layer MoTe2.
Research grant N00014-15-1-2697 and a seed grant from Stanford System X Alliance. We
Nano Lett. 15, 2336–2342 (2015).
thank Philip Kim for discussions.
33. Lin, Y.-C., Dumcenco, D. O., Huang, Y.-S. & Suenaga, K. Atomic mechanism of
the semiconducting-to-metallic phase transition in single-layered MoS2. Nat.
Nanotechnol. 9, 391–396 (2014). Author contributions
34. Ray, S. K., Mahapatra, R. & Maikap, S. High-k gate oxide for silicon hetero- Y.L., K.-A.N.D. and E.J.R. designed the simulations and the framework for thermodynamic
structure MOSFET devices. J. Mater. Sci. Mater. Electron 17, 689–710 (2006). analysis; Y.L. performed the simulations and subsequent numerical data analysis; Y.L.
35. Kang, L. et al. Electrical characteristics of highly reliable ultrathin hafnium and K.W. performed the preliminary simulations. Y.L. and E.J.R. interpreted the data and
oxide gate dielectric. IEEE Electron Device Lett. 21, 181–183 (2000). wrote the paper.
36. Lee, J.-H. et al. Characteristics of ultrathin HfO2 gate dielectrics on strained-
Si0.74Ge0.26 layers. Appl. Phys. Lett. 83, 779–781 (2003).
37. Wolborski, M., Rooth, M., Bakowski, M. & Hallén, A. Characterization of HfO2 Additional information
films deposited on 4H-SiC by atomic layer deposition. J. Appl. Phys. 101, Supplementary Information accompanies this paper at http://www.nature.com/
124105 (2007). naturecommunications
38. Shi, W. et al. Superconductivity series in transition metal dichalcogenides by
ionic gating. Sci. Rep. 5, 12534 (2015). Competing financial interests: The authors declare no competing financial interests.
39. Pradhan, N. R. et al. Field-effect transistors based on few-layered a-MoTe2.
Reprints and permission information is available online at http://npg.nature.com/
ACS Nano 8, 5911–5920 (2014).
reprintsandpermissions/
40. Chen, Y. et al. Tunable band gap photoluminescence from atomically thin
transition-metal dichalcogenide alloys. ACS Nano 7, 4610–4616 (2013). How to cite this article: Li, Y. et al. Structural semiconductor-to-semimetal phase
41. Zhang, M. et al. Two-dimensional molybdenum tungsten diselenide alloys: transition in two-dimensional materials induced by electrostatic gating. Nat. Commun.
photoluminescence, Raman scattering, and electrical transport. ACS Nano 8, 7:10671 doi: 10.1038/ncomms10671 (2016).
7130–7137 (2014).
42. Kutana, A., Penev, E. S. & Yakobson, B. I. Engineering electronic properties of
layered transition-metal dichalcogenide compounds through alloying. This work is licensed under a Creative Commons Attribution 4.0
Nanoscale 6, 5820–5825 (2014). International License. The images or other third party material in this
43. Tongay, S. et al. Two-dimensional semiconductor alloys: Monolayer article are included in the article’s Creative Commons license, unless indicated otherwise
Mo1-xWxSe2. Appl. Phys. Lett. 104, 012101 (2014). in the credit line; if the material is not included under the Creative Commons license,
44. Duerloo, K.-A. N. & Reed, E. J. Structural phase transitions by design in users will need to obtain permission from the license holder to reproduce the material.
monolayer alloys. ACS Nano. doi: 10.1021/acsnano.5b04359 (2015). To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

8 NATURE COMMUNICATIONS | 7:10671 | DOI: 10.1038/ncomms10671 | www.nature.com/naturecommunications