-
1
Supplementary Information for: Direct Measurement of Exciton
Valley Coherence in Monolayer WSe2
Kai Hao1,*, Galan Moody1,†,*, Fengcheng Wu1, Chandriker Kavir
Dass1, Lixiang Xu1, Chang-Hsiao Chen2, Ming-Yang Li3, Lain-Jong
Li3, Allan H. MacDonald1 and Xiaoqin Li1
1 Department of Physics and Center for Complex Quantum Systems,
University of Texas at Austin, Austin, TX 78712, USA.
2 Department of Automatic Control Engineering, Feng Chia
University, Taichung 40724, Taiwan.
3 Physical Science and Engineering Division, King Abdullah
University of Science & Technology (KAUST), Thuwal 23955, Saudi
Arabia.
†Present address: National Institute of Standards &
Technology, Boulder, CO 80305, USA.
*These authors contributed equally to this work.
Supplementary Note 1
Monolayer WSe2 Sample: CVD monolayer WSe2 triangular crystals
were synthesized based on
previous work1. In brief, a double-side polished sapphire (0001)
substrate (from Tera Xtal
Technology Corp.) was cleaned in a H2SO4/H2O2 (70:30) solution
heated at 100 °C for one hour.
After cleaning, the sapphire substrate was placed on a quartz
holder in the center of a 1” tubular
furnace. Precursor of 0.3 grams WO3 powder was placed in the
heating zone center of the
furnace (99.5% from Sigma-Aldrich). Se powder (99.5% from
Sigma-Aldrich) was placed in a
quartz tube at the upstream position of the furnace tube, which
was maintained at 270 °C during
the reaction. The sapphire substrate was located at the
downstream side, where the WO3 and Se
vapors were brought into contact with the sapphire substrate by
an Ar/H2 flowing gas (H2 = 20
sccm, Ar = 80 sccm, chamber pressure = 3.5 Torr). The reaction
heating zone was held at 925 °C
(ramping rate of 25 °C/min). The actual temperature of the
sapphire substrate was ranged from
750 °C to 850 °C. The heating zone was held at 925 °C for 15
minutes after which the furnace
was then naturally cooled to room temperature. The reaction
yielded monolayer WSe2 flakes
triangular in shape with a base width of ~20 μm. The thickness
was determined using atomic
force microscopy, shown in Supplementary Fig. 1, which confirms
the monolayer thickness of
the sample.
Direct measurement of exciton valley coherence in
monolayerWSe2
SUPPLEMENTARY INFORMATIONDOI: 10.1038/NPHYS3674
NATURE PHYSICS | www.nature.com/naturephysics 1
© 2016 Macmillan Publishers Limited. All rights reserved.
http://dx.doi.org/10.1038/nphys3674
-
2
Supplementary Figure 1: Atomic force microscope image of
monolayer WSe2. (a) Individual triangular monolayer flakes are
visible in the atomic force microscope image. A slice along the
dashed line is shown in (b), illustrating the monolayer thickness
of the flakes.
Linearly Polarized Photoluminescence: Linearly polarized
photoluminescence (PL) from
monolayer WSe2 is measured using a 660 nm laser excitation
source. Supplementary Fig. 2
shows polar plots of the exciton PL peak intensity (r) as a
function of the detected angle (�) for two excitation polarization
directions (indicated by the arrows). Linearly polarized PL
following
the polarization of the excitation laser has been used as the
key experimental evidence for valley
coherence in previous studies2. We repeat this linearly
polarized PL experiment to demonstrate
the consistent properties for the monolayer WSe2 investigated in
our nonlinear experiments. The
solid lines are fits using the equation � = � � (� � � � ���(� �
�)), where � is the polarization angle corresponding to maximum
exciton PL intensity. The results in Supplementary Fig. 2
demonstrate that � is determined entirely by the excitation
laser polarization angle and is independent of crystal orientation.
For both excitation angles, we measure a degree of linear
polarization��� = (�����)(�����) =�� ����, where IH (IV) is the
PL intensity parallel (perpendicular) to the excitation laser
polarization direction. PL spectra were collected in the range of 0
to 180
degrees in the detection angle, and spectra are duplicated in
the range of 180 to 360 degrees in
the polar plot.
3
Supplementary Figure 2: Linearly polarized photoluminescence
from monolayer WSe2.Normalized photoluminescence peak intensity (r)
as a function of detection angle ( ) for a given excitation laser
polarization (indicated by the arrows).
Two-Dimensional Coherent Spectroscopy: Optical two-dimensional
coherent spectroscopy
(2DCS) is an enhanced version of three-pulse four-wave mixing in
which the pulse delays are
varied with interferometric precision. A sequence of three
pulses with variable delays is
generated using a set of folded and nested Michelson
interferometers. The pulses are generated
by a mode-locked Ti:sapphire laser operating at an 80 MHz
repetition rate with a pulse duration
of ~100 femtoseconds. This setup enables femtosecond control of
the pulse delays with a
stabilization of λ/100, which allows for the four-wave mixing
signal to be Fourier transformed.
Analysis of the signal in the Fourier spectral domain, combined
with phase cycling of the pulse
delays to minimize scatter of the laser pulses into the
spectrometer, suppresses optical frequency
noise overlapping with the exciton resonance and enables us to
isolate the population
recombination and valley coherence signals.
In the experiments, three of the pulses with wavevectors k1, k2,
and k3 are focused to a
single 35 μm spot on the sample that is held at a temperature of
10 K in an optical cryostat. The
interaction of the first pulse with the sample with wavevector
k1 excites an electronic coherence
between the exciton ground and excited states. The second pulse
with wavevector k2 converts the
optical coherence into a transient population grating (in the
case of co-circular polarization) or a
valley coherence grating (in the case of alternating helicity of
the pulses). After a time t2, the
third pulse with wavevector k3 generates an optical coherence
that diffracts off the grating and is
2 NATURE PHYSICS | www.nature.com/naturephysics
SUPPLEMENTARY INFORMATION DOI: 10.1038/NPHYS3674
© 2016 Macmillan Publishers Limited. All rights reserved.
http://dx.doi.org/10.1038/nphys3674
-
2
Supplementary Figure 1: Atomic force microscope image of
monolayer WSe2. (a) Individual triangular monolayer flakes are
visible in the atomic force microscope image. A slice along the
dashed line is shown in (b), illustrating the monolayer thickness
of the flakes.
Linearly Polarized Photoluminescence: Linearly polarized
photoluminescence (PL) from
monolayer WSe2 is measured using a 660 nm laser excitation
source. Supplementary Fig. 2
shows polar plots of the exciton PL peak intensity (r) as a
function of the detected angle (�) for two excitation polarization
directions (indicated by the arrows). Linearly polarized PL
following
the polarization of the excitation laser has been used as the
key experimental evidence for valley
coherence in previous studies2. We repeat this linearly
polarized PL experiment to demonstrate
the consistent properties for the monolayer WSe2 investigated in
our nonlinear experiments. The
solid lines are fits using the equation � = � � (� � � � ���(� �
�)), where � is the polarization angle corresponding to maximum
exciton PL intensity. The results in Supplementary Fig. 2
demonstrate that � is determined entirely by the excitation
laser polarization angle and is independent of crystal orientation.
For both excitation angles, we measure a degree of linear
polarization��� = (�����)(�����) =�� ����, where IH (IV) is the
PL intensity parallel (perpendicular) to the excitation laser
polarization direction. PL spectra were collected in the range of 0
to 180
degrees in the detection angle, and spectra are duplicated in
the range of 180 to 360 degrees in
the polar plot.
3
Supplementary Figure 2: Linearly polarized photoluminescence
from monolayer WSe2.Normalized photoluminescence peak intensity (r)
as a function of detection angle ( ) for a given excitation laser
polarization (indicated by the arrows).
Two-Dimensional Coherent Spectroscopy: Optical two-dimensional
coherent spectroscopy
(2DCS) is an enhanced version of three-pulse four-wave mixing in
which the pulse delays are
varied with interferometric precision. A sequence of three
pulses with variable delays is
generated using a set of folded and nested Michelson
interferometers. The pulses are generated
by a mode-locked Ti:sapphire laser operating at an 80 MHz
repetition rate with a pulse duration
of ~100 femtoseconds. This setup enables femtosecond control of
the pulse delays with a
stabilization of λ/100, which allows for the four-wave mixing
signal to be Fourier transformed.
Analysis of the signal in the Fourier spectral domain, combined
with phase cycling of the pulse
delays to minimize scatter of the laser pulses into the
spectrometer, suppresses optical frequency
noise overlapping with the exciton resonance and enables us to
isolate the population
recombination and valley coherence signals.
In the experiments, three of the pulses with wavevectors k1, k2,
and k3 are focused to a
single 35 μm spot on the sample that is held at a temperature of
10 K in an optical cryostat. The
interaction of the first pulse with the sample with wavevector
k1 excites an electronic coherence
between the exciton ground and excited states. The second pulse
with wavevector k2 converts the
optical coherence into a transient population grating (in the
case of co-circular polarization) or a
valley coherence grating (in the case of alternating helicity of
the pulses). After a time t2, the
third pulse with wavevector k3 generates an optical coherence
that diffracts off the grating and is
NATURE PHYSICS | www.nature.com/naturephysics 3
SUPPLEMENTARY INFORMATIONDOI: 10.1038/NPHYS3674
© 2016 Macmillan Publishers Limited. All rights reserved.
http://dx.doi.org/10.1038/nphys3674
-
4
detected in transmission in the wavevector-matching direction ks
= -k1+k2+k3. The four-wave
mixing signal is interferometrically and spectrally resolved
using a fourth phase-stabilized
reference pulse while the delay t2 is varied. Subsequent Fourier
transformation of the signal
yields a rephasing zero-quantum spectrum with amplitude given by
ES(t1, ħω2, ħω3). In the
present experiments, we set t1 = 0 fs to obtain maximum
signal-to-noise; however using a value
of t1 = 100 fs, which is equal to the pulse duration, does not
lead to qualitative difference in the
data other than an overall smaller amplitude.
Polarization control of the excitation pulses and detected
signal allow for isolation of
valley coherence dynamics from population recombination. The
coherent light-matter interaction
can be modeled by the density matrix,��. Here, the diagonal
terms of � represents the exciton population in valley K and K′,
while the off-diagonal terms describe exciton coherence. Using
perturbation theory in the applied field,3 the three-pulse
excitation scheme of the K-valley
exciton for co-circular polarization can be described by the
following sequence:
�������)���� ���
�����)���� ��������)���� ���
�����)���� ���, (1)
where Ei(t) corresponds to the ith excitation pulse and t is the
pulse arrival time. It is clear from
supplementary equation (1) that the first two pulses excite an
exciton population in the K valley;
therefore by recording the four-wave mixing signal while
scanning the delay t2, we can extract
the exciton population recombination rate ΓK.
To resonantly excite and detect the exciton valley coherence, we
use a pulse excitation
scheme in which the first and third pulses are co-circularly
polarized (��) and the second pulse and detected signal are
co-circularly polarized but with opposite helicity (��). This
scheme can be described by the following density matrix
sequence:
�������)���� ���
�����)���� ���������)���� ����
�����)���� ���. (2)
As shown in Fig. 2d of the main text, the first pulse generates
a coherent superposition
between the ground and exciton state in the K valley. After a
time t1, the second pulse transfers
this optical coherence to a non-radiative coherence between
excitons in the K and K′ valleys, the
evolution of which we monitor during the time t2. The third
pulse converts the valley coherence
to an optical coherence in the K′ valley, which is detected as
the radiated four-wave mixing
5
signal field. Thus by measuring the four-wave mixing signal
while scanning the delay t2, we
directly probe the valley coherence dynamics and dephasing rate
γv.
Supplementary Note 2
Discussion of impurity or phonon scattering: Supplementary Figs.
3 and 4 schematically
illustrate the effect of impurity or phonon scattering on valley
dynamics, which include: (1)
change of the exciton center-of-mass-momentum, (2) suppression
of exciton phase coherence,
(3) scattering between bright and dark exciton states, and (4)
scattering between bright exciton
states in opposite valleys. Effect (1) is explicitly considered
in our model. Effect (2) is the pure
dephasing rate which has been measured by our experiment. Effect
(3) is part of the bright
exciton population decay process. Both effects (2) and (3) are
phenomenologically described in
the theoretical model. Process (4) is likely to be slow, since
it requires both electron and hole
spin flip as illustrated in Supplementary Fig. 4e.
An atomic-scale defect can localize excitons and lift their
valley degeneracy. Recent
experiments4, 5, including ours (Fig. 2a in the main text), have
found that impurity-bound exciton
states are energetically separated from the delocalized exciton
states, and the PL of the former
(latter) have low (high) degree of correlation between
excitation and emission circular
polarization. We conclude that the valley degree-of-freedom of
delocalized excitons is not
destroyed by atomic-scale defects. A full understanding of the
influence of atomic-scale defects
Supplementary Figure 3: Intra-valley scattering processes. (a)
The center-of-massmomentum and the phase of a bright exciton can be
changed by scattering. (b) A bright excitonis scattered to a dark
one due to the spin flip of the constituent electron. The gray dot
and emptycircle respectively represent electron and hole. The spin
splitting of conduction bands isexaggerated for illustration
purpose.
4 NATURE PHYSICS | www.nature.com/naturephysics
SUPPLEMENTARY INFORMATION DOI: 10.1038/NPHYS3674
© 2016 Macmillan Publishers Limited. All rights reserved.
http://dx.doi.org/10.1038/nphys3674
-
4
detected in transmission in the wavevector-matching direction ks
= -k1+k2+k3. The four-wave
mixing signal is interferometrically and spectrally resolved
using a fourth phase-stabilized
reference pulse while the delay t2 is varied. Subsequent Fourier
transformation of the signal
yields a rephasing zero-quantum spectrum with amplitude given by
ES(t1, ħω2, ħω3). In the
present experiments, we set t1 = 0 fs to obtain maximum
signal-to-noise; however using a value
of t1 = 100 fs, which is equal to the pulse duration, does not
lead to qualitative difference in the
data other than an overall smaller amplitude.
Polarization control of the excitation pulses and detected
signal allow for isolation of
valley coherence dynamics from population recombination. The
coherent light-matter interaction
can be modeled by the density matrix,��. Here, the diagonal
terms of � represents the exciton population in valley K and K′,
while the off-diagonal terms describe exciton coherence. Using
perturbation theory in the applied field,3 the three-pulse
excitation scheme of the K-valley
exciton for co-circular polarization can be described by the
following sequence:
�������)���� ���
�����)���� ��������)���� ���
�����)���� ���, (1)
where Ei(t) corresponds to the ith excitation pulse and t is the
pulse arrival time. It is clear from
supplementary equation (1) that the first two pulses excite an
exciton population in the K valley;
therefore by recording the four-wave mixing signal while
scanning the delay t2, we can extract
the exciton population recombination rate ΓK.
To resonantly excite and detect the exciton valley coherence, we
use a pulse excitation
scheme in which the first and third pulses are co-circularly
polarized (��) and the second pulse and detected signal are
co-circularly polarized but with opposite helicity (��). This
scheme can be described by the following density matrix
sequence:
�������)���� ���
�����)���� ���������)���� ����
�����)���� ���. (2)
As shown in Fig. 2d of the main text, the first pulse generates
a coherent superposition
between the ground and exciton state in the K valley. After a
time t1, the second pulse transfers
this optical coherence to a non-radiative coherence between
excitons in the K and K′ valleys, the
evolution of which we monitor during the time t2. The third
pulse converts the valley coherence
to an optical coherence in the K′ valley, which is detected as
the radiated four-wave mixing
5
signal field. Thus by measuring the four-wave mixing signal
while scanning the delay t2, we
directly probe the valley coherence dynamics and dephasing rate
γv.
Supplementary Note 2
Discussion of impurity or phonon scattering: Supplementary Figs.
3 and 4 schematically
illustrate the effect of impurity or phonon scattering on valley
dynamics, which include: (1)
change of the exciton center-of-mass-momentum, (2) suppression
of exciton phase coherence,
(3) scattering between bright and dark exciton states, and (4)
scattering between bright exciton
states in opposite valleys. Effect (1) is explicitly considered
in our model. Effect (2) is the pure
dephasing rate which has been measured by our experiment. Effect
(3) is part of the bright
exciton population decay process. Both effects (2) and (3) are
phenomenologically described in
the theoretical model. Process (4) is likely to be slow, since
it requires both electron and hole
spin flip as illustrated in Supplementary Fig. 4e.
An atomic-scale defect can localize excitons and lift their
valley degeneracy. Recent
experiments4, 5, including ours (Fig. 2a in the main text), have
found that impurity-bound exciton
states are energetically separated from the delocalized exciton
states, and the PL of the former
(latter) have low (high) degree of correlation between
excitation and emission circular
polarization. We conclude that the valley degree-of-freedom of
delocalized excitons is not
destroyed by atomic-scale defects. A full understanding of the
influence of atomic-scale defects
Supplementary Figure 3: Intra-valley scattering processes. (a)
The center-of-massmomentum and the phase of a bright exciton can be
changed by scattering. (b) A bright excitonis scattered to a dark
one due to the spin flip of the constituent electron. The gray dot
and emptycircle respectively represent electron and hole. The spin
splitting of conduction bands isexaggerated for illustration
purpose.
NATURE PHYSICS | www.nature.com/naturephysics 5
SUPPLEMENTARY INFORMATIONDOI: 10.1038/NPHYS3674
© 2016 Macmillan Publishers Limited. All rights reserved.
http://dx.doi.org/10.1038/nphys3674
-
6
on valley dynamics requires a detailed microscopic theory, which
we leave for future
investigation.
Theory of exciton valley dynamics: We start from the exciton
problem in monolayer TMDs. The
band extrema in valley K or K′ can be described by the massive
Dirac model, given by
�� = �Δ ������� + ���)
������� − ���) −Δ �, (3)
where the sign +/− corresponds to valley K/K′. To account for
the finite thickness of monolayer TMDs, we use the following form
of electron-hole interaction potential6:
�� = ����
���
�����, (4)
Supplementary Figure 4: Inter-valley scattering processes. In
(a), (b), and (c), either the constituent electron or hole in an
exciton is scattered to the opposite valley. In (d) and (e), both
the electron and hole in an exciton are scattered to the opposite
valley. In (a), (b), (c), and (d), the final exciton state is
optically dark. In (e), a bright exciton in valley Kis scattered to
a bright one in valley K'.
7
where � is the environment-dependent dielectric constant and ��
is a parameter that is inversely proportional to �.
The characteristic length and energy scale are respectively the
effective Bohr radius and
Rydberg energy, given by
��∗ = ��(ℏ��)�
��� , Ry∗ =��
����∗ . (5)
The effective fine structure can be defined as � =
��/(�ℏ��).
For monolayer WSe2, ℏ�� = 3.310Å × 1.19eV and Δ = 0.685eV, which
are taken from Ref. (7). We note that this value of Δ obtained from
DFT calculations underestimates the band gap, which, however,
should not noticeablely affect the quantities we are interested in
for this
study. We adjust the parameter �� so that the A-exciton binding
energy for � = 2.5, which corresponds to WSe2 lying on an SiO2
substrate and exposed to air, is 0.370 eV as measured in
the recent experiments.8 We therefore find �� = 22.02Å/� by
solving the Bethe-Salpeter equation for excitons in the massive
Dirac model.9
In monolayer TMDs, the center-of-mass motion of an exciton is
coupled to its valley
degree of freedom as described by the Hamiltonian
�� = �ℏ�� � �� � ����� � ������(2�)��� � ���(2�)��], (6)
where �� and ��,� are identity matrix and Pauli matrices in the
exciton valley space, respectively, ℏ�� is the excitation energy of
the A exciton, �� = ℏ���/2� is the kinetic energy of the
center-of-mass motion with � being the total mass of the exciton, �
and � are the magnitude and orientation angle of the center-of-mass
momentum �. The intra and inter-valley exchange interaction are
described by the ���� and ��,� terms, respectively. The coupling
constant �� is linear in the magnitude of �:
�� = Ry∗ �� �����∗ �(0)��(2���/Ry∗)�/�, (7)
where ���∗ �(0)�� is the probability that an electron and a hole
spatially overlap.10
6 NATURE PHYSICS | www.nature.com/naturephysics
SUPPLEMENTARY INFORMATION DOI: 10.1038/NPHYS3674
© 2016 Macmillan Publishers Limited. All rights reserved.
http://dx.doi.org/10.1038/nphys3674
-
6
on valley dynamics requires a detailed microscopic theory, which
we leave for future
investigation.
Theory of exciton valley dynamics: We start from the exciton
problem in monolayer TMDs. The
band extrema in valley K or K′ can be described by the massive
Dirac model, given by
�� = �Δ ������� + ���)
������� − ���) −Δ �, (3)
where the sign +/− corresponds to valley K/K′. To account for
the finite thickness of monolayer TMDs, we use the following form
of electron-hole interaction potential6:
�� = ����
���
�����, (4)
Supplementary Figure 4: Inter-valley scattering processes. In
(a), (b), and (c), either the constituent electron or hole in an
exciton is scattered to the opposite valley. In (d) and (e), both
the electron and hole in an exciton are scattered to the opposite
valley. In (a), (b), (c), and (d), the final exciton state is
optically dark. In (e), a bright exciton in valley Kis scattered to
a bright one in valley K'.
7
where � is the environment-dependent dielectric constant and ��
is a parameter that is inversely proportional to �.
The characteristic length and energy scale are respectively the
effective Bohr radius and
Rydberg energy, given by
��∗ = ��(ℏ��)�
��� , Ry∗ =��
����∗ . (5)
The effective fine structure can be defined as � =
��/(�ℏ��).
For monolayer WSe2, ℏ�� = 3.310Å × 1.19eV and Δ = 0.685eV, which
are taken from Ref. (7). We note that this value of Δ obtained from
DFT calculations underestimates the band gap, which, however,
should not noticeablely affect the quantities we are interested in
for this
study. We adjust the parameter �� so that the A-exciton binding
energy for � = 2.5, which corresponds to WSe2 lying on an SiO2
substrate and exposed to air, is 0.370 eV as measured in
the recent experiments.8 We therefore find �� = 22.02Å/� by
solving the Bethe-Salpeter equation for excitons in the massive
Dirac model.9
In monolayer TMDs, the center-of-mass motion of an exciton is
coupled to its valley
degree of freedom as described by the Hamiltonian
�� = �ℏ�� � �� � ����� � ������(2�)��� � ���(2�)��], (6)
where �� and ��,� are identity matrix and Pauli matrices in the
exciton valley space, respectively, ℏ�� is the excitation energy of
the A exciton, �� = ℏ���/2� is the kinetic energy of the
center-of-mass motion with � being the total mass of the exciton, �
and � are the magnitude and orientation angle of the center-of-mass
momentum �. The intra and inter-valley exchange interaction are
described by the ���� and ��,� terms, respectively. The coupling
constant �� is linear in the magnitude of �:
�� = Ry∗ �� �����∗ �(0)��(2���/Ry∗)�/�, (7)
where ���∗ �(0)�� is the probability that an electron and a hole
spatially overlap.10
NATURE PHYSICS | www.nature.com/naturephysics 7
SUPPLEMENTARY INFORMATIONDOI: 10.1038/NPHYS3674
© 2016 Macmillan Publishers Limited. All rights reserved.
http://dx.doi.org/10.1038/nphys3674
-
8
For WSe2 lying on sapphire substrate and exposed to air, the
dielectric constant � is approximately 5.5. The appropriate
parameter values are thenRy∗ = 75.66 meV, � = 0.66, and |��∗ �(0)|
= 0.88. Therefore, �� = 20��� × (2��/Ry∗)�/�.
The time evolution of excitons is governed by:
��(���)�� =
�ℏ ��(�� �)� ��� + ∑ ������ [�(��� �) − �(�� �)] −
�(���)� , (8)
where �(�� �) represents a 2 × 2 density matrix in the exciton
valley space at momentum � and time �. The diagonal terms of �
stand for the exciton population in valley K and K′, while the
off-diagonal terms describe the coherence between K and K′ exciton
states. The final term in
supplementary equation (8) phenomenologically captures the
effects of exciton recombination
(Γ�) and pure dephasing (��∗ ) on valley coherence. We allow the
decay rate ℏ/τ of diagonal and off-diagonal elements of the density
matrix � to differ, as explained below.
We decompose the density matrix into�(�� �) = ���(�� �)�� + �(��
�) � �. The equation of motion for the valley pseudospin vector �
makes the physical picture more revealing:
��(���)�� = �(�) × �(�� �) + ∑ ������ [�(��� �) − �(�� �)] −
�(���)� , (9)
where �(�) = 2��(���(2�)� ���(2�)�0)/ℏ. We approximate the rate
ℏ/�� for �� by the population decay rate Γ�, and the rate ℏ/�� for
���� by Γ� + 2γ�∗ . The underlying assumption for the 2γ�∗
contribution is that the pure dephasing processes for K and K′
excitons are uncorrelated.
We make the assumption of elastic momentum scattering, i.e.
����is nonzero only if |�| = |��| . Thus we can write
��(�� �� �)�� = �(�� �) × �(�� �� �) +
��� ∫ ������ �(� − ��)[�(�� ��� �) − �(���� �)] −
�(�����)� . (10)
To take advantage of the rotational symmetry, we make the
angular Fourier transformation:
�(�� �� �) = ∑ �(�)(�� �)����� , (11)
9
ℏ ��� ���(�)(�, �)��(��)(�, �)��(��)(�, �)
� =
���−ℏ��� −�2�� 0−��� − ℏ�� −
ℏ��
���0 �2�� − ℏ�� −
ℏ������
��(�)(�, �)��(��)(�, �)��(��)(�, �)
�, (12)
where �±(�) = ��(�) ± ���(�), and 1��� being the momentum
scattering rate:
���= ��� ∫ ��
��� �(�)�1 − ���(��)]. (13)
We assume δ-function impurity potentials. Then �(�) has no �
dependence. Therefore,1��� =1��� � 1���.
When � initially points to the x-direction, then Sx(Q, t)
averaged over momenta has the form:
���(�)� = 12� ∫�2��+(�, �) = ∫�����+(0)(�, �)]. (14)
The initial momentum dependence of �� is modeled by a Lorentzian
distribution given by
��(�, 0) = ��(0,0)�(1 + �������), (15) where �� is the kinetic
energy associated with the exciton center-of-mass motion. We
approximate the parameter � by the homogeneous linewidth, assuming
that finite-momentum states are able to radiate light only if their
energy is within the homogeneous linewidth. This
approximation also assumes that excitons outside the light-cone
can radiate due to interaction
with phonons and impurities. The time evolution of ���(�)� is
shown in Fig. 4 of the main text.
Valley depolarization: When � initially points to the
z-direction, the time dynamics are governed by:
ℏ ��� ���(�)(�, �)��(��)(�, �)��(��)(�, �)
� =
���−ℏ��� ��� −����2�� − ℏ�� −
ℏ��
0−�2�� 0 − ℏ�� −
ℏ������
��(�)(�, �)��(��)(�, �)��(��)(�, �)
�. (16)
We assume that the bright exciton population � also has a decay
rate 1���:
8 NATURE PHYSICS | www.nature.com/naturephysics
SUPPLEMENTARY INFORMATION DOI: 10.1038/NPHYS3674
© 2016 Macmillan Publishers Limited. All rights reserved.
http://dx.doi.org/10.1038/nphys3674
-
8
For WSe2 lying on sapphire substrate and exposed to air, the
dielectric constant � is approximately 5.5. The appropriate
parameter values are thenRy∗ = 75.66 meV, � = 0.66, and |��∗ �(0)|
= 0.88. Therefore, �� = 20��� × (2��/Ry∗)�/�.
The time evolution of excitons is governed by:
��(���)�� =
�ℏ ��(�� �)� ��� + ∑ ������ [�(��� �) − �(�� �)] −
�(���)� , (8)
where �(�� �) represents a 2 × 2 density matrix in the exciton
valley space at momentum � and time �. The diagonal terms of �
stand for the exciton population in valley K and K′, while the
off-diagonal terms describe the coherence between K and K′ exciton
states. The final term in
supplementary equation (8) phenomenologically captures the
effects of exciton recombination
(Γ�) and pure dephasing (��∗ ) on valley coherence. We allow the
decay rate ℏ/τ of diagonal and off-diagonal elements of the density
matrix � to differ, as explained below.
We decompose the density matrix into�(�� �) = ���(�� �)�� + �(��
�) � �. The equation of motion for the valley pseudospin vector �
makes the physical picture more revealing:
��(���)�� = �(�) × �(�� �) + ∑ ������ [�(��� �) − �(�� �)] −
�(���)� , (9)
where �(�) = 2��(���(2�)� ���(2�)�0)/ℏ. We approximate the rate
ℏ/�� for �� by the population decay rate Γ�, and the rate ℏ/�� for
���� by Γ� + 2γ�∗ . The underlying assumption for the 2γ�∗
contribution is that the pure dephasing processes for K and K′
excitons are uncorrelated.
We make the assumption of elastic momentum scattering, i.e.
����is nonzero only if |�| = |��| . Thus we can write
��(�� �� �)�� = �(�� �) × �(�� �� �) +
��� ∫ ������ �(� − ��)[�(�� ��� �) − �(���� �)] −
�(�����)� . (10)
To take advantage of the rotational symmetry, we make the
angular Fourier transformation:
�(�� �� �) = ∑ �(�)(�� �)����� , (11)
9
ℏ ��� ���(�)(�, �)��(��)(�, �)��(��)(�, �)
� =
���−ℏ��� −�2�� 0−��� − ℏ�� −
ℏ��
���0 �2�� − ℏ�� −
ℏ������
��(�)(�, �)��(��)(�, �)��(��)(�, �)
�, (12)
where �±(�) = ��(�) ± ���(�), and 1��� being the momentum
scattering rate:
���= ��� ∫ ��
��� �(�)�1 − ���(��)]. (13)
We assume δ-function impurity potentials. Then �(�) has no �
dependence. Therefore,1��� =1��� � 1���.
When � initially points to the x-direction, then Sx(Q, t)
averaged over momenta has the form:
���(�)� = 12� ∫�2��+(�, �) = ∫�����+(0)(�, �)]. (14)
The initial momentum dependence of �� is modeled by a Lorentzian
distribution given by
��(�, 0) = ��(0,0)�(1 + �������), (15) where �� is the kinetic
energy associated with the exciton center-of-mass motion. We
approximate the parameter � by the homogeneous linewidth, assuming
that finite-momentum states are able to radiate light only if their
energy is within the homogeneous linewidth. This
approximation also assumes that excitons outside the light-cone
can radiate due to interaction
with phonons and impurities. The time evolution of ���(�)� is
shown in Fig. 4 of the main text.
Valley depolarization: When � initially points to the
z-direction, the time dynamics are governed by:
ℏ ��� ���(�)(�, �)��(��)(�, �)��(��)(�, �)
� =
���−ℏ��� ��� −����2�� − ℏ�� −
ℏ��
0−�2�� 0 − ℏ�� −
ℏ������
��(�)(�, �)��(��)(�, �)��(��)(�, �)
�. (16)
We assume that the bright exciton population � also has a decay
rate 1���:
NATURE PHYSICS | www.nature.com/naturephysics 9
SUPPLEMENTARY INFORMATIONDOI: 10.1038/NPHYS3674
© 2016 Macmillan Publishers Limited. All rights reserved.
http://dx.doi.org/10.1038/nphys3674
-
10
����� � �����. (17)
The valley polarization can be quantified by the average of
Sz(Q, t) over momenta:
���(�)� � 12� ��2����(�� �) � �������(0)(�� �)]. (18)
If we denote the decay rate of ���(�)� by 1���∗, the degree of
circularly polarization �� is ��∗���. Using the same parameter
value for the dynamics of valley coherence ���(�)�, we calculated
the time dynamics of valley polarization ���(�)�, which is shown in
supplementary Fig. 5. �� is approximately 50% for ����= 10 meV and
57% for ����= 20 meV. These results demonstrate that ~ 100 fs
exciton valley decoherence time is fully compatible with a higher
degree of valley
polarization. This value of valley polarization in the resonant
nonlinear experiments may
increase further by reducing the parameter � in the momentum
distribution in supplementary equation (15), and adjusting other
parameters accordingly. We emphasize that the degree of
valley polarization in resonant nonlinear experiments cannot be
directly compared to that
observed in non-resonant PL experiments. It is known that the
degree of valley polarization
observed in PL strongly depends on the excitation wavelength. In
addition, the
photoluminescence signal includes contributions from incoherent
exciton population dynamics,
which can be sensitive to scattering between optically bright
and dark excitons. To achieve a
better agreement between theoretical and experimental values of
�� , the time dynamics of dark
Supplementary Figure 5: Valley depolarization. Initially,
excitons in valley K is generated bycircularly polarized light. The
solid lines represent the time evolution of Sz for
differentmomentum scattering rate ����, and the dashed line labels
the time evolution of excitonpopulation N.
11
excitons needs to be added to the theory.
Supplementary Information References
1. Huang, J.-K. et al. Large-area synthesis of highly
crystalline WSe2 monolayers and
device applications. ACS Nano 8, 923–930 (2013).
2. Jones, A. M. et al. Optical generation of excitonic valley
coherence in monolayer WSe2.
Nat. Nanotech. 8, 634–638 (2013).
3. Scully, M.O. and Zubairy, M.S. Quantum Optics (Cambridge
University Press, 1997).
4. He, Y.-M. et al. Single quantum emitters in monolayer
semiconductors. Nat. Nanotech.
10, 497-502 (2015).
5. Srivastava, A. et al. Optically active quantum dots in
monolayer WSe2. Nat. Nanotech.
10, 491-496 (2015).
6. Berkelbach, T.C., Hybertsen, M.S., and Reichman, D.R. Theory
of neutral and charged
excitons in monolayer transition metal dichalcogenides. Phys.
Rev. B 88, 045318 (2013).
7. Xiao, D., Liu, G.-B., Feng, W., Xu, X., and Yao, W. Coupled
spin and valley physics in
monolayers of MoS2 and other group-VI dichalcogenides. Phys.
Rev. Lett. 108, 196802
(2012).
8. He, K., Kumar, N., Zhao, L., Wang, Z., Mak, K.F., Zhao, H.,
and Shan, J. Tightly bound
excitons in monolayer WSe2. Phys. Rev. Lett. 113, 026803
(2014).
9. Wu, F., Qu, F., and MacDonald, A. H. Exciton band structure
in monolayer MoS2. Phys.
Rev. B 91, 075310 (2015).
10. Yu, H.-Y., Liu, G.-B., Gong, P., Xu, X.-D., and Yao, W.
Dirac cones and Dirac saddle
points of bright excitons in monolayer transition metal
dichalcogenides. Nat. Commun. 5,
3876 (2014).
10 NATURE PHYSICS | www.nature.com/naturephysics
SUPPLEMENTARY INFORMATION DOI: 10.1038/NPHYS3674
© 2016 Macmillan Publishers Limited. All rights reserved.
http://dx.doi.org/10.1038/nphys3674
-
10
����� � �����. (17)
The valley polarization can be quantified by the average of
Sz(Q, t) over momenta:
���(�)� � 12� ��2����(�� �) � �������(0)(�� �)]. (18)
If we denote the decay rate of ���(�)� by 1���∗, the degree of
circularly polarization �� is ��∗���. Using the same parameter
value for the dynamics of valley coherence ���(�)�, we calculated
the time dynamics of valley polarization ���(�)�, which is shown in
supplementary Fig. 5. �� is approximately 50% for ����= 10 meV and
57% for ����= 20 meV. These results demonstrate that ~ 100 fs
exciton valley decoherence time is fully compatible with a higher
degree of valley
polarization. This value of valley polarization in the resonant
nonlinear experiments may
increase further by reducing the parameter � in the momentum
distribution in supplementary equation (15), and adjusting other
parameters accordingly. We emphasize that the degree of
valley polarization in resonant nonlinear experiments cannot be
directly compared to that
observed in non-resonant PL experiments. It is known that the
degree of valley polarization
observed in PL strongly depends on the excitation wavelength. In
addition, the
photoluminescence signal includes contributions from incoherent
exciton population dynamics,
which can be sensitive to scattering between optically bright
and dark excitons. To achieve a
better agreement between theoretical and experimental values of
�� , the time dynamics of dark
Supplementary Figure 5: Valley depolarization. Initially,
excitons in valley K is generated bycircularly polarized light. The
solid lines represent the time evolution of Sz for
differentmomentum scattering rate ����, and the dashed line labels
the time evolution of excitonpopulation N.
11
excitons needs to be added to the theory.
Supplementary Information References
1. Huang, J.-K. et al. Large-area synthesis of highly
crystalline WSe2 monolayers and
device applications. ACS Nano 8, 923–930 (2013).
2. Jones, A. M. et al. Optical generation of excitonic valley
coherence in monolayer WSe2.
Nat. Nanotech. 8, 634–638 (2013).
3. Scully, M.O. and Zubairy, M.S. Quantum Optics (Cambridge
University Press, 1997).
4. He, Y.-M. et al. Single quantum emitters in monolayer
semiconductors. Nat. Nanotech.
10, 497-502 (2015).
5. Srivastava, A. et al. Optically active quantum dots in
monolayer WSe2. Nat. Nanotech.
10, 491-496 (2015).
6. Berkelbach, T.C., Hybertsen, M.S., and Reichman, D.R. Theory
of neutral and charged
excitons in monolayer transition metal dichalcogenides. Phys.
Rev. B 88, 045318 (2013).
7. Xiao, D., Liu, G.-B., Feng, W., Xu, X., and Yao, W. Coupled
spin and valley physics in
monolayers of MoS2 and other group-VI dichalcogenides. Phys.
Rev. Lett. 108, 196802
(2012).
8. He, K., Kumar, N., Zhao, L., Wang, Z., Mak, K.F., Zhao, H.,
and Shan, J. Tightly bound
excitons in monolayer WSe2. Phys. Rev. Lett. 113, 026803
(2014).
9. Wu, F., Qu, F., and MacDonald, A. H. Exciton band structure
in monolayer MoS2. Phys.
Rev. B 91, 075310 (2015).
10. Yu, H.-Y., Liu, G.-B., Gong, P., Xu, X.-D., and Yao, W.
Dirac cones and Dirac saddle
points of bright excitons in monolayer transition metal
dichalcogenides. Nat. Commun. 5,
3876 (2014).
NATURE PHYSICS | www.nature.com/naturephysics 11
SUPPLEMENTARY INFORMATIONDOI: 10.1038/NPHYS3674
© 2016 Macmillan Publishers Limited. All rights reserved.
http://dx.doi.org/10.1038/nphys3674